@Bed_Bug I see what you did there, since your comment is only describing what the guy you’re replying to did. What’s with all of the uninspiring meta comedy nowadays?
Would love a video about how linear-phase EQ works to add on to this. This is the most insightful video I've seen on this topic. I've had so many questions about how EQing actually works at a mathematical level - now answered in a 4 minute vid.
DSP programmer here, programmed linear phase equalizers by hand. An analog-style EQ is what we call 'causal filter'. It "reacts", so to speak, to the inputs as they happen. Without look-ahead, the EQ can only generate output using the past + present. Using info from now + before, the filter can only affect what comes out next. So, phase shifts using analog/IIR filters can only "delay" phase. Now, a "linear phase filters": Imagine an audio process that could see a bit into the "future" (know what's going to happen). This ability, so to speak, is common nowadays (DAW delay compensation, etc.). With this ability, a linear-phase filter can "shift the phase" both forward and backwards in time equally (so to speak). This technique eliminates any phase shifting all together. However, since we're able to shift the phase of frequencies *before they happen*, you may be able to hear "pre-ringing" from the filter in the output. That's maybe an over simplification, and I've left out some of the more fun analysis/stuff (impulse responses, FFT's, etc.), but yeah.
@@gerudobombshellso I have a question for you. What do you normally suggest is the optimal way to EQ - specifically, the low-end. Do you use Linear-Phase, or not? I’ve seen tons of debate on cutting frequencies below 20-40hz because you don’t “need” information below that, and then I’ve also seen the opposing argument that even if we don’t “need” that information, it’s not WORTH cutting it out because of these exact issues. This forced me to the idea that I want to present to you since you seem to know more than anyone else I’ve ever come across. So could you take an Analog Filter and gradually clamp down on the low-end? It’s a concept I’ve called “EQ Clamping”, and I created a patcher preset for it. Instead of a simple High-Pass filter down at 30-40hz. It’s MANY Low-Shelf filters gradually clamping down. Low-Shelf filters being the choice because they create significantly less Phasing. After several of these filters, THEN you can GRADUALLY add HP filters with very generous slopes. Any thoughts? Would this help to mitigate the need for using a Linear-Phase Filter?
Generally speaking, minimum-phase/analog filters are great for most needs (no problems). Linear phase filters are great for making spectral changes without affecting inter-track phase releationships (think drums, close mic + overheads). However - back to the "pre-ringing" thing: Linear phase filters will cause a variable amount/length of pre-ringing, with respect to the frequency (Hz). That is to say, the lower the frency you're affecting, the longer/more audible the pre-ringing. And, the sharper the filter (eg. imagine a 48 dB/oct hipass filter), the more audible it is yet again. Since IIR/analog/minimum-phase EQ's only can ever "delay" the phase of waves, they will not suffer the same problem. Any ringing anomolies will manifest forward in time - which is nice, since transients will "mask" the ringing (IIR filters *do* still ring, btw). My rule of thumb: Most of the time, try to use IIR. If there's a phase relationship to another track, use Linear Phase (sometimes sounds cool boosting highs too).
dan, im studying audio engineering at SAE institute and your videos have been such a massive help for me to dive deeper into the things im taught about at school. thank you so much for making your videos ❤❤
Studying electrical engineering and than going to do audio processing makes way more sense to me. I studied mechanical engineering and I'm working in the audio field. Often we work with people how studied at SAE, and they often have proplems to understand what is going on with audio devices. How basic RLC filters work and srs, easy stuff like this. Sorry for interrupting you.
@@UltimateEngineering youre right, but ive got no interest in working in the parts of audio that are that down to metal. for the most part I want to specialise into mixing/mastering and live sound
The math behind (digital) all-pass filters is relatively simple, you just need a bit of discrete linear algebra and z-transforms. The first order IIR filter implementation is just a combination of the current input sample with a scaled previous input sample and output value (feedback and feedforward loops). As a difference equation: y[n] = a * x[n] + x[n-1] - a * y[n-1], where y is the output, x is the input, and n is the sample number. Fun fact: all-pass filters are also heavily used in many reverb algorithms.
👀‼️😳 ...Good God, Dan. EVERY time I think I know something about mixing or audio you calmly show me I know NOTHING 🤯😂 So THANK YOU once again for blowing my mind and making me a better mixer. You are a Rare Creature on TH-cam -- you ACTUALLY give fantastic instruction without pushing some Patreon mixing course or 1-to-1 classes, although if you DID have such things I'd more likely be one of the 1st ppl in line 😂🤷♂️ Wishing you best & just sayin' thanks 😎👍
Dan, your parallel filters videos inspired me to discover that when you delta solo an aux with any effect on it and send a track in your mix to the aux, the effect acts as if you have it on the track you’re sending to the aux. thank you for all you do! I now use this technique to create a global sidechain pumping effect that is controlled at the mixer level at each track and the global aux. I use this for summing bus compression when I want to glue kick and bass but I don’t want to sum there outputs. I use it for de-easing and spectral resonance suppression too. I really appreciate your videos and how you think. Thanks for all you do!
