• 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!

Creating your own semi-calibrated Harman OE Curve at home

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,240
Likes
1,427
Location
California
There are lots of sites that provide headphone measurements. However, they all use measurement equipment and protocols ("rigs") that differ from those that were used by Harman to derive the Harman target curve. Sean Olive pointed this out in his twitter feed, linked on ASR
1637004242014.png

This means that the Harman target curve, as published by Harman, cannot be simply be plotted on headphone FR charts that were produced using a different measurement rig - yet this is exactly what sites are doing. The reason is that the Harman target curve would be shaped differently if plotted correctly. Perhaps this explains why following published EQ recommendations based on this invalid comparison end up making the headphones sound worse (or only marginally better).

Nevertheless, there is a strong desire to be able to compare measured headphones to the Harman target curve. One idea is to compare measurements made on the test rig vs Harman-provided measurements. Unfortunately, the conversion factor isn't simple, according to experts on this forum. Another way would be to re-measure a series of headphones on the new test rig to create a rig-specific Harman target. Harman has apparently done this, but they're not sharing the new curve (for obvious reasons).

But what if we could do what Harman is doing at home? All we would need to do is take headphones that have known Harman measurements and measure them on our own test rigs. And then to make the target curve customized for our own ears/headphones, we could use our own ears. Could this work?

To test this possibility, I did the following:
1) Utilized REW to measure a set of AKG 701 and JBL 701BT headphones, both of which have Harman-published measurements that sort of match the Harman-published target curve. Each headphone was measured 5 times, strange curves discarded, and averaged in REW. I used in-ear binaural headphones that fit at the opening of my ear canal.
1637006495533.png

The product is the Sound Professionals SP-TFB-2 (no affiliation), photo from their website.
2) I then took the Harman-published measurements of these two headphones (such as the one above), as measured on the Harman rig, and generated a CSV file to "correct" the headphone response to the Harman target.
3) Using REW's A-times-B function, I then re-created the Harman target on my own test rig. In theory, this curve could be used to make rough corrections against with any other headphone I measure, with some limitations.

Results:
Surprisingly, using this method, the two extrapolated Harman targets agreed:
50Hz-200Hz: -2dB to 1.5dB <-- and likely seal-related and something easily adjusted to taste anyway
200Hz-600Hz: -1.5dB to 0dB
600Hz-2khz: 0dB to 3dB
>2khz: pretty bad, but people usually don't equalize a whole lot above this range anyway other than perhaps applying a treble tone control or EQ'ing out a peak
Blue is based on the 710BT, and green is based on the 701.
1637005512159.png


Ideally, I would have more Harman-measured headphones that I could do this with.
However, based on this quick test, it seems like between 200Hz and 2kHz, the likely range of interest of EQ anyway, this could be a way to correct headphones to closer match the Harman curve with more customization.

Sharing this in case there's interest.
 

MayaTlab

Addicted to Fun and Learning
Joined
Aug 15, 2020
Messages
501
Likes
707
I think that it's interesting to try to use Harman's own data (with the custom pinna) instead of the more commonly used GRAS setups as a reference. It would be nice to know what average difference was found between them but as you said only Harman has that data.

1) Utilized REW to measure a set of AKG 701 and JBL 701BT headphones, both of which have Harman-published measurements that sort of match the Harman-published target curve. Each headphone was measured 5 times, strange curves discarded, and averaged in REW. I used in-ear binaural headphones that fit at the opening of my ear canal.
View attachment 165669
The product is the Sound Professionals SP-TFB-2 (no affiliation), photo from their website.

One caveat if I may of using such microphones is that, if only speaking in relative terms (ie "at 3451Hz headphones A are 1,37dB louder than headphones B), then errors may increasingly occur as frequency rises.

I've repeatedly compared the relative differences between four of my larger, open over ears over the course of several weeks with these five different microphones setups :
Screenshot 2021-11-15 at 21.16.09.png

Using a pair of HD650 + Dekoni Elite Velour pads as the reference and plotting the difference between them and the other three headphones for each microphone setup, I get something like this :
Screenshot 2021-11-15 at 21.15.32.png

The red arrows correspond to the traces for the SP mics (whether with blocked or open ear canals).
As you can see they start to deviate from either the DIY probe or the blocked ear canal entrance mics quite early (5kHz or so).

As a general rule, at least for larger, open over-ears (I haven't repeated that enough times with smaller closed backs for now to feel confident about it), I'd only feel confident with these mics' relative results up to 4kHz. Blocked ear canal entrance mics (provided they're at least flush with the ear canal entrance and don't protrude), regardless of type, tend to agree with each others and with the probes up to 7kHz or so (and then all hell breaks loose above), but may introduce moderate errors in the ear canal gain region at the least (2-4kHz).

In absolute terms (ie "Headphones A is 3,47dB louder at 5439Hz than at 1247Hz") I would't trust any of these mics, although the probe mic's results aren't too bad that they are nonsensical.

