Determining safe level of headphone volume with umik-1

[solderdude]

Frequency Weightings - A-Weighted, C-Weighted or Z-Weighted

A description of the frequency weightings used for sound level meters - A-Weighted, C-Weighted or Z-Weighted.

www.noisemeters.com

dB(A) is weighted similar-ish to the loudness contour curves as we are less sensitive to lower frequencies.
As a result the dB(A) is close to how we perceive sound. More pressure is needed at lower frequencies than for mids and highs to experience the various frequencies equally loud.
In our hearing this changes when the level changes but not so for dB(A). The curve is always the same.
When measuring music or background levels the lower frequencies thus are 'under reported' in the measurement.
You thus get lower numbers when measuring the same music.

dB(X) is basically 'flat' (normal measuring mic). As music contains more 'energy' in the lows the reported dB's in dB(X) (as well as dB(C) which is weighted with a bit of low frequency and high frequency roll-off) compared to dB(A).

dB(A) is closer to how we perceive loudness. dB(X) is the actual measured energy.

Then there is a difference between peak and RMS, average, crest factor, DR and LUFS which say something about 'average' and peak levels.
The peak levels you found using your method will probably be fairly accurate assuming the recordings used actually had peaks near 0dBFS.
The 'average' levels are recording dependent.

 

[Robbo99999]

Frequency Weightings - A-Weighted, C-Weighted or Z-Weighted

A description of the frequency weightings used for sound level meters - A-Weighted, C-Weighted or Z-Weighted.

www.noisemeters.com

dB(A) is weighted similar-ish to the loudness contour curves as we are less sensitive to lower frequencies.
As a result the dB(A) is close to how we perceive sound. More pressure is needed at lower frequencies than for mids and highs to experience the various frequencies equally loud.
In our hearing this changes when the level changes but not so for dB(A). The curve is always the same.
When measuring music or background levels the lower frequencies thus are 'under reported' in the measurement.
You thus get lower numbers when measuring the same music.

dB(X) is basically 'flat' (normal measuring mic). As music contains more 'energy' in the lows the reported dB's in dB(X) (as well as dB(C) which is weighted with a bit of low frequency and high frequency roll-off) compared to dB(A).

dB(A) is closer to how we perceive loudness. dB(X) is the actual measured energy.

Then there is a difference between peak and RMS, average, crest factor, DR and LUFS which say something about 'average' and peak levels.
The peak levels you found using your method will probably be fairly accurate assuming the recordings used actually had peaks near 0dBFS.
The 'average' levels are recording dependent.

So dB(Z), which I used in my measurements is a linear recording method across all frequencies - in that case my graphs should be more applicable to relating to distortion measurements that people like Amir creates, is that so? Would the 104dB peaks in my graph relate directly to the 104dB distortion graphs that Amir produces (re bass) or are the "units" not in the same format? Do you agree that the LZpeak red line in my graphs are corresponding to the bass parts of the track? Why are the dB figures in my graphs significantly higher by a large margin than my previously calculated max - my previously calculated max (86dB) was based on Vrms (yet I'm seeing 104dB in parts of these graphs), is that the reason.......am I measuring the absolute maximum peaks in my graphs whereas my calculations were based on Vrms? I'm asking all these questions because I don't fully understand how to interpret the graphs I showed (as well as not understanding fully each of the plotted lines on the graph), I was hoping you might have added your own interpretation of my graphs in relation to the points and questions that I had.

dB(A) seems like a better means of measuring hearing damage potential, so I should probably remeasure using dB(A) if I want to assess "hearing damage", but seems better to use dB(Z) to correlate to distortion figures.

 

[Aerith Gainsborough]

I think I listen louder than I worked out theoretically, when I calculated it last year - either I listen louder nowadays or I've misinterpreted the information.

104dB Z peak is bloody loud indeed.
One can listen to that for a track or two but after that, every normal human will have the urge to tone it down. Especially if the tracks contain a lot of loud passages wit high frequency content.