This sounds interesting but I'm not exactly certain what you do. Let me follow along, is this correct? 1. You have an aux with an effect, lets say a compressor. 2. You send a track to the aux. Lets say a kick. 3. Normally you get a compressed kick. But you delta solo. So you get only the "minus sound" that the compressor removes. 4. You also send the bass track to the same aux? So you get "minus sound' of kick+bass 5. Now what? you blend this "minus signal" back in the full track, so that you have the gain reduction (pumping) without having to route or sum the bass and kick together? That's pretty smart. Personally I don't really think I'd use this as prefer sidechaining a bass in a kick+bass combo, because I want to have at least one sound constant. For example in the context of Rock or EDM, I don't want a louder bass note making the kick more quiet, the kick needs to be constant. As for de-essing, that's (IMO!) better done on a per-track basis than on a sum of tracks, because de-essing on a sum of tracks distorts all tracks, not just the one with the harsh S in it. I can see this technique being very nice in ambient electronic or acoustic folk type music, where "naturality" and transparency is desired.
@@JDarkooJDarkoo Yes that is correct. This particular application definitely is more suited to acoustic genres. However what I like about this consept is it leaves the creative aspect of the compression (or any effect) open ended so that you can add any track you want to be affected by the same plugin. I use this for global high cut filter effects, creative transitions where I want some but not all of the parts be warped in a dynamic way. It makes the process as simple as sending to an aux and adjusting the send to taste.
I've always thought of filters in terms of an RLC circuit and the various impedances of the components. The maths to implement a digital filter just sort of falls out of that. Messing around in a circuit simulator (falstad, ltspice) is a good way to get a feel for these things
"How does the allpass filter cause the selective phase shift - some voodoo magic I suppose." Actually that's pretty accurate way to describe the math that starts from here on out. Like, in the end the equations look quite simple, but if you start looking into how those equations were derived... "Oh, it's simple. We first analysed the analogue circuit and derived it's differential equation. Than transformed the equation to give us the output voltage in terms of input voltage. Than discretized it by replacing integrals with discrete trapezoid integrator equations. Then we have resolved the zero delay feedback loops by treating the equation as an implicit equation and using it to derive intermediate results. But you have to consider equation's stability at this point. Finally we plugged those intermediate results into the main equation and that's it. Oh, also don't forget about warping the frequency range, but you can figure that out on your own."
Makes you wonder how the analog circuits came to place. With vacuum tubes no less, or old transistors, keeping those units fairly tuned must've been a nightmare.
@@BananaManPL I think this is why digital equipment often gets labelled "lifeless" or "sterile", because it doesn't have the analog grit everyone wanted rid of a few decades ago
Okay this is the perfect video to explain something I've been wondering. I understand how you can break up freqencies in a computer using an FFT, but I've been wondering for a while how analog EQ's work, and this is making more sense of that. I'm sure there's a lot more to it (for example, I nowhere near enough of an electrical engineer to understand how analog competents could introduce phase shift) but this video feels like it's getting me closer to understanding how that could work in the analog realm. Thanks Dan!
Your knowledge and understanding of this stuff from front to back, back to front, in multiple different examples, clearly demonstrated. Amazing as always, thank you.
Anyone who understands how analog filters work already understands that is the phase shift caused between the current and the voltage that causes the filtering. Lovely video Dan
Very clear and simple explanation. Great! I look forward to the next video explaining how all pass filters shift phase selectively. That was the first question I asked in my mind when you began :)
Well said Dan. I've said many times over the year that EQ is a consequence of phase shift - people look at you like you;re mad. But it really is that way round!
I really like how you talk in clear terms, cutting through the "easy tricks" that others post when advising on mixing and mastering. There is some confusion I have about some EQ elements. There is a few videos talking about mid-side EQ in order to remove overlap / mud that this causes in the stereo field. When I mix I tend to just mix on the defaults with side chaining to make room for the kick snare (to the bass and rhythm guitars) - which seemed to improve the clarity of the resulting track. But should we also focus on the mid-side equally. This was a little confusing, or perhaps badly explained.
nice, now replace the allpass filters with full on instances of disperser and you have a multidimensional 4th-wall breaking spacetime-bending hyperfilter
I thought I understood the basic workings of filters. I even built simple analog filters myself using capacitors, resistors and op-amps when I was younger. I understand the electrical concept of why they do what they do pretty well. I only recently first heard of the concept of an all-pass filter. This was after asking ChatGPT what the difference between a flanger and a phaser really was. At first hearing they even seemed useless, momentarily forgetting about that thing called phase. Functionally, they're still a bit mysterious to me, so thanks for this video. It sheds a whole new light on these things and they're really kind of interesting.
Hey Dan! Fantastic video mate! Many thanks!. I knew the concept that 'phase causes eq' for a while now, but there is just so little information out there, that or it's very difficult to find. Usually you just get articles or videos on 'how to use' a particular EQ and not how they actually work. EXCELLENT!!
I wish you released this a year ago when I was trying to learn how to make my own passive outboard gear from scratch for my audio engineering capstone project at uni. I spent forever trying to do everything with just high-pass and low-pass filters and boost the remaining signal instead. Turns out simple arithmetic was needed. Whoops.
I knew this, but it took me forever to find out cause no one ever explains it when talking about EQs. whenever you read about EQs and phase this relation should be the first to be shown.
Another amazing knowledge video from dan the question you say at the end "How does the allpass filter selectivly cause phase shift at specific frequency area " i saw in a digital signal processing that the behavior of allpass filter related to transfer function in math Interesting thank you for give us great knowledge !
@@DanWorrall - Notice I'm referring to the unwary. If you understand phase shift and are running a real-time spectrum analyzer (not just an EQ curve plotter), you can make expert use of phase cancellation. Otherwise, you're guesstimating with forces that can unexpectedly bite back. And as you point out in your linked video, linear phase filters are no panacea, they're rarely a significant factor outside of multi-pole crossovers.
