Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › OPUS: Analogue vs Digital Volume Control
New Posts  All Forums:Forum Nav:

OPUS: Analogue vs Digital Volume Control

post #1 of 15
5/18/09 at 8:54am
Thread Starter 
Hi,

I thought that I would start a new thread on this topic since it seems to be seldom discussed.

Have any of you tried a digital volume control on any of your DACs or Preamps?

The most obvious one would be the Twisted Pair Audio Volumite with the OPUS/Buffalo DACs or perhaps one of the Bel Canto digital products. I am wondering what you thought of the sound quality. My gut feeling is that the analogue control would be the best however I really have no real expeience with a digital control. Currently I am building an Opus DAC while waiting for another round of Buffalos to become available. I thought that it would be fun to experiment with the digital control but was wondering if any of you had already "gone boldly before."

Any thoughts?

Best,

Paul
Reply
post #2 of 15
5/18/09 at 1:43pm
Digital volume in a DAC is probably different from digital volume in a preamp. In a dac the attenuation is done in the digital domain, and in theory no data is lost until the attenuation is below 50 -60 dB because the DACs operate on 24 bits and the audio content of CDs is 14-16 bits.

I've tried the digital volume in OPUS and did not perceive any difference compared with a ladder attenuator. (The attenuator was not really going to 0 dB because at max, it was only outputting 2/3 of full scale).

I've documented some additional info on digital volume in my home page...
Reply
post #3 of 15
5/19/09 at 10:57am
I'm not sure all dacs use 24bits inside. do they?

if you send i2s to a dac, its 24bits no matter what the source was (it does not up-convert but it bit-pads to convert 16bit redbook to 24bit format, still with the same bit patters, just bit-shifted a bunch to the left).

if you come in via spdif, though, you get the normal 16 bits and no format changing happens inside (most dacs, I think). in that case, you are truncating the 16 bit data which is NOT good!

I'd use digital atten in a dac ONLY for listening. if you are looking to copy data from one source to another and need to level match, this isn't the way to do it. but if you are 'only' listening, you may not notice the loss in bit resolution. but its still a loss of data when you digitally attenuate.

what you save is not needing an analog variable-gain stage on the end just for level matching. to be honest, I'm not sure this is an area that -needs- to be optimized out, but ymmv.


to contrast, I'm a recent fan of the burr brown and cirrus PGA23xx style chips. they test about as good as the mid-fi and even hi-fi dacs and amps do. plus you get .5db steps (if you want) with pga style solutions.

talking serially to a dac sounds cool, of course; but I'm not sure vol control AT the dac is really what audiophiles want (?)
Reply
post #4 of 15
5/19/09 at 12:51pm
Agree that intuitively, chopping bits off seems to take data away. I don't know the details of 24-bit padding/conversion but PS Audio is using the digital volume control feature of WM9741 and Twisted Pear Audio is using the digital volume in the ESS DAC.

According to them, at least in those two DACs, there is no loss of data until you chop off 8 bits (24-16) which is 48 db of attenuation...
Reply
post #5 of 15
5/19/09 at 1:02pm
unless you have some really fancy dac chip (I know nothing about ess) - its not likely that the dac is going to CONVERT 16 bits to 24 bits other than just left-shifting data to fill 24 bits.

then it would be MORE dsp work to properly redither that new 16 bit value to a new 16bit value (24bit internal or not).

if its really dithering, that's fine. if its chopping, then I would guess (wag) that more than 1 or 2 bits and you've lost more res than a hi-fi'er would prefer.

I did casually read some of the TI and wolfson chip datasheets and it seemed like they were only truncating - which isn't what 'we' want in a volume control, really.
Reply
post #6 of 15
5/19/09 at 3:47pm
What about using something like the MCP4261 for a volume pot with digital control of an analog circuit? It's a digital pot with 257 steps of control. I believe I got the 50k ones but they might be 100k.

