Why DIRAC can correct above 500hz but you shouldn't?
ฝัง
- เผยแพร่เมื่อ 8 ก.พ. 2025
- Got questions about Home Theater, Acoustics, AV equipment, or the science behind it all? Drop them in the comments below, and I GUARANTEE to answer all the questions with a TH-cam Super Feature "$THANKS" ($30 minimum) in an upcoming videos! Here's how:
Use the TH-cam mobile app to open a Short or long form video. Under the long form video, select Super Thanks . If you’re watching a Short, tap Comments and then Super Thanks .
Choose the amount that you want to spend. Select "BUY AND SEND" to complete your purchase.
Your comment will be highlighted with a colorful chip, making it stand out in the comments section, and your support will be publicly acknowledged.
If you'd like to book a private acoustic consultation with me or are interested in purchasing top-tier Home Theater equipment, please visit our website: www.poesacoust....
We proudly carry industry-leading brands like Perlisten, Real Acoustix, Trinnov, RTJ and more.
Follow me:
TH-cam: Subscribe to this Channel / poesacoustics
Website: www.poesacoust...
Facebook: / poesacoustics
Instagram: / poesacoustics
Email: info@poesacoustics.com
You appear to misunderstand the basics of signal theory when claiming full-range EQ is achievable with "mixed-phase" filters. The steady-state response is not minimum phase in either the low or high frequencies. Moreover, the inability to equalize high frequencies has nothing to do with filter phase type. Instead, it stems from the lack of directional phase information, which an omnidirectional measurement microphone is inherently incapable of capturing.
The decision to use the so called "mixed-phase" filters by Dirac appears to have been merely influenced by constraints on filter size and computational power, aiming to achieve sharper cutoffs with lower orders than min-phase filters and all it does is to attempt to time align speaker drivers by using time-inverted allpass filters. And they apply frequency dependent windowing to the response falsely assuming that gives them the space to apply full range eq. The result is throttled sound even when they get everything else right (rare). It's a bit tragic they charged consumers extra for full range eq all these years and soon you and all other dedicated! supporters will run out of creative ideas to defend their tech.
Thanks for your comment-there’s a lot to unpack here. First, the claim that full-range EQ is unachievable with mixed-phase filters seems to misunderstand their purpose and capabilities. Mixed-phase filters (a combination of minimum-phase and non-minimum-phase filters) address not just amplitude response but also time-domain issues like group delay-something minimum-phase-only filters can’t do. While it’s true that omnidirectional microphones lack directional phase information, room correction systems don’t rely on that alone. Advanced algorithms use multiple measurement points and modeling to approximate the acoustic behavior of the room as a system, not just at a single location.
Your assertion that Dirac’s choice of mixed-phase filters is driven solely by computational limitations and filter size feels speculative at best. Mixed-phase filters aren’t a “shortcut”-they’re a deliberate design decision to address problems that simpler minimum-phase approaches leave unresolved. And while it’s easy to throw around phrases like “throttled sound,” the reality is that properly calibrated mixed-phase correction demonstrably improves time alignment and frequency response in a way that’s audible to most listeners.
You also mentioned that “everything else right” is rare-this seems more like a critique of improper setup or user error than the technology itself. No system, no matter how advanced, can compensate for poor calibration or bad implementation.
If you have alternative methods that deliver better results across amplitude and phase, I’d love to hear about them. Until then, mixed-phase correction remains one of the most effective tools for improving real-world sound reproduction.
I don’t think I have the misunderstanding here.
Here's a scenario I have encountered and it requires correcting full range. If there's a cancelation below the Transition Frequency that Dirac can't flatten, I pull the entire Target Curve down a few db until the Target is near the bottom of the cancelation. Dirac then pulls the overall response down to flat and eliminates the cancelation. This requires that you allow Dirac to work full range. The result after you load filters is that the speaker level may be too low by the amount of the Target drop ... so just raise the level trim in the processor and you have flat response at the correct level.
Thanks for the video! I'm new to Dirac BC and very pleased with the improvement over miniDSP, YPaO and Audyssey. Very interesting to hear the Perlisten and Dirac issues you've run into. (for those of us who may or may not be interested in specific combinations...🤐)
14:41 Erin uses a klippel as well
The better a speaker’s directivity is, the better Dirac and other eq systems work. Dirac does an amazing job with single point source design speakers like kef.
I've got a Genelec Coaxial center, cardiod mains and Kef LS50 Meta for rear and upper layer. There's no difference (at least to my ears) whether the correction is limited to 350 Hz or full-range. Looking at the before-correction-measurements you can also see that from about 300-350 Hz and above the FR follows the desired Harman curve without any significant dips or peaks.
It's really the speakers that are important. Before this I had cheaper on-wall speakers and thought EQ/room-correction would make it sound perfect. But no matter what I tried (no correction, Audyssey, DLBC), it all sounded horrible.
@@syntheee
You're right, ... very nice!
