|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
92.254.0.240
Came across an article by Resonessence Labs (link below)They compared analog and digital volume control.
They claim that digital volume control has a substantial worse S/N ratio with increasing attenuation.Level Analog Digital
0 96 96
-5 96 96
-10 96 94
-15 96 89
-20 96 84
-25 96 79
-30 96 74
-35 96 69
-40 96 64
-45 91 59
-50 86 54
-55 81 49
-60 76 44
S/N ratio of volume control using a DAC with a -104dB S/N ratio and 16/44.1 audio.Dou you think these numbers are correct?
Your thoughts pleaseEdit: corect link to the article
The Well Tempered Computer
Edits: 08/04/11 08/04/11Follow Ups:
Probably correct. The best way so use digital volume is to limit your use to -6 to -9 dB. More than that becomes audible IME.
Use a DAC with a good volume control or attentuation/gain setting and then use partial digital volume and you will be fine. Set the DAC volume to be normal level with a quiet track and then use digital volume to drop it for louder tracks. Most are within 9 dB.
Steve N.
As I understand it, the issue is that a well-designed analog VC attenuates noise along with the signal, while with digital VC you are stuck with the residual noise of the DAC. But if that is already inaudible (way, way, way below the noise floor of my room), why should I care? Are there other issues with DVC (apart from gain-staging issues)?
On some DACs, particularly some ladder DACs, there is low level distortion as well as noise. In addition, if the software doesn't do dithering that will cause there to be low level distortion and the sound will suddenly drop out.
On the other hand, some DACs lack headroom on loud material and will distort on intersample peaks that are over maximum. This is actually a design problem with the DAC but it happens with some DACs. This distortion can be prevented by a small amount of digital volume control reduction (typically less than 2 dB except on pathological material where it may be more).
So if you have both an analog volume control and a digital volume control the rule is, as always, experiment and select what sounds best to you. Just take care not to use too much analog gain so that you don't blow the speakers if "You've got mail!"
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
looks ok
drrd has some perplexing views. If S/N isn't important, why doesn't he swap his dCS with cheap TDA1543 variants?
Because the TDA1543 variants look like junk in comparison :) Reasons to keep the dCS gear: Power supply quality, high performance PLLs, lots of DSP, custom firmware which dCS actively develope, balanced converter with excellent *linearity*, low impedance balanced output etc. Also it sounds better than every other DAC I've tried :) Take a look at a state of the art preamp like the accuphase c3800 and it's AAVA volume control; voltage-> multiple balanced current sources then into a summing amplifier with current-> voltage conversion at the output. Sound familiar? Yes I'm aware you lose bits with a digital volume control in a DAC, the benefit though is that you don't introduce additional distortion and non-linearity. I can only say I've tried a couple of (conventional) high-end preamps and I wasn't that impressed. As I mentioned S/N figures clearly don't entirely explain a source's sound quality, otherwise the turntable manufacturers would be out of business surely? Remember the 80's with all those big japanese amps in department stores proudly displaying THD figures like it meant something? Same thing.
All other things being equal is better S/N better? Of course it is. If you have to compromise elsewhere though then maybe not.
The dCS has a very good digital volume control.
I actually don't bother with any of these; I have a made to order balanced 10k constant impedance balanced Placette and sevarl other Passive preamps in which S/N is not an issue.
I actually don't hear much difference between the dCS volume control and Windows volume in the bottom right corner. Safer to use the dCS volume control though, especially with active speakers :)
Wouldn't that depend on how the digital volume control is implemented and at which bit depth it operates?
The volume control on iTunes is abysmal while the one on my convertors is fine.
The software and thus the dvc which came with my convertors runs at 32bit floating point according to the bumph. I don't know how itunes dvc is implemented.
The assumption here is that S/N has a really significant effect on sound quality. How much S/N do you get from vinyl, even the Continuums and SMEs of this world? How does an analogue volume control compare in terms of L/R tracking, frequency and time-domain distortions? Given distortions introduced in digital systems from jitter and filters can you be sure you can even hear the effects of reduced S/N? S/N doesn't tell you much really.
The numbers don't seem unreasonable. If one uses a DAC with only 104 dB DB S/N then one shouldn't be using much digital volume reduction, if any. If one has a better DAC with more S/N ratio itself one might be able to get away with 10-15 dB of digital volume reduction. Of course it is perfectly fine to use more digital volume reduction if the volume has been turned down for background listening.
