yggdrasil technical measurements

[Argo Duck]

+1

On the 'fool oneself' thing, reminds me of one of Ashleigh Brilliant's truisms: I've given up my search for truth and am now looking for a good fantasy to believe in. Audio does it for me. The more I can fool myself I'm "really there" the better

 

[artur9]

I'm curious as to which measurements would show the time and phase coherence that Mike/Baldr talks about here?
post #4954

 

[evillamer]

Would love to see chord Hugo/2qute/TT/DAVE measurements as well. Especially on the much touted WTA filters.

 

[AndreYew]

  I'm curious as to which measurements would show the time and phase coherence that Mike/Baldr talks about here?
post #4954

The ideal measurement is probably an impulse response like Stereophile's, but the square wave graphs here are OK too, but they look a little odd to me given atomicbob's testing parameters, specifically the testing sample rate of 44.1/24:
 

 
Let me explain why, and perhaps I messed something up. The preringing in the zoomed-in view of the square wave indicates that the top frequency preserved in the square wave is 23 kHz: count the number of up bumps from -0.5 to 0, and you get 11.5 bumps, since the zero crossing of the waveform is right at t=0. That's over a 0.5 millisecond interval, so the frequency is 11.5/(0.5e-3) = 23000 Hz. That's only possible with 48 kHz sample rate, so I'm confused. 44.1 would not have preserved 23 kHz.
 
I checked my math and interpretation of the measurement units on the zoomed-out view, which is indeed a 20 Hz square wave.
 
Anyway assuming a 48kHz sample rate, the ringing is not unusually low in terms of frequency: it will ring at the highest frequency of the square wave that can pass through all the filters. It may be lower than normal in terms of its overshoot, but that's affected by the output analog low-pass filter, as well as the digital filter's antialiasing response.
 
edit: Another way to look at the time domain issue is to compute the phase response of the DAC from its frequency response. I'm not sure there's much to improve here: standard DACs have pretty much flat phase response out to near Nyquist, even ancient delta-sigma DACs like the CS4329.

 

[atomicbob]

My apologies that the resolution of the graph is limited on upload, even in the maximum size allowed. If you could see the zoomed original, you would count 10.75 zero crossings in 500uS. However, there is some slew rate limiting to the oscilloscope capture also happening so the last zero crossing of the ring is just a little before the 0 trigger threshold. Adjusted the number of cycles between -0.5 and 0 is approximately 11. Using the method you chose would result in 22 KHz rather than 23 KHz.

 

[evillamer]

On the topic of DAC warmup, Audioquest had a technical paper on the jitter effects of cold vs 15min vs 1hr vs 24hrs. They mention jitter reach stasis beyond 24hours.
 
Worth looking for those interested:
 
http://www.audioquest.com/wp-content/uploads/2014/04/Phase-Noise-Jitter-Report-0317-14.pdf

 

[AndreYew]

Thanks for the zoomed in picture of the ringing!

 

[atomicbob]

It might be worth mentioning that the ODAC recently received an update in terms of components, and thus performance.
http://blog.jdslabs.com/?p=1003

Here's the new board's jitter measurements as done by JDS Labs with their Prism dScope III (keep in mind that they're using a 24/96 signal at 0 dBFS):

That's still pretty darn good for a ~$150 DAC from what I've seen.


I'd be curious to see how the AURALiC Vega compares though seeing as it has a femto clock.

And here is a Mousai MSD192 modulation measurement. This is from the spdif input as there was some interaction between the DAC's driver and the dScope such that the test would not complete properly. However prior to completion the measurement looked essentially the same as the one shown. I need to track down the problem in the dScope script. Also note that this performance is at 44 KHz instead of requiring 96 KHz to achieve. Note how narrow the trace is down to -140 dBFS compared to the ODAC at 11 KHz, even considering the expanded frequency axis for the MSD192.

 

[miceblue]

And here is a Mousai MSD192 modulation measurement. This is from the spdif input as there was some interaction between the DAC's driver and the dScope such that the test would not complete properly. However prior to completion the measurement looked essentially the same as the one shown. I need to track down the problem in the dScope script. Also note that this performance is at 44 KHz instead of requiring 96 KHz to achieve. Note how narrow the trace is down to -140 dBFS compared to the ODAC at 11 KHz, even considering the expanded frequency axis for the MSD192.