You can let a real time analysis of the mic's input run in a second window during the SPL measurement.
Should be pretty obvious to see whether the bass or the 4KHz gets the highest peaks.

dB(A) is typically used when measuring noise in working environments in order to assess the hearing damage potential.
I just went with dB(Z) because any error I make will be on the safer side of things.

 

[solderdude]

So dB(Z), which I used in my measurements is a linear recording method across all frequencies - in that case my graphs should be more applicable to relating to distortion measurements that people like Amir creates, is that so?

Not exactly as the gain in the EARS is different and so is the compensation (if you are using it).

Would the 104dB peaks in my graph relate directly to the 104dB distortion graphs that Amir produces (re bass) or are the "units" not in the same format?

That depends on whether or not the test fixture is calibrated or not.
Yo have a HD600. The sensitivity is known. You can also measure or calculate which SPL belongs to a 1kHz 0dBFS tone and see if that matches.
I would not use a 400Hz tone as the pad condition of both measured HD600 is not known.

Do you agree that the LZpeak red line in my graphs are corresponding to the bass parts of the track?

They related to accumulated signals. It measures the total signal not just the bass. The biggest part of it will be bass. Have a look at the frequency spectrum if you want to know for each specific song (in peak hold mode)

Why are the dB figures in my graphs significantly higher by a large margin than my previously calculated max - my previously calculated max (86dB) was based on Vrms (yet I'm seeing 104dB in parts of these graphs), is that the reason.......am I measuring the absolute maximum peaks in my graphs whereas my calculations were based on Vrms?

Yo are looking at peak values which may or may not be calibrated. Peak levels are 1.4x higher than RMS for the same signal for one thing (3dB) and the average can be anything from just below RMS to 10 to 20dB below that. Depending on the recording DR.

dB(A) seems like a better means of measuring hearing damage potential, so I should probably remeasure using dB(A) if I want to assess "hearing damage", but seems better to use dB(Z) to correlate to distortion figures.

dB(A) is often used in the scales you see everywhere and the warnings that come with it are based on prolonged exposure to those average and dB(A) weighted noise levels. Noise isn't as dynamic in nature as music. Perhaps with the exception of highly dynamic compressed 'music' with DR ratings close to 1 or 3.

dB(Z) will give you the electrical levels converted to SPL. Similar to the shape of a music signal where it shows the envelop of the signal.
You must realize that there is also pinna gain so when the total music signal has a bass note near -3dB and for instance a 4kHz signal peak that makes it reach the 0dBFS line means that the measured acoustical signal at the mic. (assuming flat FR) can be higher than expected.

When the FR is wonky and lets say has a peak of +8dB at 6kHz (not uncommon) and a Harman type bass boost on top of that but you use the efficiency rating given at 400Hz, 500Hz or 1kHz or based on noise in a limited frequency range that is rated at say 100dB at 1V then the bass output from the headphone could be +5dB more efficient and when that also has a treble peaking at 6kHz (+8dB more efficient) then the actual measured peak could well be 5 or 6dB higher (in a short moment but registered by the peak hold function) than expected based on the used sensitivity number.

So it is not as cut and dry as it seems.
Measured values could also be lower than expected when bass rolls off and there is a dip/recession at higher frequencies.

In that sense, making measurements with a calibrated (important aspect) microphone and based on compensated (not Harman but probably optimum hifi) will give you the best possible efficiency numbers at each given frequency (the plot that basically is what one sees on the DAC output and not what the measurement fixture's raw signal gives you.

The OP wanting to use an UMIK should do measurements creating a sealed environment around the mic and the proper distance (so about flat from lightly compressed pad distance) and then again as the mic is used is intended for free space measurements the response may not be the same as reality. Also hard surfaces around the microphone that seal the headphone (so no leakage anywhere) and positioning opposite the driver can cause erroneous readings.

Doing accurate measurements for headphones is a biatch. Also confounded by product tolerances.

