How important is signal input level (volts) from a DAC into an active speaker in influencing perceived sound quality?

[Robbo99999]

So like the title says I'm trying to get some validation on the influence of the signal input level (V) from a DAC into an active speaker in terms of perceived (& measured) sound quality. Also, what are the objectively measured sound quality variables that can be influenced negatively or positively by the size of the signal input level (V) from a DAC into active speakers, and can this have an effect on perceived sound quality? (I'm not talking about maximum volume achievable by the speakers because that's somewhat obvious, but I'm talking about other aspects of sound quality assuming equal listening volume levels).

I ask these questions because of my own experience using active speakers (JBL 308p Mkii's) and using them with different DACS that output different max signals. I've found that my Topping E30 DAC (output 2V) seems to sound better than my SoundblasterX G6 DAC (output about 1.5V) and also my Asus Xonar U7 (output 1V). Now all of these DACS are 'good DACS' in as much as some people might say they should sound the same in perceived sound quality as any differences are likely to be above the thresholds of perceived hearing...that's why I thought it might be the different output voltages that are influencing the sound quality of my active speakers, and hence my question I'm asking in the first paragraph of this post. (I'm using up 5dB of negative preamp through software to allow for room eq if that has any bearing on this, probably not).

To round it off I'll do a poll you folks can vote in.

 

[solderdude]

When comparing did you compensate for the differences in output level ?
Did you use the E30 with 6dB attenuation when comparing to the U7 for instance ?

This seems like it would be a logical thing to do.

I voted 'yes' because a signal level that is higher in level compared to a lower level is always perceived as better sounding (unless an input starts clipping)

Another question would be... if one were to level match would there still be perceived differences. The answer to that question would depend on what DACs one would be comparing, in most cases the answer would be 'no' (when tested blind) and probably 'yes' when tested sighted.

 

[Blumlein 88]

What solderdude said. Did you carefully match levels?

To be clear, measuring with a volt meter the output so each DAC has the same signal level within 1% of each other. Level matching by ear is too coarse and you can hear a quality difference matched this way.

 

[samsa]

I guess I'm wondering the same thing as the previous 2 respondents.

• DAC A with an output voltage of 1.5V.
• DAC B with an output voltage of 2V, but attenuated by an additional 2.5 dB.

Where did you do the attenuation? Using the "Pre" mode of the E30, or using the analog volume control on your speakers?

 

[Aerith Gainsborough]

Can signal level (V) from DAC to active speaker affect perceived sound quality?

Of course it can.
-If the signal level is too low, the gain of the amplifier might not be enough to reach the desired volume or you have to turn up so much that you start amplifying the noise floor into the audible range. I actually have this problem with my AVR, because Dirac Live pushes the overall signal level down another 12.5 dB.
-If the signal level is too high, it might overload the input stage and cause audible distortion.

Outside of these rather rare scenarios: no. The only difference you will see is is a different position of the volume knob.

 

[Robbo99999]

Thanks for your replies folks. A lot of you asking the same thing, and I did kind of expect that question: did I level match in my listening tests between the different DACS? The answer to that is no, I didn't. I am familiar with the louder is better phenomenon, and I do have the ability to level match using UMIK and writing down the various volume settings I'm using, however switching between the hardware would take a long time during the listening tests, so even when level matched the tests are always gonna have a lot of 'subjective problems' because as we all know we can't really remember sound quality too finely unless we switch 'immediately' between sources. So I didn't bother level matching as I knew there was this overarching issue in the testing and felt it wasn't worth the extra hassle because of that. So it's fine to assume my subjective testing is on the extreme side of 'subjective'! So I don't need you to believe my perceptions, but I was curious what kind of theoretical sound quality differences could potentially be observed by different input signal levels into the speakers (independant of final volume of output from the speakersJ)? And what would be the measured differences be in sound quality, in terms of variables like SNR or other quality variables (I'm hazy on which variables would be influenced or important) and that's part of my question?

