Hey Guys,
Thought I’d chime in to share a little technical info that (hopefully) will help to show that published noise specs often don’t tell the whole story when comparing DACs (or any other pieces of audio gear for that matter).
Only the noise floor is better: -145 dB. But we do not need more than 110 dB.
Not to be contentious, but neither the above claim nor the quoted specs are necessarily accurate. If the specs are true from DC, across the full audio spectrum and out to infinite frequency, then I would argue that –145 is not only measurably better than –110, but audibly as well. This fact harkens to the primary benefit of the
TDSS Custom Upgrades in that they reduce all measurable noise levels, be they audible or beyond the audible spectrum. Therefore, if –110 dB represented the limit of audibility then there would be little to no benefit from the TDSS Custom Upgrade of the DAC-10, seeing its stock published dynamic range is already an excellent 115 dB (which is not exactly the same as the S/N Ratio – but close enough). Yet, we have clearly shown that the Upgrade offers a very substantial improvement. Go figure.
That said, there must be more to the story than the published specs for all of these devices. Specifically, most implementations of Sigma-Delta DACs include noise-shaping techniques that “push” the noise up higher in frequency to beyond the 20KHz limit of human hearing. Then the “A-weighted” measurement protocol gives exceptional results because it diminishes the importance of low frequency noise artifacts in favor high frequency noise, but then is limited to only those below 20KHz.
https://en.wikipedia.org/wiki/A-weightingMany (if not most) industry folks would argue that while the above may be true, the A-weighted measurement still has value because it represents a good method for comparing audio gear. True… at a cursory level of comparison, but it’s not likely to be of much help when comparing the “sound” or sonic signature between different components – and that’s what we are really talking about here.
So what’s the problem? See here:
http://www.stereophile.com/reference/292noise/index.html#iTAAI5BUWgCvOCmq.97Furthermore, it’s a known fact that while noise energy above 20KHz can’t be heard directly, its affect on signals in the audio band certainly can be. Ever wonder why an aluminum dome tweeter can measure great in every way out to 20KHz, but it still doesn’t sound all the great and/or there are few that do? Compare the speakers offering them to the number that offer soft-dome tweeters if you have any doubts about the validity of that statement.
OK, the reason for that is if you measure that same aluminum-dome tweeter out to say... probably not necessary to go beyond 30KHz… you will likely find a HUGE resonant peak somewhere close to about 25KHz. In fact, there will likely be more acoustic output at that frequency than any other frequency from 20KHz on down. “So what’s the problem? You can’t hear 25KHz anyway” – and that’s what their manufacturers will try to tell you too. What they WON’T tell you though is that the energy above 20KHz will “beat” with frequencies in the audio band and produce sum & difference sideband tones. In other words, that 25KHz resonance will be a significant source of INTER-MODULATION DISTORTION. And that’s WHY there are very few good sounding aluminum dome tweeter.
That said, using a DAC design that “pushes” audible noise into the ultra-sonic band is NOT a panacea for improving the overall sound because the exact same process occurring in the tweeter example above is going to take place to some degree in the DAC. In fact, it also occurs in virtually all Class-D/switching amplifiers as well due to the UHF noise generated by their high switching frequencies (anywhere from 200KHz up to near 1MHz depending on make & model). Again, that’s the reason the TDSS Custom Upgrade of the NuForce & NuPrime amps offers the improvement that it does. Via proper shielding and grounding techniques (among others), the upgrade reduces the level of UHF noise circulating within the amplifier circuitry, which then reduces the effects of UHF noise inter-modulating with the audio band signals (not to mention the amp’s negative feedback loop).
So anyway, noise shaping makes for great A-weighted noise specs, but you can’t really say much more than that for the process – at least with respect to its ability to predict the final sonic performance of a component or for use in comparing components with the same goal in mind.
Furthermore, while the inter-modulation issue tends to be “static” in nature (not changing with changes in output signal level), there is another concern that affects the sonic signature of a DAC using Sigma-Delta conversion. In addition, you will note in the above Stereophile article that any remaining noise below 20KHz can be modulated in frequency in direct relation to output signal level changes. That means the noise “spectrum” will change as the audio volume level increase and decreases. The article also points out that it has been shown such variations in noise of no more than 1 dB can be heard by discerning listeners.
So there ya go. Go get a DAC-10 and LISTEN to it for yourself if you wanna know for sure. Short of some artificial intelligence algorithm running on a computer-based audio analyzer, the only other way to test these gizmos is to use the “natural intelligence” analyzer that you were born with between your ears. Now, if you don’t have one of those, uh... never mind.
Take care,
-Bob