I hope I used the right words to explain this, otherwise I can try to clarify this further.

 

[Robbo99999]

Not exactly as the gain in the EARS is different and so is the compensation (if you are using it).



That depends on whether or not the test fixture is calibrated or not.
Yo have a HD600. The sensitivity is known. You can also measure or calculate which SPL belongs to a 1kHz 0dBFS tone and see if that matches.
I would not use a 400Hz tone as the pad condition of both measured HD600 is not known.


They related to accumulated signals. It measures the total signal not just the bass. The biggest part of it will be bass. Have a look at the frequency spectrum if you want to know for each specific song (in peak hold mode)


Yo are looking at peak values which may or may not be calibrated. Peak levels are 1.4x higher than RMS for the same signal for one thing (3dB) and the average can be anything from just below RMS to 10 to 20dB below that. Depending on the recording DR.



dB(A) is often used in the scales you see everywhere and the warnings that come with it are based on prolonged exposure to those average and dB(A) weighted noise levels. Noise isn't as dynamic in nature as music. Perhaps with the exception of highly dynamic compressed 'music' with DR ratings close to 1 or 3.

dB(Z) will give you the electrical levels converted to SPL. Similar to the shape of a music signal where it shows the envelop of the signal.
You must realize that there is also pinna gain so when the total music signal has a bass note near -3dB and for instance a 4kHz signal peak that makes it reach the 0dBFS line means that the measured acoustical signal at the mic. (assuming flat FR) can be higher than expected.

When the FR is wonky and lets say has a peak of +8dB at 6kHz (not uncommon) and a Harman type bass boost on top of that but you use the efficiency rating given at 400Hz, 500Hz or 1kHz or based on noise in a limited frequency range that is rated at say 100dB at 1V then the bass output from the headphone could be +5dB more efficient and when that also has a treble peaking at 6kHz (+8dB more efficient) then the actual measured peak could well be 5 or 6dB higher (in a short moment but registered by the peak hold function) than expected based on the used sensitivity number.

So it is not as cut and dry as it seems.
Measured values could also be lower than expected when bass rolls off and there is a dip/recession at higher frequencies.

In that sense, making measurements with a calibrated (important aspect) microphone and based on compensated (not Harman but probably optimum hifi) will give you the best possible efficiency numbers at each given frequency (the plot that basically is what one sees on the DAC output and not what the measurement fixture's raw signal gives you.

The OP wanting to use an UMIK should do measurements creating a sealed environment around the mic and the proper distance (so about flat from lightly compressed pad distance) and then again as the mic is used is intended for free space measurements the response may not be the same as reality. Also hard surfaces around the microphone that seal the headphone (so no leakage anywhere) and positioning opposite the driver can cause erroneous readings.

Doing accurate measurements for headphones is a biatch. Also confounded by product tolerances.

I hope I used the right words to explain this, otherwise I can try to clarify this further.

There's quite a bit to digest there, so I'll re-read it after work (and post additional points or questions), but I get the general points I think. Re "calibration", the mics in the EARS are calibrated inasmuch as I am using the RAW calibration file that is included with each unit of EARS, so the actual mics are calibrated for frequency response & SPL (perhaps in the same way that "free air mics" are calibrated - flat accross the spectrum) - I don't know if that means the entire fixture is calibrated in terms of "industry standards" for measuring SPL? Certainly if I used one of their other compensations that supposedly mimic something close to Harman then this would effect the SPL's recorded at different frequencies, so I'm not sure which one I would use to get the most "industry standard" calibrated readings for reading SPL from a headphone? I can certainly try the HD600 experiment to calibrate.

EDIT: I think I know how to apply an EARS to GRAS fixture calibration file to my EARS that is valid for my K702 only,.....would that end up with some more "industry standard" measurements of SPL if I was to create the calibration file based on a 1kHz crossover of the 2 curves before creating the calibration file, as I think the chosen crossover point might influence the SPL measured (even though it wouldn't result in any "headphone EQ differences" if used for that purpose)?

