[HN Gopher] AutoEQ: Automatic headphone equalization from freque...
       ___________________________________________________________________
        
       AutoEQ: Automatic headphone equalization from frequency responses
        
       Author : ishitatsuyuki
       Score  : 148 points
       Date   : 2021-10-08 14:47 UTC (8 hours ago)
        
 (HTM) web link (github.com)
 (TXT) w3m dump (github.com)
        
       | rokweom wrote:
       | Autoeq is not just a collection of ready-made EQ settings for
       | different headphones, it's also a database of measurements from
       | few different review sites and a tool for creating custom EQ
       | curves. Do you have a specific target that you like? You can
       | create a custom curve tuned for that target. Do you want to know
       | how a certain model of headphones sounds? You can make your
       | current ones sound almost exactly like them. Really cool project.
        
       | tanvach wrote:
       | I created and opened up the design of a standalone DSP [1] that
       | leverages AutoEQ. The DSP is optimized for our home brewed
       | headphones for VR, but I've been using it for my desktop
       | headphones too.
       | 
       | AutoEQ is not a perfect tool, but an amazing (and free) starting
       | point to get the most out of your audio systems.
       | 
       | [1] https://github.com/tanvach/prettygood_dsp
        
       | khimaros wrote:
       | would it make any sense to use this for tuning a car stereo?
       | particularly, to compensate for engine or road noise?
        
         | dsr_ wrote:
         | Not really. That's an immediate noise cancellation function,
         | which only works (a) with headphones and (b) by recording the
         | outside noise and adding its inverse to the sound being played.
        
           | khimaros wrote:
           | i don't mean necessarily doing active NC. i was more curious
           | about doing an initial tune with the car at idle or while
           | driving on the road as a way to compensate for the average.
           | is there any value in doing something like that? maybe even
           | just to tune the EQ profile of the car.
        
       | rweichler wrote:
       | If you have an iPhone X or older, you can jailbreak it and use
       | EQE (https://eqe.fm) which is system-wide and has AutoEQ
       | integration built-in
        
       | londons_explore wrote:
       | All this assumes the headphones are linear devices... which
       | clearly isn't the case!
       | 
       | None of these tools seem to do anything but rudimentary nonlinear
       | correction... Even simple things like measuring and cancelling
       | harmonics nobody seems to do...
        
       | natdempk wrote:
       | It would be interesting to compare the suggested EQ curves here
       | vs. commercial offerings that claim to introduce a neutral
       | response like Sonarworks. In theory I think they should be doing
       | quite similar things, but curious if there are any real
       | differences or if they land in the same general area.
        
         | dsr_ wrote:
         | This is exactly what Sonarworks is doing, but without a
         | specific software package to implement it (and sell to you).
        
       | Bayart wrote:
       | The precompiled AutoEQ/results/ tree has been a godsend, and so
       | is Wavelet. I honestly would think about buying a pair of
       | headphones that's not in it, unless I'm actively looking for a
       | specific sound signature (but it's really all about price and
       | comfort as far as I'm concerned).
        
       | kohlerm wrote:
       | it's great project. No idea why some people do not want to eq
       | their headphones at least a little bit. It can make a huge
       | difference. Also frequency response is IMHO at least partially a
       | matter of taste.
        
         | writeslowly wrote:
         | I've experimented with this and the results seem to vary
         | depending on the headphones. I have a set of planar magnetic
         | hifimans that don't have enough sub-bass out of the box, but
         | can easily make tons of sub-bass with eq adjustments. On the
         | other hand, some of the other headphones I tried this on sound
         | like they're just distorting more when I try to apply similar
         | levels of eq compensation.
         | 
         | There's also no way to apply a system wide eq to an iPhone, so
         | you'll need an external device.
        
         | MCllorf wrote:
         | from my experience, it's absolutely necessary if you listen to
         | a lot of podcasts because every tech-illiterate podcaster dude
         | will see an EQ setting on their microphone/recording software
         | and think "low frequencies are manly and sound good" and boost
         | the hell out of it, and it sounds absolutely awful.
         | 
         | Granted my home speakers might have something to do with it,
         | and headphones/earbuds do tend to have a high-pass filter built
         | in just from their construction, so it probably doesn't affect
         | everyone the same way.
         | 
         | I've been using equalizerAPO for desktop for years. I can only
         | think of a handful of content creators that don't pull that
         | bass-boosting garbage anymore, so the high-pass filter pretty
         | much always stays on unless I'm playing music.
        
           | dillondoyle wrote:
           | Or don't even normalize. The one podcast I regularly listen
           | too regularly has guests record over the phone or DIY and
           | they are often inaudible (to my over used ears). That'd be a
           | great feature on spotify if it doesn't exist
        
             | bee_rider wrote:
             | Somebody should really make an easy-mode podcast recording
             | app that functions as a phone call, but records the
             | speakers locally, and then sends the audio to the host. It
             | could even measure the latency on the call can cut that
             | time out. This seems like it would be pretty trivial...
        
