[ADUHF]

Now you've got me intrigued by this mystery explanation. As far as I understand minimum-phase EQ, digital just does digitally what analog EQ already does: shift phase to attenuate or boost certain frequencies. @VNandor reaffirms above what I had assumed. Digital sounds functionally identical to analog. But that doesn't mean there isn't more nuance than that, and like I said above, seemingly every recording engineer has their own philosophical dogma when it comes to EQ. Just like there are diehard analog-only audiophiles, there are probably similar divisions among recording/mastering engineers. But that's just a guess. Now you've got me super curious to hear the case against digital EQ. I'll have to do some digging...

I wish you luck on that.

I have no plans to go back to analog at this time. There are good arguments to be made on both sides of the digital/analog question though, irrespective of phase.

If your sound source is analog, like an LP or tape for example, then an analog EQ has a distinct advantage, because it means you can keep the signal in analog form from the source to your ears. A good analog EQ can run into some $$ though, whereas good digital EQs will cost you basically nuthin but time.

Either will work in my setup though, because my source is digital (YouTube music clips). But I can also easily insert an analog EQ between my DAC and amp. I designed it that way intentionally, so I had both options. And could even use them together on the same signal, if I wanted. But it's not somethin I'd probably do in my own setup.

 

[VNandor]

I think there is still something about the differences between EQ-ing in the digital and analog domain that I'm missing in this discussion. And will probably have to go back through some previous threads to maybe find it, which could take some time.

Do you remember this difference between analog and digital EQs perhaps? The frequency cramping problem is unique to digital EQs although it's not a huge problem to begin with and there are multiple ways to mitigate it (by the developers of the EQ).
https://www.head-fi.org/threads/fre...-move-thread-to-appropriate-sub.954577/page-6

Whoa. I just tried inverting phase in one channel using Roon's "Speaker setup" filter. Not what I expected at all. Not only is it audible; it is way more audible through headphones than it is on speakers! Weird. I guess that means our brains are good in the other direction. Rather than realign the two different phases, your brain makes them even more distinct than when they interact in space. And more bizarrely, the waves don't interact in space, but they do interact in your head. Complete with phase cancellations and everything. I mean, they're neural signals at that point. They're no longer sound waves, but they still interact as if they were... The effect is even more pronounced than it is in space. Wow. Speaks strongly in favor of folks who argue that our ears/brains are more sensitive to timing imperfections than we assume.

It's a very interesting effect. Something else you could try is to create two basic sine waves with a frequency of 200Hz and 202Hz with an amplitude of 0.2. You'll hear a distinct beating as the waves get in and out of phase. If you start panning one of them to left and the other one to right, you might still hear a faint beating even after you 100% panned them even though there can't be any phase cancellation between them once they are completely panned.

A trick that can be used to perceptually pan sounds is to slightly delay (<1ms so around ~40 samples typically) one of the channels. The sound will be seemingly louder on the non-delayed side even though there's no volume difference between them. As you increase the delay between the channels, at some point it will actually sound like a delay instead of just a simple volume change. Something to note is that our ears are much more sensitive in difference of timing between our ears than just difference in timing in general. Noone would notice if a drummer was 1ms early or late on the downbeats for example and of course no drummer would be that accurate anyways.

 

[Hypops]

I wish you luck on that.

I have no plans to go back to analog at this time. There are good arguments to be made on both sides of the digital/analog question though, irrespective of phase.

If your sound source is analog, like an LP or tape for example, then an analog EQ has a distinct advantage, because it means you can keep the signal in analog form from the source to your ears. A good analog EQ can run into some $$ though, whereas good digital EQs will cost you basically nuthin but time.

Either will work in my setup though, because my source is digital (YouTube music clips). But I can also easily insert an analog EQ between my DAC and amp. I designed it that way intentionally, so I had both options. And could even use them together on the same signal, if I wanted. But it's not somethin I'd probably do in my own setup.

I'm with you. I use a turntable in my living room speaker setup, but even in that case, I actually use an ADC and DAC to allow me to do DSP room correction with mixed-phase EQ (minimum phase < 300Hz and linear phase > 500Hz). It's so easy--and audibly indistinguishable to my ears--that I could never go back to analog EQ.

My desktop/headphone setup uses only digital sources, so EQ is even easier. Even better, I don't see/hear a need to bother with linear phase EQ.

A trick that can be used to perceptually pan sounds is to slightly delay (<1ms so around ~40 samples typically) one of the channels. The sound will be seemingly louder on the non-delayed side even though there's no volume difference between them. As you increase the delay between the channels, at some point it will actually sound like a delay instead of just a simple volume change. Something to note is that our ears are much more sensitive in difference of timing between our ears than just difference in timing in general. Noone would notice if a drummer was 1ms early or late on the downbeats for example and of course no drummer would be that accurate anyways.

