"You wouldn't download a car..." Turns out maybe you can? The free endgame solution for objectivists

[ChickenChaser]

So, I never really fell for the subjectivist talk around amps and DACs, much less cables and the like. But my experiences led me down the rabbit hole with headphones. I like the LCD-2 Revision 2... tried the LCD-3 and realized I liked it quite a bit more. Then I liked the LCD-4 quite a bit more, too. I liked the HD800 way more than the HD6XX line, even before fixing it with the SDR and EQ. This gave me some reason to think price and quality really correlated with each other in headphones in a way I knew they didn't with other aspects of audio.

Gradually, I'm thinking more and more about whether there's more to headphones than frequency response. Detail... soundstaging... am I really sure terms like these mean something? As I dig into the arguments, I see that CSD plots might reveal different properties of headphones (specifically different amounts of ringing at different frequencies, which show up in the frequency response, but originate from a different source). Even if headphones do differ in amount and placement of ringing, it's questionable if this is really audible by human beings. Maybe distortion could influence perception of sound in some cases, but this is usually unlikely.

Turns out experiments in audio research journals have already been done emulating one headphone's frequency response on another one, and they've found that FR makes up for the vast majority of preference ratings for headphones, if not all of it.

I speculated in a few places that someone could one day make a headphone with DSP embedded to let it emulate the sound of a variety of other high-end headphones at will. People assured me that this was a long way away from potential reality, even if they thought it was possible in theory. Little did I realize something called AutoEQ already lets you create an EQ for any headphone to tune it to the FR target of any other.

The LCD-4 is my favorite of any sound I've ever heard. It helped me identify that I like a smoother rise in the midrange compared to the Harman curve - rising earlier, and not peaking so high at 3dB. If you graph this with Harman set as baseline, you'll see a hump at 1-1.5kHz falling into a dip at 2kHz that moves back towards 0dB as it gets back to the usual peak around 3kHz. Then the laid back treble and bass, thanks to the equal loudness curve, means it sounds right/neutral at a slightly higher volume, which lets me really feel the impact of heavy guitars and so on in certain kinds of music. After the LCD-4, it's frustrating putting anything else on and not being able to get it loud enough without noticeable discomfort.

Well, I used AutoEQ to generate LCD-4 curves for two headphones. One of them is the fucking HD800, nearly the polar opposite in terms of the stereotypes around its sound. The other is a ZMF closed back. And guess what? I have 0 interest in ever paying $2-4000 with a possible $1250 to replace drivers out of warranty ever again. Not only is this "close enough" to make me happy about spending the cash on something else, I am happier than I would be with an actual LCD-4. Even though the Audeze weight never bothered me as such, the HD800 is infinitely comfier. And the closed back gives me isolation and subbass impact I couldn't get with a real LCD-4. Again, I say this having owned and used an LCD-4 for hours a day for over a year. I can literally get everything I loved about it for free. I honestly can't in good conscience advise anyone to buy audio gear anymore. I really just can't at all. This thing has officially put me at the end of my search for good, and I can't see any reason why it wouldn't be the absolute universal recommendation everywhere to anyone asking what headphone they should look to buy, to just try the house sound out with AutoEQ. No one needs to worry about demo'ing and returning gear, it making it to meets, to find their favorite sound. This feels even bigger than it felt when I discovered you could download video games off the Internet as a kid.

 

[Blumlein 88]

https://www.sonarworks.com/standard

Sonarworks has something similar. Doesn't let you create your own, but will let you fix the one you have. You can download trials for your phone.

Also looks like they are about to discontinue the software I was familiar with to replace with a new version. Still the same principle.

Of course AutoEQ is free which is much better.

 

[McFly]

Agree, for the most part. I love DSP. Now go try the same thing on say, some $80 beats headphones

 

[Blumlein 88]

Oh, the good DSP will spread. Is happening more slowly than I thought, but I was impatient.

On the other end of things there are a few efforts to do the same thing for microphones. Get a good basic flat performing microphone and DSP/EQ it to mimic the multi-thousand dollar Neumann or Schoeps you can't afford. Seems to work at least fairly well.

 

[Fluffy]

Here we go again with the Auto EQ. I've said many times that I don't believe in it, because a headphone sound is not only it's frequency response. Distortions, resonances, and the acoustic build of the cup play a significant role in how the headphone's sound is perceived.

