The Etymotic Target (R.I.P. Harman)

[restorer-john]

Only purists hated tone controls and wanted defeat buttons so the market gave it to them.

Tone controls were originally removed or bypassed not for purists, but for the elimination of frequency respones variations, lowering of noise and distortion and better channel balance in preamps and integrated amps. All those parameters were easily measureable and highly obvious. Manufacturers were striving to produce the best products for their spec sheets. The S/N differences between active tone controls at 0dB and bypassed is a number of dB. Scrap the entire active front end tone/buffer and go direct to a higher gain power stage and get even better numbers.

In the late 70s- early 80s, many audiophiles realized they rarely or never used their tone controls and the British cottage HiFi industry saw an opportunity to save a heap of money, drop the unwanted controls, proclaim them as polluting and not pure, whether they were bypassed or not. The Audiophools jumped on board and soon even the loudness contour was deemed evil by the British HiFi press. Filters were work of the devil too. Bad luck if you played vinyl- the infrasonic filter was disappearing in the rush to rid gear of useful functionality.

By the early 1980s, most integrated amplifiers from the big manufacturers had direct/bypass buttons that basically put a volume pot in front of the power stage. S/N ratios were above 110dB with respect to 65W amps. Things were looking good, just as CD came along. It was so "pure" you needed nothing between your player and your power amp- even preamps were getting banished. Manufacturers put CD-direct inputs as additional inputs on power amps.

My preamplifiers are run in direct pretty much 99% of the time. But all my preamps have tone controls, filters, MM/MC stages etc. The presense of those useful functions in no way impedes the performance when in pure/bypass/direct.

 

[Sean Olive]

Tone controls were originally removed or bypassed not for purists, but for the elimination of frequency respones variations, lowering of noise and distortion and better channel balance in preamps and integrated amps. All those parameters were easily measureable and highly obvious. Manufacturers were striving to produce the best products for their spec sheets. The S/N differences between active tone controls at 0dB and bypassed is a number of dB. Scrap the entire active front end tone/buffer and go direct to a higher gain power stage and get even better numbers.

In the late 70s- early 80s, many audiophiles realized they rarely or never used their tone controls and the British cottage HiFi industry saw an opportunity to save a heap of money, drop the unwanted controls, proclaim them as polluting and not pure, whether they were bypassed or not. The Audiophools jumped on board and soon even the loudness contour was deemed evil by the British HiFi press. Filters were work of the devil too. Bad luck if you played vinyl- the infrasonic filter was disappearing in the rush to rid gear of useful functionality.

By the early 1980s, most integrated amplifiers from the big manufacturers had direct/bypass buttons that basically put a volume pot in front of the power stage. S/N ratios were above 110dB with respect to 65W amps. Things were looking good, just as CD came along. It was so "pure" you needed nothing between your player and your power amp- even preamps were getting banished. Manufacturers put CD-direct inputs as additional inputs on power amps.

My preamplifiers are run in direct pretty much 99% of the time. But all my preamps have tone controls, filters, MM/MC stages etc. The presense of those useful functions in no way impedes the performance when in pure/bypass/direct.

Thanks for that nice historical piece of information.. I only wish my $20k Mark Levinson No 519 had some !@#$ bass controls so I could adjust the bass of my loudspeakers... But I am told audiophiles wouldn't purchase such a feature because of "compromised sound". But they will apparently listen to a 10 dB resonance at 50 Hz caused by the acoustical interactions between the speaker and the room. Complete B.S. (Before Science).

 

[Inner Space]

But they will apparently listen to a 10 dB resonance at 50 Hz caused by the acoustical interactions between the speaker and the room.

I'm going to crowbar a question in here, if I may, because I have never figured it out - let's say we use a tone control (or DSP now) to pull down 50Hz, so the resonance drops in level, to what sounds and measures about in line ... what exactly are we now hearing at 50Hz? A mix of faint signal and tamed boom, which is not actually an acoustic amplification of the signal, except at the fundamental? Is that correct? Is there math that predicts how much energy input triggers the resonance?

 

[restorer-john]

I only wish my $20k Mark Levinson No 519 had some !@#$ bass controls so I could adjust the bass of my loudspeakers

Your ML is essentially a modern preamplifier. It brings together multiple sources, both analogue and digital/disc/network, switches between them, controls playback, manages the volume and drives a power amplifier. Looks like a very cool device.

For it to not have a global PEQ or some rudimentary control of tone seems ridiculous to me, but I'm not surprised. You could always slip in an old skool graphic or parametric eq between the line outs and your power amp. Or is that too "compromised"

 

[Sharur]

The Harman Target is based on a stereo pair of anechoically flat loudspeakers measured at the DFP in a semi-reflective room, which is neither free-field nor diffuse field. The DRP was measured in the stereo seat using a +- 30 degree spatial average.

