AutoEQ Magic?

[aldarrin]

As I wrote in the first post, I wanted to have a thread dedicated to AutoEQ. Not a thread about something else that mentions AutoEQ. I mentioned that I discovered AutoEQ in a different thread, but it was mentioned in passing. This is still the first thread solely dedicated to discussing AutoEQ. That's why I searched titles only.

 

[aldarrin]

At any rate. I'm learning more about EQ than I ever wanted to. And learning that Windows likes to turn back on whatever it this "Sound Enhancements" are when I turn my DAC on/off. When it does this is makes everything sound, just off. It's honestly similar to loading the EQ for a very different headphone. Nearing "I think I've wasted entirely too much on this hobby" levels of disappointing. Grrt. I can report that I'm very happy with AutoEQ for the most part. CCA C12's sound stunning on it (probably not kila-dollar good, but very pleasantly "this is fine"). Currently tweaking the settings for the EXP/95X. I think I've almost got the dip at 10K dialed out. I understand why people are "meh" about the stock experience, but between a pad swap and good digital EQ these are fantastic (and imaging in games that support Dolby Atmos for Headphones is insane).

 

[aldarrin]

Follow up, I'm using AutoEQ for CCA C12 IEMs for casual listening,and it's fantastic. But using the same setup with no with EQ on Koss 95X for gaming and Argon Mk3 for music is much better. Long story, short I ignored Ryan's recommendation and here's the results: "They sounded like something was missing from the audio. I could hear the bass, the mids, and the treble. I could hear instruments and details, but something felt like it was missing. It was like the audio equivalent of forgetting what you were about to say without any ability to remember what it was. Whatever it is, it's gone but just out of reach. I couldn't play Overwatch with them because suddenly I couldn't hear people sneak up behind me anymore. So I ordered a pair of the ZMF pads that you recommended on the Argon Mk3 page as the ones that sound like the original Argons and that did the trick."
I don't know for certain, but it seems something important is lost w/ EQ. EQ for low quality headphones works wonders, not sure about other sets.

 

[dasdoing]

Follow up, I'm using AutoEQ for CCA C12 IEMs for casual listening,and it's fantastic. But using the same setup with no with EQ on Koss 95X for gaming and Argon Mk3 for music is much better. Long story, short I ignored Ryan's recommendation and here's the results: "They sounded like something was missing from the audio. I could hear the bass, the mids, and the treble. I could hear instruments and details, but something felt like it was missing. It was like the audio equivalent of forgetting what you were about to say without any ability to remember what it was. Whatever it is, it's gone but just out of reach. I couldn't play Overwatch with them because suddenly I couldn't hear people sneak up behind me anymore. So I ordered a pair of the ZMF pads that you recommended on the Argon Mk3 page as the ones that sound like the original Argons and that did the trick."
I don't know for certain, but it seems something important is lost w/ EQ. EQ for low quality headphones works wonders, not sure about other sets.

could be the Harman bass boost. I tryed AutoEQ on my headhpones for mixing (plus a room convolution) and I ended up with too loud low mids. they are masked by the boost

 

[dasdoing]

also minimum phase eq will cause phase shifts.
maybe linear phase eq would be better when EQing phones?

 

[rxp]

I love AutoEQ - it's saved my DT990. Horrific sounding headphone before.

The other project Jaakko has is legit magic too: https://github.com/jaakkopasanen/Impulcifer/. I'm simulating my real room and applying virtual room correction to flatten frequency response, get rid of bass ringing and have perfect level/distances in the headphones. I often prefer it to my real speakers.

 

[dasdoing]

I love AutoEQ - it's saved my DT990. Horrific sounding headphone before.

The other project Jaakko has is legit magic too: https://github.com/jaakkopasanen/Impulcifer/. I'm simulating my real room and applying virtual room correction to flatten frequency response, get rid of bass ringing and have perfect level/distances in the headphones. I often prefer it to my real speakers.

have you heard my simulations? https://www.audiosciencereview.com/...ction-spatialization.13256/page-5#post-446230

 

[jaakkopasanen]

also minimum phase eq will cause phase shifts.
maybe linear phase eq would be better when EQing phones?