         | h2odragon wrote:
         | > a matter of taste
         | 
         | Taste, yes; and individual circumstances. My hearing is so good
         | it might count as a disability, but I know where I have peaks
         | and valleys in my sensitivity and a good EQ can help with the
         | spikes taken by tinnitus ringing from youthful big boom car
         | stereo work and explosives.
         | 
         | I checked their suggestions for my headphones against my EQ
         | profile and theirs is pretty good. I like more lowfreq and much
         | less high freq than the "flat" they're correcting for.
         | 
         | They're offering a great resource for skipping the "what does
         | this set of cans sound like" stage. I probably spent 60hr or
         | more dialing these in when i got them.
        
           | mckirk wrote:
           | In which sense is your hearing 'so good it might count as a
           | disability'?
        
             | h2odragon wrote:
             | https://en.wikipedia.org/wiki/Hyperacusis
             | 
             | I don't hear "notes" and "chords", I hear frequencies. My
             | range (at near 50) still goes up to where i can hear the
             | bats talking. In a room with people, i cannot _not_ hear
             | their pulse, breathing, and other biological proceedings...
        
               | O_H_E wrote:
               | > i cannot not hear their pulse, breathing, and other
               | biological proceedings
               | 
               | Oh WOW, that is ...... wild and interesting
        
             | pvarangot wrote:
             | I had it as a kid, went away in my teen years and now I
             | only have tinnitus. For me it was only on certain
             | frequencies that were boosted and they are near the
             | principal harmonic of my always-there ringing tinnitus or
             | in the frequency of another noise I sometimes hear that's
             | more like a pure sinewave.
             | 
             | As far as I know it's more of a brain thing and not like a
             | super-ear thing.
        
           | natdempk wrote:
           | How can you find out where the peaks and valleys in your
           | hearing are? I've wanted to get basically an EQ curve for my
           | own ear hearing issues if possible as I'm sure I have some
           | minor hearing loss, but wasn't sure how to do that. Any tips?
        
             | h2odragon wrote:
             | listen to a frequency generator while twiddling its dial. I
             | dunno what would be the easiest tool to do that with right
             | now, i'd start in audacity or some "audio programming
             | toolkit".
        
               | natdempk wrote:
               | Yes, I know enough to know I can sweep a sine wave, but I
               | also know enough to know it's more complicated than that.
               | There are various curves that affect the perception of
               | sound volume at different frequencies, like the response
               | of the headphones, the varying response curve of
               | frequency perception at different volumes, the inherent
               | differing volume response curve in your brain/ears that
               | is the basis for stuff like LUFS. I was wondering if
               | there is a correct way to do this that corrects for all
               | these different effects, or something professional you
               | can do or pay for to get this measured correctly.
        
               | h2odragon wrote:
               | AFAIK "professional/medical" tests like 8 bands and may
               | have put a decibel meter to their equipment this month.
               | or not.
               | 
               |  _some_ professional audio engineers have some special
               | recordings they listen to on everything and use a faith
               | based or at least difficult to quantify internal process
               | to come up with the  "right" sound. I'm more in that end
               | of the spectrum.
        
               | natdempk wrote:
               | Ah okay, I'm surprised there isn't something more precise
               | and scientific out there. Interesting to know that the
               | theoretical gold standard I would really want might not
               | even exist.
               | 
               | Edit: I stand corrected, as the poster below mention (I
               | can't reply), I'd want an audiogram from a professional.
               | Thanks!
        
               | mh- wrote:
               | You want an Audiogram from a professional. There are
               | definitely medical facilities that can do this to a
               | sufficient degree of accuracy.
               | 
               | https://en.wikipedia.org/wiki/Audiogram
        
             | mh- wrote:
             | You can go to an audiologist that can produce this for you,
             | if it's worth it to you
        
       | serverholic wrote:
       | FYI, EQs will distort audio. Especially if the adjustments are
       | large.
       | 
       | Your average EQ will introduce phase shifts to the various
       | frequencies that make up a sound.
       | 
       | Linear-phase EQs don't phase shift but they also introduce pre-
       | ringing and post-ringing effects.
        
         | bitbang wrote:
         | Depends largely on how it's implemented. There is a big
         | difference between the quality of a well implemented
         | convolution engine and a crappy biquad filter.
        
           | l33tbro wrote:
           | Where would something like Fabfilter fall in the scale? I
           | imagine there's some pretty complex stuff going on under the
           | hood.
        
             | serverholic wrote:
             | Fabfilter isn't really doing anything magical. Its default
             | mode is a pretty normal EQ with phase shifting. Linear-
             | phase mode is like any other linear-phase mode with the
             | same drawbacks.
        
               | l33tbro wrote:
               | Probably the UI that made me think it was a Wonka
               | creation. It's really not sonically much different to a
               | stock EQ.
        
