Can You Hear the difference between 16 and 24 bit audio - take the test

[Craig Maier]

Hi folks,

Try this - - - it's a lot of fun.

There has been a longstanding debate concerning the usefulness of 24 bit audio. Do you want to determine if you need 24 bits for your audio signals? Take the following test. I did and my wifes niece of 16 years in age both took it with the same results. Run this test in a fairly low ambient noise environment.

1. Using the Make Waves generator, create a Sine Wave at 0 db, 1000 Hz 10 seconds with 44.1 kHz Sample Rate and 16 bits resolution.

2. Play that signal and adjust your sound system to a level just below at which it'a painful (called the "threshold of pain"). (Raise the volume control slowly to get there).

3. Stop the play and give yourself a 15 minute rest.

4. Using the Make Waves Generator with all settings the same as before, except set it for - 70 dB

5. Play that wavefile

6. Note that you (probably can hear that)

7. Using the Make Waves Generator with all the settings the same as before, except set it for - 80 dB. (keep in mind that you are not even using the entire dynamic range of the 16 bits available, which would be -96 dB).

8. Play that wavefile. Can you hear it? Neither my wifes niece nor I can hear it.

9. If you can hear it, repeat the test, except set the amplitude for -90 dB. Remember that -90 dB is not even using the full extent of the 16 bit system which is capabole of -96 dB. -90 dB represents a 15 bit system.

10. Can you hear that??? I doubt it, but would love to hear from others.

11. Now, delete all the temp files and then close the program.

12. Re-open it and set the make waves generator for a Sine Wave at -120 db, 1000 Hz 10 seconds with 44.1 kHz Sample Rate and 24 bits resolution..

13. Play the wavefile. You do not hear it. Turn the volume all the way up on your sound system - - - still you do not hear it, right - - - of course you don't - - - right!

So, now, I need to be taught about why we need a 24 bit audio system that has a range down to -144 dB considering that we can not even hear 15 bits worth of dynamic range.



Get the whole family involved & have fun, folks.

Craig

[Dan McDonald]

I did a 'hearing the unhearable' set of files a few months back.
I took a 24-bit, 96kHz music file and ran the high pass filter at 20000 Hz.
Then I gain-normalized the file and did successive speed change to bring the frequencies down until there was something I could hear.
Eventually, I got it all gain-normalized at about 1/8 the speed so I could hear whatever was at 96kHz.
It was just a bad hiss, which I assume does produce a minor difference in spl on the eardrums when it's part of a file. I ran the same procedure with 16/44.1 file of the same material, and the result was less hiss for 16-bit (since there was nothing up that high).

Anyway, fun stuff. I wonder what people are saying when they claim they can hear the difference. I'm guessing they picking up on the hiss in terms of sound pressure, but I've never met anyone who actually could tell it in an ABX test.

[Marc Hildebrant]

Easy answer....24 bits maintains the original "feelings" in the song.

Unless you have Golden Ears like me (I'm Special), you can't hear these bits of the song.......


All said in jest.

Marc

[Craig Maier]

Dan,

I think your test is a little different than my test; You were also listening to the effect of higher sampling rate. That could be something different in terms of results. I was just looking for the bit depth effect to confirm my long held suspicion that 16 bits is beyond the limit of my hearing dynamic range. As a matter of fact, it would appear that a 14 bit system would work fine for me.

Craig

[Dan McDonald]

Yes, I was looking mainly at sample rate, but also trying t see what was in the 24-bit/96 that was not in 16/44.1, so sort f mixing it all together since most people who claim to hear a difference claim a difference between16/44 and24/96.

[Craig Maier]

I would say that if an average 16 year old female child could not hear the difference in bit depths, than us male adults do not have a prayer of detecting the difference. Women have better hearing than men (in gneral) and children have better hearing than adults. Neither my wifes niece nor I could detect anything below -80 dB when referenced to a signal "at the threshold of pain" or 0 dB @ 1 kHz.

[Marc Hildebrant]

ENOB or effective number of bits equation relations the A to D real bits to the signal to noise ratio. For many 16 bit A to D converters the real number is about 14-15.

Even 14 bits is plenty.


Marc

[Craig Maier]

Yup - 14 real bits is certainly good enough for me and my wifes niece.

Craig

[mikeebb]

Is it coincidence that this thread just recently started, then Tom's Hardware posted an article the other day where a guy did a mostly blind subjective comparison of several DACs (including one found in a lot of better-quality motherboards). Using high-end headphones & "experienced" listeners he concluded that there was no audible difference between them on music playback. Also, in several cases he had to downsample very high-res to 44.1 to make things even between DACs (only one could handle all the formats available) - and found that even the "golden ear" of the group couldn't hear a difference between the original high-res & the resampled CD-audio. He does make the point (as has been made here a few times) that the higher-res formats are useful in studio work to provide more headroom (and as has been pointed out here for DC better performance of some filters); his article focused only on playback performance. And the Realtek mobo chip did have measurable rolloff at the low end (1.4 db @ 100 Hz), which suggests that for studio (and DC) use an upgrade would be desirable. YMMV of course, and he provides abundant caveats (and thoughts for more studies). Interesting

Mike B

[Craig Maier]

Hi Mike,

I was not familiar with "Tom's Hardware" article regarding DAC's. Anyway, it has been an ongoing topic of discussion here for years. Somewhere on this forum, some time back, I wrote a post regarding the theoretical noise floor of ADCs and DACs. Given typical source resistance values and non-cryogenic temperatures, the noise floor can not be much lower than 100 dB below the useful signal, thus anything below that value is just random (thermal) noise. Thus, 16 bits is adequate, unless one wants to record at -50 dB levels. I try to keep the peaks at around -3 dB, so there is no practical issue.

Craig