 

[solderdude]

The SPL calibration thing (at 1kHz or 500Hz if it is) could be done in the EARS. I don't know though. Never worked with it. As the output is digital I suppose the manufacturer could easily calibrate SPL. In that case there would have to be some referencing to SPL opposite the 0dBFS (max measured SPL) in the manual about it.

Making an 'EARS to GRAS' measurement based on the headphones measured by you and Oratory you could get closer to the results Oratory gets.
It would have to be a very 'smoothed' compensation and certainly not an 'exact' one.
Similar to the kind of 'average' and 'smooth' target curve as shown in Amirs measurements.
That 'smooth' target actually never is accurate (certainly not above 8kHz) so take that with a grain of salt... a pound of salt if you will.

With that 'conversion' you will get a GRAS-ish (but never exact) 'raw' signal which you can then overlay with Oratory's Harman or optimum hifi target as published by GRAS for the exact same HATS config Oratory has.

It will, probably, be closer to some reality than the EARS compensations. Although I do suspect miniDSP improved their compensation over time.
For 'investigative' reasons you could also plot the compensated measurements of one headphone of choice with the supplied compensations (from your raw, not converted, measurements and maybe obtain the SBAF calibration file as well.

See which of those compensated plots comes closest to that of Harman corrected measurements Oratory made and also include the 'converted and then Harman compensated' trace as well.

Then you'll know which compensation method to use that comes closest to the Harman target on Oratory's HATS.

 

[Robbo99999]

The SPL calibration thing (at 1kHz or 500Hz if it is) could be done in the EARS. I don't know though. Never worked with it. As the output is digital I suppose the manufacturer could easily calibrate SPL. In that case there would have to be some referencing to SPL opposite the 0dBFS (max measured SPL) in the manual about it.

Making an 'EARS to GRAS' measurement based on the headphones measured by you and Oratory you could get closer to the results Oratory gets.
It would have to be a very 'smoothed' compensation and certainly not an 'exact' one.
Similar to the kind of 'average' and 'smooth' target curve as shown in Amirs measurements.
That 'smooth' target actually never is accurate (certainly not above 8kHz) so take that with a grain of salt... a pound of salt if you will.

With that 'conversion' you will get a GRAS-ish (but never exact) 'raw' signal which you can then overlay with Oratory's Harman or optimum hifi target as published by GRAS for the exact same HATS config Oratory has.

It will, probably, be closer to some reality than the EARS compensations. Although I do suspect miniDSP improved their compensation over time.
For 'investigative' reasons you could also plot the compensated measurements of one headphone of choice with the supplied compensations (from your raw, not converted, measurements and maybe obtain the SBAF calibration file as well.

See which of those compensated plots comes closest to that of Harman corrected measurements Oratory made and also include the 'converted and then Harman compensated' trace as well.

Then you'll know which compensation method to use that comes closest to the Harman target on Oratory's HATS.

Ah, it's ok, I'm very happy with the conversion curve I've done for K702 based on Oratory's GRAS measurement and my EARS measurement of the same unit (re EQ & listening result which is not the topic of this thread) - I overlaid it at 700Hz because there was some good correlation of "specific traits" in the response both at that point and in the high treble when lined up at that point, and it also so happened to be the point that REW chooses to match the 2 measurements, which I believe is based on preserving the same overall "sound power" (if I've not misused that term) so that there is equal areas of points both above & below the target before EQ.....so intuitively that seems like the best point to convert the measurements for SPL measuring as it's doing the least amount of "digital skewing" (made up term) to the measurement. I've found that the conversion curve I've worked out works well on all 3 units of K702 without creating some crazily shaped results in the treble (they look like GRAS measurements of K702).....for instance I experimented with some other compensations that I calculated based on some "other measurements" (too much detail so being vague) and it created some obviously wrongly shaped responses in the characteristic shape of the K702 treble (and also in the bass in some instances), so I think you know to an extent if you've got it really wrong, for instance I know that my conversion curve needs to be based on the Unit#1 that Oratory measured and not the Unit#2 that he measured.....the pad differences (even though stock) of the Unit#2 K702 makes it too different from a good representative K702 (or instead Oratory's measurements he gave me for that unit aren't as accurate) , and compensations based on that unit looked obviously wrong when applied to my Unit #1 & Unit #3. Averaging them also doesn't work as well, gotta be based on Unit #1 that he measured, it's a more representative K702 of units out there and I know (both objectively via measurements on my EARS and subjectively through listening that) the measurements he supplied me with for that Unit are significantly more accurate than those he gave me for Unit #2.....gotta be based on Unit #1. Anyway, that's enough detail about that, too much!

