• 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 daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

PCM vs DSD

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
2,639
Likes
5,911
Location
Berlin, Germany
I'm saying that is not the case. It's always PCM going into the multi-bit modulator.

Besides, you haven't told us how DSD async resampling works?
By converting it to PCM just as you and I said ;-)
Simple DSD-inflation data in that case, mapping 1 to some positive high value and 0 to the same value but negative. The final DAC is PCM so nothing else can be feed to it. Obviously.

But we're nitpicking here. The key takeaway is that the DSD modulation causes no or very little signal-correlated RF glitch energy and this is why DSD mode has lower distortion.
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,130
Likes
17,609
Location
Netherlands
But we're nitpicking here. The key takeaway is that the DSD modulation causes no or very little signal-correlated RF glitch energy and this is why DSD mode has lower distortion.
But then again, why use a 1-bit modulator for DSD content that was converted to PCM? Why develop a 5th order modulator if it turns of that the 1-bitter is actually better?
 
Last edited:

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
2,639
Likes
5,911
Location
Berlin, Germany
But then again, why use a 1-bit modulator for DSD content that was converted to PCM? Why develop a 5th order ik to it modulator if it turns of that the 1-bitter is actually better?
Only ESS insiders could give us insight.
 
OP
Sokel

Sokel

Master Contributor
Joined
Sep 8, 2021
Messages
5,630
Likes
5,541
As a side note,this glitch moves proportionally to the test signal,you can already see that on the 50Hz and 100Hz signal measurements in previous posts.
(I like seeing this "dance" during sweeps,the whole grass area moving along to the right)
 
Last edited:

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,130
Likes
17,609
Location
Netherlands
Only ESS insiders could give us insight.
… or your assumptions are not correct… I’m not sure how well-filtered the current outputs are before the IV stage? One should be able to verify this easily. Should look something like this:
1688970857575.gif
 
Last edited:
OP
Sokel

Sokel

Master Contributor
Joined
Sep 8, 2021
Messages
5,630
Likes
5,541
… or your assumption are not correct… I’m not sure how well filtered the current outputs are before the IV stage? One should be able to verify this easily. Should look something like this:
View attachment 298122
You mean the actual waveforms?
I can show you this for the output but before the IV stage,how?
 

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,548
Likes
2,759
My educated guess:

In PCM mode, the ESS chip first does the usual upsampling and digital filtering, and (sometimes) resampling.
Yep @Sokel can you repeat just a couple of main measurements you do, for PCM353k vs DSD? PCM353 bypasses ESS 1st stage digital filtering

Just 100Hz tone and IMD hump
 

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,548
Likes
2,759
But then again, why use a 1-bit modulator for DSD content that was converted to PCM? Why develop a 5th order modulator if it turns of that the 1-bitter is actually better?
I think they are multiple 1-bits

1695166531494.jpeg
 
OP
Sokel

Sokel

Master Contributor
Joined
Sep 8, 2021
Messages
5,630
Likes
5,541
Yep @Sokel can you repeat just a couple of main measurements you do, for PCM353k vs DSD? PCM353 bypasses ESS 1st stage digital filtering

Just 100Hz tone and IMD hump
Yep.

96kpcm.PNG


100Hz PCM@96Khz

353kpcm.PNG


100Hz PCM @352.8Khz

DSD.PNG


100Hz DSD128 (DoP)

Instead for an IMD sweep I did something better,I measured the absolute worst point (about -30db level) so to see how it looks:



IMD96k.PNG


SMPTE 60Hz/7k 4:1 96Khz @ -30db


IMD353k.PNG


SMPTE 60Hz/7k 4:1 352.8Khz @ -30db

IMD DSD.PNG


SMPTE 60Hz/7k 4:1 DSD128 @ -30db

all IMD.PNG


SMPTE 60Hz/7k 4:1 all @ -30db

I don't know for the rest of 9038qm or ESS DACs' in general but this one is way happier with DSD (apart from some elevated noise).
 

Attachments

  • 1695170845875.png
    1695170845875.png
    316 KB · Views: 29
Last edited:

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,548
Likes
2,759
I don't know for the rest of 9038qm or ESS DACs' in general but this one is way happier with DSD (apart from some elevated noise).
HQPlayer's developer has mentioned this for years.

Might be related to my ESS slide I show above
 
OP
Sokel

Sokel

Master Contributor
Joined
Sep 8, 2021
Messages
5,630
Likes
5,541
HQPlayer's developer has mentioned this for years.

Might be related to my ESS slide I show above
We will only be sure if someone with nice gear (not like my ancient E-MU) repeat this test,my results may be repeatable for my gear but we will only know for sure if others confirm that too.
So,I take it just as a hint,please guys,someone repeat it with a nice E1DA or something!
 

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,548
Likes
2,759
We will only be sure if someone with nice gear (not like my ancient E-MU) repeat this test,my results may be repeatable for my gear but we will only know for sure if others confirm that too.
So,I take it just as a hint,please guys,someone repeat it with a nice E1DA or something!

Yes but over on AudiophileStyle forum HQPlayer's developer has been posting measurements showing such for 5+ years.

I think he showed same with Pro-ject Pre Box S2 here on ASR too (same DAC chip you measured)
 
OP
Sokel

Sokel

Master Contributor
Joined
Sep 8, 2021
Messages
5,630
Likes
5,541
Just for the fun of it that's what 20 years of evolution in engineering looks like:


IMD vs level PCM.PNGIMD vs level DSD.PNG
Khadas Tone (PCM+DSD)

emu imd.PNG
E-MU 0204

Yep,it's bad.

Lets now put it under the magnifying glass:


IMD96k.PNG
Khadas Tone


emu imd2.PNG
E-MU 0204

Yep(²),some 25db (or 10db for DSD) better IMD distortion for the 20 years old (really abused) poor interface.

Someone to change my mind that they don't "game" it for 1Khz measurements?
Or do I have to post a SINAD vs Frequency sweep too? (yes,apart for 300-2000Hz the old thing is comparable or better,specially low)

Before someone tells me that there are better DACs,I already know,is just for the fun (or for the WTF?) of it.
I mean after 20 years everything should be WAY better.
 

Attachments

  • IMD vs level PCM.PNG
    IMD vs level PCM.PNG
    149.6 KB · Views: 53

HDavidson

Active Member
Joined
Jun 1, 2019
Messages
151
Likes
113
In foobar2000 I convert 44.1/16 to DSD. What kind of transformation path does it take in the ESS ES9068ES (for example)? From DSD input to analogue output.
And does this make sense?
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,691
Have a look at this too. DSD256 not only resulted in higher noise floor, but also higher distortion than PCM.
Khadas tone board is known to have hump, but there are also many other ES9038 DACs with no hump. So deliberately use a DAC with hump for comparison cannot draw very useful conclusion.

E1DA%20%239038D6K%20-%20+3.1dBDSD%201kHz%20THD+N%20-%20DSD64-128-256.png



Specifically, Archimago converted a DSD512 signal to PCM then sent to the DAC, the figures immediately improved a lot. Pay attention to the Realtime DSD -> PCM annotation.
E1DA%20%239038D6K%20-%201kHz%20THD+N%20-%20PCM%20and%20DSD512-to-PCM.png



Also clear sign of higher distortion in DSD256 playback, similar to Archimago's previous tests on Oppo UDP205 and Topping D90SE.
E1DA%20%239038D6K%20-%201-10-Decade%20Multitone%2032%20-%20DSD64-128-256.png

E1DA%209038D6K%20RightMark%2024-69%20DSD%20vs%20PCM%20Graphs.png
 
Top Bottom