[SilentNote]

From my understanding, the diffuse field curve is the most neutral / flat. As in if you re-record what is heard at your eardrum, and played it back through the earpiece, and repeat it several times, there should not be too much difference.

Flat signal in, flat signal out.

The harman listener target curve is developed for a different reason - the intended reproduction of music. When a recording studio records music, they “target” the reproduction through a quality hi-fi system in a listening room. Note that this room is not a recording studio and adds “color” to the sound through its walls and ceiling.

The Harman listener target curve attempts to recreate this home listening condition in the IEM. Thus fulfilling the recording studio’s intended playback scenario - making it sound “best”.

Why isn’t flat signal the “best”? Because the recording studio did not record the music intending them to be listened in recording studios, but home theater systems.

Some interesting personal observation: music such as theatrical plays, recorded in the audience perspective (binaural), sound incredibly transparent (even soundstage) with flat iems like the ER4SR. However label recordings always sound better with a colored IEM.

The conclusion is that both target curves are excellent at their intended purpose. For monitoring of microphone recordings? Diffuse field curve flat. For simulating reproduction of home hi-fi? Harman listener target curve.

On other question about human ear sensitivity, our ears most likely evolved to listen to rustling leaves and crackling branches of predators. To miss that is a death sentence. These are both sounds higher than the typical vocal range.

As IEMs bypass the outer half of our ear’s natural amplification system, a good IEM needs to simulate that natural amplification. This tends to be a peak at 2.6-2.8kHz and a roll off at higher frequencies. i.e. the diffuse field curve.

Thus the next time you are auditioning a colored IEM, remember that you are really buying a listening room and a pinna.

 

[bigshot]

I'm confused by this. Maybe I'm misunderstanding.
To my understanding, the Harmon Curve is for headphones, to compensate for how flat speakers sound in an anechoic chamber, as compared to the same speakers in a non-anechoic room.

No, it's the opposite. Flat speakers in an anechoic chamber sound flat. The room alters the response and mixing stages aren't anechoic chambers. If you look at a measured response of a good listening room with speakers from the listening position, it looks very much like the Harman Curve.

I'm still finding it counter-intuitive that the target curve calls for such a big boost, centered at 3khz. to 4khz., since that's where human hearing is the most sensitive.

When you EQ and try to establish a balance for the response, you can see what is happening. The idea of a listening room isn't to eliminate all of the reflections. It's to eliminate bad reflections. The speakers are only half of the sound of a speaker system. The room is just as important as the speakers. The Harman Curve is an attempt to synthesize the effect of space on sound. That physical space is what creates true soundstage. You can match the affect on the response, but you can't recreate true soundstage with headphones... at least not yet.

 

[KeithEmo]

I think you've got pretty much the right idea.

For example, an IEM bypasses the outside of your ear, and the room itself, so we would expect an IEM to sound most natural with a binaural recording.
(A binaural recording is made using a microphone installed in place of an "artificial eardrum"; presumably this would be the exact signal we would want an IEM to deliver directly to our eardrum.)

However, thie issue of "diffuse field curve" and "anechoic curve" is far more complex.
(With speakers, we're attempting to duplicate the signal part-way on its way to our ear.)

In an anechoic chamber, there are no reflections, so all you hear is sound travelling directly from the speaker to your ear.

However, in a room, you are hearing two distinct sorts of sound.
The direct sound from the speaker arrives at your ear first (because it travels the shortest path).
Then, after various amounts of delay, sound arrives at your ear that has been reflected from one or more room surfaces.
This sound that arrives later is not only delayed - but its frequency response is altered by the acoustic properties of the room (different surfaces absorb more or less of different frequencies).
So, in a room, what you hear is a mix of the direct sound, and the sum of the reflected sound, which may itself vary over time.
To make this even more complex, our brains interpret "duplicate sounds arriving after a delay" differently depending on the length of the delay..
To slightly oversimplify things, we tend to hear reflected sounds that arrive after a long delay as "an echo", but sounds that arrive soon after the original simply make it seem louder.
(So, for example, if a room is very reflective at high frequencies, reflections from near walls make things sound "brighter", but reflections from far walls sound like distinct echoes.)