But just for heck, I tried turning my Audeze LCD2C into my Focal Clear. I've downloaded their respective parametric EQ parameters, and applied them both in APO EQ, with the parameters of the Focal reversed in sign. That should first flatten the response of the LCD2C, and then add in the response of the Focal. The result actually somewhat approximates the sound of the Focal, but I wouldn't say it sounds exactly like it. It still sounds like one headphone trying to disguise as a different one. I don't believe I would fall for the sound of the approximated Focal coming from the Audeze, the way I fell for the sound of the actual Focal.

 

[ChickenChaser]

There are some pretty good mathematical reasons why reversing the EQ values doesn't always work very well. I tried that, and the results I got were nothing like AutoEQ. If you use AutoEQ, it doesn't address narrow dips and peaks in the treble, because where that peak/dip shows up on a measurement rig is no guarantee of where it will show up on your head. The measurement rig could show a peak at 14kHz, but when you put it on your head it could be all the way at 10kHz. So AutoEQ has an algorithm for getting the overall curve pretty damn close while letting the narrow spikes be. You can use the graphs as a rough guide to help you figure out how to address those manually if you want to take it a step further: "okay, the graph shows a narrow spike at 14kHz of 5dB, so we'll start there and move a narrow -5dB peak around until we think we've found it..."

One headphone makes the LCD4 EQ sound a little "brighter," the other one makes it sound a little "darker." I suspect it's just a narrow peak in one and a narrow dip in the other, but I actually like it this way. I always liked having different sounds to switch between as my brain gets too used to a certain presentation anyway. If I was trying to make it sound as perfect an emulation as possible for someone else though, making it sound perfect to me might not work, I'd need to show them how to address the peak/dip with it on their own head.

 

[Fluffy]

I'm all for DSP, and I EQ the heck out of my headphones. But I do it solely based on how it sounds to me, not based on any measurement or target curve. What you said about the peaks and the minute details that are different in each person (or HRTF) is very much correct. You can only approximate how a different transducer might sound – never perfectly emulate it. So that's why you will never get the actual experience of listening to an HD800 when trying to emulate it in LCD4. You can however, turn a "dark" headphone into a "bright" one, that's not a problem at all.

By the way, I delved into the peak hunting thing with my Focal Clear. I tried to tame a slight sizzle in the treble for a long time, using tone seeps and what not. I got it to a point where in a frequency sweep it would sound flat, but in actual music it stole the liveliness and presence. So I ditched that approach and just applied a slow shelf to reduce the overall treble levels. It made them less fatiguing while still sounding exciting. My conclusion is that hunting down and correcting very narrow peaks (I used a q of 18) is a fools errand.

 

[JeffS7444]

MiniDSP's HA-DSP turned me into a DSP believer some time ago. Today if I were to spend more than $200 for the likes of Sennheiser HD8xx-series headphones, it would have to do more with comfort and luxury than anything else. But DSP'd Fostex T50RP II works so well once you replace the too-shallow stock ear pads (couldn't get a proper seal) that I dunno that I can be bothered.

 

[jae]

I imagine this would probably work best on planars with competent drivers to begin with, which ironically are the most expensive headphones.

 

[ChickenChaser]

I've taken the online test that shows I can hear distortion down to -26dB, which put me well above average even with an open-back in a far-from-quiet room. I'm making a very bright dynamic (HD800 SDR) sound like a very warm planar (LCD-4) without any issues. No audible distortion that I can identify (plugged into a lowly Atom -> Modi 2).

The only test I've had on hand to judge "what's the cheapest headphone I can replicate this on?" is a ZMF Atticus. That didn't sound right when put on the LCD-4 setting to me at all. However, there's a pretty big peak around 14kHz. As mentioned before, AutoEQ doesn't try to iron every one of those out, and graphs showing where the peak is on a dummy head don't guarantee where those treble peaks will show up when it's on your head. I could work on increasing the number of taps in the generated EQ and increasing the default allowed amount of pre-amp headroom, then try to manually find and reduce the peak further, and I think that could fix it.

Anyway, this leads me to think the top priority is just using something with the smoothest treble response possible, or at least without any peaks that bother you. And you'll have more issues if that peak or dip is in a part of the treble that needs to be adjusted significantly to reach the target EQ for the other headphone. Uneven bass is ridiculously easy to level out with EQ. Uneven mids are doable with some work, learning, and trial and error, and uneven treble is really difficult.

