[jswallac]

Hopefully somebody can clear up my confusion.  I am looking at CDPs that allow you to connect direct to a power amp.  The volume controls seem to come in two flavors, some in the digital domain and others analog.  What I have read is that the digital volume controls are to be avoided.  Terms like "loss of resolution" or "bits are lost" are used.  I am having trouble reconciling this with the new digital receivers such as the Pannys.  What I read about these is that one should avoid the analog stage and just connect your CDP directly to the digital in so the signal stays digital throughout.  What about the volume control?  Is it digital?  If so, why is it okay here, but bad if done in the CDP?  Thanks for any help in explaining what is going on.

[PhilNYC]

When you use a digital volume control, you are basically using a DSP to alter the audio data...reducing values as you lower the volume.  Imagine plotting the original data on a 100x100 grid...that is what your "best resolution" is.  Now imagine reducing the grid to 60x60, altering the data values proportionally (to the nearest integer), and plotting them again.  You'll see that the data plot on the smaller grid is a less-than-exact portrayal of the original data plot.  This is exactly what is happening to your data when you use a digital volume control.

In the case of Wadia (and some other companies), they use a huge upsample/oversample to increase the "resolution" before implementing the digital volume control to help maintain some of the accuracy in the digital volume control.  Imagine that same 100x100 grid blown up to 1000x1000.  You can still draw the waveform with the same accuracy that you did at 100x100, but when you now reduce the grid by 40% to 600x600, it is still relatively similar to the 100x100 grid.

An analog volume control basically attenuates an analog signal.  So while the volume gets softer, it is not limited by digital resolution, so you maintain accuracy of the original analog waveform.  That said, if a poor volume control is used, you then add distortion to the waveform.

I don't know what the Panasonics do, but if they implement a digital volume control, they have the same issues.

[jswallac]

Thanks Phil.  I posted the same question over at AudioAsylum and received the following answer (like yours, it uses a great analogy):

"Digital volume control lops off the real soft bits of the music (called least significant bits) and sort of pushes it down a step (to put it simply) Think of it as chopping the rungs off a ladder from the bottom , you lose a rung and make the ladder shorter. with analog you are really just pushing the ladder down one rung but the ladder doesnt get shorter.
So you lose info totally when using digtal attenuation as well as introducing truncation distortion. Most modern or higher end dedicated digital preamps or digital attenuators do it a lot better than cheap ones found on most cdps as they convert the signal to a much higher resolution one and apply noise shaping"

I was also given the link to how Wadia does it:

http://www.wadia.com/technology/dvcontrl/sld001.htm

It appears the takeaway is that neither analog, nor digital, are perfect.  Each can be done poorly or well, but you need to be a lot more careful with digital since it is a lot harder to do well.

[PhilNYC]

jswallac,

Thanks for the explanation from AA, although I'm not sure how accurate it is.  If all a digital volume control did was lop off the least significant bit and push down the remaining bits, you'd only have 16 volume levels based on red book's 16-bit data word...and you'd get huge volume changes at the start of the volume knob range and tiny increments towards the end.  In addition, the level of distortion you'd have at low volumes would make things almost unlistenable.

I'm not an engineer, so I'm not 100% sure the following is the right explanation, but I'm guessing it's pretty close.  To get a relatively smooth attenuation digitally, a digital volume control probably uses a DSP to scale down the 16-bit words by some logical percentage with each increment on the volume knob.  To get 100 positions, the DSP might subtrack 1% with each increment.

[jswallac]

Phil,

The Wadia link does a pretty good job of explaining what is going on in layman's terms.  It also explains why doing it poorly does cause a large loss of resolution.

[PhilNYC]

jswallac,

Thanks.  I do understand why there is loss of resolution.  I'm just confused as to whether their description of a "basic" digital volume control is actually accurate...because, as I said, if bit-shifting were all that was happening, you'd only have 16 increments of volume...and certainly there is need for more than that (practically speaking).

[denverdoc]

Wadia want but my money? Gimme a break--pretend to first expand the size of the picture then attenuate w/o info loss? How does one get something from nothing? More exactly, how is this qualitatively different from post processing the original signal with small scale digitally implemented attenuation ladders?

Not to suggest that DSP pitfalls don't exist--have to use every bit thru the chain then attenuate, but not all that different in principle from keeping the analog VU meter almost in the red. Nor to suggest Wadia gear is anything less than state of the art. But this seems more a case of marketing honchos in hyperbolic mode--laws of energy conservation and information conservation are immutable for now.

In the meantime, let your ears decide!!!!!!!!!!!!
J

[denverdoc]

Wadia want but my money? Gimme a break--pretend to first expand the size of the picture then attenuate w/o info loss? How does one get something from nothing? More exactly, how is this qualitatively different from post processing the original signal with small scale digitally implemented attenuation ladders?

Not to suggest that DSP pitfalls don't exist--have to use every bit thru the chain then attenuate, but not all that different in principle from keeping the analog VU meter almost in the red. Nor to suggest Wadia gear is anything less than state of the art. But this seems more a case of marketing honchos in hyperbolic mode--laws of energy conservation and information conservation are immutable for now.