I'll measure again with an EARS to GRAS K702 conversion file and see what I come up with for the dB(Z) / dB(A) measurements of those tracks I posted (the graphs)....I'll also be re-reading your earlier detailed reply (https://www.audiosciencereview.com/...adphone-volume-with-umik-1.32594/post-1147466 ) to see if need to ask you anything about that or for incorporating into my listening loudness measuring of my tracks. I'll also have a look in the EARS manual re the bit where you say "referencing to SPL opposite the 0dBFS (max measured SPL)" - I think I remember noticing that there is a peak level that the mics can measure, that I saw when I was doing frequency sweeps in REW for distortion checking at high SPL's, and the ceiling is 120dB if I remember rightly....if that's what you're getting at. Won't be doing anything on this today, maybe at the weekend.

(The main thrust of what I'm wanting to achieve with my contributions in this thread is just to get the EARS setup most accurately that I can to measure my typical music listening levels.....and to compare that to what I theoretically calculated last year based on headphone impedance & sensitivity & listening tests).

 

[Robbo99999]

104dB Z peak is bloody loud indeed.
One can listen to that for a track or two but after that, every normal human will have the urge to tone it down. Especially if the tracks contain a lot of loud passages wit high frequency content.

You can let a real time analysis of the mic's input run in a second window during the SPL measurement.
Should be pretty obvious to see whether the bass or the 4KHz gets the highest peaks.

dB(A) is typically used when measuring noise in working environments in order to assess the hearing damage potential.
I just went with dB(Z) because any error I make will be on the safer side of things.

Thanks re your 2nd paragraph there, I'll see if I can do that in REW to help me determine for sure the cause of the 104dB peaks, almost 100% sure it's in the bass, but I'll check it out. Don't worry about the 104dB peak though, it's pretty much for sure in the bass at around 30Hz where my EQ boost sits, so 104dB in bass I don't think is dangerous or particularly loud - people can correct me if they know I'm wrong about that. Re dB(A), and dB(Z), I'll probably end up measuring both during my next experiments.

 

[solderdude]

Note that REW also needs to be calibrated (needs to know what input level is a specific SPL).
This requires an SPL meter (I have 2 of those) but perhaps the EARS is already calibrated and sends along data REW can do something with.

 

[Robbo99999]

Note that REW also needs to be calibrated (needs to know what input level is a specific SPL).
This requires an SPL meter (I have 2 of those) but perhaps the EARS is already calibrated and sends along data REW can do something with.

I think it knows this through collaboration between REW & EARS, it recognises it on a hardware level, and you can change the analog gain of the miniDSP via the DIP switches, and then REW can see which Gain you have set for your EARS and sets itself accordingly. So miniDSP have told REW the SPL level for each of the DIP switch Gain levels. That's my interpretation after looking in the manual, here are three screenshots from the manual that leads me to this conclusion:






In the calibration text files there's nothing but frequency and corresponding dBFS level: e.g -1, +1, -0.5, that kind of thing vs frequency, so REW is being told the SPL to relate to the input from the microphone by the hardware gain setting in the DIP switches of the EARS perhaps along with some settings in the REW software that miniDSP has communicated to them. So it must be calibrated - how proficiently and using what standard I don't know (& I don't know if there are standards). But for sure it is SPL calibrated in some sense!

