Sennheiser HD 650/HD 6XX

[dc655321]

Is it the technique that is the problem

IMO, yes, this.

or does 'some' notified body has to establish a 'standard' first to which manufacturers have to comply ?

That may help.
And a measurement standard does not need establishment by a recognized body: I don't care where the technique/standard arises.
Although endorsement by a recognized body may lend credence.

Speakers measurements often are designed to strip away environmental, controllable variables to get to the point of how the speaker performs in isolation. Performance in-room cannot be precisely accounted for given the variability of rooms, so I would argue the same reasoning be applied to ears. I'm sure that's not an original proposition, but I've yet to see any real attempt either.

 

[solderdude]

Can I ask what you mean with the technique as this is kind of broad.
(Not attacking you just curious about your ideas and attempting to learn and debate)

Looks like Olive and Welti made an attempt for the latter. I don't know if the choice of 'listeners' and averaging results is the way to go.
I mean those listening/preferring rock will probably prefer another compensation as those enjoying classic recordings. Assuming the recording quality is good. This may be born out of referencing to loudness by those people and references to actual instruments.
A horrible thing ... the ears thing as there is preference which should not belong in any standard.

How different do various HATS actually measure (RAW) with the exact same headphone ?
I know Tyll visited Sean with his HATS but 'remember' the raw measurements differed as well.
I assume both HATS are built to standards.

 

[pkane]

A question... (Have wrestled with similar issues)

Are the measured differences between the ears the result of:
A: The Pinnae/HRTF
B: The L-R difference between the HE-560 drivers
C: the exact angles/position of the mics differing
D: differences between mics
E: different positions of the drivers on the head
F: differences in gain between L and R mic-pre-ADC ?
G: Is the mic measuring 'flat' when inside an ear cavity ?

or ... a mix of any or all of the above ?

There are so many variables/unknowns here and differences of 2 dB are quite audible. I see differences well over 6dB here and there.
What I have found is that when creates 'sharp' filters to fill in dips or lower peaks that this rarely leads to worthwhile compensations.
Heavy 'smoothing' or using wide filters seems to produce the best (sonic) results, when one can rely on the measurement gear being 'flat' as the reference. And only then when the corrections needed are small.

Did you swap L and R driver (and the pads as they are angled ?) to check for differences between the drivers ?
What FR differences do you get with different mic orientations ?
How does one check FR differences in the mic circuit ?
Can the mics themselves be swapped between L and R earpiece or are they 'fixed' to the ear clamp ?

I have played with in-ear mics (different ones) as well and found them highly unreliable and incapable of producing repetitive results within a few dB over the entire FR range so gave it up.
It proved impossible to get meaningful results for me especially in distortion measurements due to mechanically coupled sounds.
How can one build a reliable compensation on this is a question that comes forward ?

Lots of good questions. Let me answer the easy ones, first.

I measured left/right cans of my HE560s with a calibrated mic. The measurements showed they were pretty much the same.

It’ll take a bit more effort to calibrate the in-ear microphones, although they are claimed to be pretty flat by the manufacturer.

The ADC was calibrated through a loop back connection (Behringer Midas-based unit)

Different mic positions mostly result in a level difference not in different frequencies.

The mics can be swapped, but will not fit the same in the wrong ear due to how they are shaped.

I tried to reposition the mics in the ear until I could get a reasonably good match between measurements. A 1dB or so difference variation is going to be hard to eliminate.

After I measured and created DSP filters, I measured again. The result was much more flat, but not quite. I then applied additional filters to correct for the differences and remeasured. The result was as flat as I would normally expect the corrected response to be — I tend to use as few corrections as possible, and eliminate any that are too narrow or too large.

As to what I’m correcting for, that’s just a speculation. I suspect it’s pinna effect, maybe a little of the HRTF since the ear canal is blocked, and mostly the HE560s frequency response. For me, this makes more sense than correcting just for the headphone response, or using a Harman ‘average’ for HRTF.

 

[solderdude]

A question would be why one would want to strive for a flat measurement at the ear entrance ?

