The most useful video I have watched in my whole DIY audio career (+10 years). Thank you, Sir. I'm usualy modest and reserved (especially towards audiophile lingo with subjective impressions) but this tool just simply transformed my DIY full range speakers. So grateful.
You can combine as many filters as you want. You will need to vector multiply each filter with the previous one. You can use "generate response from filter" in the EQ window of REW to create a filter from your EQ bands ready to be used as a convolution file. The suggestion depends on where and how much you will be EQ'ing. The inversion shoudl normally take care of eveything that needs to be corrected.
@@ocaudiophile When doing vector avg of multiple readings, lets say 8 readings, the spl level after 200Hz goes down for 7 db. Won't be better to perform RMS + Phase Avg? I think we get better results for both SPL and phase corrections. Do you agree? Thks!
@@ocaudiophile yes I have, but when you average many measures, it is normal this behaviour in rew, I understand because of the entropy of phases at higher frequencies. It is documented
You were talking about a dip at 200Hz though. Depends on which response you're aligning the others to. If measurements around a central measurement are cc aligned to the center vector average will show the correct frequency and phase response in the measured area and there will not be major dips anywhere. I regularly average 8 Audyssey measurements that way.
Thank you so much for this quality content. I’ve been searching for more in depth videos on REW. Seems like the algorithms prioritize the most basic stuff. More please!! New subscriber here.
I just did an experiment with this convolution with inversion method, near field on a CD horn to see if it would create a phase perfect passband including an octave into a 12 dB / Octave high pass filter. It worked pretty much perfectly. The phase and frequency response measured near perfectly flat! I spent a little time listening. It's hard to say without more listening if there are any audible problems, but initial impression is very positive. I'd say more smooth and natural than before. I think I might have to do all th drivers this way.
Precisely central (09:45). Unlike how your narration emanates from the left side of the stereo field exclusively. 😄No big deal. Thank you so much for sharing your time and expertise with us. Much appreciated.
I've actually used an Audyssey mic to record the videos which is mono. I am not a content creator or have any monetary concerns for the channel so the videos are unedited. It's just information sharing.
@@ocaudiophile And great information it is. I have moved on to your "phase shift" tutorial and your voice comes to me in full stereo. Keep up the great work. Have a wonderful day.
@@ocaudiophile Thanks, I like both but your vids are almost explanatory enough esp in the context of OCD :-) Ive been doing acoustics prob 30 years and just lately having some real breakthroughs in room stuff that I havent seen anywhere else...but not had time to delve into the 'cream' region of final tweaking which is exciting to finally get there...although an eternity still to go ;-)
I want to thank this man for investing so much time in this tutorial and sharing his knowledge. It took me some hours to reread, measure and correct but it was 100% worth it! I'm blown away with the results, my speakers sound so much better as I had a lot of troubles with the lower frequencies in my room.
Instead of dividing L and R by Target LR and then inverting, I did it vice versa - Target LR (A) divided by L and R (B) respectively for each channel. Did this due to some changes in the calculation options for the inverse action. Seemed to reach the same result, just one step less. Otherwise, a great video for dummies - much appreciated!
Correct, trace arithmetic division options have improved a lot in REW since that video. You've done the right thing. You may wanna watch "mastering digital room correction" video for updated and a bit better sounding tech.
Thank you for the great tutorial, OCA! yours is the only one that I've found; to correct the phase of the speakers. When I listening my Left seems to be off compare to the Right. This fixed it!
@@ocaudiophile I do. One hiccup with this method is that the convolution file created ends up being really long. On my desktop, this resulted in an audio latency of 0.341 seconds. I took the wav file into Izotope RX and trimmed the empty space from beginning and end, which brought my latency down to 3.1ms. Equalizer APO seems to indicate the same corrections. Now that I'm running convolution, my right speaker is slightly louder than my left. When I took the measurements, the calibrated mic was on a tripod in the center of my main listening position. Did I miss a step somewhere?
@@wyleyrabbit There have been some revisions to A/B trace arithmetic recently in REW. Depending on the version you're using you might need to use target level at auto and adjust left and right 1/A levels. There're instructions at the end of the updated pdf guide.
@@ocaudiophile I downloaded and followed along with the PDF. I'm using 5.20.13. The impulse files end up with what could perhaps be best described as having an echo. I opened the file in RX and there was most definitely more than one impulse, so I manually removed the echo part of the impulse wav file. It sounds great, but I'm not sure how the echo is getting in there. Any ideas about what I'm doing wrong?
@@wyleyrabbit You need to use an early access version of REW (5.20.14 onwards) for best results. There should not be any echo with this method but I cannot gurantee that for older REW versions.
Thanks a bunch, looks like a method that can yield great results. I will give absolutely be testing this method. I will say, your video only has audio on the left channel, which on an audio engineering video is a big crime, the punishment being you can only own bose speakers. Please surrender your good speakers to the audio police.
Thanks for your kindness shareing.I follow the tutorial and complete my second room correction attemp.The result is impressive especially compared it with previous autoEQ,though my secondhand speakers are not in their best condition, L&R FR dosen't match quite well(and they are bass limited model,so I change some settings).After inversion,I can easily tell the different when gain at 16k,but the day before correction,gain at 63 or 16k seems nothing change in my system.
I love this channel. Since i came to know about your channel and videos, I am learning REW and calibration. Even I calibrated my Yamaha entry level AVR which has no auto room calibration. I calibrated and measured. It improved sound quality a lot.
Hello hi. First things first, thank you for your tutorials... Anybody tried to remeasured the sweep signal with the FIR correction? All you need is two convolution engines to apply the first FIR and then the second. I know it sounds a little bit crazy, but it really works.
I have to be honest, this took me a while 😂. But first impression is excellent! I tried the regular EQ filters before (with the Harman Target curve). But I think the phase part in your tutorial played an important part in why it sounds a lot better now. SPL and Time were allready almost perfect here. Thank you for taking the time! to create this video 🎉
This was very helpful! Your explanation brought clarity to my weak tuning understanding!! I now better understand the reason for using certain values in the correction process. Now don’t get me wrong but I will need to go back to study this many times. My prior efforts always resulted in huge cuts in amplitude (make it flat approach) I know that you will get great support to this video and you will continue to educate us. I to use Roon and solving timing issues in different tools in Roon is going to be very helpful. We’ll done
Incredibly useful video! Have you looked at how this functions on the latest V5.20.14 early access build? Seems like the Trace arithmetics have changed. Need to use |A| / |B| instead of A / B and you no longer have the ability to set Target level on 1 / A causing the end result to be offset from the target line.
@@ocaudiophile man, i used your tutorial to create the shebangs including the rephase stuff for my headphones, christ all mighty what a diffrence.... not to mention now i could use that convolution file as reffrence no matter what song i play or movie and create an auto eq with tdr nova eqs smart ops feature... genius stuff :D
Thank you for all your helpful videos. but which method should be used to measure your living room cinema? (5.1.4) MultiEq app available miniDsp2x4HD is available for the subwoofer. 🤙🏼
Hello. I am grateful to you for introducing an amazing method. Not until I saw your video did I realize the computer could generate great music. I am Japananese. I'm sorry that I don't have much vocabulaly to express my thanks in English. I have never used REW and I don't have any knowledge of accostics. But thnaks to your video and PDF, I couild make my own convolution files. I have some questions. Sorry for my poor English and no knowledge. 1 Now I use REW v5.20.11 according to your PDF (Umike1 asio-4all). When I saw your video , I used v5.20.13 and measured IR for the first time in my life. When I used v5.20.13, most of the first dips were 100%, the next peaks were 70-90%. So I thought the first dips were the immdiate impulse. But with v5.20.11, most of the first dips are 80-90% and the next peaks are 100%. So which should I think the immediate impulses were? I think I'd like to use v5.30.9. I have 3 questions. 2-1 As you have written in Update of your PDF, Is it OK that I skip Correction Process 5-8 and use only "Align SPL" command to left and right speakeer measurements? 2-2 With v5.30.9, in correction process 20, there is no button of "Apply windows". What should I do? 2-3 With v5.30.9, should I use |A|*|B| instead of A*B? I'm very sorry to trouble you.
Download the latest REW Beta from this link: www.avnirvana.com/threads/rew-api-beta-releases.12981/ Trace Arithmetic and especially inversion operations have been highly optimized as of yesterday. Start using this version and let me know, if you run into problems.
So to confirm, if I'd like to combine this with the xo phase correction of the other tutorial, I just TA multiply the inverted_MP results with a phase correction impulse from rephase?
Hello. Awesome video, thanks! I would like to test this principle with my LXmini speakers (no subs), but have not figured out yet where to upload those wav correction files. My source is Apple Music on a Macbook Pro, via USB digital to external DAC. Is there a recommended plug-in or free (cheap) software for MacOS/Apple Music?
@@ocaudiophile thanks, I installed Audio Hijack (trial for now), loaded my convolution files to test. There is an improvement in sound when listening, but when measured the graphs are not that different without FIR and with FIR filter applied. And definitely not that close to the implied or calculated graphs (for example clarity, impulse) for "corrected" responses that are shown in REW during the making of the filters. Is that normal?
@@mantasj6678 I'm not familiar with audio hijack, why don't you use equaliser Apo? Frequency response shoukd be very similar, clarity may change significantly depending on measurement distortion.
@@ocaudiophile Equaliser APO is only for Windows, while I am using MacBook with an Apple M1 chip :/ I could share my REW file from which I made the filter, and the measurement file after I tested with filter and without, if you don't mind looking at them. It's ok if you don't want to bother, I can hear the improvement and will continue listening anyway :)
Ciao, una curiosità. se ti mando la risposta in frequenza delle mie casse monitor sotto forma di file Wave puoi farmi avere i filtri generati da Reweq per linearizzare i miei monitor?
Thank you very much. A great work. Is there a more recently and usefull version of REW for this tutorial? Is it necessary a powerfull PC or could be sufficient also an old asus?
@@ocaudiophile Maybe Audiovero Acourate. With the trial version what can I do? Is there a tutorial for using Acourate for measuring distance and exact position of LP and Speakers? Thank you
@@Angelos58 Acourate has a tool that helps centre the mic position between left and right speakers and it's available in the trial version. You can also measure each speaker using one of the speakers as acoustic timing reference in REW and adjust mic position accurately by checking their impulse peaks in Overlays/Impulse graph.
Hi, thanks for. your video on how to use Convolution. It works like a charm, brings out so much details that I have not noticed before. There is only one problem I encountered is the latency of the convolution, which is a bit large. I have read through the comments and saw this: At the very end of "Excess Phase Inversion" video, I've explained a method for trimming down convolution files... I would like to see if I can trim down the convolution file, but the video seems to be nowhere? Which video is that? Or can you explain in simple procedures that I can try? Thanks a lot.
@@ocaudiophile Thank you for your quick response. I have watched the video and followed the step and cut quit a bit of the windows. It works better now. One thing I have noticed that you are using a newer version of REW 5.20.14 which is not available yet. And the difference is that my 5.20.13 version do not have the "Trim IR to windows" button. But I figured out if I export the IR wav file with option "Apply IR Window before Export", that may do the job, is it right?
That will accelerate the loading time of the filter but will not decrease the number of taps. Google "REW Early Access version" and you can download the newest one.@@trumankong7053
The best video I have seen this year! Two things as a feedback though - would be great for follow-up video. 1) You might have made a mistake at 9 minutes in your video when SPL matching speakers. If you're focused on low frequencies 50-60Hz, I think you might calculate room modes into the SPL average! Not sure ROON is capable enough to remove them completely - or is it? In my L-shaped room with pillar behind one of the speakers - the bass frequency response is *completely* different (in bass department) left vs right. To get around this, I would not calculate anything below 250Hz - or am I missing something? 2) I think I would introduce less time smearing, phase issues and precision going through a simple EQ calculated for minimum phase - correcting 2 offending frequencies in my case. I wouldn't like to touch *every* single frequency during DSP calculation if my speakers are already playing fine - and I just have issues at 10KHz region and one in bass. I would welcome clarification video, why in your oppinion EQ (correcting just 2 bands) will sound worse and introduce more problems that adjusting entire 20Hz-20KHz band, even in places where it doesn't need adjusting. People, I really do not need to touch 1-2db differences, can live with that! If I stand up or move away from speakers, differences are far greater. 3) Was vector RMS average fixed in latest ROON, or you still advise to go with generic one? 4) Where can I find latest Harman curve? Is your description still valid? The idea behind would but require some clarification - as if I listen at 83db where Harman curve is not required, why would I still need to apply it? 5) I would welcome phase follow up video or and updated one - as again - this is best video I have seen this year or last year. Any time I touch equalizer on my Mojo2, *even by 1db on single band* I feel it destroyed the precision of 3D space I was hearing previously. So Rob Watts loose-less EQ sounds like a lie to me. I would need to try approach in your video. Thank you very much - you're beyond knowledgeable and I love Focal Sopra No2 speakers very much.
L-shaped room is a challenge. You've to aim for max symmetry at least with the remaining 2 near walls and deal with the major assymetry of the 3rd by DSP. RMS + phase average gives more accurate representation of the high frequency response when averaging multiple mic positions but for bass response averaging, time aligned vector averaging rules. The method in the video is new and very powerful. You may need to use minimum phase versions of 1/A divisions for limited FIR taps as they "might" cause pre-echo otherwise. I prefer Dr Toole curve, find Harman bass a bit too much and not inline with my speakers natural response.
My friend thank you for all the knowledge you are sharing. I have learned so many things! However I am still a noob and I need your advice about which one of your videos is the best for my case... I have a 2.1 system in a small, asymmetric room for music listening. PC is the only source and it's connected to a DAC/preamp through USB. All three speakers are active and they are connected through XLR (DAC > Sub > Monitors). I followed the first 10 pages on the attached pdf to the letter and I'm very satisfied with the results in EQ APO. But although the room is partially treated, I have a bid dip in 1.2khz area and the house curve doesn't really help because the SPL graph is so uneven around that area. Should I continue the pdf guide and use rePhase? I'm asking because I read in one of your comments that it's not necessary in an active speaker setup. You have uploaded so many wonderful videos after this one and I would be grateful if you could point me to the right direction, as the next step...