Minimum phase filters create the same phase changes as would be done by tuning the headphone by acoustical / mechanical means. Linear phase filters are only very, very specific use cases, not for equalizing headphones. Minimum phase filters are exactly what is needed for headphones.

 

[dasdoing]

Minimum phase filters create the same phase changes as would be done by tuning the headphone by acoustical / mechanical means. Linear phase filters are only very, very specific use cases, not for equalizing headphones. Minimum phase filters are exactly what is needed for headphones.

So the phaseshifts of the EQ will actualy correct existing phase shifts in the heaphone?

 

[dc655321]

I'm not disputing that minimum phase filters are appropriate here, but why say this?

Linear phase filters are only very, very specific use cases, not for equalizing headphones.

Also, this makes little sense to me:

Minimum phase filters create the same phase changes as would be done by tuning the headphone by acoustical / mechanical means

Are you trying to say it's a more "natural" approach (homeopathic filtering? )?

Minimum phase means (among other things) that the magnitude and phase are mathematically related and derivable from each other in isolation. Practically, if you change the magnitude of a minimum phase system, the phase must also change. In terms of headphone EQ, this may or may not be what is required (audibility of phase changes aside).

 

[xykreinov]

i'm a little confused. the topic is about AutoEq, and it incluedes the whole Oratory database: https://github.com/jaakkopasanen/AutoEq/tree/master/results/oratory1990

The database includes oratory1990's measurements, but, not his eq settings. The eq settings are done by AutoEq, as opposed to by hand like Oratory.
I agree with @tankas that Oratory's settings usually sound better. While I'm fuzzy on the subject, it seems to be because the program can't best a trained engineer in figuring out what resonances are unnecessary to correct and what are, yet. All I know is that AutoEq has gotten a lot better in recent months- I prefer the new AutoEq settings for my Sony 1000XM3 than the old settings from months ago. Looking at the graphs of the fr and eq, it seems the new settings tamper less with certain aspects of the fr in comparison to the old settings for the old graphs.

 

[jaakkopasanen]

So the phaseshifts of the EQ will actualy correct existing phase shifts in the heaphone?

I'm not disputing that minimum phase filters are appropriate here, but why say this?


Also, this makes little sense to me: Are you trying to say it's a more "natural" approach (homeopathic filtering? )?

Minimum phase means (among other things) that the magnitude and phase are mathematically related and derivable from each other in isolation. Practically, if you change the magnitude of a minimum phase system, the phase must also change. In terms of headphone EQ, this may or may not be what is required (audibility of phase changes aside).

Headphones are minimum phase devices so whatever frequency response corrections are done by eq must also do the phase "corrections" required to keep it minimum phase device. I keep corrections in quotes because the headphones don't have any phase problems being minimum phase devices. Using minimum phase filters ensures that the frequency response and phase response stay in sync and the result is the optimal one. Unless one enjoys non-minimum phase behaviour with headphones but I see no reasone why anyone would.

The only use case I've heard of (but there could be more) for linear phase filters are mixing drums recorded with multiple mics because the phases need to be correct which are naturally distroted due to different delays for mics in different locations. Or something like that. I'm not a recording engineer so take this with a grain of salt. Room acoustic corrections can use mixed phase filters to fix the frequency response and phase problems in the room (which are mostly but not always minimum phase) but here both need to be controlled separately and linear phase filters won't do.

 

[rxp]

Good to see you here Jaakko! (john29 at Headfi).

xykreinov, I actually sent my XM3's into oratory to measure which is where that EQ was from. It was actually one of the first corrections he did for me that I didn't enjoy, which was odd because I know he spent a lot of time with the headphone in trying to figure out the Sony EQ app. Compared to the other corrections he did for my WI-1000x or DT990 it just doesn't sound right to me. I've since moved onto the Bose 700s for my over-ears and mostly use Impulcifer anyway but worth seeing if the corrections work better on any other cans you have.