When we use some sort of meter or software to measure the frequency respons eof a speaker in a room we get to choose how we do so.
We can take a measurement very quickly, after the direct sound has had time to arrive, but before the reflected sound can arrive at the microphone.
(This is virtually the same as the response would be in an anechoic chamber.)
Or we can take a measurement over a period of several seconds, and average the total together.
(This is sometimes called "the power response" or 'the average response" and is influenced by the acoustics of the room; this is what you measure with a classic "SPL meter".)
Or we can measure both, and use the difference to tell us a lot about the acoustics of the room.
HOWEVER, it's pointless to argue that one or the other is "more correct", because our brain hears and analyzes both, and derives information from both, and from the difference between them.
This is why you cannot make "a live room sound dead", or vice versa, by using EQ.... because you can adjust EITHER the direct sound, or the total direct + reflected sound....
But the ratio between the two is directly tied to the PHYSICAL acoustic properties of the room.
(You can add some phony echoes by creating them electronically and adding themartificially... but you cannot remove excess room reflections other than by preventing them.)

From my understanding, the diffuse field curve is the most neutral / flat. As in if you re-record what is heard at your eardrum, and played it back through the earpiece, and repeat it several times, there should not be too much difference.

Flat signal in, flat signal out.

The harman listener target curve is developed for a different reason - the intended reproduction of music. When a recording studio records music, they “target” the reproduction through a quality hi-fi system in a listening room. Note that this room is not a recording studio and adds “color” to the sound through its walls and ceiling.

The Harman listener target curve attempts to recreate this home listening condition in the IEM. Thus fulfilling the recording studio’s intended playback scenario - making it sound “best”.

Why isn’t flat signal the “best”? Because the recording studio did not record the music intending them to be listened in recording studios, but home theater systems.

Some interesting personal observation: music such as theatrical plays, recorded in the audience perspective (binaural), sound incredibly transparent (even soundstage) with flat iems like the ER4SR. However label recordings always sound better with a colored IEM.

The conclusion is that both target curves are excellent at their intended purpose. For monitoring of microphone recordings? Diffuse field curve flat. For simulating reproduction of home hi-fi? Harman listener target curve.

On other question about human ear sensitivity, our ears most likely evolved to listen to rustling leaves and crackling branches of predators. To miss that is a death sentence. These are both sounds higher than the typical vocal range.

As IEMs bypass the outer half of our ear’s natural amplification system, a good IEM needs to simulate that natural amplification. This tends to be a peak at 2.6-2.8kHz and a roll off at higher frequencies. i.e. the diffuse field curve.

Thus the next time you are auditioning a colored IEM, remember that you are really buying a listening room and a pinna.

 

[bigshot]

With multichannel, you can alter the timing between the front and rear channels to create cancellation or synthetic reverb that would make the recording sound either dead or live. It would be hard to totally control that in the real world though.

 

[KeithEmo]

Exactly - but only within limitations.
It's realtively simple to create "fake room reverberation" - although the results often sound less than perfectly natural.
However, you cannot reduce room reverberation effectively, because you cannot "cancel out" existing sound using additional sound.
(Some software, like Izotope RX, can actually do this to a degree, by "clipping off the reverb tails".... but can only do under limited circumstances.)

Many people have the idea that, using some sort of advanced room correction, you can "cancel out sounds you don't want using out-of-phase signals".
(So that, for example, you could "correct for an excessively live room.)
The reality, of course, is that this is impossible.
If you measure a certain sound at a certain spot, and you wish to "remove" it, you can indeed generate an out-of-phase signal that exactly cancels it out.
You will have deliberately generated a "null".
However, that cancellation will only work as intended at one single precise spot in three-dimensional space.
At every other spot in space, not only will the original signal remain, but you will have added a second signal to it, which will add or subtract from it in complex patterns.
(You might as well try to cancel out a reflection from a flashlight by using another flashlight.)