Thank you. Your set up is capable of sophisticated high tap FIR filters so most problems can be fixed without side effects although asymmetric room can be a bit of a problem. It's hard to tell the problem around 1200 hz. Can you share your measurements? You can paste Google drive links here.
@@ocaudiophile Yes of course, thanks for helping! This is the link of "Step 2", according to the pdf. Inside the folder you'll also see 4 photos of my space... drive.google.com/drive/folders/1UjOMFWnFZ3aRuZ-Uo6ZY2Tel0U13ZUGk?usp=share_link
@@cnkosm6536 The dip around 1200Hz is caused by a reflection surface around 7-8cm away from your tweeters (I don't know your speaker crossover points and assume it's your tweeters producing 1200Hz) and a tiny bit longer distance for your right speaker. Sound waves bounce back from that surface, travels an extra 14.68cm distance which takes it 0.1468/343=0.000428 seconds and reach you and interact with the original sound wave causing a peak at 1/0.000428=2336hz and a dip at half of that. I could fix most of it with excess phase inversion. I created a stereo convolution .wav file in the link below also included the mdat: drive.google.com/file/d/1EtCL2GrK9S3E4WrKx2BPVf1L8zHSlNmH/view?usp=share_link I applied a slighly better sounding EQ technique and combined phase inversion with that but you can also add the phase filters to your own inversion filters if you so wish, they are all in the mdat. I will be waiting to hear your feedback.
@@cnkosm6536 The dip is caused by an extra 15cm distance traveled by the sound wave relative to the direct sound. Check for surfaces from which when sound will need to take this much extra length when it bounces and arrives at your ears . The new filters are in the link along with the mdat file which contains eveything. Filter1 is no pre-echo guaranteed but Filter2 additionally fixes some of the dip around the 200Hz area but could cause ringing. If both of them cause ringing then the problem is in your right speaker which has some pre-ringing even in the default measurement. Let me know. drive.google.com/file/d/1KB2G_aaUM3xfu9AYmy7sqF4el-3ms8mt/view?usp=sharing PS "Beat It" -Michael Jackson will show you instantly and clearly if there's any pre-ringing in the system!
Great video. I followed these instructions very carefully, but used the alternate room curve in lieu of the Harman, which results in a flat downward sloping target. My L, R, and LR plots looks as expected, but my inverted plots are very "squared off", with a hard stop at 0 SPL. Not sure what I may have done wrong.
@@ocaudiophile A couple of times during the final maths calculations, I received Java errors which crashed REW. I WAS able to produce the files without crashing, but I wonder if there's a processing issue that's creating faulty files without an outright crash.
I have followed your instructions and have achieved pretty good results, better than using PEQ. There is one item I am not sure about, which is the regularisation setting when doing 1/A conversion. What does the regularisation do? I used 8, and the result is boost to overall volume, so I tired a few different values and if I set to 15, the overall level matches the original signal. So, I am a bit confused.
In the book Accurate Sound Reproduction Barnett, Mitch. Accurate Sound Reproduction Using DSP (p. 6) I read that is not convenient to using a USB mic. I purchased a Minidsd UMIK-1 Angelo Scozzarella you advised .
USB mics have clocking issues but REW's clock adjustment with acoustic reference feature has greatly solved that problem. They have advantages like already calibrated SPL level and plug and play usability.
Hi Serkan, many thanks for this video! What would be your prefered way adjusting Stereospeaker with active xover via minidsp flex eight? EQ in REW to Target curve and then rephase?
You do not need crossover phase correction when using digital active crossovers but you may still benefit from speaker box (port) phase correction in rePhase if you know the port frequency of your speakers. You can do a near field port measurement with REW to learn that. It should be a phase linearization of either "vented low Q" or "vented std Q" at the port frequency. Export it with less than 6128 taps as bin file from rePhase and upload it in MiniDSP. Inversion method only produces FIR taps and you cannot easily limit their total tap count in REW. So for frequency correction, use REW Auto EQ filters on the target curve and export them as MiniDSP biquad filters. For best results different frequency bands should be equalzied with different settings. I am very close to launching a new video showing all that. Until then you can play around a little and compare your results.
Thank you very much for the great video. I am looking forward to trying this on my system but I have a question - I have two subs in my setup one for left and right that are also running off same signal as the left and right speaker. I assume the phase alignment will be a key issue in my particular setup. Do I treat the left a right channels like a 4 way speaker with the crossover from the mains at the sub crossover - (in my setup that is 100hz)?
Time align subs between each other and also with the speakers. Try to use speaker's spec bass roll off frequencies as your XO frequencies, Once all is set, you can apply inversion methods to the combined speaker/sub system as one.
@@ocaudiophile Following your instructions has elevated my setup beyond my expectations!! I don't think I could have ever achieved this level of sound spending thousands more on gear. I am running Audirvana with Hang Loose Convolver through RPi4 to Benchmark DAC3L, feeding Benchmark AHB2 AMP to Revel F226Be Speakers and Rythmik E15s. I tried the Roon + HQPlayer combo and Audirvana + HLConvolver spanks it IMO. Highly recommend giving Audirvana and Hang Loose Convolver a trial run - would love to hear your opinion.
Thank you for making this video/pdf and for your attentiveness to your viewers! I have two questions if you don't mind: 1) I like my results, but sometimes it feels too smooth and clinical, and sometimes I get a strange echo effect, most notable with spoken word or internet radio (try "Room 29" by Jarvis Cocker, where his voice just sounds really echo-y and bizarre with some attempts at this method). Could this be due to a recording issue, a mistake during the REW process, or an irreconcilable issue between my two speakers? 2) As mentioned, the results are sometimes too smooth, where deeper male voices in particular (like Jarvis's above) feel robbed of their character and depth. Is there a way to to do this where the peaks in the frequency response are just reduced, rather than are chopped off by the Harmon curve? Right now I've settled on a correction that only works up to 200Hz, but I'd like a solution that covers the entire range. Thank you in advance!
You can try experimenting with the regularization setting during division trace arithmetic operation. It determines the maximum boost that will be applied with a filter i.e 8% is for 5dB maximum boost, 25% is for 0dB maximum boost. Usually, less or no boost filters will sound less "clinical" in your terms. Also check for clipping and apply a headroom if necessary. The echo you sometimes hear could be related to clipping.
@@ocaudiophile Thank you for your advice and quick reply. I took a few more stabs at this this morning, and really found that the process is quite sensitive to the configuration of the target curve in EQ, and the regularization settings during the 1/A calculation: I was able to achieve a better result once I tinkered with the Target db of the curve, and used a regularization of 4 percent rather than 8 during 1/A. I messed with headroom, but I don't think this was the source of my original echo issue. Male voices, cellos, etc, still feel a little thin, with just a HAIR of that echo effect remaining, but bass and high frequencies sound really smooth and clear, with less dramatic oscillation between bass peaks and nulls in the room. I'm going to see how much more refinement I can bring to the midrange. Thanks again for all your help and making this process accessible to laymen.
Thanks you! your video's are very helpful! before i dive into this, is this video good for room correction for obtaining flat frequency response for mixing and mastering? will i gain anything sonarworks wont provide regarding hearing music and not the room?
This is an efficient and easy way to correct the room and sound will not end up boxy or throttled inherent to auto calibration systems. However, REW only produces the filters and it's not a convolution engine that applies these filters to your mixes on the go. You can use Equalizer APO for that which is another free tool.
Hi, thank you for the tutorial, someone suggested me to follow it on AV Nirvana and I will, although I'm just a guy who wants to listen to good music with 0 audio knowledge, so in the current state I'm sure I will mess up this for two reasons, the first would be that I don't know what I am to do with that correction file inside Equalizer APO, can you explain that to me or point to a video that shows it? I assume I need to click on "Convolution with impulse response" and load that file, there is anything else that needs to be done? The second point of failure for me is the fact my setup is a PC with 2 speakers on the sides and a subwoofer centered on the back 2 meters away (Kali LP6-v2 and SVS SB-1000 Pro) therefore I don't have the slightest idea on what I should do in order to include my subwoofer in that correction and time alignment, can you make a video for that as well? (or point me to a video that explains it) ^_^' Thank you !
I wouldn't worry about Equalizer APO, it is quite simple to use and intuitive after couple of minutes of use. Here's an old tutorial on it: th-cam.com/video/Cyb1r5E9K3I/w-d-xo.html Subwoofer is also not a problem, you should measure your system with everything on and connected and create correction files for that. How you connected your sub and how you cross it over doesn't really matter as long as you measure the system in REW as you listen to it.
thank you this is one of the best videos i have seen! But as far as i understand the convolution files are not time and amplitude aligned and I have to set it in the convolution engine (for example in Jriver).... is there a way to write the time and amplitude alignment data directly into the convolution wave file?
You don't need to do anything. Just direct JRiver config.txt file to the .wav file directory. For a REW created wave file at 48,000 sampling rate for example something like: 48000 2 2 0 0 0 0 0 convolution48.wav 0 0.0 0.0 convolution48.wav 1 1.0 1.0 If you mean windowing and normalization of the convolution impulses, you can select these options while exporting impulse response from REW. They don't have audible effects but exporting IR windows may increase speed of convolution with some engines.
How real these small picks and dips? I mean, whether they will stay on their places if we take another measurement(s) from slightly different mic position(s). I suspect most of them are kind of random, which means that we correcting nothing. This seems to me similar to "over-learning" issue in AI.
Measurements at different places will result in very similar response up to around 2,000Hz. Higher frequencies will differ but dips and peaks will be similar relative to each other.
Hi there, Is this procedure only for a stereo situation ? I have a 7.2 system and are in the process of turning the Subs as per your optimal subwoofer alignment. Once I have finished that I would like to do my mains and was wondering if this would be the best procedure for the other 7 speakers (LCR, Sides & Rears) All my source material is played back via Jriver 30
These are FIR filters and are more powerful than any other so you can of course correct each speaker in a surround system with it. I am sure you can use convolution files with surrounds as well with Jriver with the right config files although I've not done myself.
O.K, I have finally managed to Optimize my dual subs as per your Optimal Sub Alignment with sub correlation instructions and are very happy with the outcome. I am now following your "REW (Room EQ Wizard) top tricks: Convolution with Inversion (no EQ filters just FIR)" for my 7.2 Home theater setup. I hope I am heading in the right direction with this, I have completed the Impulse alignment for all speakers including the Sub. Not sure if I needed to do the Sub ? but they are all adjusted and set at zero Would this normally be the next step to complete in turning a system having completed the Subs Thanks in advance Neil
@@neilpage1185 Once you phase align dual subs, you should treat them as a single subwoofer and yes, the impulse peak of that single sub should be at time zero along with all the other speakers. To apply the inversion correction technique, you need a processor with convolution capabilities. No receivers can do that as far as I know. If you are not using a FIR filter capable solution I suggest you follow the method in "Manual DOlby Atmos Calibration" videos.
Thanks for replying. One more question. In SPL alignment, the info window shows align offset -1.18db for L and +1.18db for R. Are those the numbers I put in for gain in Roon Speaker setup or do I reverse the signs?
Hello, this is a great video thank you. I followed it carefully and did a test sweep once I had applied the filter and it did a great job (particularly at the lower end of the frequency spectrum). Where the filter inverse curve is above 0 db it goes up to just under 5db which I assume is the amount by which the filter will boost the response at the relevant frequency? I have read a lot about limiting the amount by which a filter boosts (to avoid driver damage). If I am right when I say a 5dB point on the filter means a 5db, is that amount of boost OK?
+5dB is fine below room transient frequency which is around 200Hz. You can adjust boost limit with "regularization" during vector division. 8% allows up to 5dB, 25% allows 0dB (no boost)
Thank you, all clear. I also meant to ask about your house curve tip - is that instead of the Harmon curve or is it in addition to the Harmon curve (i.e. you have both)?
Recently, I prefer to use my speakers' anechoic response as the target curve. Although this will lack a bit in the bass department, it yields the clearest results. Bass you will get above the limits of your speakers will make use of room reflections which are by definition delayed waves and causes the bass waves to rain into following higher tones. Still, a lot of people love deep bass and don't mind the group delays introduced. Harmon has deeper bass then Dr Toole's curve.
Great tutorial! Thanks! Can you please advise how in REW do I cut only selected frequency range for the filter once it is ready e.g. I want to only affect 20 - 200hz
Do the A/B division as shown in the method with 0%regularisation adn no frequency limits ticked. The when you're applying "1/A" to the result of the previous division "A over B" with 8% regularisation, tick the "Lower" and "Upper" frequency limits and enter 20 and 200Hz. Leave Target level at Auto.
Would you recommend doing this for home theater. If you ran Dirac after this process, would it ruin everything you just did? Is there an 11 channel dsp you recommend?
The process is also good for surround systems of course and can be done with JRiver if you use a HTPC for example. Trinnov in fact is a16 channel HTPC. Dirac is doing something quite similar to this but user satisfaction is varied. . I never used one so cannnot recommend but Minidsp has multichannel FIR filter capable units.
Thanks for this approach to DSP. In checking my results I see that L Corrected and R Corrected are mostly above the alternative house curve that you suggested. I thought they should be straddling the house curve. Any ideas why this is happening? Thanks.
REW trace arithmetic division may add up to 3dB offset to be able to calculate Fourier transforms and avoid division by zero, it doesn't effect the actual filter.
Hi OCA, First i would like to say thank you for sharing all your knowledge in these matters 🙏👍 I tried this tutorial with Apo EQ and it gave nice results, but I use a Minidsp 2x4 HD and I would like to not be dependent on PC as a source. Do you have any suggestions? Would it be possible to use the 1024 taps FIR and the 10-band PEQ together or something like this? Sorry for the noobish question! 😊
REW cannot produce filters with just 1024 taps, too low. You will need to use REW's equalizers for correction instead of the inversion method and send them to MiniDSP. rePhase can produce filters with 1024 taps and maybe you can adapt the crossover phase shift corrections to work with MiniDSP although the resoltuion might be too low. Select rectangular windowing and energy centering for the best possible low tap filter.
Cheers 👍 I use the Minidsp as active crossover for my diy bookshelf speakers at 1700hz 24db LR. So firstly i should use REW to create correct overall response? and then I can try Rephase to create a fir filter adjust phase further? 🤘 /David
Thank you. You tell me that umik-1 have to positioning horizontally, parallel to floor. But in yout pdf it seem to me that mic is vertically positioning. Can you help me, please, to understand?