I have a few of them I wanted to use for the volume control of my tube surround sound rack. It would allow for independent volume control of the channels so I could calibrate the system digitally for a room and never have to mess with it again.
Reply
post #7 of 15
5/20/09 at 6:40am
i did some test on this in a way by bypassing my preamp and running directly from my pc to the dac and onto the amp. Sound quality improved by quite a bit especially since my preamp is pretty poor. Personally i think if done right attenuation in the digital domain could have real benefits.
Reply
post #8 of 15
5/20/09 at 9:42am
You're losing data no matter how you turn down the volume. If you can't hear it, the data's gone. If you're listening at 70dB, then anything at -71dB and lower is gone, whether it was taken out by dropping the bits, or by attenuating the electrical impulses.
Reply
post #9 of 15
5/20/09 at 9:53am
Quote:
Originally Posted by grawk View Post
You're losing data no matter how you turn down the volume. If you can't hear it, the data's gone. If you're listening at 70dB, then anything at -71dB and lower is gone, whether it was taken out by dropping the bits, or by attenuating the electrical impulses.
Don't think it works that way, attenuation is done equally throughout the volume range, the signal should be simply reduced in voltage proportionally to the rest, not as an absolute attenuation which can drop certain notes (played at lower volumes) to 0V.
Reply
post #10 of 15
5/20/09 at 9:59am
That's true whether the attenuation is analog or digital.
Reply
post #11 of 15
5/20/09 at 10:35am
Well, if you are talking about attenuating the lower power frequencies less than the higher power frequencies that would be coloring the sound, which is a no no.

Now, if the speakers/headphones do that non-linear attenuation, they would also be coloring the sound. Regardless, shifting everything one, two, etc bits would only color the sound if the bits that are dropped are not always zeros anyway. when you start dropping bits that may have been ones you wind up with rounding errors and hence a coloring of the sound.
Reply
post #12 of 15
5/20/09 at 1:27pm
try it and see, if you use a computer as your source, connect directly to you amp bypassing the pre and volume level some music a lot lower than you would normally or just use the volume control if your using kernal streaming to attenuate the sound (inbuild volume control apparently is very poor). listen to it and do an a/b between analog and digital attenuation to see the difference. I found that cutting the preamp out of the signal path improved the sound dramatically. I do agree that if you can't hear something then it may as well never have been there so in a way either form of attenuation is still making sound dissappear.
Reply
post #13 of 15
5/20/09 at 2:39pm
Take a look at the AD5220 - it's a digital potentiometer that takes very simple input (it can even be used with buttons and no microcontroller).
Reply
post #14 of 15
5/21/09 at 2:42pm
The problem with using a regular digital pot is popping (DC spikes), which occurs when you switch an AC signal that is not at 0V, especially if there is any DC bias component (there is always some). the BB PGA chips address this (specifically for audio) by only making volume transitions (basically throwing switches) as the AC signal is crossing 0V. Check out the datasheet for more information.

Analog attenuators are typically just voltage dividers, so you are reducing the output voltage of the signal to achieve lower volume.

Digital signals are numeric representations of voltage. When a DAC attenuates, doesn't just chop off bits; that would not sound very good at all, and would offer only 16 or 24 steps. It alters the digital representation of a voltage to be a lower voltage, then outputs the new number in binary format. Completely hypothetical example, take in 0110011010001101 (26253d), calculate the voltage it represents, say 40% of max signal, drop that to 39% of max, then calculate the binary number (25559d or 0110001111010111). Again, this is just an illustration, and is does not account for dB levers (logarithmic math), etc.

Either way, voltage in (or representation of voltage), lower voltage out.

In defense of digital volume, is does have better inter-channel matching (typically) and less noise (miller noise from resistors).
Reply
post #15 of 15
5/21/09 at 9:16pm
I didn't think about the zero crossing stuff and it makes sense the audio chips would do this (go to extra effort, in fact!) whereas 'pot chips' don't always need this. good catch!

as for the inter-channel matching, the digital volume controls all do this, no? there's no 'ramp' on the separation graph on digital attens (or relays) like there is on wiper based pots.

I've been living with a PGA bb chip the last month or so, as my main volume control. I'm loving it perfect balance at any atten point, instant access to any setting (unlike motor pots) and you can do fancy slow ramping and stuff with them, under software control. I don't know if I'll ever go back to mech pots again, on decent DIY gear.
Reply
New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › OPUS: Analogue vs Digital Volume Control