@@QuicksilverSG there's nothing to be wary of with bell and shell filters. The phase cancellation will cause a tiny change in the shelf frequency or the bell width, both of which you set by ear anyway so you'll dial that difference out if it matters. You only need to be a bit careful with parallel high or lowpass filters.
Haha! Hmmm solving for x has a way of being so simply complex, doesn’t it! Out the door or window, into the rabbit hole, breach the subterranean trenches to find yourself standing in China in front of a rickshaw selling fried rice with shrimp to go and as you walk thousands of miles and who knows how many hours you find yourself back and square one and all you had to do was cross the street… to find the answer. Nooooo, it couldn’t be that simple; are you $hittake mushrooming me! 😂 Thanks Dan. Hope you are recovering well. Your video’s… I always learn and I always laugh. Appreciate you and your channel.
Please do get back to me on that food for thought you left us with at the end of the video. I’d be very interested to dive more in depth to the mechanics of all-pass filters.
Firstly, I'm not disagreeing, I'm only adding detail. Secondly: For digital filters: technically, your using delays in the range of one or two samples, feedback, scaling and summation to create those curves, but these delays can be seen as frequency dependent phase shifts, so 6 one way, half dozen the other. I know this is a distinction that nobody except me cares about but there you have it. If one would be interested one could have a look at a biquad filter structure. For analog filters: it's more complex as the frequency dependent components (capacitors and inductors), even if you'd only look at ideal version of them, are always complex impedances, basically amplitude and phase manipulation wrapped into one neat inseparable little package.
Phase shift is more perceptible from 20Hz to about 1~2kHz, and more crucial for lows and low-mid freq., then above that it acts like a phaser (who would have thought?!) and chorus in the highs. I sometimes cripple a pink\white noise into high-gain amp sim to comb-filter the nasty harsh hiss by its random affect on the "digital noise" above 6~7kHz (trying to invert it thus causing phase cancelation). Sometimes it works, other times not so much. After NAM (Neural Amp Modeller) came out there was no need to do it and I got rid of all amp-sim plugins ever since.
Yea thats the off axis mic thing, ab volume is introducing the oblong pickup pattern, or 45 degree. And you can then like have either 2 bidirectional, or 3 omnidirectional mic pointing like >^
Probably the most elegant answer to why you use your ears, not your eyes, to EQ. It’s a 2D drawing of a cube- it may be the headline, but not the whole story. Excellent work on this.
Hello dan! Would you find it interesting to make a video about the main differences between digital consoles live mixing strategys against studio recordings mixing strategys in a daw? Its good to see your videos again
Oh wow.... I never surmised that second order all pass is just 2 first order all pass in series.... It's hard to wrap my head around it, but I guess it makes sense.
Dan I have a question: How does this phase shift affect impulse/transient response? Would a high pass filter for example "smear" the attack of a bass note? I've tried a number of EQs and pedals and it seems that something like this must be happening but I'm not entirely sure either.
I think a filter can potentially smear a transient, but I don't think it's the phase shift that's to blame, rather the ringing. And you usually need steep filters or resonance to make the ringing audible. If you used a linear phase filter because you were scared of the phase shift, well now you moved half the ringing in front of the transient, and made it more likely to be audible.
@@DanWorrall I read that all-pass filters are used in algorithmic reverbs exactly because they can soften the 'reflections'. HPF and LPF might have half of this effect according to the experiment in the video i suppose.
@@DanWorrall Sounds like the result of ringing to me too. That said, filters affecting the upper harmonics of a sound definitely can smear transients, just maybe not in a particularly audible way in most cases. you might find kilohearts' disperser intersting in this regard, its an all-pass filter that takes this idea to an extreme using tens of thousands of degrees of phase shift. if you put a single impulse through it with a Q of 0, I believe the result would look like a short sine sweep.
It's not math magic, it's actually very easy to understand if you look at the flow diagram of an all pass filter. It has a feed-back as well as a feed-forward stage, that even each other out. The fun with all pass filters doesn't stop there. Put them in series to create phasers. Cascade them for control over Q, and more!
Great video as usual, and very informative. It maybe would have been a more fortunate starting point to build up an allpass from a lowpass, and explaining, that a hipass filter "hp(x))" is just a lowpass subtracted from the input "hp(x) = x - lp(x)", and an allpass is the result of the subtraction of lp from a hp filter (or vice versa) "allpass(x) = lp(x) - hp(x)", since "hp(x) = x - lp(x)", we can say that: "allpass(x) = lp(x) - (x - lp(x))", so this simple formula also works: "allpass(x) = 2 * lp(x) - x". Your shelving filter works by adding the lp/hp filter to the input: "lowShelf = x + lp(x) * gain ", so you have to invert the lowpass to cut. When using a dB scale, you have to use logics ("if" statement probably) to obtain a proper scale, since 0 represents -oodB in this case instead of 1. It is more convenient to subtract the filter from the input, and add it to the filter multiplied with the gain: "lowShelf = hp(x) + lp(x) * gain", or "lowShelf = x + lp(x) * (gain - 1)"... it's the same thing. This way you have a shelving filter, where 0 gain represents -oo dB, 1 represents 0 dB, and 2 represents +6dB boost, so a simple Log->Lin conversion does the job. For the bell, you'd have to create a notch and adding the band multiplied with the gain to it, this way you can keep the gain in the 0..n range to keep things simple. So deriving everything from the Lp instead of the Ap represents the relationship between lp, hp, ap and the more advanced stuff a bit more obvious, while your explanation keeps the allpass as a black-box that does magic, while most of your math drives things backwards, and makes the intuition, that lp, hp, ls and hs are being derived from an allpass, while the reality is the opposite (lp is an inverted feedback loop around an integrator and a gain(the cutoff)). "x = input" "lp(x) = lowpass" "hp(x) = highpass" "lp(x) - hp(x) = allpass" "lp(x) * gain + hp(x) = lowShelf" or "lp(x) * (gain - 1) + x = lowShelf" "lp(x) + hp(x) * gain = highShelf" or "hp(x) * (gain - 1) + x = highShelf" "allpass(alpass(x)) = allpassx2" "x - allpassx2 = bandpass" "x + allpassx2 = notch" "bandpass * gain + notch = bell" Note, that these bandpass, bell and notch filters are 2pole (or second order) filters, while the lp and hp are 1pole (1st order). The bell/bp would work fine for a 3 band EQ with 2pole lp and hp filters. This setup is not capable of resonance, so changing the Q is not possible without an additional feedback loop, and complications involved (the best way probably is to build an SVF filter, and derive all the EQ stuff from it - at least, this is what I did).
