• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Unit to Unit Variation

concorde1

Active Member
Joined
Mar 15, 2021
Messages
261
Likes
167
Location
New Zealand
I have 3 headphones currently, all EQ'd to Harman/AutoEQ target.

Fostex TH900mk2 sounds really great with EQ.

Klipsch HP-3 sounds tinny with EQ.

Audeze LCD-3 sounds dull with EQ and a member on ASR has done a specific AutoEQ for my frequency response graph supplied by Audeze, and the result has a 6dBr dip at 5 kHz relative to the target.

My HP-3 probably measures quite differently from Crinacle's measured one, and my LCD-3 definitely measures very differently from Oratory's measured one.

This all leads me to wonder, what are the headphones out there with low unit to unit variation in frequency response?

I believe the LCD-2 (which I used to own) has good consistency these days - Audeze claims the variation is low.

I may have simply gotten lucky with my TH900mk2 - or maybe it's just a consistent measuring headphone. Unfortunately while it sounds great its build quality is poor.

I partly want to sell the HP-3 and LCD-3 and get either an LCD-2 or HD800S.
 

Helicopter

Major Contributor
Forum Donor
Joined
Aug 13, 2020
Messages
2,669
Likes
3,835
Location
Michigan
I think companies are probably the only ones with much real statistical data on this, and I am sure they want to keep their tolerances for acceptable sound quality secret in most cases. Sean Olive had mentioned Harman buying multiple pairs of some headphones to analyze this, and found that unit variation does tend to be pretty high. He may have even mentioned a couple brands, but I don't remember them if he did.
 
OP
concorde1

concorde1

Active Member
Joined
Mar 15, 2021
Messages
261
Likes
167
Location
New Zealand
Well that means two frightening things.

1. It's luck if you find a headphone unit that AutoEQ suits perfectly (I'm lucky with my TH900mk2, but HP-3 sounds bad). [Doomed if you don't have a means to find measurements for your unit, unless you are very good at EQ-ing by ear]

2. For a headphone which has been measured with AutoEQ to fit Harman target nicely, your unit potentially may not be able to fit well even if you have the ideal EQ for your specific unit (ie my LCD-3). [Just doomed]

My HP-3 may have an ideal EQ that would allow it to fit Harman target, whereas my LCD-3 does not. But I probably cannot access a GRAS thing.

It's funny how LCD-2 is the one with little variation but it's one of Audeze's less expensive products, makes no sense.
 

staticV3

Major Contributor
Joined
Aug 29, 2019
Messages
1,643
Likes
2,316
My recommendation would be to make sure there are many reputable measurements available for the headphone you're considering. The more there are, the higher the chance that one of them matches your particular unit.

In my case, it was ReferenceAudioAnalyzer's measurement of the HD600 that best matched my unit, despite RAA's custom HATS showing terrible consistency compared to GRAS.

Of course, one way to avoid all his would be to learn to EQ by ear
 

staticV3

Major Contributor
Joined
Aug 29, 2019
Messages
1,643
Likes
2,316
Regarding that AutoEQ preset I created for you, I can send you a modified version with full adherence to the target if you'd like. But it would be a high band GEQ for use with EQApo, not a PEQ.
 
OP
concorde1

concorde1

Active Member
Joined
Mar 15, 2021
Messages
261
Likes
167
Location
New Zealand
Regarding that AutoEQ preset I created for you, I can send you a modified version with full adherence to the target if you'd like. But it would be a high band GEQ for use with EQApo, not a PEQ.
That would be great!

After a bit of listening I was thinking my LCD-3 actually is good enough with the preset you made that I'll keep it, but I would like to try this GEQ sure.
 
OP
concorde1

concorde1

Active Member
Joined
Mar 15, 2021
Messages
261
Likes
167
Location
New Zealand
Newbie question no doubt:

To determine if I'm using AutoEq correctly, as a "control test", I'm trying to AutoEq off Crinacle's result on the HP-3, but . . .

When digitizing, how do I make the graph here:
crinaclehp3.jpg

match the black line on:
Klipsch Heritage HP-3.png

Do I just digitize the red/blue lines AND the black line on first graph, and add the graphs together?
Regarding that AutoEQ preset I created for you, I can send you a modified version with full adherence to the target if you'd like. But it would be a high band GEQ for use with EQApo, not a PEQ.
Also I really would appreciate that full adherence GEQ :)
 

Attachments

  • crinaclehp3.png
    crinaclehp3.png
    33.7 KB · Views: 27

staticV3

Major Contributor
Joined
Aug 29, 2019
Messages
1,643
Likes
2,316
That would be great!

