Effective number of bits---or why "you have to keep software at full volume" is nonsense

[stv014]

Quote:

Yes, but there is nothing to prevent a naiive user turning down the digital volume and then turning up the volume at an external analog post-processing stage.

 
In which case they will hear the noise, and know that there is a problem. But with a decent (not even expensive) DAC, the digital volume needs to be turned down a fair amount (>20 dB) before the noise becomes a real issue (try these files to get a rough idea how many bits the resolution can be reduced to before audible hiss appears during music playback).
 
Of course, 100% digital volume may be a "safe" option for naive users, but in fact it is not always the case. It can result in clipping (e.g. the input stage of the amplifier cannot handle the full scale output of the DAC) or channel imbalance (with analog volume pots that need to be set to a very low volume because of the high input signal level), for example.

 

[mikeaj]

Nah, but there are legitimately plenty of users with no analog volume control other than on the amp that are brainwashed into not touching digital volume. This is sometimes even the case with a source, amp, and headphones such that they're dealing with channel imbalance at usable listening volumes, not to mention being in the hypersensitive part of the pot rotation.
 
How many people are really going to set volume in software too low and have gain and volume cranked up somewhere else down the chain? I think we encounter more of the former than the latter here.

 

[ab initio]

Quote:

Nah, but there are legitimately plenty of users with no analog volume control other than on the amp that are brainwashed into not touching digital volume. This is sometimes even the case with a source, amp, and headphones such that they're dealing with channel imbalance at usable listening volumes, not to mention being in the hypersensitive part of the pot rotation.

 
This is exactly the point. I have a modi+magni (i.e. loud) and HD 280pros (i.e. fairly sensitive and isolating). With software volume maxed, and the magni volume knob at 9 o'clock (which is where channel imbalance disappears) the audio is loud. At maxed volume software volume, I have serious concerns regarding the magic smoke leaving my headphones if I bump the volume knob too high.
 
In a quiet listening environment, it takes -35dB to get 9 o'clock to be a good "quiet" level. If I'm working and I want the music further in the background, I'll use another -15dB. The point is, if I want my music quiet, I don't need full resolution because I'm not concentrating on it. But if I am listening to music, even with -35dB, I still have more than 65--75dB of dynamic range and I can actually use the volume knob on the Magni between 9 and 3 oclock. The noise levels just aren't a problem. I'm more limited by my audibility threshold and ambient SPL levels than by the noise floor of the source+amp+headphones.
 
65dB + ambient SPL levels in a very calm room = enough dynamic range to reach  hearing damage levels before the noise is audible. The best thing is that I can always turn the software volume up if I want more juice or have a quite song---and there will be no sound degradation as I turn up software volume.

That's exactly the point---you can use software volume control to achieve a comfortable listening level because the artifacts are below the threshold of audibility for all but the very most sensitive configurations. Keeping software volume at 100% forces people to listen at higher volume levels unless they have a really really fancy analog volume control (or inefficient headphones, in which case, by all means max out software volume). Either way, the only time using 100% volume gives you a dynamic range advantage is when peak SPL levels are well above damage threshold levels.
 
Of course, I have assumed that everybody is using a 24bit soundcard/DAC. If you have a 16bit soundcard/DAC, then you probably don't want to use much more than -6 dB in software if possible.
 
Cheers

 

[xnor]

Let's face it. The real problem is the stupidly high fixed gain in the amp(s).
May be easier to just use higher gain for the manufacturer instead of doing proper compensation, but I find that to be ruthless. Half of the posts I've read about the magni deal with exactly that problem. And what I find exceptionally questionable is saying that an amp with such high gain works well with IEMs as well.

 

[Joe Bloggs]

Quote:

Let's face it. The real problem is the stupidly high fixed gain in the amp(s).
May be easier to just use higher gain for the manufacturer instead of doing proper compensation, but I find that to be ruthless. Half of the posts I've read about the magni deal with exactly that problem. And what I find exceptionally questionable is saying that an amp with such high gain works well with IEMs as well.

 
I've been given to think that an amp works best when the volume knob is maxed out on whatever gain setting is used--so I had an idea--why not make two knobs, one for the volume and one for the gain?  You keep the gain pegged at 1x most of the time, and only use the volume knob.  If and when you don't get enough volume with the volume knob maxed out, you start turning the gain knob.  Such a design should maximise the available SNR at all times, right?  Any drawbacks to such a design?

 

[ab initio]

Quote:

Let's face it. The real problem is the stupidly high fixed gain in the amp(s).
May be easier to just use higher gain for the manufacturer instead of doing proper compensation, but I find that to be ruthless. Half of the posts I've read about the magni deal with exactly that problem. And what I find exceptionally questionable is saying that an amp with such high gain works well with IEMs as well.

 
I half agree and half disagree on this point. Yes, the magni is loud, but the modi+magni combo is quiet enough that I have absolutely no problem whatsoever getting the full useable range out of the volume pot by using a little bit of software volume control, and the guys from Schiit are very honest about that. By comparison, my laptop's soundcard blasts me with tons of noise compared the Modi+Magni (which I don't hear any noise until 1--3 o'clock on the volume (or even not at all) depending on the ambient noise. So in that respect, the M+M stack are a big step up from my laptops soundcard.
 