Only matters for high frequency response which you don't want to correct anyway. In general, if you are measuring a surround system with speakers all around the mic, use it vertical. If you are measuring a stereo set up, keep it horizontal.
Thanks for the video. Much appreciated. I was wondering if the final results you shared were new measurements or just predictions of how it would look after the adjustments?
This is a great tutorial! I fllowed the steps and it does make a big difference. One thing I would like to ask is that after applying the Convolution in Roon I hear cracking sounds when the treble is high. I turned the convolution off the cracking sound is gone. May I know if you have any idea how to prevent the cracking sound? Thanks!
It could be clipping. Arer you also using upsampling in Roon ? This is known to increase clipping. Watch my new tutorial (launched yesterday in two parts), there's a lot of info there on clipping as well as some minor updates on the inversion method.
@@ocaudiophile Hello may I know the two videos you referred to are "Supreme Digital Room Correction with REW & rePhase" and "Excess Phase Inversion"? Not sure if I missed anything but I didn't find the content talking about "clipping". Maybe you used another word which is equivalent to "clipping"? I am new to this apologize for the ignorance! :-)
Hello sir and thanks for the great video. I followed your pdf guide and I have some questions. Firstly in the written guide you apply ir windows to l inverted mp and r inverted mp wich makes the correction files active after 1 kHz only. Second question: I use eq apo on windows and loaded the corrections to convolution filter. Sadly the correction starts after a second or so after the original sound and it makes an echo effect that lasts for seconds after I stop playback. Is it a fault in eq apo or maybe I made a mistake somewhere? Thanks for the answers.
If you're using one of the recent early access versions of REW, you're better off exporting filters in stereo with left and right directed to Linverted-mP and Rinverted-MP, 48kHz sampling rate and "nothing else" ticked for EQ APO. To remove the delays you mention, apply "Trim IR to windows" to each *-MP filter before exporting. You'll find this under "Measurement Actions". Hope this helps!
Hi, thanks for the tutorial. Some questions 1. I have Kali LP6v2, I had previously corrected by taking MMM of L and R seperately. Since my L speaker is in a corner, my EQ to Harman trained listeners target was about -17-18dB at some bass frequencies. Would this level of negative gain be too much for this method? Is there anything I could do while keeping the speaker in the same spot? 2. Although conflicting it seems a lot of advice is to only EQ up to Schroeder frequency. What is your opinion here? I notice this tutorial applies full range correction Thanks
1. The inversion method in this tutorial will invert any peak regardless of the size but will limit boosting dips to a max of 5dB (which is also adjustable). 2. You correct for the room up to Schroeder's and you correct for the speaker above 500-600Hz and kind of correct for both in between. In other words, your filters counteract room reflections below the room transient frequency and above that they should correct for what's wrong at the source after eliminating all reflections in the room. Frequency dependent windowing used in the method aims to do just that so it's safe although far from perfect. a) Most speakers today have almost perfect frequency response at the source. b) There are better ways to eliminate reflections but it gets very complicated very quickly. By just inverting the response below Schroeder's, you get 90% of the way to optimal correction anyway hence the tendency to not EQ above Schroeder's IMO.
I've tried, and it works great, thanks a lot! Is it possible that the offset values in the SPL alignment get inverted in the "real world"? Or I did something wrong? I had to lower the L side by 0.3 SPL, and to make higher the R by the same value in the alignment process, but, when I used the result in Convology XT, the R side of the master was a little bit higher than the L side.
Which version of REW did you use? Vector division basics were changed in recent version of REW (Feb 17). "A over B" now appears 110dB below the target level. Although applying "1/A" afterwards still produces the same result, I didn't have time to check if the SPL adjustments advised in this video are still valid. You can adjust SPL offset of "1/A-MP" for each speaker until the AxB operations yields results with similar levels. You can also use volume adjustment in your convolver. A neat way of eqaulizing left and right speaker levels volumes to each other is to use "SPL Alignment" tool in REW for the final results of L & R. If you choose 160Hz as centre freq and 2 Octaves width, it will adjust their volumes between each other from 80Hz to 320Hz. This pretty much covers all vocals so the singer's voice will always appear exactly at the center. You can check in the Info window the adjustment REW applied after SPL Align operation and dial in the same amount in your convolution engine.
Hello, great video that helps me further to understand how to deal with creating filters by really understanding what to do and have full grip on the results. I first was using the latest stable version of REW 5.20.9. Then at the part where I needed to use the “Trace arithmetic” windows I missed some options. After downloading and installing version 5.20.11 I was able to use these options. Later I will continue using version 5.20.11 and try to finish it with support of your video. I am really curious about the results. I was first thinking of using the EQ option in REW to “automatically” create the filters. I have the impression that this method is able to generate better filters, but I am not sure. At least I am learning by doing this. I will be using the filters with HQ Player on a self-build high end computer streamer (Jcat cards, linear power supply etc.). Thanks for the great work.
Although automatic EQ does a descent job, it cannot be as accurate as the inversion method. I suggest, you also download the text tutorial linked in the comments as it has a lot more detail and additionally includes phase correction and test songs.
I did download the pdf and I am reading it. I will start creating new measurements using some different settings like you suggest in the pdf (Length: 4M - 87.4 s for example). I will use the filters in HQ player on my streamer. Jussi Laako (developer HQ player Signalyst) indicted to create filters with a sample rate of 192k if the highest sample rate of the source is also 192k. Of course, I could do this when I export the filters but I thought it would make sense to create REW measurements also with a sample rate of 192 to avoid converting the sample rate later during export. What do you think? What makes sense?
@@JeroenDortmans I would rather leave the upscaling to the HQ Player and just produce 48kHz and 44.1kHz convolution files in REW. Upscaling is quite a complex process to do properly. However, in my experience the corrections usually make such a positive difference that sampling rate differences become far less audible in comparison. Also make sure "Apply IR Windows before export" option is selected when producing the .wav files. It will increase the processing speed of HQ Player massively.
In HQ player I up-sample PCM to 1536 kHz and DSD to DSD 512. Jussi suggested to create filters in the same sample rate as the max. sample rate of the sound files used which is 192 kHz. HQ player will first apply the convolution filters and after that up sample to the selected rate. I will create measurements by REW in 48kHz like you suggested. Later I can always export them in different sample rates and test if it makes a difference.
First of all thanks for your excellent work. But can you please explain why the target level during the inversion needs to be set to 3.6? When I'm doing this, there is an offset between the target curve and the product of the correction and the original speaker curve.
It's due to offsets applied during fast fourier transforms to be able to perform vector arithmetic operations correctly. The resulting offset from the target curve will not have any effect on the actual sound. Also I think depending on the REW version you're using, the offset might change a little because the very latest early access versions of REW had some minor changes there. But as I said, none of them will effect the sound, just visual.
I just found the video and pdf last night, and I got to admit…. I’ve been searching for something like this for a long time. Very excited to try it out. Thank you Quick question …. I have the buchardt s400 speakers and your ears are not supposed to be at the Tweeter level, but right in the middle of the speaker in between the two drivers. When I set up the mic, that is probably the appropriate position for height, correct? Many thanks!!!
So….just out of curiosity, because I trust your opinion. What would you say is the best “automatic” room correction out there right now? Dirac, Room perfect….etc. Thanks!!
At the moment there's no automated software I know of which will realistically improve a well setup stereo system with symmetrically and carefully placed speakers and a central LP in a rectangular shaped room. The sound will have flat freq. response but will sound throttled. But almost all of them will make improvements in systems with major fundemental placement problems and suffering badly from standing waves. For HT systems, both Audyssey and Dirac Live will make major improvements because even time and volume alignement of so many speakers will result in acoustic benefits. But I wouldn't fully rely on neither for my own setup. If you want to calibrate your system manually; REW, DRC (free) & Acourate, Audiolense (paid) are all programs with capacity to achieve optimal results but they have quite steep learning curves.
Two comments: a) At 5 minutes +/- you talk about moving the speaker for arrival times. It is far more likely that the microphone is positioned off the absolute center. One way of finding the center is to play pink noise through both speakers and use the RTA function (or you can measure, but it takes longer) to find the spot where are no comb filtering, simply by moving the microphone. This is the position where your speaker are summing. If this physical location of the mic doesn't make sense, you need to double-check the speaker location. b) it would be nice to have audio on both channels of your video.
You can use Mic Alignment Tool of Acourate Trial software for precise mic centering but the idea here is that user has already placed the mic at the LP and speakers need to be moved accordingly. I used a mono Audyssey mic for the video hence one channel only :(
@@ocaudiophile Does moving a speaker by 1mm really change the end result audibly, given that one's actual listening position changes by far more than that every time one tries to return to that same position?
@@stephenjarzombek2903 You are absolutely right that it does not make an audible difference. The reason for the requirement of very accurate alignment of left and right speakers is accuracy of trace arithmetic vector calculations.
@@ocaudiophile Because the correction is applied in the software phase of the reproduction chain where there is no delay between left and right signals?
Hi. You made a great tutorial. Respect. I have a question. At which stage of the pdf tutorial can I improve the correction impulse so that there is no distortion around 2'00`` and 2'30`` (vocal in the upper frequency range)? The track is "Statements (Acoustic)" by Loreen. Thank you in advance. Best regards.
Thank you. It's hard to know what is causing the distortion without seeing the actual measurements but since the method will normally filter any peaks around that region, it could be phase related. If that is the case, you can try correcting the phase manually in the 200-230Hz area by using "paragraphic phase equalizers" in rePhase, just be careful not to use high Q and high phase shifts (Q>2, shift>45 can cause pre echo effects). In the guide you can do that in Step 29.
@@ocaudiophile Thank you for your quick reply. I will try to change something during preparation "phasefix.wav". Temporarily fixed by increasing the L and R attenuation (2dB) in the Brutefir settings. I use volumio and brutefir plug. The distortions are gone. Music sounds great. Thanks :)
@@mariuszz3376 I found something similar. When following the tutorial, and exporting the convolution into rephase to check digital levels/headroom, REW had the file sitting at +6db. This would cause signal clipping. To solve this I exported the convolution into Rephase, and set the highest peak to 0db.
Thank You fot this very informative video, however I can't get through on last step, namely to load the correction file to minidsp, it says: " you are not allowed to upload more than 1024 coefficients"? What am I doing wrong? Can someone help please?
@@ocaudiophile Thank you for your very fast answer! So I can only use my minidsp 2x4HD with IIR correction using REW or I have to use other software, can I?! I'm very dissapointed...in spite of this, I'm very glad for Your videos, they are very informative!
@@SteveSoldier13 That depends on your source. If you're using a PC as a player for example, you can use the free program Equalizer APO which will accept all the convolution files you generate in REW. Roon and JRiver both have capable convolution engines as well but they are paid software.
@@ocaudiophile Thanks for the tip, I've already known this app but not about this exact feature! However I'll rather stick to using minidsp (physical tool) so this way it's always in the way of audio line independently from whatever source I'm using. Anyway it's a very good tip, I thought I can only use EAPO for IIR correction!
Fantastic! I had an amplifier with Dirac and wanted to upgrade. I also use Roon and did not understand how to use REW until now!! What an IMPROVEMENT with my new amp and this convolution filter! The only thing for me is that the vocals sound a little "thin". Is there a way I can let the vocals sound a bit fuller? I use Roon equalizer to correct that a little but there is probably a better way to do within REW?
Thank you for the digital equalization video. I'm encountering an implementation roadblock: When plotting Overlay:Impulse response, REW shows (what looks like) the filtered IR, not the step response. I'm using V5.20.13. I can't see a way to time align using the filtered IR. Is there a setting I need to change to get Overaly:impulse response to plot the step response? Thank you
Very interesting work. I will try it out myself. QUESTION: Instead of using a real time convolution engine I assume I could convolve a few of the actual audio files and play them directly to the speakers for an A/B comparison?
@@ocaudiophile - Any Reverb plugin that accepts IR's should, in theory I think, be able to do this. Sounds like a fun experiment to run if I ever get some time.
It has a 32bits 192Hz interface so resolution and dynamic range is higher but as long as there's a calibration file in place, you wouldn't see significant differences in measurements. They both have the timing issues inherent to USB interfaces. I'd rather keep the Umik-1 and add a descent analog mic if I were you because USB interface is so practical.
Hi, Curious to try out this method. Would I get much better results with your "Supreme Digital Room Correction with REW & rePhase" video or how would this process compare (I see in this video you are just staying in REW and not needing re-phase)?
This video comes with a written guide and in that you'll find details of rePhase crossover phase correction stage. The Supreme correction video has additional 2 more filters on top of that but it's harder to get right.
I am just getting started with your tutorial. I am using 5.20.14. I have both L an R channel in the SPL display window and I clicked "Align SPL". L channel Shows .13 dB offset and R channel shows -.13dB offset. Does REW apply those offsets or do I need to "Add offset to data"? Is it normal that they would be the same except for positive and negative? Thanks for any advice.
Wonderfull tutorial! I tried in the past and it works flawless. Now with a Goldnote I got an output issue. L channel doesn´t work (John from REW told me it could be a muting issue of some DAC). I got some result in this way: I did measurement for R ch. and then I switched R with L line output cablea in the dac in order to let REW think that I was using always R ch. The issue is that I did it with "no reference timing". With reference timing acousti REW blocks at 1% even if I turn up ref level. Now the convolution filter with inversion that I created seems to work quite good but for 44.1MQA. It works for 44.1 standard and 96 MQA, but with 44.1 MQA the sound is accelerated (about 10%). is that related to "no timing reference" mode or it´s something else? Do you think you could help me a bit? please!!😇😇😇
@@psycofurious interesting, never heard something like that. Sounds like a DAC sourced problem, unlikely to have anything to do with correction. Try to set player to play silence before memory or any similar settings to get the async working
@@ocaudiophile sorry, I lost you, I use Roon and I don´t know exactly what you mean and how to do it. BTW thanks! Could you please tel me if using or not a ref timing acoustic could change something? is it necessary only to check distance LP to speakers? do you think it could be up to the DAC even if I got faster speed only from some freq.?
Sorry this comment was being held as spam by TH-cam!!! I really don't have a clue why the accelration is happening but it's your DAC causing it somehow.
