strange SMSL M500 high 3rd harmonic

[Jimbob54]

Not the first oddity with this dac. The thermal performance of the dac chip itself was wayward.

Could you test rca out too?

 

[liu]

Is it REW? The UI looks different from mine. Do you update it to the latest?

It's already the latest. probably you are not using Mac?

 

[liu]

Not the first oddity with this dac. The thermal performance of the dac chip itself was wayward.

L+R both playing

just L playing.

ignore the elevated noise level here as there is a ground loop in RCA.

 

[liu]

BTW, a super senior guru PMed me to measure the crosstalk. I measured the crosstalk on each and could not see any issues. both channel's crosstalk values are super low, near the limit of my ADI-2. So it is possible that I'm measuring ADI-2's own crosstalk rather than M500's. so crosstalk is not an issue here.

I packed up both units and will return them to amazon tomorrow. It has already been taking me a lot of time. So I won't do any more measurements.

 

[Veri]

BTW, a super senior guru PMed me to measure the crosstalk. I measured the crosstalk on each and could not see any issues. both channel's crosstalk values are super low, near the limit of my ADI-2. So it is possible that I'm measuring ADI-2's own crosstalk rather than M500's. so crosstalk is not an issue here.

I packed up both units and will return them to amazon tomorrow. It has already been taking me a lot of time. So I won't do any more measurements.

Appreciate your reporting though. Does seem like an awkward implementation 'bug'.

 

[solderdude]

yes, it does not seem to be a crosstalk thing though the square wave on the other channel seems to suggest something is up.
I suspect it is something in the digital domain that doesn't handle certain things well.
Maybe some layout error but that would also be visible in channel separation tests.

Some other test could be to use say 996Hz or something close to but not exact 1kHz. @amirm found weird behavior at slightly different frequencies in some Schiit DAC.

It's a good thing you caught this and gives the digital guys at SMSL something to do for the coming weekends.

The issue also seems to be amplitude dependent given the drastic improvement at -20dB

Does this only happen with USB input ?

 

[liu]

> Some other test could be to use say 996Hz or something close to but not exact 1kHz. @amirm found weird behavior at slightly different frequencies in some Schiit DAC.

Didn't try 996. I tried 440Hz (which is the A note frequency) and problem was there too.

> It's a good thing you caught this and gives the digital guys at SMSL something to do for the coming weekends.

I don't have an SMSL SU-9 but I guess it has the same issue --- their hardware are similar and the firmware are exactly the same.

I wonder if they have such "digital" guy. Probably not. As I said before, Chinese Hifi companies have little dsp knowledge. Their firmwares/drivers are all outsourced.

While this means they have little chance to mess it up, when it failed, usually they have no idea why.


> The issue also seems to be amplitude dependent given the drastic improvement at -20dB

right.

> Does this only happen with USB input ?

Tested coax input as well. same thing.

Oh, btw, I think it's the first ChineseHiFi product I tested that gets COAX and SPDIF jitter right.

 

[Veri]

"gets jitter right" as in what, good performance?

 

[liu]

"gets jitter right" as in what, good performance?

yeah. all other dacs do not have clean jitter performance. none of the device with AK449x chips performs well, even in USB mode. topping's es9038q2m devices also perform super bad in jitter test. So this really amazed me.

 

[solderdude]

Given the pics of the analog section I would rule out the analog part. If I would had to investigate it I would route the analog out from L and R of the DAC chip(s) to the other channel's analog section (or just measure the DAC chip output) to find out if it is digital issue of sorts.
When performance is always tested with 2 channels doing the same thing they will never find this kind of error.

 

[restorer-john]

Distortion compensation/cancellation in the DAC may have been incorrectly implemented by SMSL? Perhaps they are using the D/A in a wierd differential/dual parallel situation where the harmonics reinforce one another.

"Residual distortion from suboptimal PCB components and layout can be minimized using ES9038PRO’s unique THD compensation circuit".

My bet is they are using the THD compensation co-efficients to get the stellar numbers, but there is some mutual coupling (sub-optimal layout) between channels that manifests in the absence of the right channel signal (ie it doesn't cancel, it reinforces) and the fact that the co-efficients for cancellation are not channel specific (they can't trim one differently to another)

So, do we regard these distortion compensation "features" as the equivalent of VW DieselGate? When the unit is playing a 'test signal"(identical in both channels) as opposed to real world music...

 

[Bullwinkle J Moose]

stop this meaningless troll --- otherwise there's no reason for this site to exist.

We are investigating an important issue. this product is clearly fucked up.

I'm not the Troll here

If you do not hear a problem, then why do you believe there is a problem?

 

[solderdude]

I would agree with liu that there is a problem. It is repeatable, should not be there, is in one channel only and significant, seen in more than one M500 and not in other designs.
It may well remain below audible limits and may not be a really big issue when playing music but that doesn't mean the device doesn't have a problem.

The only way to determine how much of a problem it is with music is nulling and listening to the null. Once with only 1 channel and once with both channels.

 

[restorer-john]

When performance is always tested with 2 channels doing the same thing they will never find this kind of error.

D/A converters were always tested right into left and left into right for channel separation and all tests for THD were L,R and L+R, L-R back in the day. Amir's AP can do it all of course.

 

[Bullwinkle J Moose]

D/A converters were always tested right into left and left into right for channel separation and all tests for THD were L,R and L+R, L-R back in the day. Amir's AP can do it all of course.

But does it remain within the stated spec (THD)

If it is within the spec for both channels driven, there is no problem other than a perceived one

 

[solderdude]

But does it remain within the stated spec (THD)

If it is within the spec for both channels driven, there is no problem other than a perceived one

specs say: 0.00015%
measured 0.015% on the left channel only when the other channel is not driven.
Yes, it is well below 0.1% but still the device should not do this. It is clearly not intended to do this otherwise both channels would do the same.

 

[MusicNBeer]

It's all screwed up. Wonder if my V1 does that. I'll have to check when I get time.

 

[half_dog]

yeah. all other dacs do not have clean jitter performance. none of the device with AK449x chips performs well, even in USB mode. topping's es9038q2m devices also perform super bad in jitter test. So this really amazed me.

look at SMSL M300 coax, for instance. input is pure 12k without lsb.

View attachment 126230

on M500 this is just super super clean.

When I bought a M500 in 2019 (a little after its release), I did some measurements with a US-366 (it isn't the cleanest interface, but that what I have) and M500 showed a strange behavior at its right channel (high THD at sweep test) I returned it because of this. About jitter, I can get better jitter and THD measurements when the interface is connected to a notebook running on batteries and the measured device connected to another machine (desktop or notebook) which only this is grounded. The measurements are much cleaner this way.

 

[WolfX-700]

I bought a used M500 and verified it after it arrived in a few days.

 

[MusicNBeer]

My V1 does it too, but in the right channel! I verified it by physically disconnecting the left channel to ensure I didn't have channels reversed somewhere. If I have both channels going, perfect. A really interesting thing is if I lower the volume, the 3rd order harmonic reduces by 3-5 times that of the fundamental, like an IP3 should, but moreso. If I lower volume to 31, harmonic is 100dB down, or 0.001%. I'm going to test with different frequencies on each channel now.