DAC and digital volume level mystery

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
On some DAPs (iPods, etc.) that have both Line Out (LO) and Headphone Out (HO) I've noticed that using variable-level LO (volume), the sound is best about 66% instead of the usually (??) recommended/suggested 100%. The improvement in SQ is significant.
Not sure why sound is best below "0 dB" gain. I've posed this query on a number of forums and have gotten no agreeable response. Is it because lower line-level out forces one to use the receiver/integrated amp's volume pot on higher (more "linear") level? Or is it a combo of factors?
Some have suggested the issue is related to the 2vrms output of CS4398 (or wm8740) used in some of the better DAPs. Here's a direct quote from one responder: "[High performance in] OpAmp LPF [low-pass filter -- this is just the output OPA and its caps] in [a portable, battery-powered DAP] simply cannot be achieved without distortion [at] 2Vrms with limited supply voltage."

The CS4398 has built-in volume. So if you were going to put it in a portable device, what would be its "ideal" Line Out level?

Thanks!!
 
Any takers on this query?
A "kit/modder" on another forum literally noted in a PM: " I found a solution and not just going to discuss it with you (especially considering that the answer is more than simple)."
It's curious .... that indiv. is willing to discuss every other part of the project in detail. I'd actually caught the indiv. in a lie ... and you know how that goes ... to save face the string of cover ups progressively gets more and more absurd until your end up with remarks like the above quote.
Hey ... I'll stand corrected ... so that's why I'm posing the query here ...
"Thank you .... thankyouverymuch"
 
Administrator
Joined 2007
Paid Member
Variable level line out might use the class D headphone stage and so be dependant on loading. The headphones often form part of the filter.

Why not try plugging phones in while using LO and seeing if it has any effect.

(Just a guess at a possibility)
 
Lower DAC digital volume sounds better

Variable level line out might use the class D headphone stage and so be dependant on loading. The headphones often form part of the filter.

Why not try plugging phones in while using LO and seeing if it has any effect.
(Just a guess at a possibility)
Thx for your reply.
BOTH the LO (LPF) and HO stages use AD8397 opamps.
The internal config is:
Headphone: CS4398-->LO(LPF)-->HO-->jack
Line level: CS4398-->LO(LPF)-->jack
Where LO is line out stage (with low-pass filter) and HO is headphone amp out.
I have software control over the CS4398's slow/fast DF and vol. control. I also have EQ, but that is done in the main DSP.
Phones + LO simultaneously makes no diff. I have always held the belief that digital vol. decreases bit depth, unless done very carefully (Wadia?). I'm going to assume that lowering the DAP's volume still degrades sound, but MAYBE lower output from the DAC more than makes up for bit-depth loss because other components downstream act more linearly (e.g., output OPA) when the DAC isn't "shouting" at it. And/or my integrated amp's behaves better with lower input signal (e.g., its vol. being operated over entire range -- does an analog pot sound better at the higher end where there is less resistance?)

Oh ... in case anyone cares ... this is one of the DAPs ...
An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.
 
Ok, first I need to clear up a somewhat prevalent misunderstanding about digital audio, in a correctly designed system the word length sets how far down the noise floor is, and THAT IS ALL. Critically the system is LINEAR all the way down to (and even below) the noise floor!
This assumes of course that everything is dithered correctly and that the math is done with sufficient precision (Not a given in the portable market).

With a 16 bit system noise floor is 93dB below full scale output for example, and this is independent of the number of bits in use, so turning down by 20dB in the digital domain leaves your noise floor 73 dB below the level corresponding to the input full scale, but of course you can hear right down into it, and the noise remains uncorrelated with the signal.
In a 24 bit system the noise floor is set by the Boltzmann constant (Analogue stage noise always stomps the required dither noise level).

Again, bit depth sets the available dynamic range only, it does not set the 'resolution' (Whatever that means when applied to audio, I have never seen a cognent explanation), and you can hear well down below the LSB in a correctly designed digital system.