Wow. Yeah. My mind is pretty blown. I had always assumed that headphones were more or less immune to subtle timing differences. Now I see that the reverse is true. They are in fact even more revealing of timing differences than speakers are. With speakers, those differences create the sublte effect of "phase smear" (in which the sound-image gets a little blurry), but on headphones you can hear a a very pronounced "pulse," like you say (more like a double-exposure of the sound-image). The human brain is amazing. Totally wild.

 

[bagwell359]

Great thread. Had a 10 band Soundcraftsman analog fixed frequency EQ and found that whenever adjacent bands from 240 Hz and up were audibly weird if they were more than 5 db apart (-3db @240 & +3db @480 for instance) and learned that it was phase issues. Sold it and took that Calvinist audiophile set against EQ for 40 years.

Later in the mid/late 80's got into absolute phase getting into it so much with two detachable headshells with the same cartridge in both. Then was into the ARC SP-15 which had a absoluite polarity switch. One thing that became clear that the further you got from purist analog recordings (3 mics, direct to disc) the more you would find mixed polarity recordings, and I gradually got off that.

Fast Forward: LG-v40/UAPP Toneboosters/TIDAL - and keeping the changes tight by db and Q, doesn't seem to be negative audibly, and I was addressing FR as well as ringing in plnars which is not always tied to a rise in FR.

I've also been looking at phase graphs of drivers from Quad ESl-57 to Rall CA-1a, and it makes the fascination/adherance of FR analysis/review/alteration even less important than many have it. Check out this graph of the phase response of the Rall CA-1a (accuratesound.ca) - can't see any of the light handed PEQ changes I favor stacking up against this. Toggling between +180 and -180 degrees... whoa. Outside of the floppy bass (THD) it sounds pretty amazing stock, with the filter it's incredible sounding.

 

[Hypops]

Fast Forward: LG-v40/UAPP Toneboosters/TIDAL - and keeping the changes tight by db and Q, doesn't seem to be negative audibly, and I was addressing FR as well as ringing in plnars which is not always tied to a rise in FR.

This is where I've landed, as well. Now I often start with an aggressive tuning (based on measurements), then I gradually mellow it out (based on listening). By the time I get to the end of my headphone EQ process, I'm usually down to just a handful of filters with Q values all under ~2. I also typically only tune above 5kHz with a single high freq shelf (since FR measurements above that range are pretty inconsistent due to fit, HRTF, etc).

I've also been looking at phase graphs of drivers from Quad ESl-57 to Rall CA-1a, and it makes the fascination/adherance of FR analysis/review/alteration even less important than many have it. Check out this graph of the phase response of the Rall CA-1a (accuratesound.ca) - can't see any of the light handed PEQ changes I favor stacking up against this. Toggling between +180 and -180 degrees... whoa. Outside of the floppy bass (THD) it sounds pretty amazing stock, with the filter it's incredible sounding.

This is where my brain starts doing backflips. I know how to read these graphs, but I have hard time wrapping my head around what's actually happening with the drivers. If I understand right, when graphs show phase response changes like this across the frequency spectrum, this basically just means that a driver is (suddenly) faster or slower at different frequencies, right?

 

[bagwell359]

This is where I've landed, as well. Now I often start with an aggressive tuning (based on measurements), then I gradually mellow it out (based on listening). By the time I get to the end of my headphone EQ process, I'm usually down to just a handful of filters with Q values all under ~2. I also typically only tune above 5kHz with a single high freq shelf (since FR measurements above that range are pretty inconsistent due to fit, HRTF, etc).

I start out with a Harman curve as well as looking for waterfall charts - and end up with mostly high Q and low db, I posted this a few weeks ago re: HE-6 SE v1

48 Hz +2.6 Q 1.0 bass shelf
70 Hz +1.2 Q 1.0 bass shelf
840 Hz -1.8 Q 4.8 (regular or digital bell curve if you have it - all settings below here)
1.8k +3.8 Q 2.2
3.2k -2.6 Q 6.4
3.8k -2.0 Q 7.4
4.4k -2.8 Q 8.4
7.5k +2.0 Q 3.6
9.8k -3.6 Q 8.0 (ringing - could be anywhere from 9.1-10.2k - may have to hunt around

This is where my brain starts doing backflips. I know how to read these graphs, but I have hard time wrapping my head around what's actually happening with the drivers. If I understand right, when graphs show phase response changes like this across the frequency spectrum, this basically just means that a driver is (suddenly) faster or slower at different frequencies, right?

180 degrees away from 0 degrees means you have a driver that is producing highest air pressure when at some other frequency its producing its least. That graph has -180, +180, and 0. My confusion is -180 and +180 is 360 apart - so does that mean it is correct at 0? But if so why not make it clear? It is a graph from a company selling a product that purports to address these issues, and a quick look is alarming. The FR graph is also alarming looking too with the corrected curve looking quite excellent.

Hopefully someone can make it clear.

 

[Hypops]

I start out with a Harman curve as well as looking for waterfall charts - and end up with mostly high Q and low db, I posted this a few weeks ago re: HE-6 SE v1

48 Hz +2.6 Q 1.0 bass shelf
70 Hz +1.2 Q 1.0 bass shelf
840 Hz -1.8 Q 4.8 (regular or digital bell curve if you have it - all settings below here)
1.8k +3.8 Q 2.2
3.2k -2.6 Q 6.4
3.8k -2.0 Q 7.4
4.4k -2.8 Q 8.4
7.5k +2.0 Q 3.6
9.8k -3.6 Q 8.0 (ringing - could be anywhere from 9.1-10.2k - may have to hunt around

In my experience, fixing those narrow treble peaks/valleys never works out. Even if I find a peak in one listening session, if I just slightly readjust my headphones (or even worse, if I take them off and put them back on), the peak will have moved. Add in things like headband adjustments, head and facial hair growth, glasses, hats, etc., and those peaks get really squirrely. My HE6se v2 in particular is pretty heavy, so sometimes just shifting a bit in my chair can change the position of the headphones on my ears. I'd have to change the frequency of that narrow filter every time I put the headphones on again, or even if I just slightly readjusted them. For that reason (and others), I just use a high shelf instead. Glad those narrow filters work out for you, though. Whatever works!

180 degrees away from 0 degrees means you have a driver that is producing highest air pressure when at some other frequency its producing its least. That graph has -180, +180, and 0. My confusion is -180 and +180 is 360 apart - so does that mean it is correct at 0? But if so why not make it clear? It is a graph from a company selling a product that purports to address these issues, and a quick look is alarming. The FR graph is also alarming looking too with the corrected curve looking quite excellent.

Thanks for the explanation. And, yeah, you should be able to “unwrap” a phase measurement. Not sure if that tool lets you do it. I’m more familiar with things like REW and rephase where you can toggle between wrapped and unwrapped phase measurements. It's easier for me to understand what's happening when I see it unwrapped.

 

[VNandor]

There are two things going on that makes it hard to read that phase response graph. The first one is that the graph shows the wrapped phase of the headphones, not the unwrapped phase. Wrapped phase is usually presented between -180 degrees and +180 degrees or sometimes between 0 degrees and 360 degrees. If you think of phase as traversing a circle, it makes sense why wrapped phase is often being used. In a sense, being off from a reference angle by 5 degrees is the same as being off by 365 degrees or being off by -355 degrees so there's no need for the plot to cover more than 360 degrees.

Here's a picture for the different representations. The top graph plots the wrapped phase while the bottom one uses unwrapped phase. They both show the same phase just in different ways. It also uses radians instead of degrees. The +pi and -pi wrapping points are the same as -180degrees and +180degrees.

I used my mad paint skills to make a graph of the unwrapped phase which I think is much easier to read.


The other thing that makes the phase response hard to read is that the group delay is what ultimately matters for a headphone. A constant horizontal line would look super satisfying on the phase response graph but a graph like that would indicate that the different frequencies are being delayed by varying amount of time (except if the phase response sits at 0) which is what causes the "smearing".

 

[Hypops]

There are two things going on that makes it hard to read that phase response graph. The first one is that the graph shows the wrapped phase of the headphones, not the unwrapped phase. Wrapped phase is usually presented between -180 degrees and +180 degrees or sometimes between 0 degrees and 360 degrees. If you think of phase as traversing a circle, it makes sense why wrapped phase is often being used. In a sense, being off from a reference angle by 5 degrees is the same as being off by 365 degrees or being off by -355 degrees so there's no need for the plot to cover more than 360 degrees.

But if I'm understanding correctly, being 360 degrees out of phase isn't exactly aligned. It's one full cycle off, so while you won't hear a lot of nasty phase interactions, wouldn't you still notice that something is a bit "off" in terms of timing? Again, this is where my ignorance starts showing with respect to phase. Like with complex math, I can do the calculations no problem, but I don't understand the theory behind them.

The other thing that makes the phase response hard to read is that the group delay is what ultimately matters for a headphone. A constant horizontal line would look super satisfying on the phase response graph but a graph like that would indicate that the different frequencies are being delayed by varying amount of time (except if the phase response sits at 0) which is what causes the "smearing".

Huh, yeah. This is also true of speakers. Knowing group delay is central to being able to do good EQ on speakers and keeps you from chasing impossible EQ corrections. Surprised that headphone review measurements only seem to show frequency response, but never things like an impedance curve, phase, or even distortion v. frequency. Those would all be super useful to know. Those are all standard measurements for speakers. Unfortunately, everyone in the headphone world seems fixated on FR measurements. I always figured it was because that other stuff doesn't matter as much with headphones, but I'm quickly learning that it does matter (and possibly matters more).

 

[Angel III]

I have had the DBX1231 for a few months and am extremely satisfied with it. I use it with the Violectric V200 and DAC. Now the DBX has placed itself between Schiit MB and Atoll IN80 - Dynaudio Evoke 20 speakers. I tested with my closed Sony MDR-1AM2 headphones on the Violectric V200 with a track paused and volume at 3 o'clock (to be hurt if you deactivate the pause), I didn't hear any noise or hum coming from the DBX. I have a lot of fun with this equalizer and it costs very little.

 

[bagwell359]

In my experience, fixing those narrow treble peaks/valleys never works out. Even if I find a peak in one listening session, if I just slightly readjust my headphones (or even worse, if I take them off and put them back on), the peak will have moved.

After my mods that what was left - and I wanted to have the smallest footprint - and I use reference recordings I've listened to hundreds or thousands of times so I don't adjust for a narrow band of music.

Add in things like headband adjustments, head and facial hair growth, glasses, hats, etc., and those peaks get really squirrely.

I agree, I always take of my glasses, make sure the VAC is steady and in range, no headaches, or lack of sleep - if I am evaluating anything - casual music or listening doesn't matter.

My HE6se v2 in particular is pretty heavy, so sometimes just shifting a bit in my chair can change the position of the headphones on my ears.

Yeah that's a thing. I listen reclined with my ears the same relative to each headphone where is sounds most accurate.

I'd have to change the frequency of that narrow filter every time I put the headphones on again, or even if I just slightly readjusted them. For that reason (and others), I just use a high shelf instead. Glad those narrow filters work out for you, though. Whatever works!

I agree, it took me many iterations to find the exact location of that 9.8k setting I use - planars over about 2.5k have lots of unit-to-unit variation because the tension on the mylar has such a huge impact. I would never have so much patience back in the old days, but, I've got to max what I have in retirement.

Thanks for the explanation. And, yeah, you should be able to “unwrap” a phase measurement. Not sure if that tool lets you do it. I’m more familiar with things like REW and rephase where you can toggle between wrapped and unwrapped phase measurements. It's easier for me to understand what's happening when I see it unwrapped.

Yes I just saw some, thanks to your comment, much easier that way.

 

[bagwell359]

I have had the DBX1231 for a few months and am extremely satisfied with it. I use it with the Violectric V200 and DAC. Now the DBX has placed itself between Schiit MB and Atoll IN80 - Dynaudio Evoke 20 speakers. I tested with my closed Sony MDR-1AM2 headphones on the Violectric V200 with a track paused and volume at 3 o'clock (to be hurt if you deactivate the pause), I didn't hear any noise or hum coming from the DBX. I have a lot of fun with this equalizer and it costs very little.

I had a relative back when DBX started with companders/expanders like the 127, 128, 2BX, 3BX. Didn't know they were still around. Interesting.

Nice speakers!

 

[VNandor]

But if I'm understanding correctly, being 360 degrees out of phase isn't exactly aligned. It's one full cycle off, so while you won't hear a lot of nasty phase interactions, wouldn't you still notice that something is a bit "off" in terms of timing? Again, this is where my ignorance starts showing with respect to phase. Like with complex math, I can do the calculations no problem, but I don't understand the theory behind them.

Sometimes being 360 degrees out of phase can be considered as "aligned". If you're familiar with math, maybe you're also familiar (or at least could quickly get a a grasp on) control systems? When you want to know the phase margin of the system to determine if it's stable or not, you don't really care about the phase being n*360 degrees off. If the system uses negative feedback, you want to make sure the feedback is actually negative, positive feedback causes the system to become unstable. In this case, all that matters is that the negative feedback does not become a positive one due to phase shift and a 360 degree phase shift would not change that.

Surprised that headphone review measurements only seem to show frequency response, but never things like an impedance curve, phase, or even distortion v. frequency. Those would all be super useful to know.

I think phase is a kind of red herring for headphones. They just don't have the kind of phase shift that's actually bothersome/audible IMO. How much the phase shift matters for audio playback depends much more on the steepness of the phase response, a large total amount might even be unnoticeable while a steep but comparatively smaller change could be especially bothersome. I haven't seen headphones with this kind of phase shifts.

 

[CallMeDaddy]

I use evga software to eq the optical out of my nu audio sound card. Need to tame the treble on my Ananda nano

 

[KaiFi]

Never. I am pretty staunchly anti-EQ and have been since I first got into this hobby in 2017. I get the sound I want from using headphones that have sound signatures I like and pairing them with the right amps (and changing the pads if necessary). I certainly create my own tuning that works for me, but not with EQ.