Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Any tests show audible noise floor? (Read 18644 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Any tests show audible noise floor?

Reply #25
because as with the generally accepted human hearing range (frequency)  there is also generally accepted human hearing range of loudness (generally quoted as 0 db - 160 db but point being there is a "quietest" people can here as well as a "most high pitched" etc etc). So surely it's just comparing that range to the noise output of the device though obviously it depends where you have the volume control.


It seems to me that different dB measures are getting mixed up here.

The human hearing range is (unsurprisingly) measured in dB(SPL), which is a measure of sound power, so it is nothing electrical. The dynamic range of a DAC is an electrical measure, obtained at its analog output. You can express it in % or in dB, and it means the ratio between the power of a full-scale sine wave signal, and the power of the background noise when there is no signal.

Just because you can express both with a dB figure, doesn't mean they can be exchanged willy nilly.

The limits of the dynamic range of human hearing are usually seen in the Fletcher Munson curves (google that if you don't know it already). It is apparent that there is a strong dependency on the frequency of the sound signal. When assessing the audibility of background noise, those curves have to be taken into account. As a consequence, it is customary to apply a frequency dependent weighting curve when measuring background noise. The resulting measure is expressed as db(A), because the weighting curve A (according to some international standard) has been used, which is optimized for this purpose.

Note that DAC background noise is typically measured unweighted, which is one of several reasons why you can't directly derive an audibility estimate.

Other obstacles lie in the fact that there is additional gear between the DAC and the ear, which has its influence. The DAC output, at the very minimum, has to be amplified, and fed to a loudspeaker, where the electrical signal gets converted into sound. How the DACs dynamic range in dB translates into db(SPL) depends on the gear in between, including the listening room with the positioning of speakers and listener.

So, as a simple example, any DACs background noise becomes audible if you crank up the volume knob of the amplifier far enough, or if you get your ear to the speaker close enough.

Hence, dynamic range in this situation only makes sense as a relative measure between the loudest possible sine wave and the loudness of the background noise, given a fixed setting of the system. You need to specify this explicitly, because there isn't a standardised or self-evident setting to work from.



Thanks for a very clear explanation. I was indeed getting db and dbv etc mixed up. I have done the loopback test with REW (for calibrating sound-cards) this basically puts out a sine sweep from the output of the DAC to the input of the ADC - so any other influences are removed (ie speakers/amps etc) . Obviously the transfer function is reading noise on the preamp and the DAC. It will give you a reading of total harmonic distortion, accumulative spectral decay times and frequency response. What I found was that in terms of frequency response, and decay times there was almost no difference at all between the behringer ADA8000 (a cheap model) Neve and Studer preamps (hi end) and even a laptop DAC output ! THD was somewhat different between the units - but not vastly and theinput of the laptop had no bass - its DAC was flat as a pancake though. So my conclusion is that since there appeared to be no great "noise " difference between DACs - and that it was largely snakeoil. I did not try it on different volume levels though and of course a better test might be to do the loopback without the sinesweep - to measure just the noise at different volume/gain settings . Life is short though.

Any tests show audible noise floor?

Reply #26
So my conclusion is that since there appeared to be no great "noise " difference between DACs - and that it was largely snakeoil. I did not try it on different volume levels though and of course a better test might be to do the loopback without the sinesweep - to measure just the noise at different volume/gain settings . Life is short though.

The noise difference between DACs can be substantial (I'd say a difference of 20 dB would be quite substantial, for example). It is an entirely different question whether it is audible, however. Actually, the better question would be under what circumstances it might become audible, because the usual use cases of DACs in cosumer playback gear do not require a particularly low noise floor, as the noise contributed by other factors tends to dominate.

Even a fairly basic audio DAC these days has a low enough noise floor for ordinary use. You would need an exceptionally quiet listening room for the DAC noise to become prominent in a system that is adjusted for normal listening levels. Most people haven't got access to such a listening environment.

Any tests show audible noise floor?

Reply #27
I'm wondering if there's any information on the audibility of fluctuations in the noise floor rather than steady state?
Seeing as the ITU-R 468 noise weighting shows a sensitivity to noise which is >12dB greater ( +12.2 dB at 6.3 kHz) than the Fletcher-Munson curves & "Dependent on spectral content, 468-weighted measurements of noise are generally about 11 dB higher than A-weighted", I wondered if there's any knowledge here about noise fluctuation audibility & JNDs?

Any tests show audible noise floor?

Reply #28
As part of Meyer & Moran (2007)'s tests, where they used an A/D/A loop to compare the 'native' output of SACDs to CD-quality conversions of same, they added 14dB of gain to their signal chain's reference level,  with no other input (i.e., all you hear is the noise floor, if anything). Two subjects were tested and both  identified the CD loop '+14' versus 'reference' 10/10 times in ABX (though one of the wasn't sure he/she had heard any difference).  At >14dB gain,  detection of the A/D/A loop's noise floor was 'easy, with no uncertainty'.