Assuming your player is runnig from a single lithium cell (Reasonable guess), the supply rail will be somewhere around 3.5 - 4V or so, so it is quite possible that the issue is running out of headroom in the analogue stages, (for all that that would be poor design), dropping the output 6dB would extablish this easily, as would recording a .wav file containing a clean two tone test pair and doing an FFT of the signal from the lineout while fiddling with the volume.

It is also quite possible that you will find that the digital gain control implementation leaves a lot to be desired, this math is easy to stuff up if you are hard of thinking.

Regards, Dan.
 
I have always held the belief that digital vol. decreases bit depth, unless done very carefully (Wadia?).

your believe is wrong and there are many modern dacs (last 5 years or so) that now make it pretty trivial as far as the control (calculated in the DAC itself), supplying 32-64bit floating point calculation and dither of course. you still have to maintain a low noise floor, but it doesnt 'remove bits' and you would hope to maintain a low noise floor regardless.

it would be really good if we could really try not to sustain old, somewhat cliched out of date information about digital here on this forum.

hollowman, do you get the same problem if you disable the EQ? I just wonder if the 2 are combining to cause problems, or if perhaps you are causing the signal to be too hot with EQ and then when you reduce the gain this ceases to be as much of a problem.
 
Last edited:
sorry what I mean is SNR is reduced, once you start eating into the signal. even with correct dither thats unavoidable yes?, you cant turn the analogue noise floor down digitally ;) now that would be handy

provided decent gain structure thats not going to present much of a problem, moreso with headphones if they are really sensitive; than speakers perhaps. I use digital attenuation on everything I own including my headphone rigs and I dont notice any issue at all even with my IEMs (JH13) and if anything is going to show it, its them, I wouldnt have it any other way. matching that level of performance with balanced analogue controls is pretty darn hard, HUGE and extremely expensive
 
Last edited:
That's about the truth of it, getting good interchannel matching in an analogue control scheme is hard, it is utterly trivial in the digital domain.

SNR is of course reduced in just the same way as you turn down an analogue system digital is not unique in this.

Regards, Dan.
 
Assuming your player is runnig from a single lithium cell (Reasonable guess), the supply rail will be somewhere around 3.5 - 4V or so, so it is quite possible that the issue is running out of headroom in the analogue stages....
The battery is 7.4v LiPO which is doubled via dc-dc converter. The stock OPAs were all spec'd with supply voltage (MIN) =5V. I replaced stock OPAs with AD8397 which has slightly lower supply voltage (MIN)= 3V. (But the lower MIN supply is NOT why I chose this OPA; in any case, the better perf. at lower vol. affects all OPAs I've swapped)

Yes, higher dither may be a possibility.
 
NOT headphone out (HO)!

On the issue of EQ ... EQ is always "off". But on or off, the volume issue I note is unaffected.
provided decent gain structure thats not going to present much of a problem, moreso with headphones if they are really sensitive; than speakers perhaps. I use digital attenuation on everything I own including my headphone rigs and I dont notice any issue at all even with my IEMs (JH13) ...
I'm not talking about the Headphone Out (HO) of this device. Only Line Out (LO), which is subsequently connected to an outboard headphone amp or integrated amp/receiver or even a preamp.
 
Something else to consider is that certain devices, like these DAPs, are actually at unity gain (100% volume) when their GUI (the displayed level on the screen) is at about the 66% "level" I orig. noted. And above or below that 66% is when ugly volume/signal processing kicks in.

Ref: Benchmark Media topical page notes: "When the digital volume control is set to 100% (or 0.0dB, or Unity Gain, depending on nomenclature), no multiplication is performed, and this distortion is not present."

About dithering ... well, if the volume control is done in the DAC as the CS4398 allows, one would think the designers at Cirrus would've been careful so the "dither factor" is minimized for internal vol. control (and for most if not all its range).
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.