The differences you see up to 10 kHz between ours and the others can be explained by those differences. If you look at the predicted-in room curve (PIR) of the loudspeaker which is a weighted sum of the direct sound, early-reflected and sound power it pretty well lines up with the Harman Target Curve. No one knows what the exact shape of the curve should be above 10 kHz because there are so many measurement errors in the IEC couplers, and the subjective and objective measurements themselves.

The B&K 5128 and the latest GRAS RA0045 (?) couplers have lower measurement errors above 10 kHz (the resonance at 13.5 kHz in the GRAS RA0045 has been damped making measurements more repeatable particularly distortion) and the 5128 is based on a large study that scanned human ear canals -- so there is promise that the measurement device better represents average humans.


For those reasons mentioned above, we don't even calculate predicted preference scores of headphones using data beyond 10 kHz. Add to the fact that most adults have greatly elevated hearing thresholds above 10 kHz, the task of defining what is the ideal HF target a fuzzy, grey area that is very dependent on the individual and the music signal. This is an area for more research.

I suppose my question should have been targeted towards the Harman Flat In-Room response which was based on anechoically flat speakers eq'd to be flat in a semi-reflective listening room? @JohnYang1997 brought this up a while back, but I've done some basic calculations to compare it to the average of diffuse field and free field at the ear canal gain frequency, 6 kHz, 8 kHz, and 10 kHz.

* Harman target pushes the ear canal gain frequency to 3.1 kHz (why?)
* Harman target is shy of the predicted value by 1.75 dB at ear canal resonance (because of the ear canal gain frequency shift?)
* Harman target exceeds the predicted value by 1.5 dB at 6 kHz
* Harman target exceeds the predicted value by 1.25 dB at 8 kHz
* Harman target is shy of the predicted value by 0.5 dB at 10 kHz

EDIT: I'm not sure what the +- 30 degree spatial average means. How was it decided which frequencies will exceed the predicted values and others be shy of it?

 

[JohnYang1997]

Sine tones are good for identifying buzzes and rattles... and perhaps peaks/dips within a local frequency region where loudness sensitivity is constant. But there are more efficient ways to determine if the headphone has a smooth response/accurate spectral balance.In order of efficiency:
1) Measure it
2) listen to broadband pink noise -- the most sensitive signal for detecting low/medium/high Q resonances.

If you are trained you can identify the frequencies at which the resonances/anti-resonances occur. The more it rings or whistles the higher the Q

Measurements work very well(for in ears) relatively well(for over ears) from 500Hz to 6khz. The rest is better with in the ear "measurements" using tone generator.
Music works for overall balance and mid range.
Trained listeners can identify the place of the resonance but not exactly. Especially not to the point you can point out what exact frequency, magnitude and Q value.

 

[Sean Olive]

I suppose my question should have been targeted towards the Harman Flat In-Room response which was based on anechoically flat speakers eq'd to be flat in a semi-reflective listening room? @JohnYang1997 brought this up a while back, but I've done some basic calculations to compare it to the average of diffuse field and free field at the ear canal gain frequency, 6 kHz, 8 kHz, and 10 kHz.

* Harman target pushes the ear canal gain frequency to 3.1 kHz (why?)
* Harman target is shy of the predicted value by 1.75 dB at ear canal resonance (because of the ear canal gain frequency shift?)
* Harman target exceeds the predicted value by 1.5 dB at 6 kHz
* Harman target exceeds the predicted value by 1.25 dB at 8 kHz
* Harman target is shy of the predicted value by 0.5 dB at 10 kHz

View attachment 153419
View attachment 153420
View attachment 153423

EDIT: I'm not sure what the +- 30 degree spatial average means. How was it decided which frequencies will exceed the predicted values and others be shy of it?

Someone has gone to a lot of trouble here to try and discredit the Harman Target Curve..I wonder if they happen to be in the headphone business selling a different headphone flavor?

There are 19+ papers available for reading in the AES E-library explaining the subjective and objective measurements behind the research. We included both free-field and diffuse field targets in our tests, and many others, and they did not score as well as the Harman Target. Other researchers have come to similar conclusions about the DF and FF as well, including Fraunhofer.

I would direct you and Mr. Yang to those papers. The basic answer to his question: "why deviate from DF and FF targets?" is as follows: because the research indicates our target sounds better to the majority of listeners tested. Better sound quality is the driving goal behind all our research and the best reason why.

 

[JohnYang1997]

didn’t suddenly change.