 

[jaakkopasanen]

The database includes oratory1990's measurements, but, not his eq settings. The eq settings are done by AutoEq, as opposed to by hand like Oratory.
I agree with @tankas that Oratory's settings usually sound better. While I'm fuzzy on the subject, it seems to be because the program can't best a trained engineer in figuring out what resonances are unnecessary to correct and what are, yet. All I know is that AutoEq has gotten a lot better in recent months- I prefer the new AutoEq settings for my Sony 1000XM3 than the old settings from months ago. Looking at the graphs of the fr and eq, it seems the new settings tamper less with certain aspects of the fr in comparison to the old settings for the old graphs.

oratory1990 adjusts the parametric eq filters by ear because he has the headphones at hand. I can't do that since everything in AutoEq is, well, automatic and based only on frequency response measurements. AutoEq also uses a bit less bass boost than what the Harman target asks for. I made this decision since not all headphones can do a large bass shelf like that without distorting and I have no way of checking if the headphone would distort. Another thing is that AutoEq's parametric eq filter optimization algorithm doesn't have safeguard against narrow bandwidth filters with positive gain which can introduce ringing. I intend to implement this eventually though.

There are differences in the parametric eq filters between AutoEq and oratory1990's own, but a lot of the differences people have reported have also been caused simply by having different pre-amp gains to avoid clipping and we all know how dominant effect on preference even a small difference in overall volume can have. If you can pinpoint what difference there is when the levels are matched, please tell me because I'm always eager to find out new ways of improving the algorithm.

 

[dasdoing]

Headphones are minimum phase devices so whatever frequency response corrections are done by eq must also do the phase "corrections" required to keep it minimum phase device. I keep corrections in quotes because the headphones don't have any phase problems being minimum phase devices. Using minimum phase filters ensures that the frequency response and phase response stay in sync and the result is the optimal one. Unless one enjoys non-minimum phase behaviour with headphones but I see no reasone why anyone would.

The only use case I've heard of (but there could be more) for linear phase filters are mixing drums recorded with multiple mics because the phases need to be correct which are naturally distroted due to different delays for mics in different locations. Or something like that. I'm not a recording engineer so take this with a grain of salt. Room acoustic corrections can use mixed phase filters to fix the frequency response and phase problems in the room (which are mostly but not always minimum phase) but here both need to be controlled separately and linear phase filters won't do.

So you are the man behind AutoEQ? it's a pleasure.

I am no expert either (btw, I have never seen any real expert, besides perhaps Bob Katz, the rest are trying there best).
From what I understand minimum phase is limiting phase shifts to the ones caused by magnitude shifts. excess phase would be what is caused by reflections. So far so good.
Does that mean phase shifts caused py magnitude shifts are "good phase shifts"? I don't thinks so, they are delay, and delays are not got.
linear phase filters correct magnitude without causing phase shifts on the other hand.
audio engineers use minimum phase EQs on sources(!) because the shifts are actualy an "desireble effect" (they also have much lower latency btw).
Now in case of reproducing devices we don't want those colorizing effects (in theory). We want to reproduce as transperent as possible.
I am not saying that liniar phase correction is better; I actualy don't know. But you have not offered a real explenation why minimum phase filters are the way to go.
Anyways, (natural) minimum phase shifts seam to be always single digit ms and therefore not audible. when I mentioned MP vs LP in this topic I responded to the before mentioned cases of people missing something from the sound after EQing, and sugesting this COULD be caused by some extreme phase shifting in case of extreme MP EQing. also I want to make clear that I mentioned this with a "?". I actualy bet on the Harman boost for the cause, as it masks the rest of the spectrum.

EDIT: if we correct a speaker to flat, we are actualy counter-shifting the existing phase shifts with MP filters, right? that's why MP makes sense for correcting the direct sound. Now in case of heaphones we are actualy going the oposite way: a "huge mountain" of magnitude shift is percieved flat. So we are not counter-shifting here

 

[Aldoszx]

Hi all.
I have a question, @jaakkopasanen how can I use it with Roon?
I tried to import convolution filter (.wav files) but it sems not working.
Thanks
Later edit:
My bad, it is working !

 

[jaakkopasanen]

So you are the man behind AutoEQ? it's a pleasure.

