PDA

View Full Version : Does digital volume control cost resolution?



gdg
2010-02-07, 03:07
The old rule of thumb was that for each 6db of digital attenuation one lost 6 db of resolution. I believe that no longer holds with devices that are processing 16 bit data with an internal resolution that is higher (i.e. 20 or more bits). I have a squeezebox 3 and use the digital output into some pretty good equipment starting with a big ben to re-clock the data stream. The SB3 is being pressed into double duty as a volume control for a little while. Am I degrading the signal?

Phil Leigh
2010-02-07, 03:44
The SB3 uses a 24-bit internal implementation for it's level control. Provided you keep the volume fairly high you shoudn't notice too much loss of resolution.

m1abrams
2010-02-07, 20:30
The SB3 uses a 24-bit internal implementation for it's level control. Provided you keep the volume fairly high you shoudn't notice too much loss of resolution.

If you notice a loss in resolution with such a low level then maybe you should not use so much gain ;)

Soulkeeper
2010-02-08, 03:15
So in other words, you control the volume by choking the DAC, and you need to keep the SB volume on 100% in order to get full resolution?

There is no volume controlling amp/circuit after the DAC?

Is this the case for all SB models?

KMorgan
2010-02-08, 04:04
Having nothing better to do I've been pondering this.

If a 16bit source is made up to 24bit by adding eight 0s on the end, then isn't it fair to say that there will be no loss of resolution if all you do is remove a few of them? Of course the signal will be closer to the noise floor, but what about resolution? e.g. (assuming 50% digital level control = an arithmetic division by 2) then simply removing one of the 0s will not reduce resolution.

Is the level control arithmetic?

Take this a bit further. Remove all the 0s and the level will now be 1/256 with no loss in resolution. You're just back where you started. Since whole number multiplication does not have any effect on resolution then the original signal can be kept intact as long as the digital level control is kept to a multiple of 1/256.

Perhaps the volume control should be made and marked as 1/256 -> 256/256, or at least do that in the background. In whole number % terms you only hit one of these "magic" numbers at 25, 50, 75, 100%. Any intermediate volume will introduce rounding errors. At the top end there will be eight spare bits to accomodate this, at the lower end this will be less.

So does an arithmetic volume of 75% have better resolution than, say, 95%?

Please note the Numpty Warning at the top. Please assist by spotting the logical/technical whoopsie. Looking forward to being set straight :)

Keith

DaveWr
2010-02-08, 04:22
You have already been through this debate before....

Just search for Sean Adams + Volume Control you will find the answers to you questions - as you previously posed them.

For further complex data Patrick Dixon + Volume Control will explain all the issues about the arithmetic calculations an effects on resolution.

Dave

KMorgan
2010-02-08, 04:34
For further complex data Patrick Dixon + Volume Control will explain all the issues about the arithmetic calculations an effects on resolution.

Thanks

K

Phil Leigh
2010-02-08, 04:35
So in other words, you control the volume by choking the DAC, and you need to keep the SB volume on 100% in order to get full resolution?

There is no volume controlling amp/circuit after the DAC?

Is this the case for all SB models?


The use of the SB volume control is optional.
The control operates in the digital domain, by changing the bits sent to the DAC or spdif

The idea is to run your SB device into a setup that has correct gain staging.
This means that you should have an active or passive preamp with level control between SB and amp/speakers. You then don't NEED to use the SB control at all.

Alternatively use in-line attenuators into just a power amp to lower the level to the loudest you ever want to hear and then use the SB control for day-to-day tweaking of level. This way, you will generally never lower the control enough for loss of resolution to be a big issue.

With the 24-bit control and a 16-bit original signal (say redbook CD), any lowering will at minimum impact the value of the 16th original bit - this is why HDCD won't be passed correctly via SPDIF unless vol is on MAX and replaygain is disabled.


There's nothing really bad about (24-bit+) digital level control when used wisely.

Sean has explained the ins and outs of digital level control several times - it's in the Audiophile forum.

gdg
2010-02-08, 16:31
The SB3 uses a 24-bit internal implementation for it's level control. Provided you keep the volume fairly high you shoudn't notice too much loss of resolution.

