Hearing limits experiments

Absolute threshold of hearing

I hope you find it entertaining. It's not really meant to be science, rather me sharing an experiment I did.

Now this is the penultimate test of ADHD and 'tism hyperfocus. Here's how to do it

1. Find a particularly good day. You have to be in the mood for staying up hours past your bedtime in a situation that is both highly stressful and totally boring
2. Also it has to be a day of particularly low noise, both external and internal. On some days you won't be able to concentrate, won't be able to hear the quietest sounds simply due to psychological reasons. Likewise if someone near you is throwing a party - forget about it. It's much easier if you live in a quiet area like I do.
3. (If need be take stimulants or GABAergics to take the edge off. Don't overdo it, though.)

Now the preparation:

1. Prepare a rig with multiple ways of measuring the SPL where you sit. Measure the actual voltage going out of the amp, too. Measure the frequency response close to where you sit, while you are sitting at the listening position. Average many measurements to get an idea of the FR you're dealing with and the limitations of your ability to measure the threshold of hearing.
2. Find a way to blind test what you're hearing so you don't fool yourself.
3. Prepare something to drink and some snacks since this will take a while.

As for the actual testing:

1. After setting up everything wait for it to get dark and for everything to quiet down outdoors.
2. Make sure your Laptop is charged. Make sure to eliminate as many sources of noise and distractions as possible. A phone charger is still plugged in that makes a whine? Unplug that shit. Set your phone to silent, etc.
3. Prepare yourself. Put in some earplugs while waiting for ten to twenty minutes. There's not really any dark adaptation for the ears, this is for mental preparation.
4. Prepare REW: Set the generator to a narrowband noise. Since we're measuring with speakers in a room we can't use sine waves.
5. When you're ready start with a very low SPL. Slowly use the trackpad of your laptop to silently increase the SPL until you hear something. Back off a little until you no longer notice anything.
6. Export the file and blind ABX test it against digital silence. If you fail, increase by 2dB. If you can still hear it, decrease by 1dB.
7. Do the same thing for as many frequencies as you want.

This is more or less what I did 2019 for the graph I casually posted somewhere. It took multiple days and it was awful. Now after 5 years I decided to subject myself to the same agony again. This time I used my smaller Hathor speakers instead of the large Amon OB, which I think partially explains some differences.

Here's what I got:


A couple comments:

• I don't actually believe my hearing is exceptional for someone of my age who generally takes care of his hearing. I expected it to get a little worse from the hours I spent listening to my speakers at high SPL, but this seems to not be the case.
• In other words I really don't mean to brag, that is not my intention.
• Rather I think it's a limitation with the studies. You won't find very many people as dedicated to it and willing to torture themselves for days only to get the best possible result.

• I specifically used IEMs with low distortion for the lower frequencies to make sure I was hearing only the fundamental. To be specific I used both the S12 and the RE272 IEMs. Also to avoid the room modes.
• MAP does seem to be 6dB worse than MAF as can be seen from the IEM graph being 6dB higher than the speaker graph. However IEMs also make it harder to hear quiet sounds due to the ear canal being blocked.
• I tried to compensate for the speaker FR this time around, which I didn't last time. So the 2 and 3kHz being lower makes senes given that I measured the old graph with the Voxativs. Or maybe it is some hearing loss.
  I'm not sure about the difference in the 200Hz to 1kHz region. I think the more recent graph is likely more accurate.
• I find it reassuring that I got similar results as last time indicating that I likely didn't screw up somewhere.

Notable features:

• I find 20Hz very very hard to hear. It takes about 70dB SPL for it to become audible to me.
• I can't hear pitch past 18kHz or so, but I do still hear a tone being played this high up at higher volume levels. This seems to hold true all the way up to 40kHz. I only tested up to 40kHz.
• So given the two above I'd say my hearing range is 30Hz to 18kHz.
• With a more stringent 30dB SPL limit it's only 65Hz to 18kHz.
• Between 400Hz and 10kHz it seems to be lower than 0dB.
• For the HF tests I used my microphone to measure the SPL. There did seem to be a slight bit of subharmonics at 35 and 40kHz, but it didn't seem to change the results.
  
• The quietest audible sound seems to be -15dB SPL at 3kHz.
• Between 1kHz and 5kHz the hearing threshold seems to be lower than -10dB.

I am in the process of measuring the minimum audible harmonic distortion and at some point I want to verify the equal loudness contours. However both will take a while if I decide to even complete it.