Amazing video. I am working with my measurements (and i also tried with yours downloaded) but I am having a problem when I am creating L Inverted, because the SPL is around -50dB instead of 0db. I can't see where I am doing something wrong. Do you have any suggestion? Thanks
Hey, OCA. This is very fascinating. Is there a way to do this for my home theater? I just have Audyssey so I can't use IR files. I am a complete noob to this. I am googling but can't find much yet haha. Thanks for the video!!
Thanks for the tutorial! Any tips or links to point a relative noob to regarding how to incorporate a sub into these measurements. I have a 2.1 system and with my sub I get some weird L + R readings in REW. Should I measure without the sub on? Or is there a better way?
IMO you should crossover your SUB with your front speakers at or below 80Hz and then take a combined measurement and correct that. If you want to use them altogether at all frequencies then watch this video I made yesterday. You can use the technique to time align your SUB with L+R: th-cam.com/video/ga2eOwJRtXo/w-d-xo.html
@@ocaudiophile Thank you so much for the quick response. I figured out how to get the FIR without the sub and now just need to incorporate that. This is super helpful. Cheers.
OCA - I followed your instructions above. I used REW to help me first integrate my sub: finding the “right” volume, placement, and crossover. Then I followed your guide again. WOW, BRILLIANT! I had always felt a little disappointed with prior attempts at room correction. This really takes my system t another level. One final question. Could you offer some guidance on how to to do the final speaker adjustment for distance correction? In my case: 1.05 inches for my right and 1.48 inches for my left…. Is this really going to make a big difference? I’m guessing my head moves more than 2 inches from my mic placement position. THANK YOU!! SUBSCRIBED AND LOVED
I have a question, after carefully following your videos, I have built several impulse wav filters. I have successfully exported it to Camilla DSP in Moode Player. The result is that it sounds very good in PCM, but very bad in DSD files, they lose naturalness and Stage, despite adjusting their gain. Do you have any comments about this correction on DSD files (Camilla resamples them to 352 k 32 bits)? Thank you!
I'm not very familiar with Camilla's filter upsampling procedures. Roon can match 48kHz wav filters fine to DSD tracks although it slows down and stutters with anything above DSD 128 in my Intel i5. Probably Camilla is not upsampling the filter for that reason. Try exporting the same filters at 352Hz from REW and see if they sound better with DSD.
Thanks for such informative video! I tried to follow the step by step with this vid but some function button are not the same with the latest version (5.30). Would you please kindly help if I upload my mdat? Thanks in advance!
@@ocaudiophile here you go mate! I live near a noisy street so it's literally impossible to do a 4M measurement without cars / scooters going by so I did a 2M instead! drive.google.com/file/d/1QOLJ805q5sLe-rx51F8OkWhndtGB2IUr/view?usp=sharing
Great video, however when I do this, my final result seems to be about 4 db over the target curve and doesn't fit underneath it like your example. I am wondering if it's a difference in the version of REW? I have the newest version and some things are a bit different.
Some convolvers ie Roon normalise filters and add/remove gains. As long as you check for clipping, that's fine. The method has only cut filters no boosts.
@ what actually happened in REW is that when at the end we show the target curve and then overlay the correction, your correction fit nice underneath the target - your target line was the top and everything came below, but my correction was above the target line. Do you think the new REW maybe has other things happening behind the scenes or different standard settings that I have to change, to make the calculations work correctly?
@@ocaudiophile yes, I followed step by step process in your video, there were slight differences in the menus of REW because my version was more recent than what you used in your instructions, also it seemed there was an automatic alignment tool for the impulse response, where your video shows how to manually move it, my room’s impulse response also looked a lot different than your example but everything else I followed exact.
You cannot adjust number of taps with REW generated convolution files and the default values are much higher than 1024. Your easiest option is to use REW EQ filters for correction, export them to rePhase and generate the filter with desired 1024 taps in rePhase. I think you can even save it as .bin. One advantage is, you can also align crossove phase shift, etc in rePhase in addition to the REW filters. To reproduce descent filters with so few taps, you may need to try different windowing types like rectangular and energy.
Hi Luca, have you by chance found a way to do this on Mac? Just spent some time doing everything in this video, only to find that MiniDSP won't load my fir :(
@@angeloboltinimusic Sound Forge can do this but it's not free although there's a free trial period. Load the WAV then save as "Raw" and in the "Custom Settings," select "32 Bit, IEEE float.". I am not sure if you can match the number of taps to Minidsp standard though. Another way could be writing a Matlab (or the free version Octave) script to convert the files. It could be something like: [x,fs] = audioread('LR.wav'); out('L.bin',x(:,1)); out('R.bin',x(:,2)); function out(filename,x) id=fopen(filename,'wb','ieee-le'); fwrite(id,x,'float'); fclose(id); end But I have "zero" Mac experience and my Matlab skills are very limited but I guess this is the right direction.
@@ocaudiophile I use 1024 taps Fir filter for hi and mid driver. This means in my case down to something like 200Hz. But anything under 1000Hz will not produce a filter that can do accurately what the video is showing you. So depending on your setup try to produce 1024 tap fir filters with RePhase that works in the Hi region from 1000Hz up. Then do with REW the normal target curve EQ generation for the signal under 1000 Hz. This is the way to go if you are restricted on taps in your system. The fir filters will insert much more air into the sound in the upper region. IIR filters can perform well in the lover region. If you are playing music from a computer then you can measure your setup with the filters explained here before. Then do the FIR filters that the video does explain. Import the filters into RePhase and adjust the phase. Export the filters and convolve these multitaps (long filter) in the PC on top of the MiniDSP filters (Fir + IIR). This makes it possible to correct or finetune the full bandwidth.
@@spalmgre9167 Excellent explanation. I don't think it's possible to do with MiniDSP but an alternative to correct low frequencies with 1024 taps is frequency warping. I think Audyssey is using this technique with their 1024 tap FIR filters. The idea is to replace the delays with first order allpass filters. That makes the overall contraption an IIR filter, but it maintains the overall structure of a FIR filter. Instead of a chain of identical delays you have a chain of identical all pass filters. The output in both cases is a weighted sum of each node in the chain. Warping maintains the overall shape of the filter (peaks and dips, etc.) but it moves it to different frequencies depending on which direction you warp.
That's a great tutorial and it actually produces brilliant results!! Thank you very much for this! May I ask a couple of questions though, being a newbie to this? 1. When I follow your pdf tutorial and make the MP copy, the windows limits must be opposite than what you describe for it to work i.e. 1 on the left channel and 125 on the right. Does this matter? 2. Why is the corrected vector average so much different (worse) than the RMS? Thanks again!
Thank you and happy to hear you had great results. 1. In the pdf, the suggestion for the MP's are: “Left Window: 125 ms” / “Right Window: 1,000 ms” which is a good chocie and there's not any in the video? Left cannot be 1ms, that would compeletely ruin the correction as it will massively change the magnitude response of the filter. You can check for the effects of left and right windows under the "Impulse" tab by moving the L and R (green and red squares) with your mouse over the time axis. If you cannot see them, make sure "windowed" is ticked and you have zoomed out enough in the impulse graph (the graph on top). The bottom graph is the magnitude response of the filter. 2. Because of phase differences between your left and right speakers. Ideally RMS and vector averages should be the same in an ideal room. Think of phase as time delay (it's not exactly that) and the same frequency coming from the two speakers with such a time difference at certain frequencies that they cancel each other out and causes dips there or vice versa. RMS on the other hand just averages the SPL values.
Thank you very much again! I understood 1=1,000 by mistake, hence the strange results I got when I followed the pdf. I'll revise that and review again. Perhaps that could affect the vector average as well.
@@ocaudiophile I got this right now; comma vs dot does make a difference! One more question please: if I want to do further adjustments to the generated responses, can I just EQ the final L and R "Corrected" response measurements in REW and export new IR or will I just ruin everything? Thank you!
I don't have any experience with Helix and from what I quickly searched, looks like it accepts IR files in wav format but only with 1024 or 2048 taps. I might be wrong. You can use its parametric equalizer to copy REW EQ settings and probably can also fit a phase correction within the 1024 tap limit.
Thanks. It worked. Hard to master tho. But, why using Harman curve? The studios are using flat curves. So the mixers could already have added bass respons similaer with Harman.
Harmon (short for Harman Kardon) curve is in fact developed for headphones. I find it a bit bass heavy and lately prefer Dr Toole's curve. These curves are in fact flat in their perception by the human brain but that's volume dependent. So what curve you will apply also depends on what volume are you measuring and what volume are you going to listen to your system. That's why I lately advise to measure at the volume you listen to your system. Check equal loudness contours and you will understand better: en.wikipedia.org/wiki/Equal-loudness_contour
Hey OCA, Is FDW best used at 15 cycles? Does this seem to be a good value for all frequencies? Also, does it matter if I smooth first and then FDW or do you FDW first and then smooth? Thanks.
15 is quite well suited to most cases but the actual number depends on how reflective the room is and how far is the LP from the speakers. For highly reflective rooms lower numbers can yield more realistic results at high frequencies whereas higher fdw values (or none at all) can result in more effective filters in the low frequency region. A better fdw function in theory would be one which can apply varying degrees of fdw throughout the frequency band IMO. I'd use VAR smoothing to disable over correction in the high freq area.
I have a simple request, I'm trialling down firing sub and effect of distance... How do you do a measurement, calibrate the eq and plain it back and resmeasure using the applied eq?
New REW Beta has an option to run a measurement with the applied EQ: www.avnirvana.com/threads/rew-api-beta-releases.12981/ www.roomeqwizard.com/help/help_en-GB/html/eqwindow.html#:~:text=Measure%20with%20these,clipping%20the%20output.
How may I positioning the UMIK 1 for The first step with REW for aligning the speakers ? Horizontally or Vertically ? And what file I have to use 0° or 90°? Thank you
![[LIVE] : ONE ลุมพินี 88 | คู่เอก "ป้อมเพชร vs อัสลามจอน"](http://i.ytimg.com/vi/1ZqTVVbMgbo/mqdefault.jpg)
The most useful video I have watched in my whole DIY audio career (+10 years). Thank you, Sir. I'm usualy modest and reserved (especially towards audiophile lingo with subjective impressions) but this tool just simply transformed my DIY full range speakers. So grateful.
You can combine as many filters as you want. You will need to vector multiply each filter with the previous one. You can use "generate response from filter" in the EQ window of REW to create a filter from your EQ bands ready to be used as a convolution file. The suggestion depends on where and how much you will be EQ'ing. The inversion shoudl normally take care of eveything that needs to be corrected.
@@ocaudiophile When doing vector avg of multiple readings, lets say 8 readings, the spl level after 200Hz goes down for 7 db. Won't be better to perform RMS + Phase Avg? I think we get better results for both SPL and phase corrections. Do you agree? Thks!
If the vector average SPL is dropping, you probably didn't "cross corr align" the responses and phase differences are causing SPL dips.@@Estorki2
@@ocaudiophile yes I have, but when you average many measures, it is normal this behaviour in rew, I understand because of the entropy of phases at higher frequencies. It is documented
You were talking about a dip at 200Hz though. Depends on which response you're aligning the others to. If measurements around a central measurement are cc aligned to the center vector average will show the correct frequency and phase response in the measured area and there will not be major dips anywhere. I regularly average 8 Audyssey measurements that way.
Thank you so much for this quality content. I’ve been searching for more in depth videos on REW. Seems like the algorithms prioritize the most basic stuff. More please!! New subscriber here.
I just did an experiment with this convolution with inversion method, near field on a CD horn to see if it would create a phase perfect passband including an octave into a 12 dB / Octave high pass filter. It worked pretty much perfectly. The phase and frequency response measured near perfectly flat! I spent a little time listening. It's hard to say without more listening if there are any audible problems, but initial impression is very positive. I'd say more smooth and natural than before. I think I might have to do all th drivers this way.
Precisely central (09:45). Unlike how your narration emanates from the left side of the stereo field exclusively. 😄No big deal. Thank you so much for sharing your time and expertise with us. Much appreciated.
I've actually used an Audyssey mic to record the videos which is mono. I am not a content creator or have any monetary concerns for the channel so the videos are unedited. It's just information sharing.
@@ocaudiophile And great information it is. I have moved on to your "phase shift" tutorial and your voice comes to me in full stereo. Keep up the great work. Have a wonderful day.
This is the most concise and explanatory video on this subject I have found. Kudos! Although a bit late
There's also a pdf manual.
@@ocaudiophile Thanks, I like both but your vids are almost explanatory enough esp in the context of OCD :-) Ive been doing acoustics prob 30 years and just lately having some real breakthroughs in room stuff that I havent seen anywhere else...but not had time to delve into the 'cream' region of final tweaking which is exciting to finally get there...although an eternity still to go ;-)
Simply - Thank You. Your work is Great.
So nice of you
I want to thank this man for investing so much time in this tutorial and sharing his knowledge. It took me some hours to reread, measure and correct but it was 100% worth it! I'm blown away with the results, my speakers sound so much better as I had a lot of troubles with the lower frequencies in my room.
You're very welcome!
How did you applied corrections in to AVR?
@@BhavikPatelhr i'm using Roon + HQplayer with a Devialet Expert amp. I don't know how to add these in an AVR sorry
Instead of dividing L and R by Target LR and then inverting, I did it vice versa - Target LR (A) divided by L and R (B) respectively for each channel. Did this due to some changes in the calculation options for the inverse action. Seemed to reach the same result, just one step less.
Otherwise, a great video for dummies - much appreciated!
Correct, trace arithmetic division options have improved a lot in REW since that video. You've done the right thing. You may wanna watch "mastering digital room correction" video for updated and a bit better sounding tech.
@@ocaudiophilehearing from a master - a must read then!
Thank you so much! This is an amazing new field im getting into and you truly inspired me!!!
Fantastic!
Thank you so much for your excellent work
👍👍👍
Thanks so much for creating this content in such an easy way to follow. I’m really enjoying my new sound.
👍
I would never have thought that EQ could make so much difference. I'm shocked by how much the quality improved in my case. Just wow. Thanks for this
Thank you so much for both your video and the pdf, you explain very clearly, great work 😀
DAMN !
Thanks, thanks & thanks again !
I directly felt the difference, man, I was thinking buying 8k$ speakers but I won't.