Results:
Surprisingly, using this method, the two extrapolated Harman targets agreed:
50Hz-200Hz: -2dB to 1.5dB <-- and likely seal-related and something easily adjusted to taste anyway
200Hz-600Hz: -1.5dB to 0dB
600Hz-2khz: 0dB to 3dB
>2khz: pretty bad, but people usually don't equalize a whole lot above this range anyway other than perhaps applying a treble tone control or EQ'ing out a peak
Blue is based on the 710BT, and green is based on the 701.

I would encourage people to spend a bit more time EQing stuff above 2kHz :D. Years ago I used low pass filters at various frequencies to understand how important higher frequencies were to my subjective impressions and it was, to me, an eye-opener.
 
OP
preload

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,240
Likes
1,427
Location
California
I think that it's interesting to try to use Harman's own data (with the custom pinna) instead of the more commonly used GRAS setups as a reference. It would be nice to know what average difference was found between them but as you said only Harman has that data.



One caveat if I may of using such microphones is that, if only speaking in relative terms (ie "at 3451Hz headphones A are 1,37dB louder than headphones B), then errors may increasingly occur as frequency rises.

I've repeatedly compared the relative differences between four of my larger, open over ears over the course of several weeks with these five different microphones setups :
View attachment 165791
Using a pair of HD650 + Dekoni Elite Velour pads as the reference and plotting the difference between them and the other three headphones for each microphone setup, I get something like this :
View attachment 165792
The red arrows correspond to the traces for the SP mics (whether with blocked or open ear canals).
As you can see they start to deviate from either the DIY probe or the blocked ear canal entrance mics quite early (5kHz or so).

As a general rule, at least for larger, open over-ears (I haven't repeated that enough times with smaller closed backs for now to feel confident about it), I'd only feel confident with these mics' relative results up to 4kHz. Blocked ear canal entrance mics (provided they're at least flush with the ear canal entrance and don't protrude), regardless of type, tend to agree with each others and with the probes up to 7kHz or so (and then all hell breaks loose above), but may introduce moderate errors in the ear canal gain region at the least (2-4kHz).

In absolute terms (ie "Headphones A is 3,47dB louder at 5439Hz than at 1247Hz") I would't trust any of these mics, although the probe mic's results aren't too bad that they are nonsensical.



I would encourage people to spend a bit more time EQing stuff above 2kHz :D. Years ago I used low pass filters at various frequencies to understand how important higher frequencies were to my subjective impressions and it was, to me, an eye-opener.
This is really interesting data, thanks for sharing it, @MayaTlab. So in other words, when the pinna is constant, changing the in-ear microphone doesn't really result in much measurement variation below 4k or so.
 

MayaTlab

Addicted to Fun and Learning
Joined
Aug 15, 2020
Messages
501
Likes
707
So in other words, when the pinna is constant,

Well that's something that I can't test as I'm not exactly into self-mutilation.

changing the in-ear microphone doesn't really result in much measurement variation below 4k or so.

For these four headphones, and the mics I tested, indeed the relative difference between headphones is nearly non-existent between 100Hz and 1.5-2kHz and moderate up to 4kHz. The absolute results, however, are completely different, particularly above 800Hz or so (obviously).

I am less certain that this holds quite as well for smaller, closed over-ears (and even more so on-ears).

The difference in the 2-4kHz band is moderate, but not necessarily inaudible.

Also, it's precisely because I'm repeatedly getting concordant results with different types of in situ microphones that I get increased confident in the measurements I'm performing in the range(s) where they agree, as it means that any concerns I would have logically need to apply to all microphones types. This was the crucial "bingo" moment for me.
So I don't think that this should be seen as an endorsement of the idea that any microphone is good enough to perform relative on-head measurements, but rather an encouragement to systematically perform multiple measurements with different types of mics and compare their relative results.
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
5,684
Likes
5,524
Location
Brussels, Belgium
I thought each measurement rig is custom calibrated to all the targets (Harman, Diffuse field .etc), if that's not the case then why the hell do they cost a small fortune?
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
37,479
Likes
157,091
Location
Seattle Area
>2khz: pretty bad, but people usually don't equalize a whole lot above this range anyway other than perhaps applying a treble tone control or EQ'ing out a peak
I do a ton of equalization above 2 kHz. Above around 7 to 8 kHz we can let go but not at 2 kHz.
 

TheTalbotHound

Active Member
Forum Donor
Joined
Nov 23, 2020
Messages
180
Likes
119
I think that it's interesting to try to use Harman's own data (with the custom pinna) instead of the more commonly used GRAS setups as a reference. It would be nice to know what average difference was found between them but as you said only Harman has that data.
I wish harman made their custom pinna more widely available. Seeing more measurements with it would be very interesting.
 
OP
preload

preload

Major Contributor
Forum Donor
Joined
May 19, 2020
Messages
1,240
Likes
1,427
Location
California
I thought each measurement rig is custom calibrated to all the targets (Harman, Diffuse field .etc), if that's not the case then why the hell do they cost a small fortune?
Negative, Captain.
 
Top Bottom