This is what we are talking about. Etymotic target vs original Harman Target in 2013 has very little difference to begin with >200hz. So it's already splitting hair or basically the bass is in the question.

For the IE target it’s higher

It may very well be, or it's precisely caused by mismatch in leakage in human ear vs measurements. There shouldn't be any difference if the seal is perfect although this doesn't say much for recommending a target in practical sense. ER4S mentioned numerous times has deeper bass, better bass than many good quality speaker systems. It just needed to be inserted properly. A couple of dB boosting is ok. But more than that actually would sound boomy/hammering the brain and uncomfortable.

 

[Sharur]

There are 19+ papers available for reading in the AES E-library explaining the subjective and objective measurements behind the research. We included both free-field and diffuse field targets in our tests, and many others, and they did not score as well as the Harman Target. Other researchers have come to similar conclusions about the DF and FF as well, including Fraunhofer.

I would direct you and Mr. Yang to those papers. The basic answer to his question: "why deviate from DF and FF targets?" is as follows: because the research indicates our target sounds better to the majority of listeners tested. Better sound quality is the driving goal behind all our research and the best reason why.

My comment was about the average of diffuse field and free field, not one or the other.

 

[JohnYang1997]

Someone has gone to a lot of trouble here to try and discredit the Harman Target Curve..I wonder if they happen to be in the headphone business selling a different headphone flavor?

There are 19+ papers available for reading in the AES E-library explaining the subjective and objective measurements behind the research. We included both free-field and diffuse field targets in our tests, and many others, and they did not score as well as the Harman Target. Other researchers have come to similar conclusions about the DF and FF as well, including Fraunhofer.

I would direct you and Mr. Yang to those papers. The basic answer to his question: "why deviate from DF and FF targets?" is as follows: because the research indicates our target sounds better to the majority of listeners tested. Better sound quality is the driving goal behind all our research and the best reason why.

The comparison is done to raw FF and DF instead of modified DF like etymotic.
Like you said before apart from major discrepancy in the bass <200hz. There's very small difference between etymotic target and Harman target.
The question was simply what's neutral instead of what's preferred.
Would you steer the people at studios to adopt preferred bass boosted targets if they are to use headphones?
And we should agree most people like a little bit of smily face type response(mild V shape). This is what sounds good.
Harman target is not too far off to neutral compared to the current state of performance in most headphones and earphone. In fact should be godly rather. This isn't the question. The question is actually what is the last 2dB of improvement what's the more preferred target than current one. The historical change/variants of Harman Targets expressed just that. It's not precise but a roughly there guide.
I joined AES in the year 2018(membership expired in 2019). I have and have read a few papers. In fact I looked for all headphone related papers. This is like a long time ago for me and the weird dynamics between different people is seriously not positioning me the right way.

 

[Sean Olive]

I'm going to crowbar a question in here, if I may, because I have never figured it out - let's say we use a tone control (or DSP now) to pull down 50Hz, so the resonance drops in level, to what sounds and measures about in line ... what exactly are we now hearing at 50Hz? A mix of faint signal and tamed boom, which is not actually an acoustic amplification of the signal, except at the fundamental? Is that correct? Is there math that predicts how much energy input triggers the resonance?

You are essentially putting less energy into that room mode. The mode still exists but you are not exciting it to the same degree.

 

[solderdude]

The latest GRAS RA0045 (?) couplers have lower measurement errors above 10 kHz (the resonance at 13.5 kHz in the GRAS RA0045 has been damped making measurements more repeatable particularly distortion) so there is promise that the measurement device better represents average humans.

For those reasons mentioned above, we don't even calculate predicted preference scores of headphones using data beyond 10 kHz. Add to that, the fact that most adults have greatly elevated hearing thresholds above 10 kHz makes the task of defining what is the ideal HF target a fuzzy, grey area that is very dependent on the individual and the music signal. This is topic that needs more research.

I have a question about this. I assume it is about a choice made based upon averaging of many measurements.
Having seen hundreds of measurements made with the GRAS RA0045 and then looking at the corrected plots one thing is painfully obvious.
All of them have a dip at 10kHz and start to deviate above 10kHz.

When the (extremely averaged ?) Harman target is applied all measured (and decently extended) headphones appear to have a dip at 10kHz where there isn't any. In fact many headphones are peaking there which helps when EQ'ed in that range.
Why isn't that familiar dip in 10kHz not reflected in the target for the RA0045 ?

Above 10kHz all headphones measured with RA0045 appear to have +10dB at 20kHz which none of them actually have. I assume this is because of averaging the mess above 10kHz but was wondering since there is more info available and based on that one could adjust the target for RA0045 to have a dip and change the average between 10kHz and 20khz would make the overall plots not more realistic ?