The dimensions involved are related to the wavelength of the frequency involved.
At 20 Hz, which has a wavelength of about 50 feet, you could create a null several feet across.
However, at 10 kHz, where the wavelength is about one inch, cancellation would create a "comb effect", with peaks and nulls about an inch apart.

With multichannel, you can alter the timing between the front and rear channels to create cancellation or synthetic reverb that would make the recording sound either dead or live. It would be hard to totally control that in the real world though.

 

[gargani]

No, it's the opposite. Flat speakers in an anechoic chamber sound flat. The room alters the response and mixing stages aren't anechoic chambers. If you look at a measured response of a good listening room with speakers from the listening position, it looks very much like the Harman Curve.



When you EQ and try to establish a balance for the response, you can see what is happening. The idea of a listening room isn't to eliminate all of the reflections. It's to eliminate bad reflections. The speakers are only half of the sound of a speaker system. The room is just as important as the speakers. The Harman Curve is an attempt to synthesize the effect of space on sound. That physical space is what creates true soundstage. You can match the affect on the response, but you can't recreate true soundstage with headphones... at least not yet.

Yes, I understand now, that the ideal is not to recreate the sound of a speaker in an anechoic chamber.
I guess ,as Steve999 said, that the called for boost, peaking at 3khz. to 4khz., is just coincidental to human hearing being the most sensitive in that area.

 

[castleofargh]

a bunch of things to correct here before continuing the discussion:
1/ the development of the Harman curve was made to find a target that is preferred by most listeners. the all thing about using flat speakers in a room as reference to tune the HD800 is true but it was only the initial curve to start one of the experiments they conducted at the beginning. in that experiment alone the listeners had deviated from it by choosing something warmer. and they ran many more listening tests over several years in different places to help fine tune people's preference(because some of the first EQs were just a bass knob and a treble knob, or multiple headphone choices) to lead to their target, one for headphones one for IEMs.

2/ the curves you usually find are stuff someone made himself. the one previously posted is an older version of the curve made by Tyll(the real innerfidelity). and even then it apparently didn't work on his own measurements as he felt compelled to alter the curve himself by ear before presenting so called examples of how a few headphone graphs would look like under that compensation curve. something I found troubling at the time because IMO it gaves a false idea about that curve and also about the type of measurements we could use to just go and apply the compensation to get the real Harman response. basically we have a general idea of what the curve looks like, which is a somewhat tilted toward warm DF compensation with some more bass boost. but that's about it for me, I don't have their measurement gears or calibration, my graphs may or may not give the Harman curve after I apply one of the graphs you find online that often doesn't actually come from Harman. and yet, I find myself going in that general direction of warmer DF+bass. so I want to support their curve if only for being a step in the right direction, IMO.

3/ some of you are discussing apples and oranges when talking about the midrange boost. speaker measurements are usually done with a mic that has nothing to do with a human ear. when you see flat(electrically flat), we call the response flat.
now when you measure the same speakers with a dummy head, you get the midrange bump on a RAW measurement. because a dummy head was acoustically designed to mimic an average human head and ears. I'm going chicken and the egg about this to say that we have a midrange boost naturally installed in our body. you can also see it on the DF compensation for headphones, which has been one of the most used standard in headphone measurements. so criticizing that boost makes no sense as it's only visible on a graph because the measurement coupler is imitating a human ear.
of course the actual boost may be more significant for one person and less for another. it might not be centered exactly at the same frequency(it will still usually be within about 1kHz in that area), but on average the midrange boost is as accurate as we know it to be(yes, it did involve cadavers). and a little clue for those who doubt what I'm saying, what do you imagine to be the main cause for our superior midrange sensitivity in the first place?


all this to say that one doesn't have to prefer that curve(if you happen to know what it is and which graphs it can actually be applied onto), different people are indeed physically different in some ways and the numerous Harman studies were aiming at a general preference, not at a unanimous one.