 

[Robbo99999]

Actually, @solderdude , I've been running the DIP switches at +18dB gain as that was the advice in the manual - is that why I'm seeing around 18dB more peak output than my calculated levels?! 104dB peak -18dB = 86dB! Do I need to run the DIP switches at 0dB Gain to measure SPL accurately? I know that might sound like a stupid thing to ask when I come to think of it. (Maybe miniDSP didn't want people blowing up their headphones hence the advice to run +18dB gain on DIP switches and then set 84dB as 300Hz reference tone before doing the frequency sweep - they didn't want people to blow up their headphones, ha!?

EDIT: actually, disregard what I was saying above in this post.......... did a quick experiment with changing the DIP switch position from +18dB to 0dB, it didn't change the reported dB value at all, it just allowed for a higher SPL during measurement sweeps without clipping - so this means that the EARS are "intelligently" changing the gain and adjusting the reported dB internally - as the device isn't changing how it appears in REW (it still appears as "Input Device: EARS Gain 18dB"), so I'm assuming the changes aren't being calculated by REW unless it's doing it invisibly, as otherwise I would have expected the Input Device to change to "EARS Gain 0dB" visible in REW. So it looks like the DIP switch changes in Gain are being handled internally by the EARS and it changes it's internal calculations of SPL so that always the same dB is being reported for any given DIP Switch position. I suppose REW therefore only needs to be given one set of SPL reference data by miniDSP. Regardless of how it's being done, it shows that the dB listening levels I showed for my tracks are not due to me setting the DIP switch to the wrong position, as the reported SPL doesn't change with DIP gain switch position.

 

[Aerith Gainsborough]

Don't worry about the 104dB peak though, it's pretty much for sure in the bass at around 30Hz where my EQ boost sits, so 104dB in bass I don't think is dangerous or particularly loud - people can correct me if they know I'm wrong about that.

Don't know about your ears but mine actually hurt from the bass if I venture beyond 100ish dB for more than a few minutes. Pretty sure pain is a sign that the body does not like what you are doing.

That being said: from what I've seen so far, people here seem to be split on the matter. Some say it is dangerous, others claim it is not. Frankly, given how precious our hearing is, I'd rather err on the side of caution. At least during long listening sessions.

 

[Robbo99999]

Did an additional experiment to feed 1V to my K702 (Unit#3) headphone during a 1kHz tone, and it came back with an average of 102.2dB between the two channels (Unity Gain on headphone amp, max volume knob position, 100% Windows Volume, 2V DAC, -6dBFS tone):


Crazily, the sensitivity that Oratory measured for my K702, albeit a different unit (Unit#1), came back with a 102.2dB sensitivity at 1kHz too! I would have put my Unit#1 headphone on there to do the same test, but it doesn't have pads on it at the moment. Anyway, I think that goes towards proving that my miniDSP EARS is measuring SPL accurately. What do you think @solderdude , was that a good experiment to check the validity of reported SPL?

EDIT: and to help throw away fears that my Unit#1 & Unit#3 are not comparable at 1kHz, they have the same dB level at 1kHz almost, from an earlier comparison I did of all 3 units of my K702 with the same pad swapped between all of them:

 

[Robbo99999]

Don't know about your ears but mine actually hurt from the bass if I venture beyond 100ish dB for more than a few minutes. Pretty sure pain is a sign that the body does not like what you are doing.

That being said: from what I've seen so far, people here seem to be split on the matter. Some say it is dangerous, others claim it is not. Frankly, given how precious our hearing is, I'd rather err on the side of caution. At least during long listening sessions.