A well executed coax possesses inherent advantages in this regard, ... ie., the lateral uniformity of the soundscape.
Spectral uniformity (superb polars) and the coherence (point source) of lateral reflected energy ... all add up to a solid playback experience.
@@syntheee I’ve found this to be roughly true in most systems I do as well.
It’s always amazing how good a response is above the modal region when the speakers used are excellent. Perlistens, Kefs, Revels, and Genelecs (among others) often yield amazing results and these correction systems often don’t do much above a certain point.
Of course the room plays a role and sometimes even these amazing speakers struggle in certain rooms without a bit of help. None the less, I would generally prefer to start with a good speaker.
I am aware he uses a Klippel. My point related to this known HF error that shows up in some measurements of the Perlisten. Some correction systems show this as well. We have been investigating the source of it. It’s not in the native response of the speaker and seems to come from some strange interaction between the method of measurement and nature of the HF response of the Perlisten that high. A few theories have been thrown around.
What we know for a fact is that the rolloff isn’t really there.
Yes you should absolutely correct full range. You simply cannot have phase correction without applying EQ full range.
Thanks for the video !
Hi Matt, very nice Video! I made an interesting experience in our L shaped living room on dirac. The linear Dirac chosen target curve sounded quite lame and dull. So i worked with the target curve and modeled the range between 2 khz and 15 khz. Step by step i could reach a very very similar behaviour in respect of clarity and transperancy as on the manual setup with my AVR, i did before. But of course much more controled with a very good base behaviour and a surprisingly great seat2seat consistency on 4 sofa seats. So... I think, the decision on which frequency range to apply the correction on is much more complex that just focussing on spl linearity, as you also said if i got it well)
Another great video. Thanks! I'm looking forward to more of these calibration-type videos. I do wonder how potentially needed compensation for an acoustically transparent screen factors into this discussion for folks who use Dirac. For example, would adjustments for an acoustically transparent screen increase the argument for going to a frequency above 500 Hz?
I left a comment in another video regarding issues with dirac.
With some speakers it doesnt seem to understand what its measuring and giving some very strange corrections.
This can happen from 250hz up to 20 kHz.
Seems one should do a control measurement with rew to make sure the curve is whats wanted.
Interesting stuff, thoughtful presentation imo. I personally find it hard to trust a system that exhibits inconsistent anomalies when the workaround requires more expertise to figure out than the original objective. What other unpredictable or hidden anomalies, artefacts or side effects are happening that require troubleshooting? At which point does taming your black box become a rabbithole, simply because of the sheer number of variables and unknowns, on top of the "known goods" like the quality of your speakers. So then it's back to trial and error and testing by ear, coming full circle it seems.
I guess this is the trade-off: it is a bit of a gamble to trust dsp's and algorithms and the way the data modeling is applied to the signal, the actual process is near impossible to fully analyze, but there is no real way that I could do any of it by hand. And of course even the most basic system reacts in more complex ways than we can ever dream to fully map out.
The win is you can indeed get a better sounding system, for the purist who prefers to deal with "raw" imperfections as "character", it's all way too much, and it's back to the cleanest possible signal fed into the room.
One fun aspect of exploring the depths of audio is: it never ends :)
Great video, don't take this the wrong way but you're a very high end guy. With your knowledge it would be cool to hear what your home cinema would be with a 20k budget.
@Palindromeguy1 my recommendation for a $20k theater?
I could probably come up with something for that. But honestly it would involve a bit of speculation.
It might be interesting to show a few options and point out the compromises you end up having to make. $20k is a lot of money in absolute terms. But when it comes to home theaters it’s not and so there is no way to make a really high performance system without serious compromises. Especially at retail pricing. Obviously used products and special finds might help here.
@PoesAcoustics it was a lot harder than I thought, with your knowledge it could take a while. Good luck lol
To your point about the perlisten significant roll-off in the high frq. that we see in some of the measurements online, just to be clear i'm by no means professional nor i own professional equipment, all i have is the umik-1, cal. files and REW, i just wanna say i do agree with you on that, i have measured my Perlisten R5T and i didn't see that significant roll-off either, my R5T measured flat up to 24K. which was the end frq. in REW measurements preferences prior to taking the measurements.
@@mdelhaj8858 yes we did extensive testing to identify the source. Not only did four different people take measurements, but two of us used very expensive lab grade mics. The last measurement I did to confirm that this was not a reality involved measuring an S4B in a small anechoic chamber with a 1/4” GRAS capsule.
I also measured with PCB and GRAS 1/4” capsules in my room and outside on numerous Perlistens. I could never make that shape appear that I saw in the Trinnov measurement.
I tried different angles as well and I didn’t see it until a good bit off axis. I don’t personally have any idea the source. Erik from Perlisten had some theories but we never really got anywhere. I just know it’s best not to let Optimizer or Dirac correct a Perlisten up that high.
13:05 There is a way to do it.