And regarding @Aerith Gainsborough 's post where the variable of "noise floor amplification" comes in if you use a lower signal level, however @solderdude is intimating at the end of his post that this would not really be a noticeable variable between DACS that are outputting different max signal levels. What you guys think on this?

 

[Robbo99999]

I guess I'm wondering the same thing as the previous 2 respondents.

• DAC A with an output voltage of 1.5V.
• DAC B with an output voltage of 2V, but attenuated by an additional 2.5 dB.

Where did you do the attenuation? Using the "Pre" mode of the E30, or using the analog volume control on your speakers?

(SoundblasterX G6 run at 79% windows volume to prevent distortion as noticed in Amir's review of the G6. Topping E30 DAC run in preamp mode at 0dB and 100% windows volume. Asus Xonar U7 run at 100% windows volume. Also note there is an additional 5dB of attenuation in software preamp to allow for room EQ which is done through Equaliser APO)

 

[wwenze]

A.k.a. SINAD vs output level graph

With most DACs today SINAD improves as output level increases, and then degrades slightly from distortion as max level is reached. Good DACs probably don't even have that rise at the top.

 

[Aerith Gainsborough]

but I was curious what kind of theoretical sound quality differences could potentially be observed by different input signal levels into the speakers

My reply is to be understood in that theoretical context.

This should not be of any concern if you work with standard consumer line-out devices that work within 1.5V - 2V output. Professional gear can achieve higher line out voltages, so you need to pay attention to choosing the correct settings.

 

[Robbo99999]

A.k.a. SINAD vs output level graph

With most DACs today SINAD improves as output level increases, and then degrades slightly from distortion as max level is reached. Good DACs probably don't even have that rise at the top.

Interesting, sounds promising....what's that? Is that the SINAD of the DAC you're talking? So, you're saying a graph that shows how SINAD of the DAC is affected for different output signal levels?

 

[Robbo99999]

My reply is to be understood in that theoretical context.

This should not be of any concern if you work with standard consumer line-out devices that work within 1.5V - 2V output. Professional gear can achieve higher line out voltages, so you need to pay attention to choosing the correct settings.

Cool, so you're saying if the DAC is outputting anything from 1.5V - 2V then this is unlikely to be causing any noise floor amplification problems or any associated sound quality loss? I suppose if you used a large software negative preamp for something like room eq then this could lower it below the 1.5V-2.0V that you're referring to.....I'm running 5dB of room EQ which I think equates to reducing the output to a factor of 0.56 so I'm nearly halving my output.....so that would mean my E30 effectively outputs 1.12V, my G6 0.84V, and Xonar U7 0.56. Hmm, could this be the explanation for why the E30 sounds better I wonder?

 

[samsa]

(SoundblasterX G6 run at 79% windows volume to prevent distortion as noticed in Amir's review of the G6. Topping E30 DAC run in preamp mode at 0dB and 100% windows volume. Asus Xonar U7 run at 100% windows volume. Also note there is an additional 5dB of attenuation in software preamp to allow for room EQ which is done through Equaliser APO)

I'm sorry, if you're not even approximately matching levels, I'm not sure what the point of the comparison is.

As I said in my post,

• DAC A with an output voltage of 1.5V.
• DAC B with an output voltage of 2V, but attenuated by an additional 2.5 dB.

The E30 needs to be attenuated by ~2.5dB to even be crudely compared to the SoundblasterX G6.

If you're not even going to make a crude attempt at level matching, what's the point of discussing possible mechanisms for the residual difference after level-matching?

 

[wwenze]

Interesting, sounds promising....what's that? Is that the SINAD of the DAC you're talking? So, you're saying a graph that shows how SINAD of the DAC is affected for different output signal levels?

Amir has IMD(+noise) vs output level in his reviews, which also does the job just as well.

Spoiler

Spoiler

Unfortunately those are digital levels I think (dBFS) and needs to be converted to volts...

 

[Aerith Gainsborough]

Interesting, sounds promising....what's that? Is that the SINAD of the DAC you're talking? So, you're saying a graph that shows how SINAD of the DAC is affected for different output signal levels?

