AliExpress Cirmech WM-013 CS4398 DAC - review and measurements

[pma]

I bought this DAC board from AliExpress at $18.34 (shipment and VAT included). The product page is below:

https://www.aliexpress.com/item/32960837911.html

They call it CIRMECH Optical coaxial audio decoder CS8416 CS4398 chip 24BIT192KHz SPDIF coaxial Optical fiber DAC decode board For amplifier.

1. Info from aliexpress page:

There is almost no description of this DAC board there. There is also no documentation with the board. I had to guess the connection of the power supply input terminals. There are some basic specifications in their web page, that I am posting below.

So, we have a 24bit/192kHz DAC with coaxial S/Pdif input, Toslink optical input and RCA stereo output. Power supply voltage is unspecifed, however should be 2 x 15V to work properly.

2. My additional description

The DAC arrives as an assembled printed circuit board with through hole and SMD parts. It needs 2x15V (+/-15V) power supply, which is not a part of delivery. After connecting to power supply and S/Pdif audio signal, it starts to run immediately. The proper function is announced by a blue LED on the PCB. DAC chip is CS4398, audio op-amp is NE5532P.

3. Measurements

I connected the DAC to a 2x15V external power supply and measured the output with E1DA Cosmos ADC at 7.6V input setting. The S/Pdif signal was taken from Topping D10s coaxial output. D10s serves as a USB to S/Pdif digital bridge.

3.1. THD and THD+N at 0dBFS and -3dBFS 1kHz input sine

at 0dBFS, we get 76.3dB SINAD and 105.5dB SNR. I aprreciate that the spectrum is free from any mains spuriae, though we have only RCA SE output.

at -3dBFS, we get 85.3dB SINAD and 103.7dB SNR.

3.2. THD vs. frequency at -3dBFS and @BW45kHz (96kHz Fs) and @BW90kHz (192kHz Fs)

With BW45kHz, distortion is low at low frequencies (-100dB), but starts to rise as early as from 50Hz. Measurement with BW90kHz shows higher disortion at low frequencies (-85dB) and distortion rises above 1kHz in the same way as with BW45kHz.

3.3. THD vs. output voltage at 1kHz

Distortion is low 0.001% at 0.1V output voltage, but starts to rise to reach 0.009% at 2V.

3.4. Crosstalk

is low -110dB at 1kHz/-3dBFS

3.5. Test setup photo

4. Conclusion

This is a very cheap DAC with not very good parameters, however some nice surprise to me is its quite clean spectrum free from hum, buzz and mains spuriae. I will build it as an S/Pdif input into a case of a simple, cheap preamplifier with multiple inputs. I do not expect audible issues.

 

[AnalogSteph]

I imagine this one would actually be pretty good if you were to replace all the crummy ceramic caps involved in lowpass filtering in the signal path with NP0 or film. They clearly threw in the cheapest / smallest they could find, with no regard for the finer points of audio engineering.

 

[pma]

My words as well. The ceramics are responsible for the unwanted distortion. Cannot guess if it would be so cheap with better caps.

 

[pma]

I imagine this one would actually be pretty good if you were to replace all the crummy ceramic caps involved in lowpass filtering in the signal path

My words as well. The ceramics are responsible for the unwanted distortion.

Effect on IMD:

 

[audio_tony]

I had one of these DACs - I was unable to reduce the distortion (unless perhaps the board was substantially re-engineered).

The LJM version is far superior (although twice the price).

CS4398 Digital DAC Decoder USB Coaxial Input Audio Amplifier Board 24BIT/192K | eBay UK

CS4398 DAC Decoder Board USB Coaxial. Use CS4398 IC. D/A converter (decoding chip): CS4398 120dB dynamic range Supports 24bit/192K sampling. USB input and Coaxial input automatic selection. COAX coaxial input channel is selected.

www.ebay.co.uk

Here's the measurements of mine. Ignore the 0dB measurement - it was caused by the Windows limiter (CAAudioLimiter).

Measurements of LJM CS4398 DAC kit

Here are measurements of the LJM CS4398 DAC kit (still available on Ebay) I built this around 2018 / 2019 IIRC. The board has had some mods (specifics not shown here). My test setup: Asus Xonar STX sound card with a self built buffer / gain box in front of that. Host PC is an Intel i7 running...

www.audiosciencereview.com

 

[wwenze]

That thing above performs so badly compared to the reference clone lol (0.001% THD+N)

Measurements of LJM CS4398 DAC kit

Here are measurements of the LJM CS4398 DAC kit (still available on Ebay) I built this around 2018 / 2019 IIRC. The board has had some mods (specifics not shown here). My test setup: Asus Xonar STX sound card with a self built buffer / gain box in front of that. Host PC is an Intel i7 running...

www.audiosciencereview.com

Or actually rather, it is performing well until 0.1V, and then judging from the worse THD vs voltage as well as frequency it is suffering from slewrate limitation?

How many of the things on the PCB are real anyway? These kind of boards are known to put stuff that are completely different from the label.

 

[pma]

No. The ceramics capacitors issue. Those who ever worked with different dielectrics types know.

 

[pma]

I had one of these DACs - I was unable to reduce the distortion (unless perhaps the board was substantially re-engineered).

The LJM version is far superior (although twice the price).

