Headphones and the Harman target curve

[preload]

The scientific method is, by definition, objective.

The scientific method is a process to understand observable phenomena of the natural world and advance that understanding. No part of the scientific method requires the exclusive use of objective measurements, which is implied by the term "objectivist," that you used.

 

[preload]

The Harman Curve is, by definition, for headphones, in attempt to dimiish the issues caused by headphone listening being different to loudspeaker listening (I’m using necessary shorthand here).

Not really. The Harman curve for headphones, by definition, is the curve that tends to be preferred by listeners. It just so happens that this curve is different from the loudspeaker preference curve, and the explanation is partially explained by the translation between diffuse and free field responses.

Abbey Road mixes using B&W 800Ds, which are quite excellent.

I'll let the Harman mafia here respond to that one.

 

[Yorkshire Mouth]

The scientific method is a process to understand observable phenomena of the natural world and advance that understanding. No part of the scientific method requires the exclusive use of objective measurements, which is implied by the term "objectivist," that you used.

I’m going to stick my neck out and politely disagree.

The scientific method involves proposing a theory, performing an experiment to test the theory, then publishing the results so that others can perform the same experiment to see if the results are replicated.

By definition, knowing what we know about human nature and perception, you cannot hope to accurately reproduce test results for the reproduction of sound through electronic equipment by using subjective responses from humans.

I will concede, if the SOLE goal of the experiment is human preference, rather than accurate reproduction of a frequency response curve, then you can scientifically measure preference.

But you’ll forgive me if I’m not entirely convinced that this was Amir’s intent when he set up and named Audio Science Review.

In short, if measuring equipment says there’s a 5dB peak at 2khz, but a human listens and thinks s/he does/doesn’t like it, we’re not talking about the same thing.

 

[Chromatischism]

The scientific method involves proposing a theory, performing an experiment to test the theory, then publishing the results so that others can perform the same experiment to see if the results are replicated.

*Hypothesis

 

[Chromatischism]

I will concede, if the SOLE goal of the experiment is human preference, rather than accurate reproduction of a frequency response curve, then you can scientifically measure preference.

That was the goal, yes. To understand what humans prefer in order to design better sounding speakers.

But you’ll forgive me if I’m not entirely convinced that this was Amir’s intent when he set up and named Audio Science Review.

I shouldn't try to speak for him but he has stated that he started the website to try to summarize the research that was locked in papers behind paywalls and written in a manner not digestible for the public. It evolved from there into speaker testing - which still uses that research - electronics testing, and more.

 

[preload]

I will concede, if the SOLE goal of the experiment is human preference, rather than accurate reproduction of a frequency response curve, then you can scientifically measure preference.

Correct, now you're getting it.

But you’ll forgive me if I’m not entirely convinced that this was Amir’s intent when he set up and named Audio Science Review.

In short, if measuring equipment says there’s a 5dB peak at 2khz, but a human listens and thinks s/he does/doesn’t like it, we’re not talking about the same thing.

In your specific example, which correlates a measurement with a perceived judgment related to sound quality, I would say that this would be of interest here. Perhaps you have seen, there are tons of threads and replies that attempt to "translate" measurements to perception. With SOTA electronic equipment like modern DACs and SS amps, the focus has been on the notion that measurable differences don't necessarily translate into audible differences. Whereas for transducers, the interest has been to what degree were able to predict perceived sound characteristics and quality based on more difficult to interpret measurements.

 

[markanini]

I do find it interesting that, as soon as ‘the great unwashed’ become involved, they want more bass.

The fact that people not committed to accurate reproduction want to turn the bass up is nothing new.

Every DAC/amp test Amir does, he looks for linearity of frequency. If the line wasn’t flat, but increased bass, and Amir said “That’s a good thing because people prefer more bass”, what would our response be? I suspect the whole raison d'être of the group would be undermined.

How is following the herd unscientific with a DAC, but great with headphones?

This shouldn’t be about preference, just accuracy.

If you get accurate kit, then want to turn the bass up, you can.

Realize you are outside the realm of objectivity, by challenging the average. BTW if the majority of output devices follow a low bass target the program material will be mastered with higher bass levels.

 

[markanini]

Worth noting that Harman did prior work analyzing the characteristics of professional control rooms, which went on to define the characteristics of their reference room: http://seanolive.blogspot.com/2009/10/audios-circle-of-confusion.html

 

[Yorkshire Mouth]

There are some excellent contributions here. I'll offer an example of what I'm personally arguing.