 

Very good.
A couple of notes.
ISO Contours of Equal Loudness are an average over the population tested. Individuals vary from the average.
0 Phon contour does go below 0 dB SPL from approximately 1800 Hz to 5500 Hz.
Sugar, aspirin and other numerous other substances will have an effect on thresholds.
Achieving very quiet periods in one's ambient environment is a bitch, especially for those in urban and suburban spaces.

 

Is Serious this crazy? This is crazy serious.

This is crazy, Serious!

 

Too right. And here's me saying that blind tests, although I totally believe in the concept and reasoning, are, in practice, just too much hard work.

Kudos to Serious.

 

I’m still trying to understand how best to interpret that graph (epic work @Serious ). What are those curves relative to? Background room noise level of X? How can it be negative in the midrange?

I’m sure this is me being dumb, but some clarity might help this slowski understand better.

I’ve thought of doing something somewhat similar using my rather less impressive collection (test records of frequencies and my hand held meter).

I do have the advantage of living very rural. When the fridge isn’t running at opposite end of the house, I measure 32dBA residual background noise (day) and 30-31 evenings.

 

No, it is indeed dB SPL, so the values are relative to 20uPa. It's negative simply because in the midrange the SPL I could hear was significantly lower than the reference value of 20uPa. Since it's a logarithmic scale that results in a negative number.
The actual sound pressure level is of course still larger than 0.


What's the lower SPL limit of your meter? 30dB is a very typical figure of the noise limit on those meters. If you were to record the sound the meter picks up you would get a very distracting amount of noise. I should demonstrate that using the microphones I have here.

The way I did it was to set a digital reference SPL at a level I could reasonably measure without noise affecting the measurements. Typically 70dB will do. Using REW with the speaker cal noise stimulus works well here.
Then I lowered it digitally to get to the low SPL I needed for my test.

There are three conflicting goals for a SPL meter:

• Linear frequency response.
  This is the main reason for using 6mm electret capsules. There are larger capsules with less noise, but in my experience their frequency response is quite lackluster.
• Omnidirectional behavior.
  The smaller the capsule, the more omnidirectional it is. Again, many use 6mm electret capsules. Most use a 1/2" diameter rod the capsule is inserted into, which hurts the directivity. However this makes it possible to calibrate them against professional sound calibrators.
• Wide dynamic range.
  Especially the maximum SPL is important here. There are specialized low noise capsules, but even professional SPL meters don't use those since they have a limited maximum SPL. 130dB is a typical figure of the clipping point of those inexpensive SPL meters. Again, professional ones can exceed that figure.

Dedicated low noise SPL meters are still very expensive. That may change in the future, with lower noise ECM and MEMS microphone available nowadays (or the optical one), but I haven't found one, yet.


For those who are cheap like me and don't have the money to spend on professional measurement equipment, here are some ideas:

• I tried to improve the performance of my SPL meter, but I realized there was a lower limit of 28.2dB hard coded into it. By increasing the gain I could only eek out another 3-4dB out of it, at the cost of a lower maximum SPL. In the end I decided against it. I simply swapped the capsule for a less noisy, more linear one and called it done. (6mm PUI 2370HD)

I have two cheap low noise omni microphones with a reasonably flat FR, which I calibrated against different reference microphones:

• One is a MM-1 with its capsule swapped for an 8mm PUI 3027HD. This way I could get a noise floor of 17.5dBA. If it wasn't for the distortion of 2 pin electret capsules this would be a very good choice for a general purpose measurement microphone. It's what I use for most of my FR measurements.
  However those capsules have significant unit to unit variance. I had to cherry pick mine. Also in this circuit there's significant rolloff from 10kHz on. With a more optimized circuit you could probably improve on both aspects, but my experience is that the
• The other one is a Behringer B5. I now have three units, two of which actually seem to exceed their specs. I get about 15dBA of noise, both with the omni and the cardioid capsules.
  The advantage is you get the low distortion of a true condenser microphone. However its larger size will limit its usefulness for speaker measurements.

The two lowest noise omnidirectional small diaphragm condenser microphones I have found are the Schoeps MK2 and the Sennheiesr MKH 8020 with a specified noise level of 10dBA. That puts them in the range of most 1" measurement microphones!
Only dedicated low noise microphones like the ACO7020 @atomicbob used are much better than that.