If one chooses to use digital volume reduction instead of a preamp one must be careful that there isn't excess analog gain. Some means must be taken to ensure that a full scale digital signal does not damage one's equipment otherwise the system is going to be unsafe and/or unreliable as well as poor sounding.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
I don't buy their analog figures within the context of real world equipment.
For example, they are claiming with a signal of -45 dB, the analog S/N ratio is still 96 dB on a system that has a 96 dB S/N ratio at 0 dB. That defies logic.
The S/N ratio is measured against a max signal against the residual floor level noise of the component. If that max signal is 96 dB above the floor, a -45 dB signal cannot also be 96 dB above the floor.
If they mean something else is going on, they need to clarify.
"they are claiming with a signal of -45 dB, the analog S/N ratio is still 96 dB on a system that has a 96 dB S/N ratio at 0 dB. That defies logic"
Not really. It just means that the S/N is beyond the resolution of the 16 bit measurements with moderate analog attenuation but it isn't with poorly done digital attenuation.
Rick
> > Not really. It just means that the S/N is beyond the resolution of the
> > 16 bit measurements with moderate analog attenuation but it isn't with
> > poorly done digital attenuation.
My comment wasn't about the digital side but rather their numbers for the analog side.
If a 0 dB analog signal is 96 dB above the noise floor, it is NOT still 96 dB above the noise floor if the signal is down 40 or 45 dB. The noise floor doesn't get lower as the signal reduces in volume. Their chart incorrectly implies the analog S/N ratio stays high even with low-level signals.
"If a 0 dB analog signal is 96 dB above the noise floor, it is NOT still 96 dB above the noise floor if the signal is down 40 or 45 dB. The noise floor doesn't get lower as the signal reduces in volume."
It depends on what is establishing the noise floor. I haven't read the original but let's say that the source was from a 16 synthesized source or something that had an intrinsic S/N of -96dB. If that is considerably worse than the downstream stuff then increasing the attenuation would decrease the noise along with the signal and the measured S/N would remain constant until the downstream stuff started to dominate the system. Much the same thing would appear to happen with limited resolution of the measurement.
It's easy to measure stuff wrongly or draw iffy conclusions even if you mean well. For instance a S/N without a defined noise bandwidth is as reliable and useful as a politician's promise. If the measurement is part of an advertisement, well...
Naturally a digital attenuator is limited by it's amplitude resolution in the bandwidth of interest so they usually do better with more bits or dithering fast enough to average out more bits in that bandwidth. That comes as no surprise to any of us and it doesn't take much of a pot to do better when the effective number of bits starts getting low.
Rick
Somebody who knows a lot more about these things told me once that the theoretical minimum noise floor for analog circuits is -132dB at unity gain.
With that in mind the analog numbers do look a bit iffy.
"Somebody who knows a lot more about these things told me once that the theoretical minimum noise floor for analog circuits is -132dB at unity gain."
The absolute limit of the noise floor is the thermal noise energy which is a function of the absolute temperature and bandwidth. It's just the electrons rattling around from the heat, throw in some resistance and their motion shows up as a noise voltage.
And that's if the components are perfect, usually they aren't and have other mechanisms that make them noiser, since they don't HAVE to be we call it excess noise and that's the stuff that gets improved with better parts.
Folks that are really fussy, I'm thinking astronomers here rather than audiophiles, tweak their front-ends by immersing them in LN2 so they keep their cool. So far I have not heard of any of us taking that plunge...
I don't know what the -132 dB refers to off hand.
Regards, Rick
This sort of number is meaningless for a number of reasons and the use of it in an unqualified fashion demonstrates the someone's ignorance. There is a theoretical noise level in a circuit, but it depends on impedance, bandwidth and temperature. This is a lower level on the noise, but there is no upper limit on the signal. That will depend on the power, i.e. more power, better S/N ratio.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
See link below
The Well Tempered Computer
As I pointed out, this is not a limit on S/N ratio. One simply uses higher voltage levels. For example, if one uses 3 volts instead of 1 volt one will pick up an additional 10 dB in S/N ratio. Similarly, one can use a lower source impedance than 200 ohms. (For example, paralleling multiple DAC chips will lower the source impedance and hence the noise voltage.) One can also cool things down as well, but this is common only with high performance RF receiver front ends, not audio circuits.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: