Digital Volume Control

[Ray P]

This topic has popped up a number of times over the years but I thought that this set of slides about the 32bit ESS Sabre DACs might be worthy of a conversation...

http://www.esstech.com/PDF/digital-vs-a ... ontrol.pdf

Ray

[Nick]

Slight slight of hand. When the 32 bit number gets to the outside world and the actual 24 bit dac

0000001000010110.1000100110000100

= 533.5372

(well actually its 534.53717041)

You will get

0000001000010110.10001001

= 534.53515625

So actually, neither is 533.5372

[ed]

I put together the digital volume shown in the picture a couple of years ago. At the time the Pass B1 was my yardstick but a couple of people on here thought the sq wasn't up to it so I began to question my ability to detect between good and bad. I had a crisis of confidence and dropped it.

The picture shown is a series stepped resistor type chip(mcp4241), digitally controlled via SPI, so critics will doubtless say its not up to ladder standards. I did 2 versions, one for RC5 and one for the Japanese protocol, of which Apple remote is an example. I had a switch for 3 sources but couldn't accommodate rotary encoders(knobs) at the time because my microprocessor ran out of pins.

I've since resurrected it with a new cpu so I can incorporate 2 knobs(a pre with only IR is pretty useless if you mislay the controller). I intend to put it in a nice modushop chassis with the streamer client I'm working on. Source 1 will be from the net client and by pass the pre, straight to the output, and the rest of the channels will go through the pre.

I was thinking of investigating a circuit board but I'm not sure what kind of interest it will generate.

[Ray P]

Slight slight of hand. When the 32 bit number gets to the outside world and the actual 24 bit dac

My reading is that the DAC is working with all 32bits so the 24bit outside world isn't a factor? What am I missing?

Ray

[ed]

I know I introduced a red herring earlier on by mixing the two ideologies in one example.....so I thought I'd add a thought right on topic......

the whole concept of adjusting volume in today's music environment is saturated by exactly the thing described in the presentation......i.e digital encoding is predominantly in the integer domain..

In my humble studio endeavours I nearly always have to normalise the final mixed track. Sometimes I even normalise individual tracks before mixing. Normalising, for those not familiar, involves increasing the level of a recording so that the highest part(loudest part) of a recording comes up to the required level(usually 0db). In analogue terms this is bringing the recording up to a level just below clipping. In many cases a recording is normalised and then even further increased to satisfy the requirement for modern loudness.....

Now this goes on in every studio everywhere....so before the buying public get let lose on their new purchase, and subject it to digital volume adjustment, this process has been performed on the music countless times already. All with it's inherent integer errors.

A few more digital volume adjustments aren't really going to ruin the recipe. The only way to avoid this is to restrict use to analog recorded vinyl played through analog equipment.

ps... there are WAV files encoded with 'float' instead of integer but they are currently state of the art...well, almost

just a few thoughts