strange SMSL M500 high 3rd harmonic

[MusicNBeer]

AKM ak4452 is just $1.5, and has 107db SINAD. Super transparent.
ES9038Q2M costs 10x more. 9038Pro costs 50x more.

And the 9038 DAC is exhibiting distortion 25 dB higher than the ADC in my cheap Asus Xonar DSX sound card.

 

[half_dog]

in AP crosstalk test, one would clearly see sinad and graphs for both channels.

Example: This is the test for Topping A90, which has elevated crosstalk in balanced.
View attachment 126458

had amir did this test, he could not miss the flaw at all.

Can you tell about the procedure followed to do this test? I have tried to do something similar, but us-366 has some crosstalk at its RCA inputs, so I had to run one channel at time and than substrate them. This way I could reach the <90dB stereo crosstalk from Topping L30. I can't get reliable results capturing both channels at once with this interface.

 

[Toku]

I just got a brand new SMSL M500. it's the latest 1.7 firmware and 1.3 hardware version. (windows shows 2.06 in xmos utility).

The following measurements are done with computer -> USB-> M500 XLR output-> ADI-2 Pro input -> computer.

I use REW to generate the sine wave and plot the fft.

If I play sine wave from both L+R channels, or just the R channel, the FFT looks very normal:

View attachment 125602

However, if play sine wave only from the left channel, the third harmonics is 0.015%.

View attachment 125603

I tried different DACs (such as Topping's, or SMSL's M300), swapped cables, swapped L, R input channels, and can confirm this issue only happens with M500 left channel.

The issue only occurs when left channel is playing different audio compared to the right audio --- When L and R are both playing sine wave at the same volume, it's quite normal. When only the left channel is used, or when left is playing 90% volume and right is playing at 70% volume, the problem occurs.

Any guesses why it is like so?

And no, you won't find this test in either @amirm 's or @WolfX-700 's review.

Also, since SMSL SU-9 is basically the same hardware and software, an educated guess is this will happen in that device as well. though I'm not very sure.

Hi liu,
I have seen posts with exactly the same symptoms before. I think it was a product of SMSL. SU-8 ?, However, finding the post is difficult for me, who is not good at English.

 

[liu]

Hi liu,
I have seen posts with exactly the same symptoms before. I think it was a product of SMSL. SU-8 ?, However, finding the post is difficult for me, who is not good at English.

I don’t know. I think Amir should test their Su-9 and see if problem is there too. Sent him a pm. We will see how it goes.

 

[cursive]

Yep, exactly. It'll be transparent set to 30, but annoying for sure. It's nowhere near meeting its spec.

Is 40 the max volume on this unit, or is it 30 out of 99 volume?

 

[liu]

Is 40 the max volume on this unit, or is it 30 out of 99 volume?

40 is max. At 30it is nowhere near their spec performance. In my case as far as I remember I need to turn volume down to 20 to have acceptable thd behavior.

 

[MusicNBeer]

40 is max. At 30it is nowhere near their spec performance. In my case as far as I remember I need to turn volume down to 20 to have acceptable thd behavior.

Yea that's about what I see. I can't see much lower than 0.001% due to my budget ADC. This really is a major design blunder.

 

[liu]

Yea that's about what I see. I can't see much lower than 0.001% due to my budget ADC. This really is a major design blunder.

That’s why adi-2 is truly amazing!

 

[jruser]

I'm curious if the Sabaj D5 has this issue as well.

 

[WolfX-700]

Yesterday I asked my friend to buy a used M500 (to prevent SMSL from giving me special samples)

Then he gave me the M500 today:

Difference level high 3rd harmonic issue:confirmed:

More： Difference frequency high harmonic issue:discovered

Tomorrow (it is midnight in China now), I will try to contact SMSL

 

[liu]

Thank you Prof. Wolf! Nice work!

 

[WolfX-700]

Thank you Prof. Wolf! Nice work!

It's my pleasure

 

[liu]

@amirm , now that Prof. Wolf successfully reproduced this,
I guess all 9038Pro DACs produced by SMSL have the same problem.
You really need to retest their SU-9. Might be very broken.

 

[MusicNBeer]

Thanks Wolf! Yes exactly what we saw. It's a hot mess. I was happier in my ignorance.

 

[cursive]

I'm crossing my fingers that it somehow doesn't affect the SU-9, as that's what I have. I won't be home for probably another week or so, but I'd happily send mine to someone to be tested if it's reasonably close.

 

[lucadoc]

I'm wondering what we owner of M500 can do / obtain directly from SMSL to compensate this big flaw especially when used as pure Dac. I bought mine from Aoshida but cannot send it back to China, too costly

 

[jani80k]

RIP
Bought mine because of this forum.
Gonna sell it because of it, too.

 

[half_dog]

Not the best proof, but this has called my attetion - ignore the other measurements :v
The first pic is from the left channel - it hits the interface measurement limits;
The second one shows this strange THD behavior, third harmonic

 

[MusicNBeer]

I'm wondering what we owner of M500 can do / obtain directly from SMSL to compensate this big flaw especially when used as pure Dac. I bought mine from Aoshida but cannot send it back to China, too costly

Just turn down the volume to somewhere between 20 and 30. The bad behavior greatly reduces as you lower the volume 30 and below. Nowhere near cutting edge, but transparent still. If this measurement correlated to real world sound degradation, it would have shown in the customer reviews. Honestly, I think this DAC shootout stuff is totally pointless.

 

[jani80k]

Just turn down the volume to somewhere between 20 and 30.

Interesting point.
I picked up somewhere on this website that 38 would be a good volume level in RCA DAC only mode because it is closest to 4(?) Volts output?
So now you are recommending to turn it down to somewhere between 20 and 30 to achieve transparent results?
(Just for my understanding - by "transparent" you mean that the SINAD is high enough so that decoding errors are outside the audible spectrum?)