 

[Sean Olive]

The comparison is done to raw FF and DF instead of modified DF like etymotic.
Like you said before apart from major discrepancy in the bass <200hz. There's very small difference between etymotic target and Harman target.
The question was simply what's neutral instead of what's preferred.
Would you steer the people at studios to adopt preferred bass boosted targets if they are to use headphones?
And we should agree most people like a little bit of smily face type response(mild V shape). This is what sounds good.
Harman target is not too far off to neutral compared to the current state of performance in most headphones and earphone. In fact should be godly rather. This isn't the question. The question is actually what is the last 2dB of improvement what's the more preferred target than current one. The historical change/variants of Harman Targets expressed just that. It's not precise but a roughly there guide.
I joined AES in the year 2018(membership expired in 2019). I have and have read a few papers. In fact I looked for all headphone related papers. This is like a long time ago for me and the weird dynamics between different people is seriously not positioning me the right way.

What’s important is that the recording and consumer industries pick a relatively neutral common reference so that is he quality of recordings and their reproduction is more consistent. For many years the choice of reference monitors in the recording industry has dictated more by fashion than guided by science and common sense. . 3 inch Aurotones Yamaha NS10s and speakers that fall far below what’s considered neutral.
it turns out that one of the most popular headphones used in broadcast, television and recording for the past 10+ years is the $80 Sony MDR 7506 and while not perfect it comes very close to the Harman Target. So maybe the industry has already decided the Target is acceptable even if they weren’t aware of it

 

[Sean Olive]

I have a question about this. I assume it is about a choice made based upon averaging of many measurements.
Having seen hundreds of measurements made with the GRAS RA0045 and then looking at the corrected plots one thing is painfully obvious.
All of them have a dip at 10kHz and start to deviate above 10kHz.

When the (extremely averaged ?) Harman target is applied all measured (and decently extended) headphones appear to have a dip at 10kHz where there isn't any. In fact many headphones are peaking there which helps when EQ'ed in that range.
Why isn't that familiar dip in 10kHz not reflected in the target for the RA0045 ?

Above 10kHz all headphones measured with RA0045 appear to have +10dB at 20kHz which none of them actually have. I assume this is because of averaging the mess above 10kHz but was wondering since there is more info available and based on that one could adjust the target for RA0045 to have a dip and change the average between 10kHz and 20khz would make the overall plots not more realistic ?

There is a good reason for that. The Harman Target is based on the older GRAS coupler the predecessor to the newer RA0042. So it doesn’t have the damped resonance at 13.5 kHz. But more relevant to the 10 kHz dip our research used a modified version of the original IEC pinna and that predates the current anthropomorphic pinna. Todd Welti wrote a paper (https://www.aes.org/e-lib/browse.cfm?elib=17699 )on this modified pinna which was designed to better simulate leakage measured on humans.

The new anthropomorphic pinna produces a pretty consistent notch at 10 kHz which was not produced with Todd’s modified pinna. I will have to look at the data again because I don't recall a +10 dB gain at 20 kHz that you claim is not part of the headphone response..

I just competed analysis of 20 headphones measured on 6 different test fixtures including the original one used for our research. This tells us how the test fixtures vary relative to the original fixture so we can better interpret the measurements made on different test fixtures and derive Harman Targets for each fixture based on those differences.

I will be talking about the B&K 5128 in this regard at the upcoming HBK online seminar this month.
https://hbkworld.com/tapc-conference-program/

 

[Thomas_A]

The Harman Target is based on a stereo pair of anechoically flat loudspeakers measured at the DFP in a semi-reflective room, which is neither free-field nor diffuse field. The DRP was measured in the stereo seat using a +- 30 degree spatial average.

The differences you see up to 10 kHz between ours and the others can be explained by those differences. If you look at the predicted-in room curve (PIR) of the loudspeaker which is a weighted sum of the direct sound, early-reflected and sound power it pretty well lines up with the Harman Target Curve. No one knows what the exact shape of the curve should be above 10 kHz because there are so many measurement errors in the IEC couplers, and the subjective and objective measurements themselves.

The B&K 5128 and the latest GRAS RA0045 (?) couplers have lower measurement errors above 10 kHz (the resonance at 13.5 kHz in the GRAS RA0045 has been damped making measurements more repeatable particularly distortion) and the 5128 is based on a large study that scanned human ear canals -- so there is promise that the measurement device better represents average humans.


