• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

E1DA 9038D performance according to df-metric

That'll change the phase considerably in the lows, assuming the coupling cap is directly at the output.
 
Last edited:
That'll change the phase considerably in the lows, assuming the coupling cap is directly at the output.
The box containing the dummy load was not found. I didn't have time so I gave up.
Is 600ohm ok? If so, I might be able to do it tomorrow.
 
Most folks will load that dongle with something between 16ohm and 50ohm.
600ohm will be too high but arguably is better than 200k or even 10k.
33 ohm (standard value) makes the most sense.
 
Most folks will load that dongle with something between 16ohm and 50ohm.
600ohm will be too high but arguably is better than 200k or even 10k.
33 ohm (standard value) makes the most sense.

My HD800s(re-bought) hasn't been plugged into anything besides a dongle yet... Had it for a month or so now.:cool:
 
There are always exceptions. :)
Besides with 600ohm load (and even higher in the bass !) the phase shifts will be much lower than for most users that drive their high efficiency portable headphones with it.
That is... still under the assumption there is a capacitor in the headphone out part (which would make sense).
 
@nagster then in that case no need to bother with more testing.

It does show that a gradual phase shift of 40degrees (-3 dB) @ 2Hz, that arguably is inaudible, has a quite severe effect on nulling and number generation when small and gradual phase shifts is not corrected for in the generation of a single number that is supposed to represent 'audio quality'.
 
@nagster then in that case no need to bother with more testing.

It does show that a gradual phase shift of 40degrees (-3 dB) @ 2Hz, that arguably is inaudible, has a quite severe effect on nulling and number generation when small and gradual phase shifts is not corrected for in the generation of a single number that is supposed to represent 'audio quality'.
got it. Well, thanks to that I was able to find the dummy load.
correction. The minimum load value built into the APx was 300ohm, not 600ohm. 300ohm was not in my mind.
 

Attachments

  • hpload_01.jpg
    hpload_01.jpg
    271.2 KB · Views: 103
  • apx_termination.png
    apx_termination.png
    12.1 KB · Views: 94
Nulling is a problematic method with respect to audibility. Rather make DBT with 2 files.
 
Phase measurement of 9038D.
I don't have a Shanling device.

* All measurements were conducted with AP balanced input (200kohm) and no load.
I was just looking at the settings for the ADI-2 DAC and found that it was set to Ref 13dBu instead of Ref 7dBu.
I can't edit it, so I'll correct it here.
 
I don't know exactly how various phase shifts are perceived by the ear and I also - as many - think that the low-freq phase shift is not too important. But I also know that if we replace all the phases of some music signal with the random ones, we will get just a noise shaped according to a freq content of the initial signal. In other words, audibility of phase errors has its thresholds/limits, should be characterized and hardly researched well with real music.

In this particular case of #9038 there is another source of the degradation and even more significant - time inconsistency of the output signal. And it affects both magnitudes and phases. It is most visible on phase diffrogram with the multi-tone signal [https://audiosciencereview.com/foru...nce-according-to-df-metric.47932/post-1732473]. Here is the distribution of errors (DF,dB) for our two devices with the multi-tone signal (2min, window:50ms):

9038-multi-wf.png

#9038 DFwf(median): -32.37 dB
9038-multi-mg.png

#9038 DFmg(median): -45.49 dB
9038-multi-ph.png

#9038 DFph(median): -34.84 dB
m0-multi-wf.png

M0 DFwf(median): -75.81 dB
m0-multi-mg.png

M0 DFmg(median): -75.58 dB
m0-multi-ph.png

M0 DFph(median): -61.53 dB


The time inconsistency makes the histograms of #9038 wider and significantly worsens the median values. Thanks to that inconsistency/jitter the difference signal is pretty loud in all frequencies and those frequency components are not stable throughout the whole signal.
 
I don't know exactly how various phase shifts are perceived by the ear and I also - as many - think that the low-freq phase shift is not too important. But I also know that if we replace all the phases of some music signal with the random ones, we will get just a noise shaped according to a freq content of the initial signal. In other words, audibility of phase errors has its thresholds/limits, should be characterized and hardly researched well with real music.