But you probably can't seperate the fact that 104dB in the bass on whatever headphone you're using does not equate to the same frequency response in the rest of the frequency range as that of my headphone. And then you've also got the significant variable of different tracks. You might not actually be sensitive to the bass, it might be something else in your headphone or in your track in other parts of the frequency range that makes you mistake the above 100dB bass as being the problem. It also depends on how accurate you measurement of your headphone is.....am I right in thinking you put a UMIK through a CD and measured it that way?

You'd really have to have a comparable/accurate measurement method, then you'd have to have a headphone with the same overall frequency response, and then you'd have to listen to the same track as me, which you'd also have measured your headphones whilst playing.....that would have to be all the same to say that above 100dB bass for you is a problem, or vice-versa for me - we have to have everything controlled between us in order to make those same comparisons.

 

[solderdude]

You can safely say the reported SPL at 1kHz is correct. Calibration SPL wise seems fine.
This is RMS 1V though. When looking at peak values it might show 3dB higher...
I measured my K702 at around 101dB/V but this is a wide band eye-balled measurement and not exact at 1 frequency.

 

[Robbo99999]

You can safely say the reported SPL at 1kHz is correct. Calibration SPL wise seems fine.
This is RMS 1V though. When looking at peak values it might show 3dB higher...
I measured my K702 at around 101dB/V but this is a wide band eye-balled measurement and not exact at 1 frequency.

Cool, that's good to know. I did an "EDIT" to my post whilst you were typing yours, I put in the frequency response graph of all 3 units of K702 whilst having the same pad swapped between all of them - the Unit #1 that Oratory measured & my Unit#3 that I did this SPL experiment are essentially the same SPL at 1kHz from that frequency response graph.

 

[Aerith Gainsborough]

But you probably can't seperate the fact

I can.
Because I have a lot of tracks that basically only contain sub bass. If the track contains higher frequencies, there is no way in hell I could them turn up that far and find them enjoyable. Even with speakers, 100dB in the higher frequencies is "turn that down ASAP" to me. There is a reason why I wear headphones when out in the civilization. Despite not having the best ears, I'm oddly sensitive to noise.

It also depends on how accurate you measurement of your headphone is.....am I right in thinking you put a UMIK through a CD and measured it that way?

I did the theoretical calculations a page back. Results were within 1dB of my measurements. Keep in mind, I only looked at the bass area and used real music (very bass heavy) as a measurement signal.
Why?

• Because in most music, that's where most of the energy is. Easily seen in any spectral analysis. It doesn't really matter how the can behaves in the higher frequencies, when the source signal is already a lot quieter. At least when you are only concerned about maximum volume.
• The UMIK measurement cannot measure anything past 1KHz reliably. The main reason for my measurements were loudness calibration and designing my EQ to mimic the 35Ω output impedance of my old Soundcard, which gave my Clear a warmer signature that I've come to love. The RME just sounded to cold to my taste w/o EQ.
• Keep in mind, we are always talking Z weighted peaks here. The average loudness levels are far below that. Unless you listen to brickwalled trashmetal, I guess. :'D

we have to have everything controlled between us in order to make those same comparisons.

I know people here love to do that but in this case, I don't think we need to overcomplicate things. Your ears are different than mine, you are a different human being. We will never hear the same way. Even if math / measurements say we should.

Just how anatomy works.

 

[Robbo99999]

I can.
Because I have a lot of tracks that basically only contain sub bass. If the track contains higher frequencies, there is no way in hell I could them turn up that far and find them enjoyable. Even with speakers, 100dB in the higher frequencies is "turn that down ASAP" to me. There is a reason why I wear headphones when out in the civilization. Despite not having the best ears, I'm oddly sensitive to noise.


I did the theoretical calculations a page back. Results were within 1dB of my measurements. Keep in mind, I only looked at the bass area and used real music (very bass heavy) as a measurement signal.
Why?