 

[Steve999]

a bunch of things to correct here before continuing the discussion:
1/ the development of the Harman curve was made to find a target that is preferred by most listeners. the all thing about using flat speakers in a room as reference to tune the HD800 is true but it was only the initial curve to start one of the experiments they conducted at the beginning. in that experiment alone the listeners had deviated from it by choosing something warmer. and they ran many more listening tests over several years in different places to help fine tune people's preference(because some of the first EQs were just a bass knob and a treble knob, or multiple headphone choices) to lead to their target, one for headphones one for IEMs.

2/ the curves you usually find are stuff someone made himself. the one previously posted is an older version of the curve made by Tyll(the real innerfidelity). and even then it apparently didn't work on his own measurements as he felt compelled to alter the curve himself by ear before presenting so called examples of how a few headphone graphs would look like under that compensation curve. something I found troubling at the time because IMO it gaves a false idea about that curve and also about the type of measurements we could use to just go and apply the compensation to get the real Harman response. basically we have a general idea of what the curve looks like, which is a somewhat tilted toward warm DF compensation with some more bass boost. but that's about it for me, I don't have their measurement gears or calibration, my graphs may or may not give the Harman curve after I apply one of the graphs you find online that often doesn't actually come from Harman. and yet, I find myself going in that general direction of warmer DF+bass. so I want to support their curve if only for being a step in the right direction, IMO.

3/ some of you are discussing apples and oranges when talking about the midrange boost. speaker measurements are usually done with a mic that has nothing to do with a human ear. when you see flat(electrically flat), we call the response flat.
now when you measure the same speakers with a dummy head, you get the midrange bump on a RAW measurement. because a dummy head was acoustically designed to mimic an average human head and ears. I'm going chicken and the egg about this to say that we have a midrange boost naturally installed in our body. you can also see it on the DF compensation for headphones, which has been one of the most used standard in headphone measurements. so criticizing that boost makes no sense as it's only visible on a graph because the measurement coupler is imitating a human ear.
of course the actual boost may be more significant for one person and less for another. it might not be centered exactly at the same frequency(it will still usually be within about 1kHz in that area), but on average the midrange boost is as accurate as we know it to be(yes, it did involve cadavers). and a little clue for those who doubt what I'm saying, what do you imagine to be the main cause for our superior midrange sensitivity in the first place?


all this to say that one doesn't have to prefer that curve(if you happen to know what it is and which graphs it can actually be applied onto), different people are indeed physically different in some ways and the numerous Harman studies were aiming at a general preference, not at a unanimous one.

Okay, I think I might be almost starting to understand this. So the headphone Harman curve is what the the average preferred speaker in a room will look like when measured from inside the ear when you measure headphones? And the midrange bump is because If you stick a mic in an ear the way you measure a headphone if you are playing back from a speaker in a room with the preferred Harmon speaker in a room response that’s the frequency response you’ll get from the mic in the ear (on a average) receiving that sound? And the reason is the ear is indeed more sensitive in that area and the ear has already focused the sound in that way to form that midrange emphasis by the time the mic measures it in the ear?

 

[GearMe]

so...if a piece of music, say an acoustic jazz quartet, is made to target the Harman curve, it's not as 'Audibly Transparent' or 'High Fidelity' as it could be (i.e. it's been EQ'd for the masses)?

 

[castleofargh]

So the headphone Harman curve is what the the average preferred speaker in a room will look like when measured from inside the ear when you measure headphones?

no,that was only what they started with because of the assumption that people liked that in speakers so why not try on headphones. but then the test subjects had an EQ and changed things to what they preferred, and the trend was for something warmer, then many more experiments many more data many more options to pick their favorite signature, and at some point they ended up with their target response. the flat speakers in a room was simply the starting point for the listening tests, they had to start somewhere, they went for that. it could have been electrically flat for example, but they already knew that nobody would want that crap, so instead they tried to start with something else.