Amir has noted the following. 'Sister' companies Denon and Marantz regularly put out similar product. Usually the Marantz version is preferred by audiophiles. But when Amir has measured these products, he's found that Marantz have just swapped out a few components which introduced distortion.

So, what does he say? What do we say?

Amir says, as far as I can tell, "You may prefer the Marantz/distorted version, and that's your choice. But you're scientifically wrong. The Marantz version is introducing more distortion to the Denon. If you 'prefer' that, it's up to you. But it's not 'better' than the Denon, it's scientifically 'worse'."

Now. Amir could set about 'scientifically' measuring which types/levels of distortion which people PREFER. But he doesn't. And we applaud him for that (don't we?).

In short, Amir doesn't appear to believe that a scientific evaluation of what people prefer is preferential to an scientific evaluation of faithful and accurate reproduction. He doesn't listen to the 'preference' of audiophiles towards the Marantz, and argue that Denon should add more distortion. That is, surely, the tail wagging the dog.

And the Harman curve appears to be the tail wagging the dog.

The debate may come when we ask 'a faithful and accurate reproduction of WHAT'?

For me, we go back to the mastering. Engineers and producers create masters in a mixing studio. Sometimes, sometimes not, the artist may be invited in to that process. But the job of our kit must surely be to accurately and faithfully reproduce what the mixer heard when mixing the final product.

Discussion about what our preferences are when listening to that are secondary.

Our kit should be capable, in the first instance, of faithfully and accurately reproducing what the mixing engineer heard when mixing the track. After that, if we want to apply a preference, that's our privilege.

Subsequently, any 'target curve' should be based on what the mixing engineer heard when creating the final mix, as opposed to any averaged out, personal 'preference'.

That's my opinion, and I appreciate that others may disagree.

 

[RHO]

But what did the mixing engineer hear? It's something different for every studio and every setting on the desk for every engineer working on a record.

 

[solderdude]

Subsequently, any 'target curve' should be based on what the mixing engineer heard when creating the final mix, as opposed to any averaged out, personal 'preference'.

In that case we would need to agree that the monitors are truly reference and the mixing engineer has good hearing and ethos.
A lot of them are just that, those that are in serious music recording business.
Producers for pop music have other fish to fry and do a good job at that at the expense of listening pleasure for the more 'audiophile oriented' listener.

So the next thing will be the fact that mixing is done at around 80-85dB SPL and most likely without Harman correction.

Here's the thing... listening with leisure and relaxation in mind is different from 'active' listening (I call this comfortable loud) which is closer to studio levels.

As most folks listen at lower levels one needs to 'compensate' for this due to equal loudness contour effects. This is where my personal 'compensation' is based on.
Not on turning up bass sliders to 'desired levels' for average users (for a company that makes a LOT of sense) but based on equal loudness contours compensation for 'normal' listening levels.
That happens to be close to Harman, just a tad lower and not only the impressive bass but gently sloping.

Amir uses Harman because it is kind of a standard that is closer to the actually useless compensations used for HATS measuring speakers.
It's an understandable choice... going for a standard.

 

[preload]

There are some excellent contributions here. I'll offer an example of what I'm personally arguing.

Amir has noted the following. 'Sister' companies Denon and Marantz regularly put out similar product. Usually the Marantz version is preferred by audiophiles. But when Amir has measured these products, he's found that Marantz have just swapped out a few components which introduced distortion.

So, what does he say? What do we say?

Amir says, as far as I can tell, "You may prefer the Marantz/distorted version, and that's your choice. But you're scientifically wrong. The Marantz version is introducing more distortion to the Denon. If you 'prefer' that, it's up to you. But it's not 'better' than the Denon, it's scientifically 'worse'."

Now. Amir could set about 'scientifically' measuring which types/levels of distortion which people PREFER. But he doesn't. And we applaud him for that (don't we?).

In short, Amir doesn't appear to believe that a scientific evaluation of what people prefer is preferential to an scientific evaluation of faithful and accurate reproduction. He doesn't listen to the 'preference' of audiophiles towards the Marantz, and argue that Denon should add more distortion. That is, surely, the tail wagging the dog.