Simply delaying the signal a few samples will cause a phase shift. Averaging a signal over a time window (which is the basis for LPF) causes such a delay. So how does the all-pass work?
once i heard an explanation for all this. the claim was that most effects, eqs, filters, compressors etc are based on and deriving from the one and yours truly DELAY effect
Huh, and there I am thinking the eq just uses an FFT to break down the signal, boosts (or cuts) your desired frequencies and then slaps them back together for the final result. 😮
I see you analyzing phase a lot in your videos, Dan, as long as I've been your subscriber. I've heard more than once lately, from some producers on some forums, that Melodyne messes up phase, especially in the Polyphonic algorithm. Not only that, but I've tried it, and it looks to me, like it really does, in the high end it is highly apparent that some level dip is present. Do you have any experience with Melodyne, would like to hear your commentary, if yes?
I have Melodyne, though I think my version is out of date. I don't use it that often. Honestly I'd be amazed if it could correct pitch without affecting the phase! Perhaps you mean it causes phase shift before actually changing anything? I haven't checked for that, can't comment.
@@DanWorrall Yep, to exaggerate the effect I've loaded a famous track in it, and took an A/B with an unaffected one through a simple player, I definitely hear a difference in FR, but might be wrong, haven't done a null test to be honest.
I used to have a sign on my mixer that said: Mixing sound is not rocket science. It is probably closer to voodoo. (I think the statement came from an American sound guy, but I don't remember who.)
Yes. Steeper filters can be created by stacking gentler filters in series. Eg the Moog style ladder filter is 4x 1 pole stages plus feedback. The phase shift depends on how steep a gradient you create in the frequency response, doesn't matter how you get there.
These videos are very nice because they talk about actual dsp concepts from an audio engineering standpoint
What's up with these uninsightful comments that just say what the guy did? Why is the internet like this now?
@Bed_Bug I see what you did there, since your comment is only describing what the guy you’re replying to did. What’s with all of the uninspiring meta comedy nowadays?
@@LukeLongboneOfficial Why are you coping so hard over a bad comment?
Would love a video about how linear-phase EQ works to add on to this.
This is the most insightful video I've seen on this topic. I've had so many questions about how EQing actually works at a mathematical level - now answered in a 4 minute vid.
DSP programmer here, programmed linear phase equalizers by hand.
An analog-style EQ is what we call 'causal filter'. It "reacts", so to speak, to the inputs as they happen.
Without look-ahead, the EQ can only generate output using the past + present.
Using info from now + before, the filter can only affect what comes out next.
So, phase shifts using analog/IIR filters can only "delay" phase.
Now, a "linear phase filters":
Imagine an audio process that could see a bit into the "future" (know what's going to happen).
This ability, so to speak, is common nowadays (DAW delay compensation, etc.).
With this ability, a linear-phase filter can "shift the phase" both forward and backwards in time equally (so to speak).
This technique eliminates any phase shifting all together.
However, since we're able to shift the phase of frequencies *before they happen*, you may be able to hear "pre-ringing" from the filter in the output.
That's maybe an over simplification, and I've left out some of the more fun analysis/stuff (impulse responses, FFT's, etc.), but yeah.
@@gerudobombshellso I have a question for you. What do you normally suggest is the optimal way to EQ - specifically, the low-end. Do you use Linear-Phase, or not? I’ve seen tons of debate on cutting frequencies below 20-40hz because you don’t “need” information below that, and then I’ve also seen the opposing argument that even if we don’t “need” that information, it’s not WORTH cutting it out because of these exact issues. This forced me to the idea that I want to present to you since you seem to know more than anyone else I’ve ever come across. So could you take an Analog Filter and gradually clamp down on the low-end? It’s a concept I’ve called “EQ Clamping”, and I created a patcher preset for it. Instead of a simple High-Pass filter down at 30-40hz. It’s MANY Low-Shelf filters gradually clamping down. Low-Shelf filters being the choice because they create significantly less Phasing. After several of these filters, THEN you can GRADUALLY add HP filters with very generous slopes. Any thoughts? Would this help to mitigate the need for using a Linear-Phase Filter?
Generally speaking, minimum-phase/analog filters are great for most needs (no problems).
Linear phase filters are great for making spectral changes without affecting inter-track phase releationships (think drums, close mic + overheads).
However - back to the "pre-ringing" thing:
Linear phase filters will cause a variable amount/length of pre-ringing, with respect to the frequency (Hz).
That is to say, the lower the frency you're affecting, the longer/more audible the pre-ringing.
And, the sharper the filter (eg. imagine a 48 dB/oct hipass filter), the more audible it is yet again.