And the midrange bump is because If you stick a mic in an ear the way you measure a headphone if you are playing back from a speaker in a room with the preferred Harmon speaker in a room response that’s the frequency response you’ll get from the mic in the ear (on a average) receiving that sound?

if you stick a mic at the eardrum of somebody who doesn't mind, and measure a flat speaker, you will measure a midrange bump caused by the acoustic of the ear. when you see headphone measurements they're done most of the time with a dummy head that also has that acoustic model(a fake ear canal and fake ear), so they measure a bump. it doesn't mean it's going to feel like a bump to us, a flat speaker does sound flat because everything we hear is boosted in the midrange, that's what we call natural sound. we never really hear flat sound IRL. the measurements are just showing that because it's how our ears change the sound.

is it clearer that way?

 

[castleofargh]

so...if a piece of music, say an acoustic jazz quartet, is made to target the Harman curve, it's not as 'Audibly Transparent' or 'High Fidelity' as it could be (i.e. it's been EQ'd for the masses)?

the Harman curve is an attempt to find a preferred target for headphones, it has nothing to do with a jazz band or how it would be mixed/mastered. and even if we assume you're talking about how we'd EQ our headphone, that would not be a claim about neutrality. they never claimed that. some may assume it is so because in general we tend to like flat response on speakers, but just the lack of tactile bass in the subjective experience with headphones will tend to mess with the chance that we're looking for the same thing. many people say that we're overly boosting the low end simply because we try to get a feeling of tactile bass again and we assume that more bass is the way. I find that to be a pretty reasonable idea.

 

[Steve999]

no,that was only what they started with because of the assumption that people liked that in speakers so why not try on headphones. but then the test subjects had an EQ and changed things to what they preferred, and the trend was for something warmer, then many more experiments many more data many more options to pick their favorite signature, and at some point they ended up with their target response. the flat speakers in a room was simply the starting point for the listening tests, they had to start somewhere, they went for that. it could have been electrically flat for example, but they already knew that nobody would want that crap, so instead they tried to start with something else.


if you stick a mic at the eardrum of somebody who doesn't mind, and measure a flat speaker, you will measure a midrange bump caused by the acoustic of the ear. when you see headphone measurements they're done most of the time with a dummy head that also has that acoustic model(a fake ear canal and fake ear), so they measure a bump. it doesn't mean it's going to feel like a bump to us, a flat speaker does sound flat because everything we hear is boosted in the midrange, that's what we call natural sound. we never really hear flat sound IRL. the measurements are just showing that because it's how our ears change the sound.

is it clearer that way?

 

[GearMe]

the Harman curve is an attempt to find a preferred target for headphones, it has nothing to do with a jazz band or how it would be mixed/mastered. and even if we assume you're talking about how we'd EQ our headphone, that would not be a claim about neutrality. they never claimed that. some may assume it is so because in general we tend to like flat response on speakers, but just the lack of tactile bass in the subjective experience with headphones will tend to mess with the chance that we're looking for the same thing. many people say that we're overly boosting the low end simply because we try to get a feeling of tactile bass again and we assume that more bass is the way. I find that to be a pretty reasonable idea.

Thanks for the reply...yeah I know it doesn't have to do with a specific genre of music...giving an example.

Just wanted to clarify that an industry leading methodology sounds like it's not targeting 'audibly transparent' or 'flat' or 'high fidelity' the way many folks think/talk/write about it. In reality, it sounds like they asked/tested people's listening preferences and have constructed Headphone / IEM curves that satisfy the majority of them.

As far as the genres go, the questions I'd have are, when they sampled listener preferences, did they try to account for genre preferences, listening capability differences (young vs old), etc.?

