Notes by ff123
I purchased Digital Recording's Audio CD, and performed a self-administered hearing test. I did this test at night when there is very little background noise (almost none from my computer, which runs without fans). The test's accuracy is limited by the difference in headphones (reference headphones are AKG-K270's) and by the difference in background noise, but it's probably close enough to give a general idea of what my hearing is like.
I used two different headphone sets: Grado SR325's and Beyerdynamic DT990 Pro's. I performed the test on the same night. The biggest differences are the increased bass of the Beyers and the 10 dB "shelf" that Beyers have from about 6 to 12 kHz. This is consistent with my subjective opinions about the sound of the two headphone sets. The Beyer's sound "darker" than the Grado's but have better bass (i.e., the Beyer's sound more like my Nelson Reed 8-04B loudspeakers in terms of bass). Although I love the sound and feel of good bass, I prefer the sound of the Grado's overall.
High frequency response in my right ear is worse than in my left ear. I have long suspected that I have suffered some hearing loss in my right ear as the result of a firecracker going off near my right ear when I was a child. This pretty much confirms my suspicions. At 14 kHz and above, I suffer a great deal in comparison with the ideal.
Note: The audiograms above were taken a year apart. In between, I replaced my Soundblaster AWE 64 Gold soundcard with an M-Audio Audiophile 2496, and so lost my volume reference. I scaled the raw data for the 12/31/01 measurement by 10 dB, adjusting the curve downward, so that the data above 10 kHz remained about the same (assumes my high frequency hearing has not degraded). Interestingly, the bass portion of the curve seems to have shifted (more bass). I purchased my Grado's on Dec 9, 2000.
From the CD insert:
The gray "reference" line indicates the "ideal" threshold of hearing for very young subjects (this is preliminary data obtained with the AUDIO-CD and the AKG-K270 headphones in an Audio Booth, where the background noise is very low). The "ideal" thresholds are the same for both ears.
There are individual variations of hearing thresholds for each of the frequencies in the order of +/- 10 dB. An upward deviation greater than 10 dB from the "ideal" threshold may be an indication of hearing loss.
The following audiogram was measured on one of my co-workers, who is 37 years old.
(2-17-01) I had the chance to chat with Adam (Filburt) on DALnet recently. Adam is the model for MPEGPlus's -ltq fil setting, which is an extraordinarily sensitive setting for threshold in quiet. He had some numbers he had measured for himself, and he allowed me to publish them. Adam didn't use Digital Recording's Audio-CD to perform his measurements; I have drawn their reference curve as comparison.
Note that I had to guess about where to place his measured points with respect to the reference. The Digital Recording CD does not show the peak in sensitivity at 3.5 kHz -- I must assume that this is because the calibration tones are such that the 3.5 kHz area is actually played louder than the threshold of hearing. People are typically 10 dB more sensitive in the 3.5 kHz region than they are at 1kHz. Therefore, I adjusted Adam's measurements to be 10 dB more sensitive than the Digital Recording curve at 3.5 kHz. This process does assume that Adam hears no better or worse than the Digital Recording test group at 3.5 kHz, and that the test group ATH curve would have had a shape similar to the ISO curve. However, since the -ltq fil profile is said to be extraordinarily sensitive, I suspect the actual curve resides much lower than shown.
The figure below comes from Kevin Donahue's web page showing the frequency response of the ear. It shows a typical equal-loudness curve at the threshold of hearing [ISO R226 (1961)]. Each frequency point on the plot represents the gain required for tones at that frequency (relative to 20 micro-Newtons/meter-squared) so that each tone is heard at the same volume level. It shows the peak in sensitivity at 3.5 kHz.
Adam's test procedure: At each frequency, Adam generated pure tones in Cool Edit, with -120 dB being the lowest volume (32-bit domain) and set the amplitude to gradually rise over a 20 second period. He noted where he could just start to hear them. Measurements are averaged over three tests. The actual numbers he obtained are tabulated below (I transformed them to plot against the Digital Recordings CD).
| Frequency (KHz) | Threshold (dB) |
| 3 | -102 |
| 3.5 | -107.3 |
| 4 | -106 |
| 18 | -62 |
| 20 | -41 |
| 21 | -30 |
Adam's setup: Adam used Sony MDR-ED268LP earbuds with a Diamond Monster 300MX (18-bit D/A). His computer fans were unplugged, and with the isolation provided by the earbuds and the help of a very quiet room, he could not hear any computer or other background noise during the test. Adam says the card was nearly noiseless (S/N ratio of 108 dB). Adam couldn't say what the absolute SPL's were (as measured by a sound pressure level meter).
Other comments made by Adam: "frequencies up to around 24khz I seem to notice pretty well. And music sounds noticeably different to me when it has 21khz bandwidth as opposed to 19. Some of this stuff is tested on professional hearing test equipment. I think my hearing was actually worse back then, quite a bit. I listened to stuff a lot louder back then; it kinda reversed after I layed off for a period of 4 months or so.
[in a chat] mp3 at anything under 256 still sounds odd to me -- high frequency distortion. Hell i got a whole album the other day: lame 3.88 HQ 192kbps -- drives me berzerk. I usually wont download mp3 under 256kbps unless i absolutely have to. mp3enc is absolutely the only thing i'd use under 256kbps. Lame at 224 is okay, but it still has its oddities.