Thanks so much for this tutorial. It made a huge improvement for me. Certainly could not have come up with this on my own.
You're welcome!
Thank you for the great tutorial, OCA! yours is the only one that I've found; to correct the phase of the speakers. When I listening my Left seems to be off compare to the Right. This fixed it!
Glad it helped!
Bravo! Outstanding video. Easy to follow along with, steps explained, perfect. Subscribed!
Thank you. I hope you will enjoy the result.
@@ocaudiophile I do. One hiccup with this method is that the convolution file created ends up being really long. On my desktop, this resulted in an audio latency of 0.341 seconds. I took the wav file into Izotope RX and trimmed the empty space from beginning and end, which brought my latency down to 3.1ms. Equalizer APO seems to indicate the same corrections.
Now that I'm running convolution, my right speaker is slightly louder than my left. When I took the measurements, the calibrated mic was on a tripod in the center of my main listening position. Did I miss a step somewhere?
@@wyleyrabbit There have been some revisions to A/B trace arithmetic recently in REW. Depending on the version you're using you might need to use target level at auto and adjust left and right 1/A levels. There're instructions at the end of the updated pdf guide.
@@ocaudiophile I downloaded and followed along with the PDF. I'm using 5.20.13. The impulse files end up with what could perhaps be best described as having an echo. I opened the file in RX and there was most definitely more than one impulse, so I manually removed the echo part of the impulse wav file. It sounds great, but I'm not sure how the echo is getting in there. Any ideas about what I'm doing wrong?
@@wyleyrabbit You need to use an early access version of REW (5.20.14 onwards) for best results. There should not be any echo with this method but I cannot gurantee that for older REW versions.
Thanks a bunch, looks like a method that can yield great results. I will give absolutely be testing this method.
I will say, your video only has audio on the left channel, which on an audio engineering video is a big crime, the punishment being you can only own bose speakers. Please surrender your good speakers to the audio police.
Thanks for your kindness shareing.I follow the tutorial and complete my second room correction attemp.The result is impressive especially compared it with previous autoEQ,though my secondhand speakers are not in their best condition, L&R FR dosen't match quite well(and they are bass limited model,so I change some settings).After inversion,I can easily tell the different when gain at 16k,but the day before correction,gain at 63 or 16k seems nothing change in my system.
Great 👍
I love this channel. Since i came to know about your channel and videos, I am learning REW and calibration. Even I calibrated my Yamaha entry level AVR which has no auto room calibration. I calibrated and measured. It improved sound quality a lot.
Love from India
Thank you sir. Finally a speaker/room calibration and correction tutorial that is friendly to newbies! Appreciate all your hardwork.
Thank you
I was experimenting with EQ for quite a while trying different EQ settings and different EQ software. But this FIR method sounds so much better.
Glad it helped!
Thank You Very Much! Really Impressive... Great Job!!😃
Thanks👍
Thanks and Happy Birthday!
Here are the much faster filters at 48kHz:
drive.google.com/file/d/1qd9_2YHdUGbFpkw3WSAoBCWFnv8ULo8H/view?usp=sharing
@@ocaudiophile Perfect, no more lag... You are the man!
Hello hi. First things first, thank you for your tutorials...
Anybody tried to remeasured the sweep signal with the FIR correction?
All you need is two convolution engines to apply the first FIR and then the second. I know it sounds a little bit crazy, but it really works.
I have to be honest, this took me a while 😂. But first impression is excellent! I tried the regular EQ filters before (with the Harman Target curve). But I think the phase part in your tutorial played an important part in why it sounds a lot better now. SPL and Time were allready almost perfect here. Thank you for taking the time! to create this video 🎉
Great to hear!
Thanks!
Welcome!
Pro tip: Record your mono narration to both channels and not just the left.
Also, a high pass filter to cut the subsonic rumble on the mic.
Noted!
Also LPF a bit... there's some weird periodic high resonant stuff going on.
This was very helpful! Your explanation brought clarity to my weak tuning understanding!! I now better understand the reason for using certain values in the correction process. Now don’t get me wrong but I will need to go back to study this many times. My prior efforts always resulted in huge cuts in amplitude (make it flat approach)
I know that you will get great support to this video and you will continue to educate us.
I to use Roon and solving timing issues in different tools in Roon is going to be very helpful.
We’ll done
Where can we obtain the Harmon Kardon House curve to Input into REW? The target curve thank you
There's a link above.
@@ocaudiophile thanks !
Thanks
Thanks a bunch for the donation. Much appreciated!
Incredibly useful video! Have you looked at how this functions on the latest V5.20.14 early access build? Seems like the Trace arithmetics have changed. Need to use |A| / |B| instead of A / B and you no longer have the ability to set Target level on 1 / A causing the end result to be offset from the target line.
Yes I have. A/B target level changed and I added necessary revision notes in the video links in the description some time ago.
@@ocaudiophile Perfect, thanks for all the time you've put into this!
this is genius
Thank you
@@ocaudiophile man, i used your tutorial to create the shebangs including the rephase stuff for my headphones, christ all mighty what a diffrence.... not to mention now i could use that convolution file as reffrence no matter what song i play or movie and create an auto eq with tdr nova eqs smart ops feature... genius stuff :D
@@adriangpuiu Thank you, happy to hear that!
Thank you for all your helpful videos.
but which method should be used to measure your living room cinema? (5.1.4)
MultiEq app available
miniDsp2x4HD is available for the subwoofer.
🤙🏼
Supreme Audyssey Calibration video
Tesekkurler. Deniyecegim.
Please, can you make a video about calibrate phase of subwoofer and speaker in REW? Thanks!
It's all about aligning impulse responses. I will do a subwoofer video soon.
Hello. I am grateful to you for introducing an amazing method. Not until I saw your video did I realize the computer could generate great music. I am Japananese. I'm sorry that I don't have much vocabulaly to express my thanks in English.
I have never used REW and I don't have any knowledge of accostics. But thnaks to your video and PDF, I couild make my own convolution files.
I have some questions. Sorry for my poor English and no knowledge.
1 Now I use REW v5.20.11 according to your PDF (Umike1 asio-4all). When I saw your video , I used v5.20.13 and measured IR for the first time in my life.
When I used v5.20.13, most of the first dips were 100%, the next peaks were 70-90%. So I thought the first dips were the immdiate impulse. But with v5.20.11, most of the first dips are 80-90% and the next peaks are 100%. So which should I think the immediate impulses were?
I think I'd like to use v5.30.9. I have 3 questions.
2-1 As you have written in Update of your PDF, Is it OK that I skip Correction Process 5-8 and use only "Align SPL" command to left and right speakeer measurements?
2-2 With v5.30.9, in correction process 20, there is no button of "Apply windows". What should I do?
2-3 With v5.30.9, should I use |A|*|B| instead of A*B?
I'm very sorry to trouble you.
Download the latest REW Beta from this link:
www.avnirvana.com/threads/rew-api-beta-releases.12981/
Trace Arithmetic and especially inversion operations have been highly optimized as of yesterday. Start using this version and let me know, if you run into problems.
So to confirm, if I'd like to combine this with the xo phase correction of the other tutorial, I just TA multiply the inverted_MP results with a phase correction impulse from rephase?
Yes exactly
Hello. Awesome video, thanks! I would like to test this principle with my LXmini speakers (no subs), but have not figured out yet where to upload those wav correction files. My source is Apple Music on a Macbook Pro, via USB digital to external DAC. Is there a recommended plug-in or free (cheap) software for MacOS/Apple Music?
Try Camilla DSP
@@ocaudiophile thanks, I installed Audio Hijack (trial for now), loaded my convolution files to test. There is an improvement in sound when listening, but when measured the graphs are not that different without FIR and with FIR filter applied. And definitely not that close to the implied or calculated graphs (for example clarity, impulse) for "corrected" responses that are shown in REW during the making of the filters. Is that normal?
@@mantasj6678 I'm not familiar with audio hijack, why don't you use equaliser Apo? Frequency response shoukd be very similar, clarity may change significantly depending on measurement distortion.
@@ocaudiophile Equaliser APO is only for Windows, while I am using MacBook with an Apple M1 chip :/ I could share my REW file from which I made the filter, and the measurement file after I tested with filter and without, if you don't mind looking at them. It's ok if you don't want to bother, I can hear the improvement and will continue listening anyway :)
Maç users use Camilla DSP as far as I know. I can have a look of course if you can post a Google drive link here
Ciao, una curiosità. se ti mando la risposta in frequenza delle mie casse monitor sotto forma di file Wave puoi farmi avere i filtri generati da Reweq per linearizzare i miei monitor?
It's possible but if you share the REW mdat file, it'll be better. You can ONLY share google drive links here!
@@ocaudiophile grazie, provvederò, e mi occuperò di inviarti il link, ci provo, intanto grazie!
Thank you very much. A great work. Is there a more recently and usefull version of REW for this tutorial? Is it necessary a powerfull PC or could be sufficient also an old asus?
th-cam.com/video/qoFZPlXrTeM/w-d-xo.html
@@ocaudiophile thank you again. What is the software for calculate precisely di distanc between LP and speakers?
@@ocaudiophile Maybe Audiovero Acourate. With the trial version what can I do? Is there a tutorial for using Acourate for measuring distance and exact position of LP and Speakers? Thank you
@@Angelos58 Acourate has a tool that helps centre the mic position between left and right speakers and it's available in the trial version. You can also measure each speaker using one of the speakers as acoustic timing reference in REW and adjust mic position accurately by checking their impulse peaks in Overlays/Impulse graph.
@@ocaudiophile thank you. Would be very appreciated a video tutorial.
Hi, thanks for. your video on how to use Convolution. It works like a charm, brings out so much details that I have not noticed before. There is only one problem I encountered is the latency of the convolution, which is a bit large. I have read through the comments and saw this: At the very end of "Excess Phase Inversion" video, I've explained a method for trimming down convolution files... I would like to see if I can trim down the convolution file, but the video seems to be nowhere? Which video is that? Or can you explain in simple procedures that I can try? Thanks a lot.
th-cam.com/video/NbgsiXpvz_s/w-d-xo.htmlsi=92EYq4H-zKC2D_UB&t=1219
@@ocaudiophile Thank you for your quick response. I have watched the video and followed the step and cut quit a bit of the windows. It works better now. One thing I have noticed that you are using a newer version of REW 5.20.14 which is not available yet. And the difference is that my 5.20.13 version do not have the "Trim IR to windows" button. But I figured out if I export the IR wav file with option "Apply IR Window before Export", that may do the job, is it right?
That will accelerate the loading time of the filter but will not decrease the number of taps. Google "REW Early Access version" and you can download the newest one.@@trumankong7053
@@ocaudiophile Thanks a lot for your help.
The best video I have seen this year! Two things as a feedback though - would be great for follow-up video.
1) You might have made a mistake at 9 minutes in your video when SPL matching speakers. If you're focused on low frequencies 50-60Hz, I think you might calculate room modes into the SPL average! Not sure ROON is capable enough to remove them completely - or is it?
In my L-shaped room with pillar behind one of the speakers - the bass frequency response is *completely* different (in bass department) left vs right. To get around this, I would not calculate anything below 250Hz - or am I missing something?
2) I think I would introduce less time smearing, phase issues and precision going through a simple EQ calculated for minimum phase - correcting 2 offending frequencies in my case. I wouldn't like to touch *every* single frequency during DSP calculation if my speakers are already playing fine - and I just have issues at 10KHz region and one in bass. I would welcome clarification video, why in your oppinion EQ (correcting just 2 bands) will sound worse and introduce more problems that adjusting entire 20Hz-20KHz band, even in places where it doesn't need adjusting. People, I really do not need to touch 1-2db differences, can live with that! If I stand up or move away from speakers, differences are far greater.
3) Was vector RMS average fixed in latest ROON, or you still advise to go with generic one?
4) Where can I find latest Harman curve? Is your description still valid? The idea behind would but require some clarification - as if I listen at 83db where Harman curve is not required, why would I still need to apply it?
5) I would welcome phase follow up video or and updated one - as again - this is best video I have seen this year or last year. Any time I touch equalizer on my Mojo2, *even by 1db on single band* I feel it destroyed the precision of 3D space I was hearing previously. So Rob Watts loose-less EQ sounds like a lie to me. I would need to try approach in your video.
Thank you very much - you're beyond knowledgeable and I love Focal Sopra No2 speakers very much.
L-shaped room is a challenge. You've to aim for max symmetry at least with the remaining 2 near walls and deal with the major assymetry of the 3rd by DSP. RMS + phase average gives more accurate representation of the high frequency response when averaging multiple mic positions but for bass response averaging, time aligned vector averaging rules. The method in the video is new and very powerful. You may need to use minimum phase versions of 1/A divisions for limited FIR taps as they "might" cause pre-echo otherwise. I prefer Dr Toole curve, find Harman bass a bit too much and not inline with my speakers natural response.
My friend thank you for all the knowledge you are sharing. I have learned so many things! However I am still a noob and I need your advice about which one of your videos is the best for my case...
I have a 2.1 system in a small, asymmetric room for music listening. PC is the only source and it's connected to a DAC/preamp through USB. All three speakers are active and they are connected through XLR (DAC > Sub > Monitors).
I followed the first 10 pages on the attached pdf to the letter and I'm very satisfied with the results in EQ APO. But although the room is partially treated, I have a bid dip in 1.2khz area and the house curve doesn't really help because the SPL graph is so uneven around that area.
Should I continue the pdf guide and use rePhase? I'm asking because I read in one of your comments that it's not necessary in an active speaker setup.
You have uploaded so many wonderful videos after this one and I would be grateful if you could point me to the right direction, as the next step...
Thank you. Your set up is capable of sophisticated high tap FIR filters so most problems can be fixed without side effects although asymmetric room can be a bit of a problem. It's hard to tell the problem around 1200 hz. Can you share your measurements? You can paste Google drive links here.
@@ocaudiophile Yes of course, thanks for helping! This is the link of "Step 2", according to the pdf. Inside the folder you'll also see 4 photos of my space...
drive.google.com/drive/folders/1UjOMFWnFZ3aRuZ-Uo6ZY2Tel0U13ZUGk?usp=share_link
@@cnkosm6536 The dip around 1200Hz is caused by a reflection surface around 7-8cm away from your tweeters (I don't know your speaker crossover points and assume it's your tweeters producing 1200Hz) and a tiny bit longer distance for your right speaker. Sound waves bounce back from that surface, travels an extra 14.68cm distance which takes it 0.1468/343=0.000428 seconds and reach you and interact with the original sound wave causing a peak at 1/0.000428=2336hz and a dip at half of that.