After a bit of listening I was thinking my LCD-3 actually is good enough with the preset you made that I'll keep it, but I would like to try this GEQ sure.
Here's how to use GEQs with Equalizer Apo:
  1. Download EQApo: https://sourceforge.net/projects/equalizerapo/
  2. Install EQApo
  3. Start Configurator and install EQApo for your DAC. Then restart your PC
  4. Start EQApo's Editor
  5. Add preamp:
    Screenshot 2022-02-12 002728.png
  6. Add Graphic equalizer with variable bands:
    Screenshot 2022-02-12 002750.png
  7. Click on import, choose your .csv file
    Screenshot 2022-02-12 002814.png
  8. Reduce you Preamp's gain to -9.70dB
Afterwards you can switch between the two presets and also toggle EQ On/Off without a change in volume.

I've attached two GEQ presets to this post. One with AutoEQ's conservative equalization ("Equalization"), and one with the full EQ ("Error Smoothed"). They both use the AutoEQ target, so slightly less bass than vanilla Harman.

Comparison:
LCD3 GEQs.png
 

Attachments

  • LCD3 GEQs.zip
    7.7 KB · Views: 27
OP
concorde1

concorde1

Active Member
Joined
Mar 15, 2021
Messages
261
Likes
167
Location
New Zealand
Thanks, very generous of you
 

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,370
Likes
918
concorde1,

Glad to see that you're gettin some good help here. The official Equalizer APO website doesn't contain alot of info on using either the included Configuration Editor, or graphic EQ filters. It is a somewhat clunky, but also powerful program though (once you know how to use it). So I posted what I've been able to figure out on it here, in case it may be of help to anyone else.

Alot of people seem to like using the Peace UI, and/or Room EQ Wizard with Equalizer APO though. Neither of which I have tried yet.
 
OP
concorde1

concorde1

Active Member
Joined
Mar 15, 2021
Messages
261
Likes
167
Location
New Zealand
Thanks. I've used the PEACE UI for nearly a year but not sure if you can do CSV import with it. So I'm using vanilla for the sake of just this GEQ and PEACE for my other headphones.
 
  • Like
Reactions: ADU

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,370
Likes
918
Thanks. I've used the PEACE UI for nearly a year but not sure if you can do CSV import with it. So I'm using vanilla for the sake of just this GEQ and PEACE for my other headphones.

Very good.

The Harman target is only a rough approximation of a neutral in-ear response btw. So you should not really try to rejigger all the little peaks and valleys in your headphone's response curve to precisely match the target. This is particularly true in the treble, where some peaks and valleys are normal for most headphones.

Use the Harman target simply as a general starting place for the overall shape of your headphone's response. And then you can do some additional adjustment to fine-tune that response more to you own liking by ear, with another GEQ, or PEQ filters if you want.

The GEQ filters in the EAPO Configuration Editor are stackable, btw. So you could do this by simply adding another variable GEQ to the stack, with just a few points of control, maybe something like this...

index.php


GraphicEQ: 20 0; 63 0; 200 0; 630 0; 2000 0; 6300 0; 20000 0

...And then tweak the 7 different points up or down until they sound better. I do this by selecting them with a dragbox. And then inching them higher or lower with the up/down arrow keys on my keyboard, which moves them in 1 dB steps.

You can see the combined result of all the filters in your stack in the Analysis Panel at the bottom. (And that is then your final "EQ Curve" for the headphones.) PEQ and GEQ filters can even be combined in the same stack, if you so choose.

The image above is an example of one of my correction curves for a different headphone btw. So the shape of the curve would not be applicable to your particular headphones.
 
Last edited:

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,370
Likes
918
I've used the PEACE UI for nearly a year but not sure if you can do CSV import with it.

I'm sorry I don't know the answer to that btw.

Another way to import and export GEQ curves is using the textual version of the filter, like the "GraphicEQ" example in my post above.
 

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,370
Likes
918
Unit to unit variation is only one of several possible explanations for why your headphones might not all sound the same with the AutoEQ corrections applied, btw.

