Software that emulates phase in headphones?

[Fluffy]

I'm curious to understand what impact does the phase plot of headphone has on sound. I wonder if there is a way through software to make my Audeze Lcd2C:

Behave like my Focal Clear:

In terms of phase.

Or maybe I'm not understanding this right and the phase has nothing to do with the sound and is only meaningful to the electric behavior of the driver. Logically, I assume that if there is a phase difference between frequencies, there is some smearing of the sound wave, because they are not arriving at the phase they are intended to.

Suggestions? Corrections?

 

[solderdude]

Electrical phase and acoustical phase are not the same thing.
Innerfidelity measures electrical phase only.
Rtings measures acoustical phase.
Here are some plots (also of LCD2) with phase response

 

[andreasmaaan]

Yes, as @solderdude said, electrical and acoustical phase are two entirely different things.

If you want to experiment with listening to modelled acoustical phase effects, you can use software like RePhase to create an all-pass FIR filter and then (among others) Foobar or Equalizer APO to pass your music through it.

I realise that's a very skeletal explanation, so let me know if you'd like some more details.

 

[solderdude]

What I understood from J_J is that a gradual phase shift isn't very problematic only sharp phase shifts in a very narrow band seem to be audible.
Have a look at the Rtings measurements and observe how the phase swings in all directions in the treble.

Personally I don't worry about phase that much, there is almost nothing you c an change about it.

Electrical phase may not be very problematic but depends on the amplifier of course. Never really investigated it. I noticed PMA was doing some tests with speaker amps.
Most headphones do not have a very worrysome phase response IMO.

 

[andreasmaaan]

@solderdude it's interesting you bring up the high-frequency phase swings in headphones. I wonder if you've ever measured headphones in the absence of any HATS test fixture or other major acoustical obstruction between the driver and the microphone, and if so whether you've observed this behaviour under those conditions too? I would have thought a lot of that erratic phase behaviour (and indeed amplitude behaviour) in the high frequencies in headphones would be attributable to the acoustical environment in which the measurement takes place, which is usually some kind of HATS rig or similar, or at least a sealed or semi-sealed cavity of some kind (if I'm not mitaken).

 

[andreasmaaan]

Additional thought: I guess the cups are also significant factors in this...

 

[Fluffy]

Electrical phase and acoustical phase are not the same thing.
Innerfidelity measures electrical phase only.
Rtings measures acoustical phase.
Here are some plots (also of LCD2) with phase response

Thanks for the clarification, I guess I was way off on this one.

If you want to experiment with listening to modelled acoustical phase effects, you can use software like RePhase to create an all-pass FIR filter and then (among others) Foobar or Equalizer APO to pass your music through it.

I use Eq Apo regularly, so I downloaded that RePhase, and it looks pretty complex. I don’t even understand how to do what you proposed or what would it do. Question – wouldn't applying any kind of filter change the frequency response?

 

[andreasmaaan]

I use Eq Apo regularly, so I downloaded that RePhase, and it looks pretty complex. I don’t even understand how to do what you proposed or what would it do. Question – wouldn't applying any kind of filter change the frequency response?

An all-pass filter can change phase response without affecting the frequency response. This is one of the wonderful things about digital audio

To set up such a filter, go into RePhase and click on the "paragraphic phase EQ" tab. You can then enter your phase equalisation filters into the console at the bottom of the screen in much the same way you'd use a conventional paragraphic (amplitude) EQ, in other words by setting the frequency, Q, and magnitude (in degrees rather than decibels):

To "correct" a non-phase linear headphone (or any audio device for that matter), what you'd do would be to take the measured phase response and to then create an inverse of it using these filter settings. (For example, if your measured phase response had a 90° peak at 1000Hz with a Q of 1, you would create a filter with a 90° dip at 1000Hz and with a Q of 1.)

But to do this correctly, you'd need to have an accurately measured phase response of the headphone or other device you were trying to correct. For some of the reasons I mentioned in my previous two posts, I'm not sure it would be valid to assume that the phase response of a pair of headphones measured on a HATS or similar test rig would match the phase response of the same headphones on your head.

Anyway, assuming for argument's sake you did have an accurate measurement, once you'd used RePhase to create the correction filter, you would move across to the "impulse settings" tab on the right and plug in your desired settings. In your case, I would begin with the default tap no. and FFT length of 16384 and 65536, respectively. I'd suggest selecting "Blackman-Nuttal" as the windowing function, and set optimization to "extensive" to -100dB. Then under format, select "24 (or 32) bits LPCM mono", and under sample rate, select whatever sample rate your playback system normally operates in.

Finally, select the desitnation directory where you want to save the filter and hit "generate". The filter will be saved as an impulse response in 24-bit WAV format (at whatever sampling rate you selected).