For those reasons mentioned above, we don't even calculate predicted preference scores of headphones using data beyond 10 kHz. Add to that, the fact that most adults have greatly elevated hearing thresholds above 10 kHz makes the task of defining what is the ideal HF target a fuzzy, grey area that is very dependent on the individual and the music signal. This is topic that needs more research.

Is there some influence of stereo 30° vs. a true frontal mono source on the Harman Target curve?

 

[Sean Olive]

Is there some influence of stereo 30° vs. a true frontal mono source on the Harman Target curve?

Yes of course there will be a difference compared to a single measurement at 0 degrees.. But people move their heads when they listen and the spatial averaging will also smooth out any interference effects from room reflections,etc ..

 

[JohnYang1997]

What’s important is that the recording and consumer industries pick a relatively neutral common reference so that is he quality of recordings and their reproduction is more consistent. For many years the choice of reference monitors in the recording industry has dictated more by fashion than guided by science and common sense. . 3 inch Aurotones Yamaha NS10s and speakers that fall far below what’s considered neutral.
it turns out that one of the most popular headphones used in broadcast, television and recording for the past 10+ years is the $80 Sony MDR 7506 and while not perfect it comes very close to the Harman Target. So maybe the industry has already decided the Target is acceptable even if they weren’t aware of it

Oh come on.
Large studios have dedicated room for mastering. Those are with large most of the time in the wall real deal monitors. Those are seriously no joke. The use of NS10 or auratone is to pick out nasty things to eq out. And ns10 translates well to bigger systems. And the key is that there really were bigger systems to make final decisions.
When we are splitting these small variance in targets. 7506 is closer to raw not modified DF target than anything else. And the bass was said to be muddy all the time. This is not a good candidate to represent Harman target in any way. Apart from that any remotely ok actual studio monitors from focal, genelec, adam, yamaha(not ns10) are miles better than 7506. So this argument really is pretty bad imo.

 

[restorer-john]

And ns10 translates well to bigger systems. And the key is that there really were bigger systems to make final decisions.

The NS-10M doesn't remotely translate to bigger systems. It never did. I sold them all including the NS-1000M, the NS-2000, the later X series (up to the NS-1000X) and even the rare NS-10000 centennary series). None of them were remotely comparable on the salesfloor to much of the bass-reflex two and three way designs coming out of the US and Europe at the time. Yes, they were brutally incisive, but lacked everything that made you want to listen to, and move with the music. They simply didn't sell well and were killed by Yamaha by the early 90s.

I have several "classic" Yamaha loudspeaker products, and my father still has a 10 year (badged) anniversary pair of NS-1000M Japanese home market speakers on ultra rare SPS-500 stands at his place which get sporadic use. Their midrange is absolutely flawless. Everything else is a compromise.

The NS-10M was an aberration. A cute little Yamaha, back when the NS-1000 was causing waves in the studio space. they were the equivalent of the JBL powered 305 series when powered/active speakers were non-existant.

Anyway, the brochure for the NS-1000X is uploaded here, by me, to HiFi Engine if you want to see it:

https://www.hifiengine.com/manual_library/yamaha/ns-1000x.shtml

From one of my brochures:

 

[JohnYang1997]

The NS-10M doesn't remotely translate to bigger systems. It never did. I sold them all including the NS-1000M, the NS-2000, the later X series (up to the NS-1000X) and even the rare NS-10000 centennary series). None of them were remotely comparable on the salesfloor to much of the bass-reflex two and three way designs coming out of the US and Europe at the time. Yes, they were brutally incisive, but lacked everything that made you want to listen to, and move with the music. They simply didn't sell well and were killed by Yamaha by the early 90s.

I have several "classic" Yamaha loudspeaker products, and my father still has a 10 year (badged) anniversary pair of NS-1000M Japanese home market speakers on ultra rare SPS-500 stands at his place which get sporadic use. Their midrange is absolutely flawless. Everything else is a compromise.

The NS-10M was an aberration. A cute little Yamaha, back when the NS-1000 was causing waves in the studio space. they were the equivalent of the JBL powered 305 series when powered/active speakers were non-existant.

Anyway, the brochure for the NS-1000X is uploaded here, by me, to HiFi Engine if you want to see it:

https://www.hifiengine.com/manual_library/yamaha/ns-1000x.shtml

From one of my brochures:
View attachment 153504

There are some weird/good characteristics of NS10 that made them that way. It never was claimed to sound like bigger systems. But when one mix on them and of course knowing how they sound one can make reliable decisions to sound good on bigger systems. If it's not the case no one would be using them.

 

[Sharur]

When we are splitting these small variance in targets. 7506 is closer to raw not modified DF target than anything else. And the bass was said to be muddy all the time.

Perhaps this has to do with distortion as well?