There reason you appear to have set up a false paradox is that the term "preference" is being used to mean completely different things. Let's clean that up. Harman papers are using the term "preference score" to refer to the numerically reported subjective "score" given by listeners to a loudspeakers under defined, BLINDED conditions with controlled levels, a specific acoustic environment, defined music tracks, and scoring instructions. The use of the Harman preference score should be considered a validated instrument to quantify perceived SQ differences in loudspeakers and headphones specifically. WHEREAS, if you're using the term "preference" colloquially, to essentially include a youtuber telling you on his vlog that he prefers Speaker A over Speaker B while he was rocking out last night to Mariah Carey, then that is completely different. The Harman preference score is a scientific tool with many biases removed or controlled for. The latter is a subjective, uncontrolled, opinion that may or may or not correlate with blinded listening preference.

The debate may come when we ask 'a faithful and accurate reproduction of WHAT'?

For me, we go back to the mastering. Engineers and producers create masters in a mixing studio. Sometimes, sometimes not, the artist may be invited in to that process. But the job of our kit must surely be to accurately and faithfully reproduce what the mixer heard when mixing the final product.

Discussion about what our preferences are when listening to that are secondary.

Our kit should be capable, in the first instance, of faithfully and accurately reproducing what the mixing engineer heard when mixing the track. After that, if we want to apply a preference, that's our privilege.

Subsequently, any 'target curve' should be based on what the mixing engineer heard when creating the final mix, as opposed to any averaged out, personal 'preference'.

That's my opinion, and I appreciate that others may disagree.

I hear you and I would say that it's really not quite as simple as it may seem at first glance. Mixing decisions aren't necessarily made using the same set of monitor speakers/headphones across studios. So your "mixing engineer" is going to have adjustments made based on that studio's acoustic idiosyncrasies. So, unless you happen to know what monitors were used for the track you're listening to, and you can change up your room's loudspeakers to match each track, and your room acoustics are similar to those in that studio, you're more or less at the mercy of how your stereo reproduces the mixes produced by various studios.

And to make matters worse, loudspeakers need to be designed for reproduction in a variety of room types, with various amounts of early reflections (similarly, headphones need to account for various head sizes, ear shapes, etc.). As a result, because of all of this variation, it's going to be pretty hard to say that your goal is to match "what the mixing engineer heard." And how could you even measure that and create loudspeaker engineering design goals around that?

 

[Yorkshire Mouth]

Yes, mixing rooms will sound different, and studio monitors will sound different. But similarly, with ‘great speakers in great rooms’, great speakers will sound different and great rooms will sound different.

 

[Yorkshire Mouth]

Preload, I’ve read your comments on ‘preference’, and I don’t see your definition as being something completely different.

The point remains, however you dress it up. Harman start with one curve (measurement in the room), then ask people if they like it. For me (and I understand not everyone agrees), I’d either just stick with the original curve, or if I asked people to comment at all, it’d be to check if the curve was accurate to the room*. And yes, I’d maybe consider the room.

But what might get lost is that I’m agreeing with the need for a curve, and that Harman’s approach has got a lot right.

* Actually, fair enough, ask them. Then create a number of curves. One ‘accurate’ for people wanting accuracy (I presume us), then maybe one ‘preferred’ for marketing purposes. Or 2 or 3; the results suggest people can be split into 3 categories.

 

[BrEpBrEpBrEpBrEp]

Preload, I’ve read your comments on ‘preference’, and I don’t see your definition as being something completely different.

The point remains, however you dress it up. Harman start with one curve (measurement in the room), then ask people if they like it. For me (and I understand not everyone agrees), I’d either just stick with the original curve, or if I asked people to comment at all, it’d be to check if the curve was accurate to the room*. And yes, I’d maybe consider the room.

But what might get lost is that I’m agreeing with the need for a curve, and that Harman’s approach has got a lot right.

* Actually, fair enough, ask them. Then create a number of curves. One ‘accurate’ for people wanting accuracy (I presume us), then maybe one ‘preferred’ for marketing purposes. Or 2 or 3; the results suggest people can be split into 3 categories.

IIRC the Harman curve doesn't significantly deviate from the in-room measurements, even after the preference adjustments. See Dr. Olive RE: this issue.

The bass is the significant deviation, and it's largely preferential. You'll never have the same bass effects with headphones and speakers anyways, just due to room geometry and full-body effects - therefore it's a bit of a moot point. The goal with Harman was for the bass to be perceived as having the same tone/timbre. Oratory1990 has a good post on this.

 

[Yorkshire Mouth]

IIRC the Harman curve doesn't significantly deviate from the in-room measurements, even after the preference adjustments. See Dr. Olive RE: this issue:

https://twitter.com/i/web/status/1212480010503614464

The bass is the significant deviation, and it's largely preferential. You'll never have the same bass effects with headphones and speakers anyways, just due to room geometry and full-body effects - therefore it's a bit of a moot point. The goal with Harman was for the bass to be perceived as having the same tone/timbre. Oratory1990 has a good post on this:

https://www.reddit.com/r/oratory1990/comments/eik71c/_/fcugn75

Thanks for that, and that’s astonishingly revelatory.