My reasoning.... if one were to consider a 'flat loudspeaker' in an anechoic room and measure that speaker at 1m away then the FR of the sounds arriving just in front of the ear would (theoretically) be 'flat' and measurably will be.
At the ear entrance the sound would be far from flat. There would be changes in FR due to Concha gain and Pinna shape which differ substantially (as in +10dB) at the ear entrance. (not counting the ear canal on purpose).
Still, despite that 'boost' we perceive the speaker as flat and the sounds just in front of the ear are 'flat' too.
If one were to EQ the speaker to the opposite of what is at the ear entrance the speaker would sound anything but flat.

One can repeat the same but with the speaker to the side of the head.
Due to the Concha gain not being nearly as much the speaker will sound different then when heard from the front (obviously) but why would one compensate to get that flat at the ear entrance ?
Would one not have to compensate for the difference between sounds from the front and the side ?

 

[dc655321]

(Not attacking you just curious about your ideas and attempting to learn and debate)

No worries. I'm curious about these things myself.

Can I ask what you mean with the technique as this is kind of broad.

I've tried to reason that adding any sort of "compensation" to headphone frequency responses doesn't make sense (to me). Given anatomical variations, any generalized compensation is a guess. At best. So, by that reasoning, measuring headphones through any sort of anatomical simulation likewise makes little sense.

So, what does make sense to me?
Free-field (anechoic) measurement.

This would at least indicate how the device performs prior to convolution with the ear.
Show me frequency response, distortion measurements, usable range, etc. garnered from such testing.
With the knowledge of how only the headphone or iem performs, I can at least pretend to make an informed decision about quality.
If I have to mentally deconvolve frequency response from compensation curve, I'm just guessing. At best.

 

[andreasmaaan]

So, what does make sense to me?
Free-field (anechoic) measurement.

This would at least indicate how the device performs prior to convolution with the ear.
Show me frequency response, distortion measurements, usable range, etc. garnered from such testing.
With the knowledge of how only the headphone or iem performs, I can at least pretend to make an informed decision about quality.
If I have to mentally deconvolve frequency response from compensation curve, I'm just guessing. At best.

Would you suggest measuring truly free-field, i.e. without creating any kind of seal with the headphone cup? This would seem to me also somewhat fraught, as the cup will interact significantly in the frequency range in which wavelengths are smaller than or equal to the dimensions of the cup, whereas wavelengths significantly longer will tend to wrap around the unsealed cup omnidirectionally (with a lot of erratic interference at and above the transition frequency).

I think this is worth thinking about further though, as I'd like to see some kind of HRTF-independent measurement system for headphones. Just trying to think through possible problems here...

 

[solderdude]

So, what does make sense to me?
Free-field (anechoic) measurement.

This would at least indicate how the device performs prior to convolution with the ear.
Show me frequency response, distortion measurements, usable range, etc. garnered from such testing.
With the knowledge of how only the headphone or iem performs, I can at least pretend to make an informed decision about quality.
If I have to mentally deconvolve frequency response from compensation curve, I'm just guessing. At best.

That is basically what I am doing. Measuring 'free-field' (but sealed and partially sealed) and somewhat anechoic from the side/distance where the actual sounds come from not hindered by Pinnae, just the speaker itself.
The bass correction I apply has little to do with HRTF but has other causes. HP50 is an example as when it is measured without bass compensation it has huge amounts of bass (+10dB) yet some people perceive it as 'realistic'.

The above picture is from HERE and know no 'bass correction' has been applied for certain.
They measured quite a few headphones but IMO their compensation around 3kHz is a bit 'off' and too much smoothing is applied to my taste.