Since IIR/analog/minimum-phase EQ's only can ever "delay" the phase of waves, they will not suffer the same problem.
Any ringing anomolies will manifest forward in time - which is nice, since transients will "mask" the ringing (IIR filters *do* still ring, btw).
My rule of thumb:
Most of the time, try to use IIR.
If there's a phase relationship to another track, use Linear Phase (sometimes sounds cool boosting highs too).
In the simplest layman's terms Linear Phase performs the phase shift, then puts it back in time (latency) to compensate.
I have almost no idea what you guys ate talking about but i do understand that eq isnt really eq
Why do you make everything so easy to understand. Thank you Dan for another incredible video!
As the famous quote goes: "if you can't explain it simply you don't understand it well enough." The converse of that is definitely true for Dan.
Dan Worrall once again dropping some crazy knowledge on us without any warning whatsoever. Me likes it.
These technical bits are always such a delight!
I feel like I opened an ancient tome and found knowledge long forgotten
dan, im studying audio engineering at SAE institute and your videos have been such a massive help for me to dive deeper into the things im taught about at school. thank you so much for making your videos ❤❤
Yooo, at which sae?
Studying electrical engineering and than going to do audio processing makes way more sense to me. I studied mechanical engineering and I'm working in the audio field. Often we work with people how studied at SAE, and they often have proplems to understand what is going on with audio devices. How basic RLC filters work and srs, easy stuff like this. Sorry for interrupting you.
@@UltimateEngineering youre right, but ive got no interest in working in the parts of audio that are that down to metal. for the most part I want to specialise into mixing/mastering and live sound
The math behind (digital) all-pass filters is relatively simple, you just need a bit of discrete linear algebra and z-transforms. The first order IIR filter implementation is just a combination of the current input sample with a scaled previous input sample and output value (feedback and feedforward loops). As a difference equation: y[n] = a * x[n] + x[n-1] - a * y[n-1], where y is the output, x is the input, and n is the sample number.
Fun fact: all-pass filters are also heavily used in many reverb algorithms.
Really loving how many new videos we’ve been getting recently. Thanks, Dan :)
👀‼️😳 ...Good God, Dan. EVERY time I think I know something about mixing or audio you calmly show me I know NOTHING 🤯😂 So THANK YOU once again for blowing my mind and making me a better mixer. You are a Rare Creature on TH-cam -- you ACTUALLY give fantastic instruction without pushing some Patreon mixing course or 1-to-1 classes, although if you DID have such things I'd more likely be one of the 1st ppl in line 😂🤷♂️
Wishing you best & just sayin' thanks 😎👍
My thoughts exactly! (although maybe with a few less emojis XD)
@@talktokale 😂🤣👍🤷♂️
Dan, your parallel filters videos inspired me to discover that when you delta solo an aux with any effect on it and send a track in your mix to the aux, the effect acts as if you have it on the track you’re sending to the aux. thank you for all you do! I now use this technique to create a global sidechain pumping effect that is controlled at the mixer level at each track and the global aux. I use this for summing bus compression when I want to glue kick and bass but I don’t want to sum there outputs. I use it for de-easing and spectral resonance suppression too. I really appreciate your videos and how you think. Thanks for all you do!
Ooh that's really good! I've been wondering about how to get a similar outcome for a while, so thankyou!
This sounds interesting but I'm not exactly certain what you do. Let me follow along, is this correct?
1. You have an aux with an effect, lets say a compressor.
2. You send a track to the aux. Lets say a kick.
3. Normally you get a compressed kick. But you delta solo. So you get only the "minus sound" that the compressor removes.
4. You also send the bass track to the same aux? So you get "minus sound' of kick+bass
5. Now what? you blend this "minus signal" back in the full track, so that you have the gain reduction (pumping) without having to route or sum the bass and kick together?
That's pretty smart.
Personally I don't really think I'd use this as prefer sidechaining a bass in a kick+bass combo, because I want to have at least one sound constant. For example in the context of Rock or EDM, I don't want a louder bass note making the kick more quiet, the kick needs to be constant. As for de-essing, that's (IMO!) better done on a per-track basis than on a sum of tracks, because de-essing on a sum of tracks distorts all tracks, not just the one with the harsh S in it.
I can see this technique being very nice in ambient electronic or acoustic folk type music, where "naturality" and transparency is desired.
@@JDarkooJDarkoo Yes that is correct. This particular application definitely is more suited to acoustic genres. However what I like about this consept is it leaves the creative aspect of the compression (or any effect) open ended so that you can add any track you want to be affected by the same plugin. I use this for global high cut filter effects, creative transitions where I want some but not all of the parts be warped in a dynamic way. It makes the process as simple as sending to an aux and adjusting the send to taste.
@@lawinter1949 Very interesting snd very creative. Going to give this a whirl on an acoustic folk song I'm working on.
I've always thought of filters in terms of an RLC circuit and the various impedances of the components. The maths to implement a digital filter just sort of falls out of that. Messing around in a circuit simulator (falstad, ltspice) is a good way to get a feel for these things
That's a C I V I L way of putting it 😜
Yup it's all that Leading and Lagging of reactive components being modelled mathematically innit.
"How does the allpass filter cause the selective phase shift - some voodoo magic I suppose." Actually that's pretty accurate way to describe the math that starts from here on out. Like, in the end the equations look quite simple, but if you start looking into how those equations were derived... "Oh, it's simple. We first analysed the analogue circuit and derived it's differential equation. Than transformed the equation to give us the output voltage in terms of input voltage. Than discretized it by replacing integrals with discrete trapezoid integrator equations. Then we have resolved the zero delay feedback loops by treating the equation as an implicit equation and using it to derive intermediate results. But you have to consider equation's stability at this point. Finally we plugged those intermediate results into the main equation and that's it. Oh, also don't forget about warping the frequency range, but you can figure that out on your own."