That said, would there be better a better choice for a headphone amp for IEMs? Well, I'm sure there is (although I don't have any experience otherwise). No, I don't need 15 volt peaks into earbuds (or my full sized headphones for that matter). To be fair it was intended to be an inexpensive jack-of-all-trades, and the Schiit guys do say "magni is quiet enough for many IEMs" and not "Dead silent for all IEMs". But, if somebody asked me what amp to get for their SE535's, I wouldn't start by recommending the magni!
 
 
Quote:

 
I've been given to think that an amp works best when the volume knob is maxed out on whatever gain setting is used...

 
What makes you think this? Is there some reasoning behind it?
 
Quote:

so I had an idea--why not make two knobs, one for the volume and one for the gain?  You keep the gain pegged at 1x most of the time, and only use the volume knob.  If and when you don't get enough volume with the volume knob maxed out, you start turning the gain knob.  Such a design should maximise the available SNR at all times, right?  Any drawbacks to such a design?

 
I think more components in the way introduce more places for things to go awry, and they add costs for extra parts and complexity. There are lots of amps with gain switches---I'm guessing they don't use continuous adjustment pots is because the preamp will be more sensitive to component quality and it's much easier/cheaper to get and switch between a couple of precision resistors than it is to get precision 2-channel pots.
 
If amps really only "worked best" at full bore and it was easy to implement continuously adjustable gain over a usable range, then why wouldn't the amps be designed with "max volume" hard-wired and the volume knob actually a gain knob?
 
I would guess that the amp should have some "sweet spot" that represents a compromise between signal to noise coming in and the amp's linearity coming out.
 
Cheers!

 

[xnor]

Quote:

What makes you think this? Is there some reasoning behind it?

The reasoning behind this is that the least amount of gain that gets the job done is the best: generally less noise, less distortion, proper usable volume control range.
 
 
@Joe:
There are a couple of problems with using a pot to adjust the gain (used as volume control). Some I can think of:
- most of the time you'll need far less than 1x gain and min gain with non-inverting topology is 1x
- pots can get noisy
- op amps are not necessarily stable with all gains
- DC through the pot causes noise
- probably longer feedback loop if you mount the pot on the front plate, which can cause all kinds of problems
 
It's not impossible though.

 

[Happy Camper]

It's the "inexpensive" gear that this is a problem with. The newer top end amps have both but the gain select is a toggle, not a pot.

 

[sonitus mirus]

One thing about the Schiit Modi+Magni is that the DAC only provides a maximum output level of 1.5 V.  Paired with a standard 2 V source, the Magni would be far more difficult to use with highly sensitive headphones.  Using a Modi with the Magni or connecting directly with one of several popular portable players might be ok for many iems and other efficient headphones, but it seems that cutting corners on cost was a deciding factor on the choice for designing the Modi with a 1.5 V output.   This was a way to balance the relatively high, fixed gain of the Magni.

 

[xnor]

20*log10(2 V / 1.5 V) = "only" 2.5 dB difference
 
Sure it helps, but the gain (5x) is about +14 dB.

 

[sonitus mirus]

You and your fancy math. 

 
I had to go look up the specifications of my iPod Touch's (4th gen.) LOD output.  I was going on the assumption that it was 1.4 V for some odd reason, but after working the numbers, a 3 dB difference between my 2 V DAC did not seem audibly correct to my ears.  Turns out my iPod is only capable of 1 V at full gain.  I suppose it is not that significant either way.

 

[xnor]

Half the voltage is -6 dB which is significant. You'll still get like 5V max (instead of 10V with a 2V source if we ignore clipping) which is still in excess of what many (especially low impedance) headphones, let alone IEMs, need to get loud.
 
See http://www.head-fi.org/t/668238/headphones-sensitivity-impedance-required-v-i-p-amplifier-gain

 

[sonitus mirus]

Very informative stuff.  I've been using my Nuforce Icon HDP as a DAC with the RCA output, and the specs claim a max output of 2.6 Vrms.  After reading the recent comments in this thread, I've been lowering my DAC's volume setting from around 75% of max to about 50%.  Even with my typical 16-bit 320 kbps mp3 streams, I've not been able to hear any hiss during playback when I raise the volume level with my amps.  Every DAC I own is now on the lowest gain setting they have (where applicable) and I'm able to make much better use of my amp's volume control while avoiding the dreaded channel imbalance issues at more comfortable listening levels.
 
Thanks, this has been extremely helpful.

 

[xnor]

I'm really wondering where this idea and trend - at least that's what I've observed - "the more gain the better" came from. Maybe an alteration of the idea that more power = better?

 

[ab initio]

I'm really wondering where this idea and trend - at least that's what I've observed - "the more gain the better" came from. Maybe an alteration of the idea that more power = better?

It's all about overdriving those tewbs!