I could fix most of it with excess phase inversion. I created a stereo convolution .wav file in the link below also included the mdat:
drive.google.com/file/d/1EtCL2GrK9S3E4WrKx2BPVf1L8zHSlNmH/view?usp=share_link
I applied a slighly better sounding EQ technique and combined phase inversion with that but you can also add the phase filters to your own inversion filters if you so wish, they are all in the mdat. I will be waiting to hear your feedback.
No need to listen I know exactly what it is, pre-echo caused by low frequency ringing. I must have gone too low. I will fix it.
@@cnkosm6536 The dip is caused by an extra 15cm distance traveled by the sound wave relative to the direct sound. Check for surfaces from which when sound will need to take this much extra length when it bounces and arrives at your ears .
The new filters are in the link along with the mdat file which contains eveything. Filter1 is no pre-echo guaranteed but Filter2 additionally fixes some of the dip around the 200Hz area but could cause ringing. If both of them cause ringing then the problem is in your right speaker which has some pre-ringing even in the default measurement. Let me know.
drive.google.com/file/d/1KB2G_aaUM3xfu9AYmy7sqF4el-3ms8mt/view?usp=sharing
PS "Beat It" -Michael Jackson will show you instantly and clearly if there's any pre-ringing in the system!
Great video. I followed these instructions very carefully, but used the alternate room curve in lieu of the Harman, which results in a flat downward sloping target. My L, R, and LR plots looks as expected, but my inverted plots are very "squared off", with a hard stop at 0 SPL. Not sure what I may have done wrong.
Are your measurements from 0Hz to 24000Hz? If only 20-20kHz, some mathematical absurdities may occur during inversion.
@@ocaudiophile yes, the measurements were 0-24K. They were done with FDW active...could that have an impact?
@@ocaudiophile A couple of times during the final maths calculations, I received Java errors which crashed REW. I WAS able to produce the files without crashing, but I wonder if there's a processing issue that's creating faulty files without an outright crash.
I have followed your instructions and have achieved pretty good results, better than using PEQ. There is one item I am not sure about, which is the regularisation setting when doing 1/A conversion. What does the regularisation do? I used 8, and the result is boost to overall volume, so I tired a few different values and if I set to 15, the overall level matches the original signal. So, I am a bit confused.
Regularisation adjusts the maximum amount of boost to be applied to dips in the response. 8% regularisation will limit boost to 5dB.
Thank you for your answer.
Thank you. Hoping to apply your alignment approach using minidsp SHD. Should I have Dirac on when taking measurements if using DIRAC.
No, Dirac or any kind of sound processing should be off.
In the book Accurate Sound Reproduction
Barnett, Mitch. Accurate Sound Reproduction Using DSP (p. 6) I read that is not convenient to using a USB mic. I purchased a Minidsd UMIK-1 Angelo Scozzarella you advised .
USB mics have clocking issues but REW's clock adjustment with acoustic reference feature has greatly solved that problem. They have advantages like already calibrated SPL level and plug and play usability.
@@ocaudiophile thank you very much. So I can using umik-1 usb. Also for measuring position of mic with Acourate? I don’t find trial of Acourate.
Hi Serkan, many thanks for this video! What would be your prefered way adjusting Stereospeaker with active xover via minidsp flex eight? EQ in REW to Target curve and then rephase?
You do not need crossover phase correction when using digital active crossovers but you may still benefit from speaker box (port) phase correction in rePhase if you know the port frequency of your speakers. You can do a near field port measurement with REW to learn that. It should be a phase linearization of either "vented low Q" or "vented std Q" at the port frequency. Export it with less than 6128 taps as bin file from rePhase and upload it in MiniDSP.
Inversion method only produces FIR taps and you cannot easily limit their total tap count in REW. So for frequency correction, use REW Auto EQ filters on the target curve and export them as MiniDSP biquad filters. For best results different frequency bands should be equalzied with different settings. I am very close to launching a new video showing all that. Until then you can play around a little and compare your results.
Thank you very much for the great video. I am looking forward to trying this on my system but I have a question - I have two subs in my setup one for left and right that are also running off same signal as the left and right speaker. I assume the phase alignment will be a key issue in my particular setup. Do I treat the left a right channels like a 4 way speaker with the crossover from the mains at the sub crossover - (in my setup that is 100hz)?
Time align subs between each other and also with the speakers. Try to use speaker's spec bass roll off frequencies as your XO frequencies, Once all is set, you can apply inversion methods to the combined speaker/sub system as one.
@@ocaudiophile Following your instructions has elevated my setup beyond my expectations!! I don't think I could have ever achieved this level of sound spending thousands more on gear. I am running Audirvana with Hang Loose Convolver through RPi4 to Benchmark DAC3L, feeding Benchmark AHB2 AMP to Revel F226Be Speakers and Rythmik E15s. I tried the Roon + HQPlayer combo and Audirvana + HLConvolver spanks it IMO. Highly recommend giving Audirvana and Hang Loose Convolver a trial run - would love to hear your opinion.
Thank you for making this video/pdf and for your attentiveness to your viewers!
I have two questions if you don't mind:
1) I like my results, but sometimes it feels too smooth and clinical, and sometimes I get a strange echo effect, most notable with spoken word or internet radio (try "Room 29" by Jarvis Cocker, where his voice just sounds really echo-y and bizarre with some attempts at this method). Could this be due to a recording issue, a mistake during the REW process, or an irreconcilable issue between my two speakers?
2) As mentioned, the results are sometimes too smooth, where deeper male voices in particular (like Jarvis's above) feel robbed of their character and depth. Is there a way to to do this where the peaks in the frequency response are just reduced, rather than are chopped off by the Harmon curve? Right now I've settled on a correction that only works up to 200Hz, but I'd like a solution that covers the entire range.
Thank you in advance!
You can try experimenting with the regularization setting during division trace arithmetic operation. It determines the maximum boost that will be applied with a filter i.e 8% is for 5dB maximum boost, 25% is for 0dB maximum boost. Usually, less or no boost filters will sound less "clinical" in your terms. Also check for clipping and apply a headroom if necessary. The echo you sometimes hear could be related to clipping.
@@ocaudiophile Thank you for your advice and quick reply. I took a few more stabs at this this morning, and really found that the process is quite sensitive to the configuration of the target curve in EQ, and the regularization settings during the 1/A calculation: I was able to achieve a better result once I tinkered with the Target db of the curve, and used a regularization of 4 percent rather than 8 during 1/A. I messed with headroom, but I don't think this was the source of my original echo issue. Male voices, cellos, etc, still feel a little thin, with just a HAIR of that echo effect remaining, but bass and high frequencies sound really smooth and clear, with less dramatic oscillation between bass peaks and nulls in the room. I'm going to see how much more refinement I can bring to the midrange. Thanks again for all your help and making this process accessible to laymen.
Thanks you! your video's are very helpful!
before i dive into this, is this video good for room correction for obtaining flat frequency response for mixing and mastering? will i gain anything sonarworks wont provide regarding hearing music and not the room?
This is an efficient and easy way to correct the room and sound will not end up boxy or throttled inherent to auto calibration systems. However, REW only produces the filters and it's not a convolution engine that applies these filters to your mixes on the go. You can use Equalizer APO for that which is another free tool.
@@ocaudiophile Amazing, thanks for the reply, just subbed will look into APO EQ 🙂
Hi, thank you for the tutorial, someone suggested me to follow it on AV Nirvana and I will, although I'm just a guy who wants to listen to good music with 0 audio knowledge, so in the current state I'm sure I will mess up this for two reasons, the first would be that I don't know what I am to do with that correction file inside Equalizer APO, can you explain that to me or point to a video that shows it? I assume I need to click on "Convolution with impulse response" and load that file, there is anything else that needs to be done?
The second point of failure for me is the fact my setup is a PC with 2 speakers on the sides and a subwoofer centered on the back 2 meters away (Kali LP6-v2 and SVS SB-1000 Pro) therefore I don't have the slightest idea on what I should do in order to include my subwoofer in that correction and time alignment, can you make a video for that as well? (or point me to a video that explains it) ^_^'
Thank you !
I wouldn't worry about Equalizer APO, it is quite simple to use and intuitive after couple of minutes of use. Here's an old tutorial on it: th-cam.com/video/Cyb1r5E9K3I/w-d-xo.html
Subwoofer is also not a problem, you should measure your system with everything on and connected and create correction files for that. How you connected your sub and how you cross it over doesn't really matter as long as you measure the system in REW as you listen to it.
thank you this is one of the best videos i have seen!
But as far as i understand the convolution files are not time and amplitude aligned and I have to set it in the convolution engine (for example in Jriver).... is there a way to write the time and amplitude alignment data directly into the convolution wave file?
You don't need to do anything. Just direct JRiver config.txt file to the .wav file directory. For a REW created wave file at 48,000 sampling rate for example something like:
48000 2 2 0
0 0
0 0
convolution48.wav
0
0.0
0.0
convolution48.wav
1
1.0
1.0
If you mean windowing and normalization of the convolution impulses, you can select these options while exporting impulse response from REW. They don't have audible effects but exporting IR windows may increase speed of convolution with some engines.
How real these small picks and dips? I mean, whether they will stay on their places if we take another measurement(s) from slightly different mic position(s). I suspect most of them are kind of random, which means that we correcting nothing. This seems to me similar to "over-learning" issue in AI.
Measurements at different places will result in very similar response up to around 2,000Hz. Higher frequencies will differ but dips and peaks will be similar relative to each other.
Hi there, Is this procedure only for a stereo situation ?
I have a 7.2 system and are in the process of turning the Subs as per your optimal subwoofer alignment.
Once I have finished that I would like to do my mains and was wondering if this would be the best procedure for the other 7 speakers (LCR, Sides & Rears)
All my source material is played back via Jriver 30
These are FIR filters and are more powerful than any other so you can of course correct each speaker in a surround system with it. I am sure you can use convolution files with surrounds as well with Jriver with the right config files although I've not done myself.
O.K, I have finally managed to Optimize my dual subs as per your Optimal Sub Alignment with sub correlation instructions and are very happy with the outcome.
I am now following your "REW (Room EQ Wizard) top tricks: Convolution with Inversion (no EQ filters just FIR)" for my 7.2 Home theater setup.
I hope I am heading in the right direction with this, I have completed the Impulse alignment for all speakers including the Sub.
Not sure if I needed to do the Sub ? but they are all adjusted and set at zero
Would this normally be the next step to complete in turning a system having completed the Subs
Thanks in advance Neil
@@neilpage1185 Once you phase align dual subs, you should treat them as a single subwoofer and yes, the impulse peak of that single sub should be at time zero along with all the other speakers.
To apply the inversion correction technique, you need a processor with convolution capabilities. No receivers can do that as far as I know. If you are not using a FIR filter capable solution I suggest you follow the method in "Manual DOlby Atmos Calibration" videos.
Thanks for replying. One more question. In SPL alignment, the info window shows align offset -1.18db for L and +1.18db for R.
Are those the numbers I put in for gain in Roon Speaker setup or do I reverse the signs?
In Roon speaker gain settings, you've to use them as is.
Wow, very nice tutorial!! Thank you very much! Is this so good as a Trinnov system?
Yes, it is!
Hello, this is a great video thank you. I followed it carefully and did a test sweep once I had applied the filter and it did a great job (particularly at the lower end of the frequency spectrum). Where the filter inverse curve is above 0 db it goes up to just under 5db which I assume is the amount by which the filter will boost the response at the relevant frequency? I have read a lot about limiting the amount by which a filter boosts (to avoid driver damage). If I am right when I say a 5dB point on the filter means a 5db, is that amount of boost OK?
+5dB is fine below room transient frequency which is around 200Hz. You can adjust boost limit with "regularization" during vector division. 8% allows up to 5dB, 25% allows 0dB (no boost)
Thank you, all clear. I also meant to ask about your house curve tip - is that instead of the Harmon curve or is it in addition to the Harmon curve (i.e. you have both)?
Recently, I prefer to use my speakers' anechoic response as the target curve. Although this will lack a bit in the bass department, it yields the clearest results. Bass you will get above the limits of your speakers will make use of room reflections which are by definition delayed waves and causes the bass waves to rain into following higher tones. Still, a lot of people love deep bass and don't mind the group delays introduced. Harmon has deeper bass then Dr Toole's curve.
Thanks, tried it now works great. But where do I put in my +0.3 and -0.3 to compensate for the SPL in the speakers. Can I use Roon for this ?
Yes you can. It's under "Speaker SetUp" in Muse.
In some tutorials the speakers are NOT measured individually . What advantages and disadvantages does this approach bring with it?
Great tutorial! Thanks! Can you please advise how in REW do I cut only selected frequency range for the filter once it is ready e.g. I want to only affect 20 - 200hz
Do the A/B division as shown in the method with 0%regularisation adn no frequency limits ticked. The when you're applying "1/A" to the result of the previous division "A over B" with 8% regularisation, tick the "Lower" and "Upper" frequency limits and enter 20 and 200Hz. Leave Target level at Auto.
@@ocaudiophile Thanks!
Would you recommend doing this for home theater. If you ran Dirac after this process, would it ruin everything you just did?
Is there an 11 channel dsp you recommend?
The process is also good for surround systems of course and can be done with JRiver if you use a HTPC for example. Trinnov in fact is a16 channel HTPC. Dirac is doing something quite similar to this but user satisfaction is varied. . I never used one so cannnot recommend but Minidsp has multichannel FIR filter capable units.
@@ocaudiophile thank you very much. I just found your channel and have been enjoying all of your videos. I’m subscribed and will continue giving 👍
Thanks for this approach to DSP. In checking my results I see that L Corrected and R Corrected are mostly above the alternative house curve that you suggested. I thought they should be straddling the house curve. Any ideas why this is happening? Thanks.
REW trace arithmetic division may add up to 3dB offset to be able to calculate Fourier transforms and avoid division by zero, it doesn't effect the actual filter.