They also report that their chain had no noise shaping 'and was therefore less than optimal', though there is no explicit reference to dither...

Any tests show audible noise floor?

Reply #29
At >14dB gain,  detection of the A/D/A loop's noise floor was 'easy, with no uncertainty'.

They also report that their chain had no noise shaping 'and was therefore less than optimal', though there is no explicit reference to dither...

And I'd posit that their listening environment was better than average.

This suggests that even 14 bit would be sufficient for most playback situations. And if you exploit noise shaping, the figure drops further. Or, to put it the other way round, 16 bit is ample, with some margin.

Of course this has been clear for decades.

Any tests show audible noise floor?

Reply #30
At >14dB gain,  detection of the A/D/A loop's noise floor was 'easy, with no uncertainty'.

They also report that their chain had no noise shaping 'and was therefore less than optimal', though there is no explicit reference to dither...

And I'd posit that their listening environment was better than average.


Looks that their main environment wasn't anything superb, but there were evaluations with some serious setups. However this doesn't really seem to matter as it indeed looks that most of the records they used originated from old analog tapes.

Any tests show audible noise floor?

Reply #31
However this doesn't really seem to matter as it indeed looks that most of the records they used originated from old analog tapes.

I understand that they used what was recommended to them as examples of high-res recordings that showed the superior quality of SACD/DVD. What else should they have done?

Any tests show audible noise floor?

Reply #32
However this doesn't really seem to matter as it indeed looks that most of the records they used originated from old analog tapes.

I understand that they used what was recommended to them as examples of high-res recordings that showed the superior quality of SACD/DVD. What else should they have done?


I suppose, use records that have been initially recorded in high resolution, not in something well within 16/44 capabilities, and that do make use of the high dynamic range.

Any tests show audible noise floor?

Reply #33
At >14dB gain,  detection of the A/D/A loop's noise floor was 'easy, with no uncertainty'.

They also report that their chain had no noise shaping 'and was therefore less than optimal', though there is no explicit reference to dither...

And I'd posit that their listening environment was better than average.


Looks that their main environment wasn't anything superb, but there were evaluations with some serious setups. However this doesn't really seem to matter as it indeed looks that most of the records they used originated from old analog tapes.



Which has no bearing whatsoever on the particular test I was describing: a test of the audibility of the noise floor of their A/D/A signal chain without any SACDs playing.

But to your point, it was so amusing to see how suddenly, once it was determined that not all of M&M's SACDs were DDD, analog-sourced SACDs simply *weren't hi rez enough to demonstrate SACDs virtues* anymore...regardless of the gushing testimonials to their awesomely superior  SQ that had appeared in reviews  in Sterophile, TAS, etc..and had been attributed to the format.

It was patent and pathetic goalpost moving, but as I said, from a certain perspective, amusing.

Any tests show audible noise floor?

Reply #34
Which has no bearing whatsoever on the particular test I was describing: a test of the audibility of the noise floor of their A/D/A signal chain without any SACDs playing.


Yes, sorry, it's alright for the thread title question.

Any tests show audible noise floor?

Reply #35
I suppose, use records that have been initially recorded in high resolution, not in something well within 16/44 capabilities, and that do make use of the high dynamic range.

Which commercially available recordings use a dynamic range that isn't within the capabilities of 16-bit? Are you aware of any, besides discs with test signals?

Note: I don't refer to recordings made with 24-bit gear. There are plenty.

Any tests show audible noise floor?

Reply #36
I suppose, use records that have been initially recorded in high resolution, not in something well within 16/44 capabilities, and that do make use of the high dynamic range.

Which commercially available recordings use a dynamic range that isn't within the capabilities of 16-bit? Are you aware of any, besides discs with test signals?

Note: I don't refer to recordings made with 24-bit gear. There are plenty.


No, I don't know any. Not sure if there is any either. Maybe some obscure electronic band using test signals as 'music'. Anyway one would have to be in a very specific environment to be able to appreciate that.

IMO, the study should have used true high dynamic range samples, maybe newly generated, maybe using dynamic expanders, so it could flawlessly show that hi-res has any sense only in lab conditions on extreme test signals.

Any tests show audible noise floor?

Reply #37
I suppose, use records that have been initially recorded in high resolution, not in something well within 16/44 capabilities, and that do make use of the high dynamic range.

Which commercially available recordings use a dynamic range that isn't within the capabilities of 16-bit? Are you aware of any, besides discs with test signals?



As close as I've found can be found in the catalog of Bis recordings.