           | nitrogen wrote:
           | Note that not all biquad filters are crappy.
        
         | ishitatsuyuki wrote:
         | Analog distortion of signals also affects the phase. A digital
         | filter might or might not compensate for it, but it's not
         | always bad.
         | 
         | Also, human ears tend to be more sensitive to amplitude than to
         | the phase. Especially if the same filter is applied on both
         | channels, which leaves the phase difference unchanged.
        
           | serverholic wrote:
           | It's more than just human ear sensitivity. Phase shifting
           | changes the relationship between frequencies in a sound.
           | 
           | For example, phase shifting can change the amplitude of a
           | sound. Analog hardware is especially sensitive to changes in
           | amplitude so you might be introducing distortion just by
           | shifting phase.
        
         | willis936 wrote:
         | The primary measurement is called group delay, which is the
         | derivative of phase with respect to frequency (since a linear
         | phase is just a uniform delay). Fortunately we have studies
         | from the 70s publicly available that establish a group delay
         | audibility threshold.
         | 
         | FIRs add vanishingly small amounts of distortion. Their real
         | drawback is in the added delay. For asynchronous music playback
         | this is nothing to worry about, but for video or anything
         | interactive (communications or gaming) it's going to be a real
         | tough pill to swallow.
        
         | praash wrote:
         | Phase shifts make no practical difference when you're
         | equalizing headphones. You really shouldn't need to use notch
         | filters there.
         | 
         | EQ phase is relevant if the dry signal has a chance to mix with
         | the EQ output.
         | 
         | Changes in phase is not "distortion". When you end up making
         | large boosts (especially to make sub bass audible), that's when
         | your headphone drivers might start to distort sound. This is
         | fixed with less volume, obviously.
        
           | serverholic wrote:
           | An EQ that cuts frequencies can actually increase the
           | amplitude of the waveform due to phase shifts. It's not just
           | boosting. What I'm talking about is more relevant on analog
           | gear which is more sensitive to volume.
        
         | h2odragon wrote:
         | At headphone sizes, does your DAC have a good enough clock to
         | even speak about phasing?
        
       | thih9 wrote:
       | It even has settings for Apple's Ear Pods [1] (the wired
       | earbuds). I remember being surprised how applying these values
       | makes them resemble regular headphones.
       | 
       | [1]:
       | https://github.com/jaakkopasanen/AutoEq/tree/master/results/...
        
       | scblock wrote:
       | I would caution against applying an EQ curve to headphones
       | without review beforehand. If EQ includes large boosts (which the
       | example image and table shows) that can easily introduce clipping
       | distortion if applied linearly without appropriate headroom
       | adjustments.
       | 
       | Additionally, depending on transducer performance large bass
       | boosts may simply increase transducer distortion in the bass
       | frequency. Some headphones respond well to EQ, and others not so
       | much.
       | 
       | I am not opposed to EQ, but it should be applied judiciously, and
       | at least partially by ear. I don't believe trying to exactly
       | match a target curve will necessarily provide good results
       | compared to a more judicious approach.
        
         | sanjiwatsuki wrote:
         | AFAIK, every tool that integrates with these things apply a
         | negative boost equal to the peak added to avoid clipping.
        
           | dsr_ wrote:
           | And every recommendation on the site starts with the
           | appropriate amount of negative gain.
           | 
           | But if you miss it, somehow, then setting it to the next
           | whole number of decibels larger than the largest gain will
           | do.
        
         | scblock wrote:
         | As an example of my point for using a judicious approach above,
         | applying either of the sets of Sennheiser HD8XX measurements
         | and resulting AutoEQ curves from the Crinacle and oratory1990
         | folders in the Roon DSP system collapses the headstage and
         | tilts the perceived tone of the headphones from relatively full
         | and engaging with a dip in the 2-3 kHz region to thin and
         | hollow, with an overly bright, brittle top end.
         | 
         | Also, these are two sets of measurements of theoretically the
         | same headphones, each with 10 adjustment points, some quite
         | broad, some very narrow. The overall shape of the resulting
         | curve for the same target compensation is similar, but they
         | have some significant differences in the specifics, and they
         | sound different. We have to remember that measurement systems
         | and individual measurement setup can vary quite a bit, so
         | settings from this tool will bake all of that in as well.
         | 
         | As a comparison, based on review of the measured curves in more
         | of a a big picture way, applying a much simpler EQ with a broad
         | 1.5 dB bass lift up to about 100 Hz, another broad lift of
         | about 3 dB around 2 kHz, and a slight drop of about 1 dB
         | centered around 8 kHz brings the bass and vocals up a little
         | but keeps headstage and overall tone intact. Maybe it's this
         | particular DSP implementation, but I would be wary about trying
         | to use any of the precompiled results directly. EQ can have
         | real benefits, but is going to be more personal than automated
         | settings will capture.
        
       ___________________________________________________________________
       (page generated 2021-10-08 23:00 UTC)