Very interesting indeed. How does one interpret these jitter "skirt" plots? I know that lower jitter measurements typically mean that the "skirt" is more narrow, but I don't know what this indicates in respect to the definition of jitter: timing discrepancies.

 

[atomicbob]

Very interesting indeed. How does one interpret these jitter "skirt" plots? I know that lower jitter measurements typically mean that the "skirt" is more narrow, but I don't know what this indicates in respect to the definition of jitter: timing discrepancies.

The wider skirt for the ODAC means that the jitter has a greater influence at low audio frequencies than at higher frequencies. So the MSD192 has significantly less jitter affecting 100 Hz than the ODAC, for example.

 

[cspirou]

  except that jitter at sub nanosecond levels hasn't been shown to be audible in controlled listening tests - the audiophile jitter reduction sub industry hasn't published new controlled testing verifying their claims - its all hand waving, anecdotes and FUD
 
published peer reviewed jitter audibility thresholds are from decades old papers that show 100s of nanoseconds for random jitter, 10s of nanoseconds for correlated or tonal jitter
 
jitter so low it is characterized as frequency drift does improve in xtal oscillators with aging - nowhere near audio significant frequencies
 
I expect most correlated jitter is built into the circuit - electrical and mechanical construction, parasitics coupling noise to the clock
 
I don't see how sideband tones can improve with aging - you really have to have lots more measurements to show correlation with on time/aging vs other electrical environment changes - what other equipment is plugged in, loading, line noise even what other sw may be running
 
very common is day vs night - how much else is on, actively polluting the electrical environment with EMI

 
Reducing jitter may not be an issue for Delta-Sigma DAC chips. However this is a very different DAC using high precision 21-bit chips meant for industrial applications. Very low jitter and ultra-precise timing might be a necessity for this DAC to operate correctly.

 

[evillamer]

@atomicbob, a small request when you decide to do a cold boot measurement of the Ygg.

Can you kindly run a jitter measurement test again on the Yggdrasil. This time with different types of usb cables? Standard monoprice/OEM grade vs your audio quest forest and whatever you can get your hand on.

Just want to see if Yggdrasil excellent jitter rejection is influenced by USB cables.

 

[atomicbob]

@atomicbob, a small request when you decide to do a cold boot measurement of the Ygg.

Can you kindly run a jitter measurement test again on the Yggdrasil. This time with different types of usb cables? Standard monoprice/OEM grade vs your audio quest forest and whatever you can get your hand on.

Just want to see if Yggdrasil excellent jitter rejection is influenced by USB cables.

I'll try to remember. Not a matter of not wanting to measure but of an ever forgetful memory at this point in my life.

 

[jcx]

 
Reducing jitter may not be an issue for Delta-Sigma DAC chips. However this is a very different DAC using high precision 21-bit chips meant for industrial applications. Very low jitter and ultra-precise timing might be a necessity for this DAC to operate correctly.

 
actually single bit DSD delta-sigma are the most jitter sensitive type, multibit reduces jitter sensitivity
 
switched capacitor architectures that dispense a fixed charge per processing sample interval are highly insensitive to jitter - these are mostly Vout types which many dislike because you are stuck with the internal CMOS op amps that can be built on the digital/mixed signal process used for the DAC
 
full bit depth R2R Zero Order Hold types are next best for low jitter sensitivity
 
oversampling improves random jitter immunity, doesn't change the results of the more audible correlated/tonal jitter
 
 
yggy's 20 bit R2R ladder with the ZOH should be on the more insensitive end with respect to jitter, and the jitter plots are excellent - absolutely no reason to believe even the "cold/new" jitter plot is within orders of magnitude of giving a audible jitter problem

 

[coli]

   
actually single bit DSD delta-sigma are the most jitter sensitive type, multibit reduces jitter sensitivity
 
switched capacitor architectures that dispense a fixed charge per processing sample interval are highly insensitive to jitter - these are mostly Vout types which many dislike because you are stuck with the internal CMOS op amps that can be built on the digital/mixed signal process used for the DAC
 
full bit depth R2R Zero Order Hold types are next best for low jitter sensitivity
 
oversampling improves random jitter immunity, doesn't change the results of the more audible correlated/tonal jitter
 
 
yggy's 20 bit R2R ladder with the ZOH should be on the more insensitive end with respect to jitter, and the jitter plots are excellent - absolutely no reason to believe even the "cold/new" jitter plot is within orders of magnitude of giving a audible jitter problem

You know what you are talking about. I wish I know what you know before I made certain purchase in the past.