Its*
Makes you wonder how the analog circuits came to place. With vacuum tubes no less, or old transistors, keeping those units fairly tuned must've been a nightmare.
@@BananaManPL I think this is why digital equipment often gets labelled "lifeless" or "sterile", because it doesn't have the analog grit everyone wanted rid of a few decades ago
"of course, we're considering a spherical circuit in vacuum"
Okay this is the perfect video to explain something I've been wondering. I understand how you can break up freqencies in a computer using an FFT, but I've been wondering for a while how analog EQ's work, and this is making more sense of that. I'm sure there's a lot more to it (for example, I nowhere near enough of an electrical engineer to understand how analog competents could introduce phase shift) but this video feels like it's getting me closer to understanding how that could work in the analog realm. Thanks Dan!
Your knowledge and understanding of this stuff from front to back, back to front, in multiple different examples, clearly demonstrated. Amazing as always, thank you.
I really love how you can explain something rather complex in a simple yet accurate way. 👏👏👏👏👏
The first and last video you'll ever need to understand what an EQ really is! Thank you!
I think I need a video tutorial for this video tutorial. Love seeing you debunk the hot talking points floating around TH-cam.
4:15 what an evil cliffhanger!! Looking forward to part 2.
Genuinely an amazing channel for audio engineering. Perhaps the best on TH-cam
Anyone who understands how analog filters work already understands that is the phase shift caused between the current and the voltage that causes the filtering. Lovely video Dan
Great work Dan!
Thanks, Dan. Glad I asked the wrong question, but thank you for answering anyway! You are my God... PRAISE BE!
Very clear and simple explanation. Great! I look forward to the next video explaining how all pass filters shift phase selectively. That was the first question I asked in my mind when you began :)
Well said Dan. I've said many times over the year that EQ is a consequence of phase shift - people look at you like you;re mad. But it really is that way round!
I love the scientific approach to analyzing problems and understanding them. Great video Dan!
Thanks Dan, great to see you're back❤
Dan your teaching style is so straightforward and well presented, thanks your all you do
This was beautiful! Such a great demonstration.
Well, I sure do wish I'd paid attention in school. That's some mighty fine learning right there.
i knew this answer but omg i never thought it was so complicanted, it amazed me!
I really like how you talk in clear terms, cutting through the "easy tricks" that others post when advising on mixing and mastering. There is some confusion I have about some EQ elements. There is a few videos talking about mid-side EQ in order to remove overlap / mud that this causes in the stereo field. When I mix I tend to just mix on the defaults with side chaining to make room for the kick snare (to the bass and rhythm guitars) - which seemed to improve the clarity of the resulting track.
But should we also focus on the mid-side equally. This was a little confusing, or perhaps badly explained.
That is precisely the case and a more accurate description of the two phenomena.
I can't wait for the Dan Worrall Signature Phase EQ from IK Multimedia. Great video!
this guy is like a genius....love your videos and your accent!!!
I believe this is my favorite video you've done.
This question and topic has been on reddit a lot lately. Thanks for clearing this up, I learnt something.
I had a "duh" moment. Thanks Dan.
Such a simple but clear explaination! Thank you!
my mind is thoroughly blown!
nice, now replace the allpass filters with full on instances of disperser and you have a multidimensional 4th-wall breaking spacetime-bending hyperfilter
I thought I understood the basic workings of filters. I even built simple analog filters myself using capacitors, resistors and op-amps when I was younger. I understand the electrical concept of why they do what they do pretty well. I only recently first heard of the concept of an all-pass filter. This was after asking ChatGPT what the difference between a flanger and a phaser really was. At first hearing they even seemed useless, momentarily forgetting about that thing called phase. Functionally, they're still a bit mysterious to me, so thanks for this video. It sheds a whole new light on these things and they're really kind of interesting.
Hey Dan!
Fantastic video mate! Many thanks!.
I knew the concept that 'phase causes eq' for a while now, but there is just so little information out there, that or it's very difficult to find.
Usually you just get articles or videos on 'how to use' a particular EQ and not how they actually work.
EXCELLENT!!
Well that’s a cliffhanger ending. Would love a follow up about how the allpass filter works
dan, we'd love your no nonsense approach applied to some acustica plugins, we wanna know the deal!
We are so lucky to have you Dan...
This actually change my perspective on EQ...
ALWAYS ALWAYS learn something. (even when i think i know a decent bit about a subject) THANK YOU
I wish you released this a year ago when I was trying to learn how to make my own passive outboard gear from scratch for my audio engineering capstone project at uni. I spent forever trying to do everything with just high-pass and low-pass filters and boost the remaining signal instead.
Turns out simple arithmetic was needed. Whoops.
Is Dan building an EQ from scratch? This guy is clever.
Would be a waste of his time
He has built plugins before. They are available on the web somewhere.
He literally did what op said@@lucasjames8281
@@tylerdurden6992 No, Dan Worrall is not behind FabFilter...
I wish I had known of you when I had my signals classes back in uni
I knew this, but it took me forever to find out cause no one ever explains it when talking about EQs. whenever you read about EQs and phase this relation should be the first to be shown.
The man that won the loudness wars is back!! 🤟
THIS IS INCREDIBLE!!! THANK YOU!!!