To put it into simple terms with some easy numbers for clarification: Any device has a fixed noise floor and a variable signal.
You want to to amplify the variable signal into the audible range but not the noise floor but you always amplify everything the device puts out. Signal, Noise, Distortions. So you need a sufficiently strong signal and sufficiently low noise floor to begin with.

If your noise sits at -130dB and your strong signal at 0 dB you have a Signal to noise ratio of 130dB. If your very small signal sits at -80dB you have a worse S/N ratio of 50dB. Any amp will boost both, so you might end up with an audible noise floor. That's why SINAD vs output level changes, despite the noise floor being the same.

Whether you have a signal of 0dB or - 10dB won't make an audible difference, because your S/N ratio is still 120dB.

Regarding my particular experience:
I do everything in the digital domain, and only have one amp in the chain, a low performance budget one at that. That's probably the root cause of my noise issues. After Dirac Live, there is only so much digital gain I can apply before I run into clipping, so the rest is up to my AVR. If I had an RME DAC in the chain that would convert the signal and act as a pre-amp this would be a non issue. (Working on it, stupid real life currency accumulation nonsense) ._.

 

[Robbo99999]

I'm sorry, if you're not even approximately matching levels, I'm not sure what the point of the comparison is.

As I said in my post,



The E30 needs to be attenuated by ~2.5dB to even be crudely compared to the SoundblasterX G6.

If you're not even going to make a crude attempt at level matching, what's the point of discussing possible mechanisms for the residual difference after level-matching?

Yes, I'm not expecting you to believe my subjective testing, and that is not the goal of my thread. My goal is to see if there are theoretical possible reasons for differing signal levels to be perceived as different sound quality (assuming that measured output level from the speakers is equal)....and also to understand what measured sound quality variables would equate to that. So this thread is a learning objective for me, rather than trying to prove to you my DACs sound different. (It was my very subjective & very imperfect observation that my E30 sounded better than my other DACS that led me to this theory re signal level and sound quality, and I'm exploring that here.)

 

[solderdude]

however @solderdude is intimating at the end of his post that this would not really be a noticeable variable between DACS that are outputting different max signal levels. What you guys think on this?

What I am saying is that when levels are matched, so for instance each DAC has 1V exactly on their outputs with 0dB signals, and the DACs all reach the same minimum audible limit parameters then it is highly unlikely you could notice audible differences when you don't know which DAC is playing.
However, when one DOES know which DAC is playing the brain can easily create audible differences where there are none.
This is the essence of blind testing.

When the levels are NOT (carefully) level matched the louder one (even if it is just barely) will be perceived as better sounding.
This is why exact level matching is essential.

The third parameter is statistical relevance. You need to do blind and level matched comparisons enough times to ensure guessing correctly is unlikely.

These 3 things are essential for truth-finding.

When the goal is to establish a good feeling of sound and you don't care about truthfinding, just personal pleasantness all this testing is not needed.
Pick one you feel is the best for you as this gives ease of mind which is important for enjoyment. Has nothing to do with actual, technical performance.
The loudest one will most likely 'win' in this case.

 

[samsa]

Yes, I'm not expecting you to believe my subjective testing, and that is not the goal of my thread. My goal is to see if there are theoretical possible reasons for differing signal levels to be perceived as different sound quality (assuming that measured output level from the speakers is equal)....and also to understand what measured sound quality variables would equate to that. So this thread is a learning objective for me, rather than trying to prove to you my DACs sound different.

I wasn't assuming you were trying to prove anything to me.

What I was pointing out was: if there's one drop-dead obvious reason for a perceived difference in SQ, surely you want to eliminate that, before plunging down the rabbit-hole of more subtle effects.

Say, after you made the obvious correction in levels, the effect went away entirely. The two sources sounded the same. Wouldn't you feel foolish having wasted time hunting for more complicated explanations for a difference that was simply due to the discrepancy in levels?