Hi OCA, First i would like to say thank you for sharing all your knowledge in these matters 🙏👍
I tried this tutorial with Apo EQ and it gave nice results, but
I use a Minidsp 2x4 HD and I would like to not be dependent on PC as a source.
Do you have any suggestions? Would it be possible to use the 1024 taps FIR and the 10-band PEQ together or something like this? Sorry for the noobish question! 😊
REW cannot produce filters with just 1024 taps, too low. You will need to use REW's equalizers for correction instead of the inversion method and send them to MiniDSP. rePhase can produce filters with 1024 taps and maybe you can adapt the crossover phase shift corrections to work with MiniDSP although the resoltuion might be too low. Select rectangular windowing and energy centering for the best possible low tap filter.
Cheers 👍 I use the Minidsp as active crossover for my diy bookshelf speakers at 1700hz 24db LR. So firstly i should use REW to create correct overall response? and then I can try Rephase to create a fir filter adjust phase further?
🤘 /David
@@Derrvid Yes, that should be the way and would work although I haven't done it myself to this day.
Great! Thank you! 🙏👌
Thank you. You tell me that umik-1 have to positioning horizontally, parallel to floor. But in yout pdf it seem to me that mic is vertically positioning. Can you help me, please, to understand?
Only matters for high frequency response which you don't want to correct anyway. In general, if you are measuring a surround system with speakers all around the mic, use it vertical. If you are measuring a stereo set up, keep it horizontal.
@@ocaudiophile thank you very much
Thanks for the video. Much appreciated. I was wondering if the final results you shared were new measurements or just predictions of how it would look after the adjustments?
Calculated predictions but they will be very close to correct measurements
This is a great tutorial! I fllowed the steps and it does make a big difference. One thing I would like to ask is that after applying the Convolution in Roon I hear cracking sounds when the treble is high. I turned the convolution off the cracking sound is gone. May I know if you have any idea how to prevent the cracking sound? Thanks!
It could be clipping. Arer you also using upsampling in Roon ? This is known to increase clipping. Watch my new tutorial (launched yesterday in two parts), there's a lot of info there on clipping as well as some minor updates on the inversion method.
@@ocaudiophile Thanks for quick response! Yes I am upsampling. I will watch it and do some trials! Thanks again!
@@ocaudiophile Hello may I know the two videos you referred to are "Supreme Digital Room Correction with REW & rePhase" and "Excess Phase Inversion"? Not sure if I missed anything but I didn't find the content talking about "clipping". Maybe you used another word which is equivalent to "clipping"? I am new to this apologize for the ignorance! :-)
@@Pixel_n_Chip It's towards the end of the end of "Excess phase Inversion" video.
Hello sir and thanks for the great video. I followed your pdf guide and I have some questions. Firstly in the written guide you apply ir windows to l inverted mp and r inverted mp wich makes the correction files active after 1 kHz only. Second question: I use eq apo on windows and loaded the corrections to convolution filter. Sadly the correction starts after a second or so after the original sound and it makes an echo effect that lasts for seconds after I stop playback. Is it a fault in eq apo or maybe I made a mistake somewhere? Thanks for the answers.
If you're using one of the recent early access versions of REW, you're better off exporting filters in stereo with left and right directed to Linverted-mP and Rinverted-MP, 48kHz sampling rate and "nothing else" ticked for EQ APO. To remove the delays you mention, apply "Trim IR to windows" to each *-MP filter before exporting. You'll find this under "Measurement Actions". Hope this helps!
Hi, thanks for the tutorial. Some questions
1. I have Kali LP6v2, I had previously corrected by taking MMM of L and R seperately. Since my L speaker is in a corner, my EQ to Harman trained listeners target was about -17-18dB at some bass frequencies. Would this level of negative gain be too much for this method? Is there anything I could do while keeping the speaker in the same spot?
2. Although conflicting it seems a lot of advice is to only EQ up to Schroeder frequency. What is your opinion here? I notice this tutorial applies full range correction
Thanks
1. The inversion method in this tutorial will invert any peak regardless of the size but will limit boosting dips to a max of 5dB (which is also adjustable).
2. You correct for the room up to Schroeder's and you correct for the speaker above 500-600Hz and kind of correct for both in between. In other words, your filters counteract room reflections below the room transient frequency and above that they should correct for what's wrong at the source after eliminating all reflections in the room. Frequency dependent windowing used in the method aims to do just that so it's safe although far from perfect. a) Most speakers today have almost perfect frequency response at the source. b) There are better ways to eliminate reflections but it gets very complicated very quickly. By just inverting the response below Schroeder's, you get 90% of the way to optimal correction anyway hence the tendency to not EQ above Schroeder's IMO.
I've tried, and it works great, thanks a lot! Is it possible that the offset values in the SPL alignment get inverted in the "real world"? Or I did something wrong? I had to lower the L side by 0.3 SPL, and to make higher the R by the same value in the alignment process, but, when I used the result in Convology XT, the R side of the master was a little bit higher than the L side.
Which version of REW did you use? Vector division basics were changed in recent version of REW (Feb 17). "A over B" now appears 110dB below the target level. Although applying "1/A" afterwards still produces the same result, I didn't have time to check if the SPL adjustments advised in this video are still valid. You can adjust SPL offset of "1/A-MP" for each speaker until the AxB operations yields results with similar levels. You can also use volume adjustment in your convolver. A neat way of eqaulizing left and right speaker levels volumes to each other is to use "SPL Alignment" tool in REW for the final results of L & R. If you choose 160Hz as centre freq and 2 Octaves width, it will adjust their volumes between each other from 80Hz to 320Hz. This pretty much covers all vocals so the singer's voice will always appear exactly at the center. You can check in the Info window the adjustment REW applied after SPL Align operation and dial in the same amount in your convolution engine.
Hi! I used the latest version. Ok, I will apply the SPL alignment as a final step too. Thanks a lot!
Hello, great video that helps me further to understand how to deal with creating filters by really understanding what to do and have full grip on the results. I first was using the latest stable version of REW 5.20.9. Then at the part where I needed to use the “Trace arithmetic” windows I missed some options. After downloading and installing version 5.20.11 I was able to use these options. Later I will continue using version 5.20.11 and try to finish it with support of your video. I am really curious about the results.
I was first thinking of using the EQ option in REW to “automatically” create the filters. I have the impression that this method is able to generate better filters, but I am not sure. At least I am learning by doing this. I will be using the filters with HQ Player on a self-build high end computer streamer (Jcat cards, linear power supply etc.). Thanks for the great work.
Although automatic EQ does a descent job, it cannot be as accurate as the inversion method. I suggest, you also download the text tutorial linked in the comments as it has a lot more detail and additionally includes phase correction and test songs.
I did download the pdf and I am reading it. I will start creating new measurements using some different settings like you suggest in the pdf (Length: 4M - 87.4 s for example).
I will use the filters in HQ player on my streamer.
Jussi Laako (developer HQ player Signalyst) indicted to create filters with a sample rate of 192k if the highest sample rate of the source is also 192k. Of course, I could do this when I export the filters but I thought it would make sense to create REW measurements also with a sample rate of 192 to avoid converting the sample rate later during export. What do you think? What makes sense?
@@JeroenDortmans I would rather leave the upscaling to the HQ Player and just produce 48kHz and 44.1kHz convolution files in REW. Upscaling is quite a complex process to do properly. However, in my experience the corrections usually make such a positive difference that sampling rate differences become far less audible in comparison. Also make sure "Apply IR Windows before export" option is selected when producing the .wav files. It will increase the processing speed of HQ Player massively.
In HQ player I up-sample PCM to 1536 kHz and DSD to DSD 512. Jussi suggested to create filters in the same sample rate as the max. sample rate of the sound files used which is 192 kHz. HQ player will first apply the convolution filters and after that up sample to the selected rate.
I will create measurements by REW in 48kHz like you suggested. Later I can always export them in different sample rates and test if it makes a difference.
@@JeroenDortmans exactly
Hey sir, thank you for all!!! But where is the latest version (61) of rew with the new updates??
Google "REW early access version"
First of all thanks for your excellent work.
But can you please explain why the target level during the inversion needs to be set to 3.6?
When I'm doing this, there is an offset between the target curve and the product of the correction and the original speaker curve.
It's due to offsets applied during fast fourier transforms to be able to perform vector arithmetic operations correctly. The resulting offset from the target curve will not have any effect on the actual sound. Also I think depending on the REW version you're using, the offset might change a little because the very latest early access versions of REW had some minor changes there. But as I said, none of them will effect the sound, just visual.
I just found the video and pdf last night, and I got to admit…. I’ve been searching for something like this for a long time. Very excited to try it out.
Thank you
Quick question …. I have the buchardt s400 speakers and your ears are not supposed to be at the Tweeter level, but right in the middle of the speaker in between the two drivers. When I set up the mic, that is probably the appropriate position for height, correct?
Many thanks!!!
Then just keep the mic at your ear height for measurements.
Thank you so much for the lightning fast response. I can’t tell you how excited I am to try this out. I can barely sleep last night….LOL
So….just out of curiosity, because I trust your opinion. What would you say is the best “automatic” room correction out there right now? Dirac, Room perfect….etc.
Thanks!!
At the moment there's no automated software I know of which will realistically improve a well setup stereo system with symmetrically and carefully placed speakers and a central LP in a rectangular shaped room. The sound will have flat freq. response but will sound throttled. But almost all of them will make improvements in systems with major fundemental placement problems and suffering badly from standing waves.
For HT systems, both Audyssey and Dirac Live will make major improvements because even time and volume alignement of so many speakers will result in acoustic benefits. But I wouldn't fully rely on neither for my own setup.
If you want to calibrate your system manually; REW, DRC (free) & Acourate, Audiolense (paid) are all programs with capacity to achieve optimal results but they have quite steep learning curves.
Two comments:
a) At 5 minutes +/- you talk about moving the speaker for arrival times. It is far more likely that the microphone is positioned off the absolute center. One way of finding the center is to play pink noise through both speakers and use the RTA function (or you can measure, but it takes longer) to find the spot where are no comb filtering, simply by moving the microphone. This is the position where your speaker are summing. If this physical location of the mic doesn't make sense, you need to double-check the speaker location.
b) it would be nice to have audio on both channels of your video.
You can use Mic Alignment Tool of Acourate Trial software for precise mic centering but the idea here is that user has already placed the mic at the LP and speakers need to be moved accordingly.
I used a mono Audyssey mic for the video hence one channel only :(
@@ocaudiophile Does moving a speaker by 1mm really change the end result audibly, given that one's actual listening position changes by far more than that every time one tries to return to that same position?
@@stephenjarzombek2903 You are absolutely right that it does not make an audible difference. The reason for the requirement of very accurate alignment of left and right speakers is accuracy of trace arithmetic vector calculations.
@@ocaudiophile Because the correction is applied in the software phase of the reproduction chain where there is no delay between left and right signals?
Hi. You made a great tutorial. Respect. I have a question. At which stage of the pdf tutorial can I improve the correction impulse so that there is no distortion around 2'00`` and 2'30`` (vocal in the upper frequency range)? The track is "Statements (Acoustic)" by Loreen. Thank you in advance. Best regards.
Thank you. It's hard to know what is causing the distortion without seeing the actual measurements but since the method will normally filter any peaks around that region, it could be phase related. If that is the case, you can try correcting the phase manually in the 200-230Hz area by using "paragraphic phase equalizers" in rePhase, just be careful not to use high Q and high phase shifts (Q>2, shift>45 can cause pre echo effects). In the guide you can do that in Step 29.
@@ocaudiophile Thank you for your quick reply. I will try to change something during preparation "phasefix.wav". Temporarily fixed by increasing the L and R attenuation (2dB) in the Brutefir settings. I use volumio and brutefir plug. The distortions are gone. Music sounds great. Thanks :)
@@mariuszz3376 I found something similar. When following the tutorial, and exporting the convolution into rephase to check digital levels/headroom, REW had the file sitting at +6db. This would cause signal clipping. To solve this I exported the convolution into Rephase, and set the highest peak to 0db.
Do not tick normalise ir peak during rew export
Thank You fot this very informative video, however I can't get through on last step, namely to load the correction file to minidsp, it says: " you are not allowed to upload more than 1024 coefficients"? What am I doing wrong? Can someone help please?
You're not doing anything wrong but Minidsp is not capable of using REW generated convolution files. It will only accept REW biquad EQ filters.
@@ocaudiophile Thank you for your very fast answer! So I can only use my minidsp 2x4HD with IIR correction using REW or I have to use other software, can I?! I'm very dissapointed...in spite of this, I'm very glad for Your videos, they are very informative!
Thank you.
@@SteveSoldier13 That depends on your source. If you're using a PC as a player for example, you can use the free program Equalizer APO which will accept all the convolution files you generate in REW. Roon and JRiver both have capable convolution engines as well but they are paid software.
@@ocaudiophile Thanks for the tip, I've already known this app but not about this exact feature! However I'll rather stick to using minidsp (physical tool) so this way it's always in the way of audio line independently from whatever source I'm using. Anyway it's a very good tip, I thought I can only use EAPO for IIR correction!
Fantastic! I had an amplifier with Dirac and wanted to upgrade. I also use Roon and did not understand how to use REW until now!! What an IMPROVEMENT with my new amp and this convolution filter! The only thing for me is that the vocals sound a little "thin". Is there a way I can let the vocals sound a bit fuller? I use Roon equalizer to correct that a little but there is probably a better way to do within REW?
Find me in Roon community and send me your REW mdat ;)
@@ocaudiophile
Hi i cannot send a pm in Roon community or do i miss something:)
You can post publicly?
Thank you for the digital equalization video. I'm encountering an implementation roadblock: When plotting Overlay:Impulse response, REW shows (what looks like) the filtered IR, not the step response. I'm using V5.20.13. I can't see a way to time align using the filtered IR. Is there a setting I need to change to get Overaly:impulse response to plot the step response? Thank you
You need REW early access version v5.20.14. Just Google it.
It took me a couple of weeks, but I figured out the issue - it was a failing USB cable.@@ocaudiophile
Very interesting work. I will try it out myself.
QUESTION: Instead of using a real time convolution engine I assume I could convolve a few of the actual audio files and play them directly to the speakers for an A/B comparison?
Yes, that's possible. I can't remember the name but there are free softwares that can do this.