Another amazing knowledge video from dan
the question you say at the end "How does the allpass filter selectivly cause phase shift at specific frequency area "
i saw in a digital signal processing that the behavior of allpass filter related to transfer function in math
Interesting
thank you for give us great knowledge !
Mr dan is Mr end all arguments💯🙏
best tagline ever
Great demonstration of why the unwary would best avoid parallel EQ.
No! th-cam.com/video/RL4KDVFlkUg/w-d-xo.htmlsi=ApNpbretZVJkOeWb
@@DanWorrall - Notice I'm referring to the unwary. If you understand phase shift and are running a real-time spectrum analyzer (not just an EQ curve plotter), you can make expert use of phase cancellation. Otherwise, you're guesstimating with forces that can unexpectedly bite back. And as you point out in your linked video, linear phase filters are no panacea, they're rarely a significant factor outside of multi-pole crossovers.
@@QuicksilverSG there's nothing to be wary of with bell and shell filters. The phase cancellation will cause a tiny change in the shelf frequency or the bell width, both of which you set by ear anyway so you'll dial that difference out if it matters. You only need to be a bit careful with parallel high or lowpass filters.
video liked the second he told Native Instruments to take a hint.
Thank you, Dan. This is excellent. Next step... capacitors, inductors, resistors and op-amps on a breadboard?? :)
Thanks for all these videos.
Haha! Hmmm solving for x has a way of being so simply complex, doesn’t it! Out the door or window, into the rabbit hole, breach the subterranean trenches to find yourself standing in China in front of a rickshaw selling fried rice with shrimp to go and as you walk thousands of miles and who knows how many hours you find yourself back and square one and all you had to do was cross the street… to find the answer.
Nooooo, it couldn’t be that simple; are you $hittake mushrooming me! 😂
Thanks Dan.
Hope you are recovering well.
Your video’s… I always learn and I always laugh. Appreciate you and your channel.
What the actual f**k... You've just blown my mind
Please do get back to me on that food for thought you left us with at the end of the video. I’d be very interested to dive more in depth to the mechanics of all-pass filters.
Firstly, I'm not disagreeing, I'm only adding detail.
Secondly:
For digital filters: technically, your using delays in the range of one or two samples, feedback, scaling and summation to create those curves, but these delays can be seen as frequency dependent phase shifts, so 6 one way, half dozen the other. I know this is a distinction that nobody except me cares about but there you have it. If one would be interested one could have a look at a biquad filter structure.
For analog filters: it's more complex as the frequency dependent components (capacitors and inductors), even if you'd only look at ideal version of them, are always complex impedances, basically amplitude and phase manipulation wrapped into one neat inseparable little package.
Love these kinds of videos, cheers Dan
Love your vids Dan. Was wondering wether you considered talking about weight (a/b-weighting etc.) in eqs and the audio world in general?
Did you see the Prism / Special Filters video I made for TDR? th-cam.com/video/tMzQVOfNVbo/w-d-xo.htmlsi=PrTxoDOkuSLwvr0s
@@DanWorrall I have now. Awesome vid.
Phase shift is more perceptible from 20Hz to about 1~2kHz, and more crucial for lows and low-mid freq., then above that it acts like a phaser (who would have thought?!) and chorus in the highs.
I sometimes cripple a pink\white noise into high-gain amp sim to comb-filter the nasty harsh hiss by its random affect on the "digital noise" above 6~7kHz (trying to invert it thus causing phase cancelation). Sometimes it works, other times not so much. After NAM (Neural Amp Modeller) came out there was no need to do it and I got rid of all amp-sim plugins ever since.
Yea thats the off axis mic thing, ab volume is introducing the oblong pickup pattern, or 45 degree. And you can then like have either 2 bidirectional, or 3 omnidirectional mic pointing like >^
nice insight. How about using the phase shifting of a speaker to construction of a small ported sub box... any thoughts?
Thanks a whole hell of a lot, Dan
Thank you, Sir!
Probably the most elegant answer to why you use your ears, not your eyes, to EQ. It’s a 2D drawing of a cube- it may be the headline, but not the whole story. Excellent work on this.
Well, perhaps use both. After all, this video was visual, with no audio running through the examples.
Hello dan! Would you find it interesting to make a video about the main differences between digital consoles live mixing strategys against studio recordings mixing strategys in a daw? Its good to see your videos again
Oh wow.... I never surmised that second order all pass is just 2 first order all pass in series....
It's hard to wrap my head around it, but I guess it makes sense.
I’ve never understood the math behind EQ’s before and this fondant got me 90% off the way there. The last 10% is my own intelligence gap
Make a video explaining sliding band compression ! Cheers Dan !
Dan I have a question: How does this phase shift affect impulse/transient response? Would a high pass filter for example "smear" the attack of a bass note? I've tried a number of EQs and pedals and it seems that something like this must be happening but I'm not entirely sure either.
I think a filter can potentially smear a transient, but I don't think it's the phase shift that's to blame, rather the ringing. And you usually need steep filters or resonance to make the ringing audible. If you used a linear phase filter because you were scared of the phase shift, well now you moved half the ringing in front of the transient, and made it more likely to be audible.
@@DanWorrall ...damn... GOOD to know 😎👍
@@DanWorrall I read that all-pass filters are used in algorithmic reverbs exactly because they can soften the 'reflections'. HPF and LPF might have half of this effect according to the experiment in the video i suppose.