 

[Robbo99999]

To put it into simple terms with some easy numbers for clarification: Any device has a fixed noise floor and a variable signal.
You want to to amplify the variable signal into the audible range but not the noise floor but you always amplify everything the device puts out. Signal, Noise, Distortions. So you need a sufficiently strong signal and sufficiently low noise floor to begin with.

If your noise sits at -130dB and your strong signal at 0 dB you have a Signal to noise ratio of 130dB. If your very small signal sits at -80dB you have a worse S/N ratio of 50dB. Any amp will boost both, so you might end up with an audible noise floor. That's why SINAD vs output level changes, despite the noise floor being the same.

Whether you have a signal of 0dB or - 10dB won't make an audible difference, because your S/N ratio is still 120dB.

Regarding my particular experience:
I do everything in the digital domain, and only have one amp in the chain, a low performance budget one at that. That's probably the root cause of my noise issues. After Dirac Live, there is only so much digital gain I can apply before I run into clipping, so the rest is up to my AVR. If I had an RME DAC in the chain that would convert the signal and act as a pre-amp this would be a non issue. (Working on it, stupid real life currency accumulation nonsense) ._.

Thanks, I see what you're saying there re how the -dB signal level subtracts from the noise floor. As you intimate, I'm on -5dB for room EQ so I should subtract 5dB from the noise floor stats of the DAC - so for the E30 that has SINAD of 112dB so that would be reduced down to an effective 107dB after my room EQ, have I got that right? And as long as that resultant SINAD is over perceptible limits then there shouldn't be any negative effect theoretically, so I'm guessing ballpark 90dB as the cutoff point?

Amir has IMD(+noise) vs output level in his reviews, which also does the job just as well.

Spoiler

Spoiler

Unfortunately those are digital levels I think (dBFS) and needs to be converted to volts...

Thanks, yep, I've noticed those graphs before, I'm not sure if they paint the whole picture, aren't there other forms of distortion that "make up the whole", for instance that's not SINAD vs dBFS or is it? Is SINAD the holy grail of measurements that relate to our discussions in this thread?

What I am saying is that when levels are matched, so for instance each DAC has 1V exactly on their outputs, and the DACs all reach the same minimum audible limit parameters then it is highly unlikely you could notice audible differences when you don't know which DAC is playing.
However, when one DOES know which DAC is playing the brain can easily create audible differences where there are none.
This is the essence of blind testing.

Ah, yes, I see what you mean. And this might relate to my reply in this post to @wwenze above, in as much as where the DAC happens to find itself in it's dBFS vs "sound quality" curve?

 

[Robbo99999]

I wasn't assuming you were trying to prove anything to me.

What I was pointing out was: if there's one drop-dead obvious reason for a perceived difference in SQ, surely you want to eliminate that, before plunging down the rabbit-hole of more subtle effects.

Say, after you made the obvious correction in levels, the effect went away entirely. The two sources sounded the same. Wouldn't you feel foolish having wasted time hunting for more complicated explanations for a difference that was simply due to the discrepancy in levels?

I don't disagree with you....I'm theorising though, I'm not hunting or wasting time....I'm theorising and learning triggered by a loose observation, nothing wrong with that.

EDIT: I also don't have the means to do ultra rigorous comparative listening tests, because as I previously described, it takes too long to change the DAC over, thereby too much time elapsed before listening to comparative samples, so felt that level matching also was made irrelevant given this flaw...I'm not gonna go to the hassle of that when I know it's got flaws above & beyond that.

 

[solderdude]

The difference between the DACs in level is just 6dB. That is inconsequential for DAC's as most DACs have very low noise levels that will remain undetectable when attenuated a few dB. In most DACs measured distortion is often even lower than at 0dB.
When the level differences were substantial (say 50dB or so) you could probably tell DACs apart by the amount of noise differing between them.
Not when level differences are merely 6dB and compensated so they all put out the exact same level at 0dB FS.

btw... SINAD often improves a few dB when the output is below 0dBFS. And no, SINAD is not the most important parameter. It says something about the ratio (in dB) of total undesired 'crap' in the audioband relative to the output level of a 1kHz sinewave at a certain level (often 0dBFS or just below)