But that is not the case here. The phase shift is gradual and not big so inaudible.
What you should do is remove that phase error from the weighing of the 'sound quality' aspect.
Of course substantial and steep phase shifts in the 400Hz to 8kHz range should be included in the weighing.

Jitter (all kinds of jitter) can be measured and shown to exist.
 
Is the 'except the phase error' here the biggest issue of the end result ?
Even if it results in very small magnitude differences that are in perceptible a phase error, when not corrected will appear as a magnitude difference in the null.
 
The magnitude error (DFmg) does not account the “magnitude differences” you mention (caused by the phase errors) because magnitude components of the output signal are not amplified due to the phase errors. For example the magnitude diffrograms of #9038 do not show inaccuracy of low frequencies while the phase diffrograms do [https://audiosciencereview.com/foru...nce-according-to-df-metric.47932/post-1732473]. So, the DFmg measures exclusively the “tonal” mistakes of a DUT.

And the tonal degradation in #9038 is pretty big: -45.5dB vs. -75.6dB for M0. I attribute this inaccuracy to the time inconsistency, which affects both phases and magnitudes, and phases are affected to a greater extent (see the diffrograms with multitone signal).
 
Could it be 'Frequency stability' mentioned here?
The resultant frequency variation causes a pernicious deterioration in perceived pitch stability, without delivering an apparent degradation in standard audio performance parameters such as THD+N when measured with conventional techniques.
In archimago's test of this unit, he mentioned that it uses 9038's builtin oscillator. So there supposed to be less possibilty it suffers from clock syncing issue. Anyway in ESS 9038 spec sheet, there's this:
1727128548028.png

DSD mode has a higher default bandwidth than PCM. I personally have a D6k unit, and when I upsample from PCM to DSD, despite some aspects of improvement, I hear a little bit of less edgy but more smearing sound. I guess it in some degree correlate to ESS deciding to have more timing tolerance in DSD. This chip probably need some kind of algrithm to set this bandwidth, I'm curious how E1da implement it. Anyway I just want to see my unit perfect in every possible way.
Is it possible, that this phenomenon will distort measurements of another DUT in the case of RME Adi-2 Pro FS R BE?
BTW, RME's newest 2/4 pro does a wonderful job in This comparison.
 
Could it be 'Frequency stability' mentioned here?
I doubt it. Like most any recent design, this is an asynchronous (UAC2) job. If it were to perform sample rate gymnastics as shown, this would be quite obvious in the jitter department.
In archimago's test of this unit, he mentioned that it uses 9038's builtin oscillator. So there supposed to be less possibilty it suffers from clock syncing issue. Anyway in ESS 9038 spec sheet, there's this:
1727128548028.png
DPLL bandwidth is not generally an issue in USB devices. It becomes relevant when paired with S/P-DIF receivers of a relatively large PLL bandwidth, notoriously when manufacturers had to switch from AKM to older Cirrus parts post AKM factory fire.
 
I doubt it. Like most any recent design, this is an asynchronous (UAC2) job. If it were to perform sample rate gymnastics as shown, this would be quite obvious in the jitter department.

DPLL bandwidth is not generally an issue in USB devices. It becomes relevant when paired with S/P-DIF receivers of a relatively large PLL bandwidth, notoriously when manufacturers had to switch from AKM to older Cirrus parts post AKM factory fire.
Thanks for reply. I don't have any hardware background, things I can tell from what I read in edn.com is that asynchronous mode is usb audio implementation. Protocol underlying it is usb isochronous mode which doesn't have retransformation. And I think by asynchronous mode, it means USB dac as clock master asks for how many samples it needs from host. But if dac's own clock is inaccurate, it sees the sample rate it requests as objective clock and neither host or device sees it as an issue. I don't know about jitter test but is it detectable as jitter? Maybe someone repeat the test in edn and shows us the correlation between jitter test and it. Or perhaps only DF test can do it?
 
Back
Top Bottom