Critics of Harman generally think there’s too much bass. These comments suggest the final curve was the original measured + bass.

I’d still love to see the original measured curve in full.

 

[DeLub]

Is this the curve you’re looking for:

Loudspeaker in room flat

 

[Yorkshire Mouth]

Is this the curve you’re looking for:

Loudspeaker in room flat

Many thanks. It’ll be very similar. I believe Harman went for a ‘great’ room (a suitable and well-treated lounge) rather than a completely flat room; I’m not sure how much difference there’d be.

 

[Yorkshire Mouth]

Here it is folks. The original PowerPoint presentation:

http://www.juloaudio.sk/Umiestnenie_reprosustav/History of Harman Target Curve.pdf

I believe the green, dotted line on Slide 10 is the ‘pre-preference’ curve.

Comments there reflect those linked to earlier, that Harman is basically:

1 - From 250hz down to 20hz Harman rises by around 6dB, and that should be flat, or at most a more modest +2dB.

2 - Harman's 'peak' at 3-4khz is around 2-2.5dB too low, and remains around 2-2.5dB low for the remainder.

What’s interesting is when we now compare something like that curve with the AKG K371 and something as diverse as the Sennheiser HD600/650 and the Beyerdynamic DT990s.

Leaving aside the heightened treble of the latter for a second. Compared to pre-preference Harman, the AKG is spot on, and the other two bass-lite.

Compared to this pre-preference curve, the AKGs roughly overshoot on bass by around the same amount as the other two come in under (more or less).

This is important for EQ, as the less you have to boost bass, the less distortion you’ll introduce.

Interesting that, even in the 'reference room', which we're sort of using as shorthand for for what it should sound like, listeners preferred more bass.

Also interesting (though we knew this before), older, more experienced listeners didn't want as much extra bass, and older listeners wanted a little more treble.

On their own, those factors are all interesting. Added together, we suddenly get a much broader picture of what's going on.

What I take from all this is that (a) Harman has too much bass, though how much too much is open to debate, and (b) if you're older (I'm 56) you might need a little more treble than Harman; this isn't a 'preference', it's down to natural degradation of high end frequency hearing over time.

 

[Jimbob54]

Here it is folks. The original PowerPoint presentation:

http://www.juloaudio.sk/Umiestnenie_reprosustav/History of Harman Target Curve.pdf

I believe the green, dotted line on Slide 10 is the ‘pre-preference’ curve.

Comments there reflect those linked to earlier, that Harman is basically:

1 - From 250hz down to 20hz Harman rises by around 6dB, and that should be flat, or at most a more modest +2dB.

2 - Harman's 'peak' at 3-4khz is around 2-2.5dB too low, and remains around 2-2.5dB low for the remainder.

What’s interesting is when we now compare something like that curve with the AKG K371 and something as diverse as the Sennheiser HD600/650 and the Beyerdynamic DT990s.

Leaving aside the heightened treble of the latter for a second. Compared to pre-preference Harman, the AKG is spot on, and the other two bass-lite.

Compared to this pre-preference curve, the AKGs roughly overshoot on bass by around the same amount as the other two come in under (more or less).

This is important for EQ, as the less you have to boost bass, the less distortion you’ll introduce.

Interesting that, even in the 'reference room', which we're sort of using as shorthand for for what it should sound like, listeners preferred more bass.

Also interesting (though we knew this before), older, more experienced listeners didn't want as much extra bass, and older listeners wanted a little more treble.

On their own, those factors are all interesting. Added together, we suddenly get a much broader picture of what's going on.

What I take from all this is that (a) Harman has too much bass, though how much too much is open to debate, and (b) if you're older (I'm 56) you might need a little more treble than Harman; this isn't a 'preference', it's down to natural degradation of high end frequency hearing over time.

But in practice, what does it really matter? Does anyone stick hard and fast to the EQ settings, for eg, Oratory1990 to Harman, or in fact plug those filters in first then tweak. I almost always lower the bass filters and often some of the others. How "accurate" the Harman target is only ever really becomes an issue if the only filters you can use are pre made convolved ones which cant be adjusted by the user.

But its interesting to refresh how we got to Harman. A more interesting one is how we get from the original HP curve to the one now.