What do you prefer between aural compensator and loudness? AFAIK, the loudness operates by boosting low and high frequencies, while the aural compensator will reduce midrange with some attenuation on overall level at the same time. Is it a good a idea to have an aural compensator and a loudness together in the amplifier?
Neither the aural compensator nor the traditional loudness contours are actually correct.
There are two primary problems. One is the method that the original curve data was collected, and the other is that human hearing response changes dynamically with SPL.
Regarding the first problem, Fletcher and Munson collected their equal loudness curve data using pure tones in an anechoic space, and with equipment of the day (1933) that had very limited accuracy. Their resulting equal loudness contours are therefore incorrect. The Aural Compensator has similar issues, but I didn't research its background to know the specific causes.
The topic of human hearing response vs SPL has been explored many times over the decades. On widely accepted paper by S. S. Stevens, "Perceived Level of Noise by Mark VII and Decibels (E)," J. Acoust. Soc. Am., vol. 51, pp. 575-601 (1972) contains data that was widely accepted pre ISO226, which now also has a curve family based on different SPL levels. There is some disagreement with the above and Fletcher-Munson, most notably at high SPL. And further, there was a 2015 paper ("A Method of Equal Loudness Compensation for Uncalibrated Listening Systems", Oliver Hawker and Yonghao Wang, AES, October 2015) that was concerned with the fact that the ISO curves do not cover the full frequency range of hearing, and the paper produced extended curves plus minor corrections to the ISO226 curves in general.
The traditional form of loudness compensation is to apply a correction curve that is variable dependent on the position of a volume control. However, without knowing what the specific SPL is at very least, the correction will not be accurate. Accuracy involves more than just knowing the total system gain and SPL (which will return at least some general data for a fixed correction curve) it involves knowing the actual moment-to-moment SPL in the room, and the application of a dynamic correction. To further complicate matters, the playback SPL is only part of the story. The rest would be the knowledge of the specific SPL at the original mix and mastering position, which for the music industry is not as well standardized as in the film industry. The net correction should be based on the differential between listening SPL an final mix SPL, as calculated on a dynamic basis. Sounds tough, right? And it is, but that was actually accomplished by a process found on AVRs called Audyssey Dynamic EQ, which the correction is level-differential based, given the known calibration of film dub stages to an industry standard reference level.
More than you wanted to know, I have no doubt. If you want to apply a non-dynamic correction, the correction should probably be variable, and set by the listener to preference rather than just fixed or predicted without any SPL consideration. If you just stuck with the ISO 226 data, or the Stevens data, you'd be far closer to right than Fletcher-Munson or the Aural Compensator would get you.
To help out, here's a look at ISO226 curves and the Hawker-Wang "corrections":
And if you want to do something much simpler, you can borrow an idea from the Apt-Holman preamp, which used a bass tone control with curves modeled after S. Stevens. It actually works quite well, and beats Fletcher-Munson by a lot.
The correction curves (from one of Holman's papers):
And the actual tone control:
The correction curves (from one of Holman's papers):
And the actual tone control:
And knowing a bit about hospital audiology testing you could get a graph of the tonal resonance response over the audio range so you can set your own ears to a near "flat " response by tone correction .
Also taking into account the use of headphones to alter the response by additional cranial sound vibration , which is actually two areas -one the ear via the auditory nerve ending and two the transmissional response via the skull structure .
Also taking into account the use of headphones to alter the response by additional cranial sound vibration , which is actually two areas -one the ear via the auditory nerve ending and two the transmissional response via the skull structure .
No, that won't work. Audiology - hearing acuity - tests for minimum audible threshold for a limited set of tones. It's a test of the minimum level of hearing at just a few spot frequencies, and has little to do with perceived loudness at various (clearly audible) SPLs.
All of which further invalidates the method. Sorry.
"Minimum audio threshold ---uh no .
Minimum to what audio response ?---- a quiet room /a busy railway station ?
If the patient left after having his hearing aids adjusted to a quiet room then within a week they would be back in again and cranial audio transmission is no myth its a scientific fact.
Minimum to what audio response ?---- a quiet room /a busy railway station ?
If the patient left after having his hearing aids adjusted to a quiet room then within a week they would be back in again and cranial audio transmission is no myth its a scientific fact.
I've actually dug into this quite a bit, with a mother who has had hearing aids for a couple of decades now.
A hearing profile is a test of the threshold of hearing, a point below which a tone cannot be detected. The tones used in the typical hearing test don't cover the full spectrum, and only from memory now, and not taking even a second to google, I think 100Hz to 8kHz, or at least some range just about that limited. The patient is in a quiet room with headphones that seal fairly well. The audiologist presents tones in each ear, the patient points to or indicates which ear the tone is heard in. If no tone is heard, the audiologist increases the level until it is, and records that level. The result is a chart of sorts that profiles the minimum hearing threshold vs frequency of each ear. That curve is used to produce a compensation curve in the hearing aid, if required. The compensation curve is not the inverse of the hearing threshold, it just covers the more important range, and modern hearing aids also have dynamics modification. For example, my mother's mid-band hearing is -80dB below "normal". Full compensation produces excessive SPL when higher acoustic signals are present, so there's a compressor in there that reduces the dynamic range. And there's more DSP going on that that, modern devices attempt to compensate for losses in the "cocktail party" ability to differentiate voices spatially. Frankly, they're still over-priced devices at $4K, but they do accomplish a lot. There are internet groups dedicated to hacking hearing aids and self-adjusting their parameters. I hope I don't get to that point, but I'm on the way.