• Because in most music, that's where most of the energy is. Easily seen in any spectral analysis. It doesn't really matter how the can behaves in the higher frequencies, when the source signal is already a lot quieter. At least when you are only concerned about maximum volume.
• The UMIK measurement cannot measure anything past 1KHz reliably. The main reason for my measurements were loudness calibration and designing my EQ to mimic the 35Ω output impedance of my old Soundcard, which gave my Clear a warmer signature that I've come to love. The RME just sounded to cold to my taste w/o EQ.
• Keep in mind, we are always talking Z weighted peaks here. The average loudness levels are far below that. Unless you listen to brickwalled trashmetal, I guess. :'D

I know people here love to do that but in this case, I don't think we need to overcomplicate things. Your ears are different than mine, you are a different human being. We will never hear the same way. Even if math / measurements say we should.

Just how anatomy works.

Ok, that's good that your theoretical calculations closely matched your measurements, which gives more credence to them. If your tracks you chose were really bass heavy without really anything else going on, then I think I agree that you can say it's the bass that you're sensitive to in your experiments. What kind of frequency response have you got going on in your headphones from 200Hz down to 20Hz, have you got a graph of the rough shape? I only ask because I'm wondering if that could be a factor in "sensitivity" to hearing. I mean I'm using Harman Bass, I wondered if yours was much different? Have you managed to track down to what bass frequency you might be most sensitive to in the tracks that tickle your ears when it's too loud?

 

[Aerith Gainsborough]

What kind of frequency response have you got going on in your headphones from 200Hz down to 20Hz, have you got a graph of the rough shape?

I still have the measurements, yes. A few of them at least.

Green + Red = L + R side of the Clear with the EQ set to off. As you can see: the values vary wildly past 4KHz mark. I remember them, being super sensitive to the position of the mic, so I chose to just ignore the high end. Bass area was very stable and easily repeated. The UMIK has ben ran with the 90° upright position calibration file.

Blue is the response with my EQ, very close to the response my Soundcard generated.
EQ is very simple too (RME settings): low shelf +3.5dB @ 120Hz with a Q of 0.7.

Have you managed to track down to what bass frequency you might be most sensitive to in the tracks that tickle your ears when it's too loud?

I never gave it that much thought but I'd wager that it's around 50-60Hz. Which would pretty much be the loudest frequency in the bass as well. Lots of movie effects love 50-60Hz, that's where I noticed it first because I ... well... it was way beyond 100dB when I got the Clear. Close to the clipping point, so 110ish? Me thinks. Since then I did the responsible thing and limited the "boom boom" a little. :'D

 

[Robbo99999]

I still have the measurements, yes. A few of them at least.

View attachment 198755
Green + Red = L + R side of the Clear with the EQ set to off. As you can see: the values vary wildly past 4KHz mark. I remember them, being super sensitive to the position of the mic, so I chose to just ignore the high end. Bass area was very stable and easily repeated. The UMIK has ben ran with the 90° upright position calibration file.

Blue is the response with my EQ, very close to the response my Soundcard generated.
EQ is very simple too (RME settings): low shelf +3.5dB @ 120Hz with a Q of 0.7.


I never gave it that much thought but I'd wager that it's around 50-60Hz. Which would pretty much be the loudest frequency in the bass as well. Lots of movie effects love 50-60Hz, that's where I noticed it first because I ... well... it was way beyond 100dB when I got the Clear. Close to the clipping point, so 110ish? Me thinks. Since then I did the responsible thing and limited the "boom boom" a little. :'D

I don't know how your rigged up U-MIK responds in the bass area vs the GRAS, but I have noticed that you've got a bit of a "hump" in the 70-100Hz area which is often thought to be the area where "slam" occurs, and maybe that's a bit fatiguing on your ears. I circled the area in red on your graph:


It might be interesting for you to try manipulating the shape of that region to look more like the Harman Curve continual upslope in the bass, the circled red area in the following graph (which I tried to scale to be similar to the scale you used, so that they're more visually comparable):

If it's easy for you to reshape the bass then it might be worth a try, maybe you just need to reduce the slam a bit. Maybe just flatten the red circled area in the first pic in this post a bit (your pic), so that it blends more into the rise of the bass.