What I find confusing is that Dr.Toole has emphasized on multiple occations that the tilted in room fr response is a result and not a target. So not sure if there is any research about forcing this tilted response with shelf filters. A speaker can measure flat anechoicly but depending also on directivity and reflections might not have an in room tilted response.
What you think of the use of Magic Beans along with Dirac?
I’m no expert, but I can tell you that I use both Dirac Live Bass Control and Magic Beans and it is fantastic!! I highly recommend it.
@PoesAcoustics What are your thoughts on room correction techniques such as Magic Beans where they use nearfield RTA as well as MLP for each speaker? Is it worth purchasing to complement or use instead of Audyssey, Ypao Dirac? Have you used or reviewed it? Your opinion on it?
Look into to A1 EVO, free of cost...
@@Rayman-expertsoundpro unfortunately I can’t test that. It hacks Audyssey and voids the warrants. I have been asked not to test it by the manufacturer.
Do you recommend for most people to curtain the correction at 500 hz or 300 hz?
I might be wrong, but if I don't correct the whole spectrum, my Ascendo speakers will be too loud higher up the spectrum as the lower part is corrected.
Can you do a video about wiim corection ?
What about Klipsch speakers that are known for their kinda harsher highs?
like Dirac is the Trinnov also a mix phase system with the ability to correct all the way to 20khz same as dirac? or does it require more manual intervention.
It’s basically the same. Different approach and a ton of manual control over the fixed factors in the algorithm. But it’s also a mixed phase system.
I should probably do a video on this now that I can capture my screen in higher resolution. All room correction systems of this nature are taking a room impulse response from multiple locations. Windows are applied in order to divide the energy into different time components. Direct sound, early reflections, and late reflections. It can then filter those separately and differently and this impacts a lot. This part of the correction includes FIR filters and so the phase linearization can impact imaging quite a bit. So with Dirac it divides it up based on common room assumptions and some automatic ability to shift some of the window parameters. But you the end user cannot change it. With Trinnov, you can manually change this. So it’s not manual correction, more like manual room size adjustment.
You can also change how many IIR filters, the filter resolution, and where it focuses its correction. All things that Dirac doesn’t allow. It’s really a pro only tool. We suggest end users don’t touch this stuff unless they are sure they know what they are doing. Some of these parameters could damage speakers if done wrong. All of them can ruin sound if set wrong.
@PoesAcoustics thank you for the detailed response, this was very helpful. It would be great if you did a video on this pointing to these controls that you do have access to on the Trinnov and where they can come in handy.
To the Trinnov treble problem (wich I also have seeb). Could the reason be that the microphones are facing 90 degrees up?
@@bernhardotto9201 the correct for that. It’s not the orientation (unless their correction is wrong)
Would you ever employ multiple mono / stereo tube based equalizer for the front left, center and right channels in between processor and amplifier. For various reasons to help rolloff on 70 and 80s multichannel music or even Lawrence of Arabia 4k UltraHd bluray that has some screeching highs when played at high DB?
I would not. That would be a euphonic approach. Tube based eqs are going to be imprecise and noisy compared to modern EQ methods relying on DSP. There is no reason to use a tube based EQ other than to apply a euphonic effect.
Check your spelling on the title. ?
Dirac is miles and miles ahead of audyssey. Audyssey is terrible with both their app and the algorithm. Dirac is incredible I compared them side by side. One major benefit is Dirac sets subwoofer latency to 0 which means tactile transducers don’t have delay. Audy doesn’t do that AND you can’t add enough negative delay to correct that which also ruins your phase correction.
Dirac takes bass out the room. Don't like it
Yeah you need to experiment with custom curves that suit your preferences. You can find them and import them, i like a 4 or 6db slope but that is all preference ofc.
If you are feeling like it took bass out it’s most likely that it flattened the bass to a level that you don’t like. As mentioned, try adjusting the target curve to achieve a bass level you want. It’s not uncommon for folks to like 6-9 dB of bass boost. I’ve also turned a handful of sysyems where the client wanted over 20dB of bass boost. In those scenarios I’ll often turn the sub level up by 6-10dB to avoid boosting that much.
@@PoesAcoustics Thanks for the hint on how to satisfy those of us with Don Dunn-level bass needs, on a modest budget. The pro tips are priceless. Would make a great video series...
Your room probably sucks and so does your seating no offense. Dirac is industry leading
I had this experience as well, particularly at the transition frequency between the mains and the subwoofers. I tried it again a few weeks later with a brand new set of measurements, and got a much better result. I really didn't do anything of significance differently, so I truly don't know why, but my second go around with Dirac was quite good.
Also, I assume you're talking about Dirac Live with Bass control? The more basic versions of Dirac are limited and far less capable in the bass regions.
What a pretentious explanation of a last name's spelling 😂 uptight much?
I would have mentioned it way sooner than he did because people misunderstand his name in the comment section of nearly every video he posts. I'm glad he cleared up the confusion.