 

[Blujackaal]

True Headphone mimicking is possible if the driver can handle shifts of 12db -/+. AutoEQ seems to have gotten the Reddit gang into a frenzy since the last 3 days since this kill the subjectivist idea on driver speed, detail/res.

 

[Fluffy]

True Headphone mimicking is possible if the driver can handle shifts of 12db -/+

That doesn't make much sense. Every driver can handle shifts of way more than 12 db to any frequency range. That's basically what music is – different frequencies played at varying amplitudes.

 

[FrantzM]

I imagine this would probably work best on planars with competent drivers to begin with, which ironically are the most expensive headphones.

I would think the <$200 HifiMan HE400i and its Drop clone the HE4xx, have "competent" drivers.. Wouldn't you?

 

[Blujackaal]

That doesn't make much sense. Every driver can handle shifts of way more than 12 db to any frequency range. That's basically what music is – different frequencies played at varying amplitudes.

The Grado's, HD6X0 drivers, Struggle more with bass boosts than other headphones. They can't mimic the +10db bass the Shure SRH1540 has without dying. Just like how some planars might struggle with DD/BA gear since they don't have a weird 4k dip that can max out at -8, Think the audeze LCDi4.

It's a common argument why does The Andro, SE846, 8BA CIEMs need more than 3 drivers. When the ER4XR & W30 can easily handle their FR tunings if not better since there will no crossover artifacts from having to handle 4 ~ 12 drivers.

 

[Fluffy]

I honestly don't know much about IEMs or why they need multiple drives, so I can't address that point.

About the bass boost – it depends on the absolute SPL and the SPL limit of the headphone. If a Grado headphone has -20 db roll off at 20 hz at normal listening level, that means you could lower the volume by 20 db and boost 20 hz up by 20 db, and it would have a flat bass response at that volume without issues. or course, if you then crank up the volume back to normal listening levels, the driver could be heavily distorting or clipping when playing bass-heavy content. Though on the other hands, some headphones like from Audeze are capable of handling very high spls without distorting, so you can compensate any dip in their response without worrying about it.

 

[Blujackaal]

I honestly don't know much about IEMs or why they need multiple drives, so I can't address that point.

About the bass boost – it depends on the absolute SPL and the SPL limit of the headphone. If a Grado headphone has -20 db roll off at 20 hz at normal listening level, that means you could lower the volume by 20 db and boost 20 hz up by 20 db, and it would have a flat bass response at that volume without issues. or course, if you then crank up the volume back to normal listening levels, the driver could be heavily distorting or clipping when playing bass-heavy content. Though on the other hands, some headphones like from Audeze are capable of handling very high spls without distorting, so you can compensate any dip in their response without worrying about it.

I remember solder doing a 6db bass boost at 150Hz on the GS2000e through output resistance on amp's, That made it's distortion go to >20%.

 

[Willem]

How about the same with speakers?

 

[Sir Sanders Zingmore]

I realize something called AutoEQ already lets you create an EQ for any headphone to tune it to the FR target of any other.

OK I'm prepared to look silly here with my question…. The AutoEQ curves seem to create EQ to make everything match the Harman Curve. How do you get one headphone to match the FR of another?

 

[Fluffy]

OK I'm prepared to look silly here with my question…. The AutoEQ curves seem to create EQ to make everything match the Harman Curve. How do you get one headphone to match the FR of another?

To my logic, applying the reverse of the correction of the HP you're trying to simulate. Just flip the +/- sign in every filter.

 

[ChickenChaser]

OK I'm prepared to look silly here with my question…. The AutoEQ curves seem to create EQ to make everything match the Harman Curve. How do you get one headphone to match the FR of another?

CTRL+F on the main page for "HD650" and you'll see a sample of how to make an HD650 profile for the HD800. Adapt the coding used for that for whatever you're trying to make instead. The "measurements" page has instructions you need to follow before it will work. Then there are some instructions on the main page for installing it too. Oratory1990's rig has the best measurements by far, followed by Crinacle but you have to toss him a few $ on Patreon to get them. Others are significantly less reliable/accurate so only use them if there's no other option. There's stuff you have to install, then you ultimately have to get AutoEQ running from your command prompt where you type in an instruction and it spits out the code to a predefined folder. It can be tricky so I'll be happy to answer questions to try to help from here.