There are other ways that Harman could have developed their target headphone response curve (like using DF compensation), which could potentially have made it somewhat easier to translate to other measurement systems than the ones they were using for their own in-house research. But unfortunately they did not do that, and chose to use the raw in-ear measurements from their partially customized measurement rigs instead. Which made it more difficult to reproduce their in-house results with other systems and measurement data.

Jaakko tried to do his best to find some ways around those problems in his AutoEQ presets. But they are far from being a perfect solution to this.
 
Last edited:
OP
concorde1

concorde1

Active Member
Joined
Mar 15, 2021
Messages
261
Likes
167
Location
New Zealand
Hmm I think I can hear distortion using the GEQ - but the custom parametric EQ will be absolutely fine.
 

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,370
Likes
918
Hmm I think I can hear distortion using the GEQ - but the custom parametric EQ will be absolutely fine.

Using a digital EQ like Equalizer APO requires some resampling of the audio data, which can potentially produce some artifacts or aliasing errors in the sound file, if the bit depth on your audio device is too low. Or the sample rate of your content does not match your audio device.

If your audio device has a 16-bit depth, for example, then the artifacts caused by resampling can be mitigated by increasing that to a higher depth, such as 24 or 32 bits.

Aliasing errors can also potentially be introduced by using a different sample rate on your audio device than the sample rate on your audio content (48 kHz vs. 44.1 kHz, for example). Ideally, you want the sample rates to be the same on both, so no rate conversion is necessary. If this is not possible, or the sample rate of the content is simply unknown, then using a higher sample rate on your audio device, such as 96 kHz, may be the best solution for hiding any potential aliasing errors caused by the rate conversion.

Very high sample rates can begin to introduce some latency into the audio though. And since most audio content does not contain information above about 20-24 kHz in frequency, there isn't much practical benefit to using sample rates significantly higher than 2x this frequency (aka the Nyquist frequency).
 
Last edited:

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,370
Likes
918
Hmm I think I can hear distortion using the GEQ - but the custom parametric EQ will be absolutely fine.

If your audio content exceeds unity or 0 dBFS in the louder musical passages, this can also result in clipping or distortion. This can happen if you do not use an appropriate negative Preamp filter setting with your GEQ or PEQ filters, to keep the output levels of your device below 0 dBFS.

You can see if your final EQ curve is potentially clipping the audio content above 0 dBFS at any frequency by looking at the Peak Gain value in the EAPO Configuration Editor's Analysis Panel. A positive value there (such as +2.0 or +3.0) shown in red would indicate that some clipping of the audio content is possible. The frequencies which are potentially being clipped will also show up as a red area in your correction curve on the Analysis Panel graph.

3806015


If this is the case, then a negative Preamp value can be used to reduce the Peak Gain value back down slightly below 0 dBFS. A Peak Gain that is in the negative range (below unity or 0 dBFS), should not clip your audio content at any frequency.

You don't want the Preamp and Peak Gain set too low though into the negative range, because that will also reduce the volume of the content delivered to your headphones. (This is explained in a bit more detail here fwiw.)

If the distortion that you think you may be hearing is related to any of the above factors (ie aliasing or artifacts caused by the resampling/conversion of the bit depth or sample rate, or clipping), hopefully some of these tips will help to correct the issue.
 
Last edited:
OP
concorde1

concorde1

Active Member
Joined
Mar 15, 2021
Messages
261
Likes
167
Location
New Zealand
I think I was just imagining it, the GEQ is fine.
 
  • Like
Reactions: ADU
OP
concorde1

concorde1

Active Member
Joined
Mar 15, 2021
Messages
261
Likes
167
Location
New Zealand
After some testing, at this point in time:

Audeze LCD-3 I use the Headphone_com EQ

Klipsch HP-3 I use the ReferenceAA EQ

Fostex TH900mk2 I use the Oratory EQ

--

So @staticV3 's recommendation of trying all the options out there is spot on I think.
 

DesEsseintes

Member
Joined
Mar 13, 2022
Messages
13
Likes
9
Pretty consistent for me that none of my headphones/earphones match published curves well enough to just take an EQ based off some third-party measurement (maybe the ER2SE are an exception). In some cases the difference is massive (eg Aurvana Live). Some of this can be attributed to HRTF/ear canal resonance, some to unit variation. What has been more successful for me is to use some measurement or autoEQ as a starting point, and then to start using sine sweeps to identify residual problems.
 
Top Bottom