Anecdotally, my son could listen to the Dave Brubeck Quartet and probably...
a. not really be interested and therefore his answers may not reflect those of a different segment of the population (i.e. acoustic jazz fans)
b. like it better on TH900's instead of HD800's -- simply because Eugene Wright's solos carry a bit more weight (drop that bass!)

Same goes for me listening to some of the music genres that he likes! I'm pretty sure headphone selection wouldn't matter to me on some popular music.

That said, we both like Dave Matthews Band, Zac Brown Band, and others. Even so, I'd bet he'd choose to listen to them through different cans than I would...for several reasons.

 

[Steve999]

Yes, I understand now, that the ideal is not to recreate the sound of a speaker in an anechoic chamber.
I guess ,as Steve999 said, that the called for boost, peaking at 3khz. to 4khz., is just coincidental to human hearing being the most sensitive in that area.

Not just a coincidence! Our counterintuitive alarms were trying to tell us something! See @castleofargh ‘s posts four and seven posts up.

 

[Steve999]

Thanks for the reply...yeah I know it doesn't have to do with a specific genre of music...giving an example.

Just wanted to clarify that an industry leading methodology sounds like it's not targeting 'audibly transparent' or 'flat' or 'high fidelity' the way many folks think/talk/write about it. In reality, it sounds like they asked/tested people's listening preferences and have constructed Headphone / IEM curves that satisfy the majority of them.

As far as the genres go, the questions I'd have are, when they sampled listener preferences, did they try to account for genre preferences, listening capability differences (young vs old), etc.?

Anecdotally, my son could listen to the Dave Brubeck Quartet and probably...
a. not really be interested and therefore his answers may not reflect those of a different segment of the population (i.e. acoustic jazz fans)
b. like it better on TH900's instead of HD800's -- simply because Eugene Wright's solos carry a bit more weight (drop that bass!)

Same goes for me listening to some of the music genres that he likes! I'm pretty sure headphone selection wouldn't matter to me on some popular music.

That said, we both like Dave Matthews Band, Zac Brown Band, and others. Even so, I'd bet he'd choose to listen to them through different cans than I would...for several reasons.

I know Harman classifies listeners into skill levels. I think there are something like twelve. Getting up to seven is supposed to be pretty darn difficult. The “Harman” trained listeners tend to get the highest scores. Big surprise. The listening seems to be based on discerning audio quality rather than skill in musical listening. Musicians had a bit of an edge over your average Joe, but not so much as you might expect. IIRC.

I don’t know how their studies play out in terms of genre.

I also understand that other companies have their own data and techniques, but they are not as transparent about it all. For all we know another company could have better data, IMHO.

I’ve seen at least one slide from the Harman presentations that showed professional reviewers as a group had the very lowest predictive value for what people will like.

I also saw one slide where Consumer Reports’ findings very highly or most highly correlated with with Harman’s ascertained user preferences for headphones.

I am with what you are intimating philosophically, there are so many variables between the recording process and mastering and speaker production and trying to cater to preferences and types of listening on different speakers while mastering and characteristics of consumer speakers intended or not and listening rooms that there is no brass ring as far as transparency of transducers. Some people in some of these fields point to the lack of standardization in the audio chain from production to consumer as a problem, arguing that we could get closer to a precise concept of fidelity if there was standardization along the chain. They compare it to calibration of TVs or movie sound. My gut reaction is to ask would that leave less room to bend the rules, for artistic license, for trying new ideas, etc.

I think it’s nice to have a system where you are getting close enough so you know you are getting in some range of a reasonable approximation of what was nebulously “intended.” I have one system kind of for that, within a pretty modest budget. Otherwise I am like you, if I remember what you’ve said correctly, the point to where my enjoying the music a lot is not necessarily dependent on a relatively high level of fidelity. The beauty is in how much we enjoy music, not chasing an imaginary ideal audio system. IMHO, YMMV, IIRC, & etc.