@@ocaudiophile - Any Reverb plugin that accepts IR's should, in theory I think, be able to do this. Sounds like a fun experiment to run if I ever get some time.
Great video, what is the UMIK2 like? I own a UMIK1, is the UMIK2 a big upgrade, what are the improvements by comparison?
It has a 32bits 192Hz interface so resolution and dynamic range is higher but as long as there's a calibration file in place, you wouldn't see significant differences in measurements. They both have the timing issues inherent to USB interfaces. I'd rather keep the Umik-1 and add a descent analog mic if I were you because USB interface is so practical.
@@ocaudiophile okay, thanks 👍.
Hi, Curious to try out this method. Would I get much better results with your "Supreme Digital Room Correction with REW & rePhase" video or how would this process compare (I see in this video you are just staying in REW and not needing re-phase)?
This video comes with a written guide and in that you'll find details of rePhase crossover phase correction stage. The Supreme correction video has additional 2 more filters on top of that but it's harder to get right.
I am just getting started with your tutorial. I am using 5.20.14. I have both L an R channel in the SPL display window and I clicked "Align SPL". L channel Shows .13 dB offset and R channel shows -.13dB offset. Does REW apply those offsets or do I need to "Add offset to data"? Is it normal that they would be the same except for positive and negative?
Thanks for any advice.
"Align SPL" automatically applies these changes.
Wonderfull tutorial! I tried in the past and it works flawless. Now with a Goldnote I got an output issue. L channel doesn´t work (John from REW told me it could be a muting issue of some DAC). I got some result in this way: I did measurement for R ch. and then I switched R with L line output cablea in the dac in order to let REW think that I was using always R ch. The issue is that I did it with "no reference timing". With reference timing acousti REW blocks at 1% even if I turn up ref level.
Now the convolution filter with inversion that I created seems to work quite good but for 44.1MQA. It works for 44.1 standard and 96 MQA, but with 44.1 MQA the sound is accelerated (about 10%). is that related to "no timing reference" mode or it´s something else? Do you think you could help me a bit?
please!!😇😇😇
Thanks for the compliments. I didn't understand what you mean exactly by 10% acceleration. Is it playing at a faster speed?
@@ocaudiophile yes, faster speed, about 1 second faster every 10 seconds.
@@psycofurious interesting, never heard something like that. Sounds like a DAC sourced problem, unlikely to have anything to do with correction. Try to set player to play silence before memory or any similar settings to get the async working
@@ocaudiophile sorry, I lost you, I use Roon and I don´t know exactly what you mean and how to do it. BTW thanks! Could you please tel me if using or not a ref timing acoustic could change something? is it necessary only to check distance LP to speakers? do you think it could be up to the DAC even if I got faster speed only from some freq.?
Sorry this comment was being held as spam by TH-cam!!! I really don't have a clue why the accelration is happening but it's your DAC causing it somehow.
Amazing video. I am working with my measurements (and i also tried with yours downloaded) but I am having a problem when I am creating L Inverted, because the SPL is around -50dB instead of 0db. I can't see where I am doing something wrong. Do you have any suggestion? Thanks
It's due to a recent revision in REW, results will not change. You can find details in the video notes.
Hey, OCA. This is very fascinating. Is there a way to do this for my home theater? I just have Audyssey so I can't use IR files. I am a complete noob to this. I am googling but can't find much yet haha. Thanks for the video!!
Watch this one:
th-cam.com/video/aZ4k04uAf2o/w-d-xo.html
and this one:
th-cam.com/video/LwORN-tSPjk/w-d-xo.html
Thanks for the tutorial! Any tips or links to point a relative noob to regarding how to incorporate a sub into these measurements. I have a 2.1 system and with my sub I get some weird L + R readings in REW. Should I measure without the sub on? Or is there a better way?
IMO you should crossover your SUB with your front speakers at or below 80Hz and then take a combined measurement and correct that.
If you want to use them altogether at all frequencies then watch this video I made yesterday. You can use the technique to time align your SUB with L+R:
th-cam.com/video/ga2eOwJRtXo/w-d-xo.html
@@ocaudiophile Thank you so much for the quick response. I figured out how to get the FIR without the sub and now just need to incorporate that. This is super helpful. Cheers.
OCA - I followed your instructions above. I used REW to help me first integrate my sub: finding the “right” volume, placement, and crossover. Then I followed your guide again. WOW, BRILLIANT! I had always felt a little disappointed with prior attempts at room correction. This really takes my system t another level. One final question. Could you offer some guidance on how to to do the final speaker adjustment for distance correction? In my case: 1.05 inches for my right and 1.48 inches for my left…. Is this really going to make a big difference? I’m guessing my head moves more than 2 inches from my mic placement position. THANK YOU!! SUBSCRIBED AND LOVED
Thank you. If your sound stage is at the centre (most vocals should be coming from the middle of the two front speakers) then no need to bother.
Hi, thanks for the great turorial! Any chance to get this REW beta version for Mac?
roomeqwizard.com/installers/REW_macos_5_20_11.dmg
@@ocaudiophile Thanks a lot!!!
I have a question, after carefully following your videos, I have built several impulse wav filters. I have successfully exported it to Camilla DSP in Moode Player. The result is that it sounds very good in PCM, but very bad in DSD files, they lose naturalness and Stage, despite adjusting their gain. Do you have any comments about this correction on DSD files (Camilla resamples them to 352 k 32 bits)? Thank you!
I'm not very familiar with Camilla's filter upsampling procedures. Roon can match 48kHz wav filters fine to DSD tracks although it slows down and stutters with anything above DSD 128 in my Intel i5. Probably Camilla is not upsampling the filter for that reason. Try exporting the same filters at 352Hz from REW and see if they sound better with DSD.
Thanks for such informative video! I tried to follow the step by step with this vid but some function button are not the same with the latest version (5.30). Would you please kindly help if I upload my mdat? Thanks in advance!
Sure, insert a google drive link here...
@@ocaudiophile I read the txt tutorial and found some fault in my measurement, I will upload them once correctly measured. Thanks again!
@@ocaudiophile here you go mate! I live near a noisy street so it's literally impossible to do a 4M measurement without cars / scooters going by so I did a 2M instead! drive.google.com/file/d/1QOLJ805q5sLe-rx51F8OkWhndtGB2IUr/view?usp=sharing
What do you use for digital correction?
@@ocaudiophile I use Roon
Great video, however when I do this, my final result seems to be about 4 db over the target curve and doesn't fit underneath it like your example. I am wondering if it's a difference in the version of REW? I have the newest version and some things are a bit different.
Some convolvers ie Roon normalise filters and add/remove gains. As long as you check for clipping, that's fine. The method has only cut filters no boosts.
@ what actually happened in REW is that when at the end we show the target curve and then overlay the correction, your correction fit nice underneath the target - your target line was the top and everything came below, but my correction was above the target line. Do you think the new REW maybe has other things happening behind the scenes or different standard settings that I have to change, to make the calculations work correctly?
Are you using minimum phase versions of 1/A filters?
@@ocaudiophile yes, I followed step by step process in your video, there were slight differences in the menus of REW because my version was more recent than what you used in your instructions, also it seemed there was an automatic alignment tool for the impulse response, where your video shows how to manually move it, my room’s impulse response also looked a lot different than your example but everything else I followed exact.
Hi. Gorgeous tutorial but how to make fir file (bin file) for minidsp 2x4 hd with online 1024 tap? Many thank's
You cannot adjust number of taps with REW generated convolution files and the default values are much higher than 1024. Your easiest option is to use REW EQ filters for correction, export them to rePhase and generate the filter with desired 1024 taps in rePhase. I think you can even save it as .bin. One advantage is, you can also align crossove phase shift, etc in rePhase in addition to the REW filters. To reproduce descent filters with so few taps, you may need to try different windowing types like rectangular and energy.
Hi Luca, have you by chance found a way to do this on Mac? Just spent some time doing everything in this video, only to find that MiniDSP won't load my fir :(
@@angeloboltinimusic Sound Forge can do this but it's not free although there's a free trial period. Load the WAV then save as "Raw" and in the "Custom Settings," select "32 Bit, IEEE float.". I am not sure if you can match the number of taps to Minidsp standard though.
Another way could be writing a Matlab (or the free version Octave) script to convert the files. It could be something like:
[x,fs] = audioread('LR.wav');
out('L.bin',x(:,1));
out('R.bin',x(:,2));
function out(filename,x)
id=fopen(filename,'wb','ieee-le');
fwrite(id,x,'float');
fclose(id);
end
But I have "zero" Mac experience and my Matlab skills are very limited but I guess this is the right direction.
@@ocaudiophile I use 1024 taps Fir filter for hi and mid driver. This means in my case down to something like 200Hz. But anything under 1000Hz will not produce a filter that can do accurately what the video is showing you. So depending on your setup try to produce 1024 tap fir filters with RePhase that works in the Hi region from 1000Hz up. Then do with REW the normal target curve EQ generation for the signal under 1000 Hz. This is the way to go if you are restricted on taps in your system. The fir filters will insert much more air into the sound in the upper region. IIR filters can perform well in the lover region.
If you are playing music from a computer then you can measure your setup with the filters explained here before. Then do the FIR filters that the video does explain. Import the filters into RePhase and adjust the phase. Export the filters and convolve these multitaps (long filter) in the PC on top of the MiniDSP filters (Fir + IIR). This makes it possible to correct or finetune the full bandwidth.
@@spalmgre9167 Excellent explanation. I don't think it's possible to do with MiniDSP but an alternative to correct low frequencies with 1024 taps is frequency warping. I think Audyssey is using this technique with their 1024 tap FIR filters. The idea is to replace the delays with first order allpass filters. That makes the overall contraption an IIR filter, but it maintains the overall structure of a FIR filter. Instead of a chain of identical delays you have a chain of identical all pass filters. The output in both cases is a weighted sum of each node in the chain. Warping maintains the overall shape of the filter (peaks and dips, etc.) but it moves it to different frequencies depending on which direction you warp.
hello is it possible to combine the two interventions to fix the speaker crossover phase and create these convolution filters?
Yes, just multiply them with each other. Trace arithmetic "Ä * B"
That's a great tutorial and it actually produces brilliant results!!
Thank you very much for this!
May I ask a couple of questions though, being a newbie to this?
1. When I follow your pdf tutorial and make the MP copy, the windows limits must be opposite than what you describe for it to work i.e. 1 on the left channel and 125 on the right. Does this matter?
2. Why is the corrected vector average so much different (worse) than the RMS?
Thanks again!
Thank you and happy to hear you had great results.
1. In the pdf, the suggestion for the MP's are: “Left Window: 125 ms” / “Right Window: 1,000 ms” which is a good chocie and there's not any in the video? Left cannot be 1ms, that would compeletely ruin the correction as it will massively change the magnitude response of the filter. You can check for the effects of left and right windows under the "Impulse" tab by moving the L and R (green and red squares) with your mouse over the time axis. If you cannot see them, make sure "windowed" is ticked and you have zoomed out enough in the impulse graph (the graph on top). The bottom graph is the magnitude response of the filter.
2. Because of phase differences between your left and right speakers. Ideally RMS and vector averages should be the same in an ideal room. Think of phase as time delay (it's not exactly that) and the same frequency coming from the two speakers with such a time difference at certain frequencies that they cancel each other out and causes dips there or vice versa. RMS on the other hand just averages the SPL values.
Thank you very much again! I understood 1=1,000 by mistake, hence the strange results I got when I followed the pdf. I'll revise that and review again. Perhaps that could affect the vector average as well.
If any of your windowing is 1ms on the left, correcting it will change the response a lot!
@@ocaudiophile I got this right now; comma vs dot does make a difference! One more question please: if I want to do further adjustments to the generated responses, can I just EQ the final L and R "Corrected" response measurements in REW and export new IR or will I just ruin everything?
Thank you!
Hi. Thank for your great demonstration. However, I can’t convert the file to my Helix DSP as only text file can be input.
I don't have any experience with Helix and from what I quickly searched, looks like it accepts IR files in wav format but only with 1024 or 2048 taps. I might be wrong. You can use its parametric equalizer to copy REW EQ settings and probably can also fit a phase correction within the 1024 tap limit.
@@ocaudiophile tried but still unable to accept wave files. Still trying to find ways to convert into text files.
Thanks. It worked. Hard to master tho. But, why using Harman curve? The studios are using flat curves. So the mixers could already have added bass respons similaer with Harman.
Harmon (short for Harman Kardon) curve is in fact developed for headphones. I find it a bit bass heavy and lately prefer Dr Toole's curve. These curves are in fact flat in their perception by the human brain but that's volume dependent. So what curve you will apply also depends on what volume are you measuring and what volume are you going to listen to your system. That's why I lately advise to measure at the volume you listen to your system. Check equal loudness contours and you will understand better:
en.wikipedia.org/wiki/Equal-loudness_contour
Hey OCA, Is FDW best used at 15 cycles? Does this seem to be a good value for all frequencies? Also, does it matter if I smooth first and then FDW or do you FDW first and then smooth? Thanks.
15 is quite well suited to most cases but the actual number depends on how reflective the room is and how far is the LP from the speakers. For highly reflective rooms lower numbers can yield more realistic results at high frequencies whereas higher fdw values (or none at all) can result in more effective filters in the low frequency region. A better fdw function in theory would be one which can apply varying degrees of fdw throughout the frequency band IMO.
I'd use VAR smoothing to disable over correction in the high freq area.
Can I hire you ?😊
I have a simple request, I'm trialling down firing sub and effect of distance... How do you do a measurement, calibrate the eq and plain it back and resmeasure using the applied eq?
New REW Beta has an option to run a measurement with the applied EQ:
www.avnirvana.com/threads/rew-api-beta-releases.12981/
www.roomeqwizard.com/help/help_en-GB/html/eqwindow.html#:~:text=Measure%20with%20these,clipping%20the%20output.
@@ocaudiophile Brilliant!
How may I positioning the UMIK 1 for The first step with REW for aligning the speakers ? Horizontally or Vertically ? And what file I have to use 0° or 90°? Thank you
Horizontal for stereo system measurement and 0 degree file.
@@ocaudiophile Thank you. When is necessary vertically position of mic and 90° file?
@@Angelos58 If you need to measure surround/ceiling speakers like in an Atmos system.
@@ocaudiophile very kind of you