As I don't have an artificial ear canal, which obviously has to be corrected perfectly afterwards anyway without success on the HATS measurements I have seen ?, the only 'corrections' I have to make is that of the mic on an infinite baffle.
This gives me plots that do not 'correlate' with (by me) perceived tonal balance.
I would agree that this is the 'raw' signal and would show equal 'loudness' over the entire audible range (and beyond) and makes sense as it is easier to compare if others also do not apply this.
However, those plots would only be interesting for those knowing what the plot means and would have to visually 'substract' their personal favorite 'compensation'.
I want a plot to show how it is perceived (primarily by me as at first I only measured for myself) and this simply does not jive with a flat measurement.
So.... I added the correction which works for me. A flat line gives me realistic sound with a full bodied but not bassy sound.

measuring IEM's is a mess. Insertion depth, seal and artificial ear canal shape and size would need to be standardized (there seems to be a standard for this) but then comes the correction part to 'translate' to perceived sound. That is a mess because all people insert and seal different.
I don't like shoving things in my ear so that's why I don't measure them.

 

[andreasmaaan]

@solderdude do you have your own measurements of the HP50 there for comparison? Finding it hard to see how your process relates to the measurements from that other source. Also a pic or diagram of your setup would be interesting if not too much to ask?

 

[solderdude]

Edit: Measured them recently..

My setup is described HERE

That HP50 measurement differs a lot from this one I trust more:

The link was about what a 'not bass compensated' measurement looks like in the lows.
Of course you can compare them to any of the shitload of measurements they made to get a feel of their measurements.
Just click on the 'plot icon' to see the measurements.

The one from Rtings IS compensated for bass and is also explained by them as why.

The Russian measurement site also is very well documented and use a HATS but a different one from Rtings with very different compensations.
Rtings have their own compensation curve which IMO is close to how they are perceived.

 

[andreasmaaan]

Thanks @solderdude . Interestingly, that second compensated HP50 measurement accords more with my subjective impression of their sound, although to me they do sound a bit heavier than neutral in the bass to me.

I'm still not 100% sure how you seal the headphone cups after reading the info on your web page. Could you explain a bit here? Enjoyed reading all the info you've posted

 

[pkane]

A question would be why one would want to strive for a flat measurement at the ear entrance ?

My reasoning.... if one were to consider a 'flat loudspeaker' in an anechoic room and measure that speaker at 1m away then the FR of the sounds arriving just in front of the ear would (theoretically) be 'flat' and measurably will be.
At the ear entrance the sound would be far from flat. There would be changes in FR due to Concha gain and Pinna shape which differ substantially (as in +10dB) at the ear entrance. (not counting the ear canal on purpose).
Still, despite that 'boost' we perceive the speaker as flat and the sounds just in front of the ear are 'flat' too.
If one were to EQ the speaker to the opposite of what is at the ear entrance the speaker would sound anything but flat.

One can repeat the same but with the speaker to the side of the head.
Due to the Concha gain not being nearly as much the speaker will sound different then when heard from the front (obviously) but why would one compensate to get that flat at the ear entrance ?
Would one not have to compensate for the difference between sounds from the front and the side ?

I agree with your line of thought. That's a question that's been bothering me for a while.

What I decided to check is the difference between the in-ear microphone + HE560s response compared to just the HE560s without pinna in the way. Here's what it looks like. Not a huge difference, so perhaps all I'm doing when correcting for pinna and headphones is primarily correcting for the headphones only. Up to about 700Hz, there's 1dB of difference or less, and then up to about 5dB from 3 to 7KHz, with HE560s requiring a larger correction to get to flat.

Blue (thick) line is HE560s response, purple thin line is my left ear pinna + HE560s:

 

[solderdude]

Thanks @solderdude . Interestingly, that second compensated HP50 measurement accords more with my subjective impression of their sound, although to me they do sound a bit heavier than neutral in the bass to me.

I found that both Rtings and Sonarworks both slightly 'overdo' the bass compensation. Just like Olive-Welti by the way.
I reckon if I were to measure one it would probably show a raised bass of around +5dB.
The treble roll-off in the Russion plot is weird.

Thanks @solderdude I'm still not 100% sure how you seal the headphone cups after reading the info on your web page. Could you explain a bit here? Enjoyed reading all the info you've posted

Seal is when the pads fully make contact with the measurement rig.
This usually isn't a problem with my rig as it is flat so most pads will have a perfect seal.
With HATS and a more solid Pinna this may not be the case.