A hearing test is not a test of the perception of relative loudness, and the resulting data is not pertinent to loudness compensation.
Well, none of that actually happens because:
1. The signal entering the hearing aid microphone is purely analog. The only effective means to transduce acoustic waves to electrical signals is purely analog.
2. The signal leaving the hearing aid is purely analog. So far, there's no digital interface to the human brain.
3. The design, adjustment and performance of hearing aids is based solely on user experience and feedback. If the user experience is worse, then it's not a successful product, and fails in the market. Hearing aid patients literally do vote with their wallet.
4. All successful modern hearing aids are based on digital technology because you can't actually build in a personalized correction curve with any degree of precision with the analog technology into a device that fits inside the hear canal, and there isn't any way to perform dynamics processing effectively in that same package. All you can get out of an analog hearing aid is gain and a little very basic response shaping. All of those new digital features, and many more, are perceived by users as improvements, and are very sought-after in a rather competitive marketplace. Many assistive hearing devices include Bluetooth capability and interface rather nicely with modern smart phones. There is no way to pull that off in the analog world.
5. The design, calibration and operation of modern hearing aids is based on actual science and research, not mysticism and baloney.
I'm sorry there seems to be a gap between the real world and how you perceive a small slice of it.
A long time ago I made a simple loudness correction circuit based on the Stevens data. Much later, I saw the ISO226:2003 curves on Wikipedia and had to conclude that my circuit was undercompensating quite a bit. The equal loudness contours changed substantially in 2003, see Equal-loudness contour - Wikipedia
One thing that has remained the same is that there is not that much need for dynamic correction. With loudness control, you try to correct for the difference between the equal loudness contours at the level the recording was meant for (the level during live recording or the level during mixing) and the level you play it back at. When you subtract equal loudness contours from each other, the result depends mainly on the level difference: 100 phon minus 80 phon looks similar to 80 phon minus 60 phon.
What also remains the same, is that the very low frequencies have to be reduced by about 10 dB when you reduce the volume at mid and high frequencies by 20 dB. I used a quadruple ganged potmeter for stereo volume control (or actually two stereo fader potentiometers that I coupled in a very primitive manner) and let the unfiltered signal pass through two sections and a correction signal that only consisted of the low frequencies through only one section, so each 20 dB volume reduction led to a 10 dB bass reduction. Another stereo potmeter was used to adjust the level of the correction signal, so you could adjust the amount of loudness compensation - music that is meant to be played loud requires more compensation than music that is meant to be soft.
One thing that has remained the same is that there is not that much need for dynamic correction. With loudness control, you try to correct for the difference between the equal loudness contours at the level the recording was meant for (the level during live recording or the level during mixing) and the level you play it back at. When you subtract equal loudness contours from each other, the result depends mainly on the level difference: 100 phon minus 80 phon looks similar to 80 phon minus 60 phon.
What also remains the same, is that the very low frequencies have to be reduced by about 10 dB when you reduce the volume at mid and high frequencies by 20 dB. I used a quadruple ganged potmeter for stereo volume control (or actually two stereo fader potentiometers that I coupled in a very primitive manner) and let the unfiltered signal pass through two sections and a correction signal that only consisted of the low frequencies through only one section, so each 20 dB volume reduction led to a 10 dB bass reduction. Another stereo potmeter was used to adjust the level of the correction signal, so you could adjust the amount of loudness compensation - music that is meant to be played loud requires more compensation than music that is meant to be soft.
Hearing aid dynamic range compressors were used long before hearing aids went digital. I know, because I graduated on one. There was a 2:1 dynamic range compressor with adjustable threshold and a limiter (infinite:1 compressor) meant to prevent excessively loud output levels. The compressor was also used to reduce the dynamic range requirements of the on-chip adjustable continuous-time analogue filters. Of course you can do more with digital processing.
Refer to ISO226 curve, I think an aural compensator together with a fixed loudness may be the final answer. Do you agree? The reason is the aural compensator would provide a Fletcher-Munson-liked curve, and a fixed loudness will help converting it to ISO226. Another doubt is that I heard an audio engineer once told that a correct loudness circuit should boost only bass frequency, no high frequency boosting.
There's no need for Fletcher-Munson. It's just wrong. ISO226 is probably as close as you need to get with possibly a bit of tweak because you don't know what level at which the mix was done.
A reference to that concept is found in "Loudness Compensation: It's Use and Abuse", by T. Holman and F. Kampmann, J-AES, July/August 1978. Their conclusion was that a "loudness-derived bass tone control" was the best compromise for the time, and that's what appeared in Holman's preamp. They further concluded that the operation of the control had to involve listener subjective judgement because of the lack of specific SPL data. When that data became available several decades later, the concept of a loudness-derived, specific SPL based compensator became realizable, and found its way into Audyssey Dynamic EQ, a system for which Holman was the Chief Scientist.
A notorious breakthrough was Killion's K-amp:
https://www.etymotic.com/media/publications/erl-0067-1993.pdf
Audiometric threshold is about as described: softest tone heard in "silence". For good-hearing folks it is hard to find sufficient quiet; fortunately someone like me does not hear background 50dB up from what I heard as a youth. Oh, and in US practice the tones are ISO centers, so the lowest is 125 if not 250Hz.
The curve of smallest audible sound is NOT duplicated at medium loud levels. Nerve damage acts like an expander or "noise gate". My ski-slope hearing actually flattens-out in 80dB SPL field. This is common but far from universal.
BTW: I have an audiometer, a real-ear probe (and aid tester), a stack of books and notes, and my Noahlink programmer.
Attachments
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.