Saturday, July 1, 2017

My objective DAC tests show my world is upside down

Well, everything I know is wrong.

First Tests (needing correction in various ways)

The best measuring DAC I own is the Emotiva Stealth DC-1, a relatively small but solidly made $499 device which uses sigma delta DACs and (by default) ASRC input.  Changing that to Synchronous (yes, the least expensive DAC I own gives a choice) has virtually no effect (neither S/N or THD, both of which pegged my Juli@ card residuals).  As Julian Hirsh said about the Marantz 10B in 1963, it is so good, I am unable to measure.  (Julian's test gear wasn't that great.  He could have easily measured high-by-todays-standards THD with the Sound Technology ST-1000a, which came out a few years later.  My test gear is aways behind my best equipment)  Anyway the S/N and Dynamic Range peg my limits at 110dB and the THD also at 0.0003%, using RightMark at 1kHz.  The actual performance is better, I just can't measure it.

The spectrum is very clean too (and I'm not sure now which if any peaks come from the input of my analyzer):

Emotiva Stealth DC-1

And that's at the unbalanced outputs.

I am as yet unable to measure balanced anything because I don't have the needed adapters, which I just ordered, for my Juli@ card.

The worst measuring Dac is my esteemed Master 7 Singularity, using default OS (which should have lowest measured distortion and noise).  I measure 0.13% THD, pretty mediocre for digital device.  However the other channel, sadly isn't working at all (it has 1.5V DC offset, hmm) and this channel might have some issues also, so bear in mind that might be a result of some malfunction.  I am hoping for warranty repair at this point, at least for the non-working channel.  (If user error, you pay for all shipping and parts, that may well apply.)

Now in this case the lions share of the distortion, reaching to about -69dB, is second harmonic.  That ought to be cancelled in full balanced mode...which I am unable to measure.  The other harmonics seem below -95dB, which would coincide with numbers like 0.005%.  I'm still hoping for something like that.  The spec was something like 0.0005%.  Go figure.

So, a big disappointment, but if it gets fixed with distortion below 0.01% I'll use it.  (It may already be there in this channel in balanced and high impedance.)

BTW these also measured with coax only choice for these tests, and PLL, which might or might not be the best.  I suspect no measureable effects, and that brings up something else.  The distortion progression, though it does show some non-harmonic distortions, is very clean.  In contrast, the Denon DVD-9000 has pretty good numbers (only about 3x worse than the Emotiva, rather than 50x worse, in distortion, but the noise is 10dB worse...and actually 3dB worse than Master 7), but the upper frequency spectrum (below -110dB) gets hashy looking, a lumpy partly correlated noise (I suspect aging parts).  Meanwhile the Master 7 shows discrete distortion spikes getting better above a thin line of noise--and no hash.

Denon DVD-9000

And, the Emotiva shows pretty much a perfect spectrum above the lowest noise of all, with just a few tiny ticks above the noise level for harmonic distortion, peaking below -130dB, no hashiness, no nothing, just noise at -135dB across the spectrum...which is my residual.

Anyway, as I said the Denon occupies a middle ground, pretty close to the Emotiva and quite acceptible I think of around 0.0012% THD and S/N about 100dB.  I don't understand why the S/N isn't better than might have to do with that hashiness.  But the S/N is still ok.

Still, the lack of hashiness would tilt me toward the Audio GD if it got down to 0.01% in balanced mode.

The DAC 19 measured 0.075% THD, not really all that great, and more 3rd harmonic than second, though just like the Master 7, no hashiness, just a bit too much distortion which looks like analog distortion

While the FPB and Master 7 are being repaired, I'm going to use the DAC 19 becuase it's very low latency matches the Emotiva, and the Denon measured the 300 msec additional latency (variable at different sampling rates) I found I had to compensate for the variable latency--and that's just too much trouble.  The Emotiva is dedicated to the subs which need the balanced connection.  I'm going to get more of those Emotiva's, however, or something just like it.  I would not be above trying something like that on the Krell instead of the subs.

Further Thoughts

I will be adding some pictures later, including the raw data, to this post.

The poor measurements of the Audio GD Dacs probably were made worse by the "punishing" 10k load of my analyzer card: the Julia@ in SE mode.  The 100k load of my Krell and Eagle amplifiers is probably going to improve those numbers by about a factor of 3, so I'd predict 0.02% distortion from the DAC 19.  I am going to try to measure this eventually.  That's as acceptable as it is sounding right now.  And balanced operation should dramatically reduce the mostly even order distortion of the Master 7 to below 0.01%, along with the higher impedance load.  I'm going to measure balanced operation later, and possibly buffered/balanced.  And analog distortion generally falls into the noise at lower levels anyway.

One problem, however, I don't get the benefits of the balanced Master 7 operation with the Eagle 2 amplifier, I might as well just use the Dac 19 for that I guess...or use some kind of intermediate balanced to SE buffer, I could repurpose my bedroom Emotive Preamp for that...except I need it in the bedroom.

I'm thinking Kingwa's discrete FET circuitry could be better.  I'm thinking I'd prefer a solution using the best of all opamps, the OPA 211, using good ole feedback, which is proving itself useful in my amplifier tests.  The OPA 211 uses hybrid germanium technology, and silicon FETs cannot do as well, and combined with top quality design and feedback, distortion should be able to actually meet the 0.0005% spec, I think that's what TI's reference amplifier does.

The graph almost seems to show the spike in 2nd order distortion for one channel in the Master 7...but actually both channels are the same channel, I had to use a Y adapter to connect to the Juli@ because otherwise it gave me 'average' numbers absurdly high.  It could be the peak is entirely due to this.  So I'm not counting the above spectrum and results as the official ones yet.

No comments:

Post a Comment