@@DanWorrall Sounds like the result of ringing to me too. That said, filters affecting the upper harmonics of a sound definitely can smear transients, just maybe not in a particularly audible way in most cases.
you might find kilohearts' disperser intersting in this regard, its an all-pass filter that takes this idea to an extreme using tens of thousands of degrees of phase shift. if you put a single impulse through it with a Q of 0, I believe the result would look like a short sine sweep.
It's pre ringing!
Very interesting. If this is the case, should we be all that worried about phase shift in the first place?
No!
Absolutely brilliant!
It's not math magic, it's actually very easy to understand if you look at the flow diagram of an all pass filter. It has a feed-back as well as a feed-forward stage, that even each other out.
The fun with all pass filters doesn't stop there.
Put them in series to create phasers. Cascade them for control over Q, and more!
You 're the the best, man!
Great video as usual, and very informative. It maybe would have been a more fortunate starting point to build up an allpass from a lowpass, and explaining, that a hipass filter "hp(x))" is just a lowpass subtracted from the input "hp(x) = x - lp(x)", and an allpass is the result of the subtraction of lp from a hp filter (or vice versa)
"allpass(x) = lp(x) - hp(x)",
since "hp(x) = x - lp(x)", we can say that: "allpass(x) = lp(x) - (x - lp(x))",
so this simple formula also works: "allpass(x) = 2 * lp(x) - x".
Your shelving filter works by adding the lp/hp filter to the input: "lowShelf = x + lp(x) * gain ", so you have to invert the lowpass to cut. When using a dB scale, you have to use logics ("if" statement probably) to obtain a proper scale, since 0 represents -oodB in this case instead of 1. It is more convenient to subtract the filter from the input, and add it to the filter multiplied with the gain: "lowShelf = hp(x) + lp(x) * gain", or "lowShelf = x + lp(x) * (gain - 1)"... it's the same thing. This way you have a shelving filter, where 0 gain represents -oo dB, 1 represents 0 dB, and 2 represents +6dB boost, so a simple Log->Lin conversion does the job.
For the bell, you'd have to create a notch and adding the band multiplied with the gain to it, this way you can keep the gain in the 0..n range to keep things simple.
So deriving everything from the Lp instead of the Ap represents the relationship between lp, hp, ap and the more advanced stuff a bit more obvious, while your explanation keeps the allpass as a black-box that does magic, while most of your math drives things backwards, and makes the intuition, that lp, hp, ls and hs are being derived from an allpass, while the reality is the opposite (lp is an inverted feedback loop around an integrator and a gain(the cutoff)).
"x = input"
"lp(x) = lowpass"
"hp(x) = highpass"
"lp(x) - hp(x) = allpass"
"lp(x) * gain + hp(x) = lowShelf" or "lp(x) * (gain - 1) + x = lowShelf"
"lp(x) + hp(x) * gain = highShelf" or "hp(x) * (gain - 1) + x = highShelf"
"allpass(alpass(x)) = allpassx2"
"x - allpassx2 = bandpass"
"x + allpassx2 = notch"
"bandpass * gain + notch = bell"
Note, that these bandpass, bell and notch filters are 2pole (or second order) filters, while the lp and hp are 1pole (1st order). The bell/bp would work fine for a 3 band EQ with 2pole lp and hp filters. This setup is not capable of resonance, so changing the Q is not possible without an additional feedback loop, and complications involved (the best way probably is to build an SVF filter, and derive all the EQ stuff from it - at least, this is what I did).
Excellent explanation!
does pairing phase shifting with gain multipliers result in a "cleaner" EQ?
Very nice Sir! Could you please test some of the native UAD plugins with your doctor? =') Thanks
Thank you, fantastic video and explanation.
Simply delaying the signal a few samples will cause a phase shift. Averaging a signal over a time window (which is the basis for LPF) causes such a delay. So how does the all-pass work?
Mind-blowing
once i heard an explanation for all this. the claim was that most effects, eqs, filters, compressors etc are based on and deriving from the one and yours truly DELAY effect
we need more of these videos
Talk about spring-mass resonance please.
Huh, and there I am thinking the eq just uses an FFT to break down the signal, boosts (or cuts) your desired frequencies and then slaps them back together for the final result. 😮
Wonderfull! Thank you Dan
I see you analyzing phase a lot in your videos, Dan, as long as I've been your subscriber. I've heard more than once lately, from some producers on some forums, that Melodyne messes up phase, especially in the Polyphonic algorithm. Not only that, but I've tried it, and it looks to me, like it really does, in the high end it is highly apparent that some level dip is present. Do you have any experience with Melodyne, would like to hear your commentary, if yes?
I have Melodyne, though I think my version is out of date. I don't use it that often. Honestly I'd be amazed if it could correct pitch without affecting the phase! Perhaps you mean it causes phase shift before actually changing anything? I haven't checked for that, can't comment.
@@DanWorrall Yep, to exaggerate the effect I've loaded a famous track in it, and took an A/B with an unaffected one through a simple player, I definitely hear a difference in FR, but might be wrong, haven't done a null test to be honest.
very well made thank u
True engineer...
I used to have a sign on my mixer that said: Mixing sound is not rocket science. It is probably closer to voodoo. (I think the statement came from an American sound guy, but I don't remember who.)
Does using multiple gentle sloped low pass filters cause the same phase shift as a single steeper cut (all other things being appropriately matched)?
Yes. Steeper filters can be created by stacking gentler filters in series. Eg the Moog style ladder filter is 4x 1 pole stages plus feedback. The phase shift depends on how steep a gradient you create in the frequency response, doesn't matter how you get there.
God i wish they would update reaktor and fm8 and not be a shitty company anymore. Great video as always