On the other hand my rig has problems with the likes of Ananda etc which have pads with a curvature. I have to use wizardry and trickery to get a good seal in that case.
A HATS may seal better as that has a more oval shape.

 

[solderdude]

What I decided to check is the difference between the in-ear microphone + HE560s response compared to just the HE560s without pinna in the way. Here's what it looks like. Not a huge difference, so perhaps all I'm doing when correcting for pinna and headphones is primarily correcting for the headphones only. Up to about 700Hz, there's 1dB of difference or less, and then up to about 5dB from 3 to 7KHz, with HE560s requiring a larger correction to get to flat.

Is the headphone itself measured with the same mic and fully sealed or using a different mic.
One would expect a slightly higher 2-5kHz range with the on.ear mic due to a slight influence of Concha gain.
Not nearly as high as when the sound were coming from the front and dependent on the angle, driver size and distance.

Here is my HE560 measurement for comparative reasons:

Here my 'compensation' is clearly visible. The raw measurement would show a horizontal line from 20Hz to 200Hz instead of a sloping line.

My 'correction' would be a few dB more lows and about 5dB 'fill-in' at 2kHz. When the fill-in would be 10dB it sounds way 'off'. A bit of the dip is 'filled in' by Concha gain.

The Sundara and HE560 measure kind-of similar on my rig (smoothed here for better tonal balance differences)

and for that one I would also recommend a +5dB fill-in and the bass boost below 200Hz.

When this is compared to Rtings Sundara measurement (which already have a Pinna in there so a slight boost around 2khz because of it the 'compensation' based on that rig would come down the same as on my rig + Pinna effects taken into account.
So a boost below 200Hz and a few dB around 2kHz. Below the Rtings Sundara plot.

 

[dc655321]

@solderdude and @andreasmaaan - thank you both for sharing your thoughts.
I have more thinking to do (clearly!) regarding headphone measurements, and not a lot of time at the moment to discuss.

 

[andreasmaaan]

Seal is when the pads fully make contact with the measurement rig.
This usually isn't a problem with my rig as it is flat so most pads will have a perfect seal.

That clarifies it, thanks. You're aiming to get a seal, but in some cases it isn't possible.

 

[Grave]

See original post.

 

[maverickronin]

I would recommend the leather Dekoni ear pads for the Sennheiser HD 650's and the Sennheiser HD 660 S's as long as you do not mind a slightly darker sound.

I have the Dekoni velours for my HD650. They make it quite a bit darker but I just EQ it back out since the comfort is much improved.

 

[JJB70]

I still have my old HD580 headphones, I never really felt the later replacements (600, 650, 660) offered any reason to upgrade beyond the idea of getting something new. And I am in that position where once you've been living with the sound profile of something for 25 years then any change is probably unwelcome just because it is so familiar and comfortable. At one point I was really tempted by the Massdrop HD58X to have a spare pair in readiness for the day when my old 580's die but decided there wasn't much point when people told me they weren't really like the original 580, if I just want a new pair of Sennheiser's I can buy new HD600/650/60's for little more than I'd pay for the 58X after shipping, adding the 20% VAT and customs handling charge.
My current open favourite is the Oppo PM-2. A lovely sound, not bass heavy, a little bright but in the range of neutral-ish sounding headpones to my ears and with superb comfort and build quality. They're easy to drive and are easy to carry around when I travel. I know the denim case wasn't to all tastes but I am finding it remarkably practical and hard wearing.

 

[DonH56]

I have Sennheiser HD-424. Very old, still work. Replaced the covers several times and cord once. Have not used them for several years, however, since I quit recording for area groups.

 

[Grave]

I have been looking at the frequency responses of possible open headphone purchases, but they all look worse than the HD 650's frequency response to me, so I don't think there is much point in trying them.

Hifiman HE-400X - treble peaks
Focal Elex - peak at 5k instead of 3k
Focal Elear - treble peaks
Phillips Fidelio X2 - treble peaks
Mr. Speakers Aeon Flow - treble peaks