Since for each 6db of attenuation one loses 1 bit does that mean that I can afford 48 db of digital attenuation on a 16 bit source before I lose resolution?

gdg
2010-02-08, 16:45
I just noticed, after reading some of the responses that my question has been answered in another thread by Sean Adams. Thanx for indulging me.

bobkoure
2010-02-10, 08:52
Seems like maybe the gist of Sean's response should go into the wiki?

gdg
2010-02-10, 13:09
The SB3 uses a 24-bit internal implementation for it's level control. Provided you keep the volume fairly high you shoudn't notice too much loss of resolution.

I can't find the post where Sean addressed this but I just want to be sure I understood...

Since for each 6db of attenuation one loses 1 bit does that mean that I can afford 48 db of digital attenuation (in the SB3) on a 16 bit source before I lose resolution?

Phil Leigh
2010-02-10, 13:20
I can't find the post where Sean addressed this but I just want to be sure I understood...

Since for each 6db of attenuation one loses 1 bit does that mean that I can afford 48 db of digital attenuation (in the SB3) on a 16 bit source before I lose resolution?

It's got nothing to do with "losing resolution" in an absolute sense - you still have 16/20+ bits. It's the value in them that is changing and thus the max SNR and dynamic range.
You really need to read this thread fully - especially Sean's comments.

http://forums.slimdevices.com/showpost.php?p=164666&postcount=1

In practice you can wind the volume down to 70/80 out of 100 without suffering any noticeable effects.

DaveWr
2010-02-10, 13:27
I can't find the post where Sean addressed this but I just want to be sure I understood...

Since for each 6db of attenuation one loses 1 bit does that mean that I can afford 48 db of digital attenuation (in the SB3) on a 16 bit source before I lose resolution?

I hope this helps, if you want further details search for Patrick Dixon + Volume, the simple answer is no. It's to do with the step sizes.

http://forums.slimdevices.com/showthread.php?t=28834&highlight=Volume+control

Dave

gdg
2010-02-10, 17:54
It's got nothing to do with "losing resolution" in an absolute sense - you still have 16/20+ bits. It's the value in them that is changing and thus the max SNR and dynamic range.
You really need to read this thread fully - especially Sean's comments.

http://forums.slimdevices.com/showpost.php?p=164666&postcount=1

In practice you can wind the volume down to 70/80 out of 100 without suffering any noticeable effects.

Thanx Phil,
I followed the discussion and Sean gives the best layperson explanation I've heard in the many years I've been trying to sort this out. The biggest problem (I think) is that people don't clarify that there are 2 issues:
1) signal to noise (dynamic range) which goes down with any level of attenuation and...
2) bit resolution

Sean clarified this nicely.
Thanx again for the patient guidance.

bobkoure
2010-02-12, 19:15
...
This has been discussed many times here, but I won't refer you to our archives as there have been far too many people still getting it wrong. Instead, here's a detailed and correct explanation from Lavry (login needed):

http://lavryengineering.com/lavry_forum/login.php?redirect=viewtopic.php&t=76


Lavry seems to have changed their forum. Did anyone save this offsite, or have a new link?
Thanks!

KMorgan
2010-02-13, 02:38
So did I accidentally get it right when I said (with regard to a 16 bit sample bumped up to 24)?


Remove all the [extra] 0s and the level will now be 1/256 with no loss in resolution. You're just back where you started. Since whole number multiplication does not have any effect on resolution then the original signal can be kept intact as long as the digital level control is kept to a multiple of 1/256.

From what I can glean, SB does indeed stick to increments of 1/256.

K

DaveWr
2010-02-13, 03:59
In the attached thread, P Dixon explains that 1/256 or multiples are used down to -30db. In the old 40 step volume days (2005) he had to argue with Sean & Dean to get it corrected to be this way). Further volume reduction after that bit resolution accuracy is compromised.

http://forums.slimdevices.com/showthread.php?t=28133&highlight=volume+steps&page=2

The thing I don't know is whether replaygain (if used) also follows these adjustment rules to protect bit resolution.

Dave