I am no expert either (btw, I have never seen any real expert, besides perhaps Bob Katz, the rest are trying there best).
From what I understand minimum phase is limiting phase shifts to the ones caused by magnitude shifts. excess phase would be what is caused by reflections. So far so good.
Does that mean phase shifts caused py magnitude shifts are "good phase shifts"? I don't thinks so, they are delay, and delays are not got.
linear phase filters correct magnitude without causing phase shifts on the other hand.
audio engineers use minimum phase EQs on sources(!) because the shifts are actualy an "desireble effect" (they also have much lower latency btw).
Now in case of reproducing devices we don't want those colorizing effects (in theory). We want to reproduce as transperent as possible.
I am not saying that liniar phase correction is better; I actualy don't know. But you have not offered a real explenation why minimum phase filters are the way to go.
Anyways, (natural) minimum phase shifts seam to be always single digit ms and therefore not audible. when I mentioned MP vs LP in this topic I responded to the before mentioned cases of people missing something from the sound after EQing, and sugesting this COULD be caused by some extreme phase shifting in case of extreme MP EQing. also I want to make clear that I mentioned this with a "?". I actualy bet on the Harman boost for the cause, as it masks the rest of the spectrum.

EDIT: if we correct a speaker to flat, we are actualy counter-shifting the existing phase shifts with MP filters, right? that's why MP makes sense for correcting the direct sound. Now in case of heaphones we are actualy going the oposite way: a "huge mountain" of magnitude shift is percieved flat. So we are not counter-shifting here

I'm afraid I cannot give you the mathematical formulation for minimum phase and why it's good for headphones but we can think of it in more practical terms. Let's take Dan Clark Audio headphones for example. They come with different felt pads you can put inside the earcup to reduce treble. These pads change magnitude response of the headphones and also at the same time change the phase response too. There is no delay caused by these pads nor do they introduce any other kind of anomalities either. They simply change the sound by reducing the treble. The heapdhone remains minimum phase device and therefore the magnitude and phase are interlinked. We can replicate the exact same behaviour with minimum phase eq filters. When chosen correctly, the eq filter will change the magnitude and phase in a similar fashion to the pads.

Minimum and linear phase filters' impulse responses look like this:

See how the linear phase filter has delay in it and the minimum phase does not? That's what we want. We don't want to introduce the delay and the pre-ringing of the linear phase filter.

Does any of this make sense?

 

[jaakkopasanen]

Hi all.
I have a question, @jaakkopasanen how can I use it with Roon?
I tried to import convolution filter (.wav files) but it sems not working.
Thanks
Later edit:
My bad, it is working !

I'm glad you got it working!

 

[dasdoing]

I'm afraid I cannot give you the mathematical formulation for minimum phase and why it's good for headphones but we can think of it in more practical terms. Let's take Dan Clark Audio headphones for example. They come with different felt pads you can put inside the earcup to reduce treble. These pads change magnitude response of the headphones and also at the same time change the phase response too. There is no delay caused by these pads nor do they introduce any other kind of anomalities either. They simply change the sound by reducing the treble. The heapdhone remains minimum phase device and therefore the magnitude and phase are interlinked. We can replicate the exact same behaviour with minimum phase eq filters. When chosen correctly, the eq filter will change the magnitude and phase in a similar fashion to the pads.

Minimum and linear phase filters' impulse responses look like this:

See how the linear phase filter has delay in it and the minimum phase does not? That's what we want. We don't want to introduce the delay and the pre-ringing of the linear phase filter.

Does any of this make sense?

well, the pre-delay is just introducing latency, which is not a problem at all (for listening). it occures equaly at all frequencies.
the pre-ringing is a problem, but we have to listen from case to case to the up and downsides (is it hearable at all?; what is more hearable if it is?).
check out this example where I step response corrected binaural impulse responses. the corrected ones have visible pre-ringing but I failed to hear it (and so seam to have others). the positive effect on the step response on the other hand was chosen to be better sounding by all on the other hand. sure, we are talking about phase shifts much more extreme here: https://www.audiosciencereview.com/...me-aligned-exess-group-delay-corrected.14489/

btw: phase shift and delay are the same. a phase shift of a given frequency causes a given delay.
for example, 180 degrees cause 1ms of delay at 500Hz and 10ms of delay at 50Hz: http://www.sengpielaudio.com/calculator-timedelayphase.htm
10ms is probably audible for everyone

 

[dasdoing]

this target realy sounds pretty flat: https://github.com/jaakkopasanen/AutoEq/blob/master/compensation/loudspeaker_in-room_flat_2013.csv
now I need to somehow put a little HF rollof above 10kHz on it to match realy match my speaker setup