In the meantime, let your ears decide!!!!!!!!!!!!
J

[PhilNYC]

Actually, most of Wadia's explanation makes the most sense to me.  They upsample and use a 22-bit word to improve the resolution, so that when digitally reducing the volume (the magnitude of the analog waveform), they aren't getting as much distortion.  Imagine drawing a 45 degree diagonal line on a 10x10 grid.  It is exactly straight.  Now try to reduce it to a 30 degree diagonal line...it loses its straightness, because the 10x10 resolution doesn't permit a straight line.  Now do the same thing on a 100x100 grid.  The 30 degree line, while still distorted, is not nearly as distorted as on the 10x10 grid.  Wadia is doing the equivalent of upsampling the line from 10x10 to 100x100, and increasing the precision from 16-bit to 22-bit...so that any change to the digital waveform can be done with more accuracy.

I still don't buy the "remove a bit and shift it down" explanation...I think that is an easy way to explain to a layman, but I don't think that's exactly the only thing that's happening.  Basically doing a bit-shift is the equivalent of cutting a byte's value in half (equivalent to a 3db change), and I don't think that's a fine enough volume change...on a 16bit world (red book), this only gives you 16 different volume levels.

[beemer]

Having owned the Wadia 270/27ix transport and DAC, and now a Levinson 390S I find that FOR ME the Levinson is the player of choice.

I found the Wadia never sounded quite right to me, while the Levinson with it's analog volume control did.

My experience as well not only in my system but in those of my friends we have most always preffered the sound with a quality preamp in the chain.

I personally will not consider Wadia again.....been there, done that.

Link to my system:

http://www.audiocircle.com/index.php?action=systems;system=427


Best to all,

Paul

[PhilNYC]

Beemer,

Do you use the volume control on your Levinson even though you have a preamp?  Btw - I agree with you in that I strongly prefer the use of a preamp over going direct; and I have preferred using an active preamp over a passive, although admittedly my experience with passives is limited.  I am also in agreement with you in your preference of the Levinson over the Wadia, although I prefer the Dodson over both of them...  

[beemer]

Do you use the volume control on your Levinson even though you have a preamp?

Hello Phil:

I had used the Wadia combo direct, and also the Levinson. Recently I picked up a Levinson 336 and started passively bi-amping. For me this required two sets of balanced outputs......thus neither the Levinson nor the Wadia would fit the bill as they have one set of balanced, one set of single-ended outputs. Single ended being 6db down over balanced outputs creates a problem. The Classe Omega is very synergistic and fits the bill.....Classe and Levinson preamps both use the same output impedance so match well together.

I'm not fond of using splitters, and I had a Bryston 10B sub crossover, however my Aerials did not respond well to the available settings on the Bryston.......passive biamping has worked well.

Best,

Paul  

[/quote]

[JuicyMusic]

Just a thought here, but sometimes the term "digital volume control" just implies, a "digital potentiometer" - which is simply a resistor network whose wiper position is changed by flipping FET switches, which in turn are controlled by a microprocessor. The input and output is analog, only the control is digital. What it really means is "the position of the wiper is digitally controlled."

In the digital domain, the output analog voltage is from the DAC. To change volume (or reference level) in the digital domain, you need only change the reference voltage on the DAC itself. But note that you are not changing anything about the digital bitstream, the bits are still entering a ladder network and changing the current in the output OpAmp. But since the DAC now has a "variable" Vref it's maximum output signal is also variable: aka Volume Control.
JM

[PhilNYC]

In the digital domain, the output analog voltage is from the DAC. To change volume (or reference level) in the digital domain, you need only change the reference voltage on the DAC itself. But note that you are not changing anything about the digital bitstream, the bits are still entering a ladder network and changing the current in the output OpAmp. But since the DAC now has a "variable" Vref it's maximum output signal is also variable: aka Volume Control. .

I'm not an engineer, but it would surprise me if a DAC chip had a "variable" Vref that could be done with any kind of precision (or at a minimum, maintaining an acceptable SNR).  Is this possible?

[denverdoc]

Thanks Juicy--that was what i was trying to say--sure oversampling is great for relaxing filter requirements, etc, but not necessary to implement "a digital" volume control. Cool thing about using digital amps (PWM), esp in this scenario, is that the volume can also be adjusted via simple attenuation of the supply rail, so that the full signal rez is preserved here as well--believe this is how Tex Inst is doing it in some of their "equibit" implementations,
John

[ekovalsky]

Just a thought here, but sometimes the term "digital volume control" just implies, a "digital potentiometer" - which is simply a resistor network whose wiper position is changed by flipping FET switches, which in turn are controlled by a microprocessor. The input and output is analog, only the control is digital. What it really means is "the position of the wiper is digitally controlled."

In the digital domain, the output analog voltage is from the DAC. To change volume (or reference level) in the digita ...

This is exactly how the TacT amplifiers work.  Seems to be a very good way to implement a volume control.