Next, go into Equalizer APO and add a new "Convolution (Convolution with impulse response)" filter, which you'll find under the "advanced filters" submenu. Hit the file icon on the filter and browse to select the WAV file you just generated in RePhase.

Assuming nothing went wrong, Equalizer APO will now convolve all audio with this filter.

However, like I said above, although this is how you would do it, I'm not sure it makes any sense given the uncertainty concerning the validity of any headphones' measured phase response vs its phase response on your head. My hunch would be that in low and low-mid frequencies, a non-specific measurement might be reliable, but not so in the higher frequncies. I reckon @solderdude might have something interesting to say on this question...

 

[andreasmaaan]

PS @Fluffy, I'd also say that, in respect of most single driver headphones (which means most headphones), any deviations from a flat phase response are unlikely to be great enough in magnitude or Q to be audible in the first place, anyway.

 

[solderdude]

@solderdude it's interesting you bring up the high-frequency phase swings in headphones. I wonder if you've ever measured headphones in the absence of any HATS test fixture or other major acoustical obstruction between the driver and the microphone, and if so whether you've observed this behaviour under those conditions too? I would have thought a lot of that erratic phase behaviour (and indeed amplitude behaviour) in the high frequencies in headphones would be attributable to the acoustical environment in which the measurement takes place, which is usually some kind of HATS rig or similar, or at least a sealed or semi-sealed cavity of some kind (if I'm not mitaken).

Yes I measured the phase response but because I also compensate the graph directly in the lows (6dB/oct) and compensate the used mic for being mounted close to an infinite baffle in the treble there are also phase shifts in the results which originate from the filters so showing those plots would be misleading.

I guess I could replace the analog filters with FR compensation as used in REW calibration file which would probably be better.
This will be a lot of work so haven't done this (yet).
Maybe in the distant future I might do this. For now consistency with older measurements is guaranteed when I keep using the rig.
I did swap microphones though and getting the new mic (which has lower hum) to measure exactly the same as the old mic was a lengthy pain in the but and also still not completely perfect but more than close enough. Distortion levels are now slightly higher because this mic too is an electret.

 

[andreasmaaan]

Yes I measured the phase response but because I also compensate the graph directly in the lows (6dB/oct) and compensate the used mic for being mounted close to an infinite baffle in the treble there are also phase shifts in the results which originate from the filters so showing those plots would be misleading.

I guess I could replace the analog filters with FR compensation as used in REW calibration file which would probably be better.
This will be a lot of work so haven't done this (yet).
Maybe in the distant future I might do this. For now consistency with older measurements is guaranteed when I keep using the rig.
I did swap microphones though and getting the new mic (which has lower hum) to measure exactly the same as the old mic was a lengthy pain in the but and also still not completely perfect but more than close enough. Distortion levels are now slightly higher because this mic too is an electret.

Fair enough I think you're right to favour consistency with older measurements, as in most cases there's probably little use in accurately measuring the phase response of a pair of headphones.

 

[Fluffy]

In conclusion, I tried to play a bunch with impulses generated from Rephase with all kinds of weird phase shifts. Some just exaggerated to see if I can hear any difference, some trying to approximately correct for the headphones phase. I also double checked myself with loopback recording in Audacity to validate that phase differences are indeed being applied. All in all, I can't say any of this made any significant sonic difference. Definitely not something I would be able to distinguish in a blind test.

Oh well, at least that's a fun experiment. Thanks for all the help!

 

[andreasmaaan]

In conclusion, I tried to play a bunch with impulses generated from Rephase with all kinds of weird phase shifts. Some just exaggerated to see if I can hear any difference, some trying to approximately correct for the headphones phase. I also double checked myself with loopback recording in Audacity to validate that phase differences are indeed being applied. All in all, I can't say any of this made any significant sonic difference. Definitely not something I would be able to distinguish in a blind test.

Oh well, at least that's a fun experiment. Thanks for all the help!

Haha, have been there myself, and so have psychoacousticians. That's the way human hearing works Mild (or even quite large) phase shifts are nothing to worry about.

 

[asr6576]

Try rePhase FIR tool. Do an ABX test and see whether you can distinguish the minimum phase filters and the linear phase filters in normal music. (the difference of which can be heard in extreme conditions). If not, you no longer need to consider phase correction.
I failed the ABX test and I always use minimum phase filters since then for the minimum delay.

 

[Fluffy]

Try rePhase FIR tool. Do an ABX test and see whether you can distinguish the minimum phase filters and the linear phase filters in normal music. (the difference of which can be heard in extreme conditions). If not, you no longer need to consider phase correction.
I failed the ABX test and I always use minimum phase filters since then for the minimum delay.

As I wrote above, I already did that and couldn't hear any differences. I conclude this really isn't an issue to be worried about.