I already posted this question on the official RMAA forum, but since we have Mr.Lukin, one of the developers of RMAA, with us at HA and most of us already know what RMAA does, I reckon I can post this here as well. (By the way, I am a big fan of RMAA! )

Several months ago, I posted a question on how to set the input mode for analyzing 16-bit/44.1kHz playback capable DAPs, and Mr.Liadov, also one of the developers of RMAA, kindly answered my question here:


So I've done some more measurements on a '16/44.1-capable DAP' such as Samsung YP-Q1, different bit-depth settings in 44.1kHz sample rate, and compared each settings with one another.

Test condition:
RMAA v6.20:
Playback: Samsung YP-Q1 [Test signal (44 kHz 16-bit).WAV] -> Recording: Lynx L22 [16-bit, 44kHz /24-bit, 44kHz]
*References were obtained by analyzing 'Test signal (44 kHz 16,24-bit).WAV' with RMAA itself

Please take a look at the following results:

Full version is here: http://sites.google.com/site/rcudauda/1-1

SAV files:
YP-Q1 - 16bit Mode 16-bit, 44 kHz.sav
YP-Q1 - 24bit Mode 24-bit, 44 kHz.sav

If you record the 16/44 test signal from Q1 with 16/44 & 24/44 input settings on Lynx L22, 24/44 actually yields better numbers in Noise Level & Dynamic Range than 16/44, and even better than the 16/44 reference. Unlike what Mr.Liadov said, recording in 24-bit yields better result than recording in 16-bit. (Even using the latest version of RMAA & Different DAPs yields same difference)

Does anyone have any idea on what might have caused these different results? (dithering?)
Which input mode setting (16/44 or 24/44) should be more trustworthy?

Your recording levels do not match your playback levels: they are off by around 10 dB. They should be calibrated to within 1 dB (or even closer) for accurate testing. Lower recording levels produce less noise.

Oh, you are absolutely right. I certainly have missed that.

Then, apart from the erroneous results above, once input-output level's matched properly, would recording 16/44 source in different bit-depths(16 or 24) yield any differences in Noise Level & Dynamic Range? In fact, I am still confused with the term 'useless' Mr.Liadov used. Wouldn't recording in higher bit-depth yield more accurate result of the DAP itself? (only if the DAP should output anything 16-bit input may not be able to analyze due to its limitations)

Yes, recording at 24 bits would probably result in somewhat better results because extra dither (or truncation) will not be added during recording.

Do you think Mr.Liadov could be wrong? So between 16-bit & 24-bit, which bit-depth would you recommend to use in order to measure everything the 16/44.1 DAP outputs?

BTW Mr.Lukin, I have always appreciated your (and Mr.Gilvazetdinov & Mr.Liadov's) contribution to the digital audio field by developing RMAA. You gave ordinary consumers (who are mostly passive) a chance to assess their own audio gears without all the expensive measuring devices, and ultimately helped a lot of them(including me) to make an active decision, which is a right one.
I certainly owe you a big thanks!

If the levels are very well matched and your sound card is of good quality, what he said is more or less correct. Recording a 16 bit signal at 24 bit precision doesn't do much. If you don't have the levels well matched or your sound card performs poorly at 16 bit, then 24 bit will work better. I would always use 24 bit since it should not hurt.

I agree with Mike.
Thanks for good words about RMAA.

Of course the other situation where that would hold is if the quality of the analog output which you were measuring was significantly worse than the theoretical limit of 16 bit digital audio. It would then make little or no difference whether you used 16 bit or 24 bit to sample it. This is quite often the case BTW.