CS4398 Digital DAC Decoder USB Coaxial Input Audio Amplifier Board 24BIT/192K | eBay UK

CS4398 DAC Decoder Board USB Coaxial. Use CS4398 IC. D/A converter (decoding chip): CS4398 120dB dynamic range Supports 24bit/192K sampling. USB input and Coaxial input automatic selection. COAX coaxial input channel is selected.

www.ebay.co.uk

Here's the measurements of mine. Ignore the 0dB measurement - it was caused by the Windows limiter (CAAudioLimiter).

Measurements of LJM CS4398 DAC kit

Here are measurements of the LJM CS4398 DAC kit (still available on Ebay) I built this around 2018 / 2019 IIRC. The board has had some mods (specifics not shown here). My test setup: Asus Xonar STX sound card with a self built buffer / gain box in front of that. Host PC is an Intel i7 running...

www.audiosciencereview.com

You have system gnd loop issues. This shows the operator’s system skills.

 

[audio_tony]

You have system gnd loop issues. This shows the operator’s system skills.

"This shows the operator’s system skills." - That's a little insulting don't you think?

Yes, there is a small 50Hz spike in my measurement - I have since sorted that (and it wasn't a ground loop).

Useful hint: Try to be a little more polite in the future.

 

[Chrysalis99]

"This shows the operator’s system skills." - That's a little insulting don't you think?

Yes, there is a small 50Hz spike in my measurement - I have since sorted that (and it wasn't a ground loop).

Useful hint: Try to be a little more polite in the future.

My first thought...What a jerk.

 

[IVX]

A bare CS43198/43131 has 2Vrms 1k THD+N -115db, I'm curious what the role of 5532 is over there? BTW, I saw a fake 5532, .7x.7mm die.

 

[pma]

A bare CS43198/43131 has 2Vrms 1k THD+N -115db, I'm curious what the role of 5532 is over there? BTW, I saw a fake 5532, .7x.7mm die.

Again - the distortion is due to high K, highly nonlinear ceramic capacitors used in the output active LPF with the NE5532. Everyone knows that the only usable ceramic caps at this position are NPO/COG. However, they are big and expensive. Thus, producers of cheap electronics often use X7R and similar ceramics even in the signal path and filters. I have shown high K ceramic capacitors nonlinearity here .

@solderdude has shown it as well in this post.

Another review of mine of Audiophonics PRE-TC10 shows ceramic cap distortion in the bass/treble EQ circuit.

Another links:

https://www.ti.com/lit/an/slyt796a/slyt796a.pdf?ts=1759917924600

https://www.edn.com/more-about-understanding-the-distortion-mechanism-of-high-k-mlccs/

and finally

https://e2e.ti.com/cfs-file/__key/communityserver-discussions-components-files/14/Signal_2D00_distortion_2D00_from_2D00_high_2D00_K_2D00_ceramic_2D00_capacitors.pdf

In this last TI link, one will see almost identical THD / level and THD / frequency plots as I have measured here in the post #1. Moreover, the plots are temperature dependent, another proof of high K ceramic capacitor action.

 

[IVX]

I guess the LPF on 5532 is redundant, and only increases the THD+N.

 

[pma]

Maybe. They probably followed this eval board:

https://www.mouser.com/datasheet/2/76/cdb4398-31704.pdf

I will check something later.

 

[solderdude]

5532 is bal-SE conversion with post filter would be my guess so essentially needed.

Pavel... do you plan on replacing the audio relevant ceramics for NP0/C0G and re-measure ?
They are probably all smaller than 10nF so they should be available.

 

[pma]

A bare CS43198/43131 has 2Vrms 1k THD+N -115db, I'm curious what the role of 5532 is over there? BTW, I saw a fake 5532, .7x.7mm die.

Again - the distortion is due to high K, highly nonlinear ceramic capacitors used in the output active LPF with the NE5532.

Hi Ivan, just to double check, I have replaced the supplied NE5532P with genuine LME4562. There is some slight improvement, but the general issues remain the same. I am sure it is a ceramic capacitor distortion.

But this slight improvement has disappeared with temperature ...

THD vs. level is the same misery, regardless 5532 or 4562:

Proof done, opamp is not the case, ceramic capacitors are. There is a problem that "generalizing" that passive parts make no difference is wrong.

 

[pma]

Pavel... do you plan on replacing the relevant caps for NP0/C0G and re-measure ?

I am afraid I do not. You know, this is a very cheap board and I have been curious what it can do. My 70 years old eyes are not good enough for SMD parts de-soldering/soldering and moreover, I do not expect it would be audible, it would just be more correct from engineering point. But I may do a DBT test, against a good DAC, possibly with files available.

 

[pma]

5532 is bal-SE conversion with post filter would be my guess so essentially needed.

Yeah. From the EVAL board:

Frequency response of Cirmech DAC:

 

[solderdude]

I am afraid I do not. You know, this is a very cheap board and I have been curious what it can do. My 70 years old eyes are not good enough for SMD parts de-soldering/soldering and moreover, I do not expect it would be audible, it would just be more correct from engineering point. But I may do a DBT test, against a good DAC, possibly with files available.

That saves you having to de-solder each tiny cap ensuring you don't break the cap nor the tiny pads of the cheap board, measuring each of them, making notes where they have to go on the board, looking for similar sized C0G/NP0 caps, ordering a bunch.
Then you can't use the board and have to wait till the parts arrive.
Then check if the distributor actually sent you the proper values, solder each one in the correct position and re-measure the whole lot only to find out the distortion is (marginally ?) lower and equally below audible thresholds with music.

The DBT might be fun for some. I know I can't ace that so won't spend time on that

 

[capslock]

What is DBT, please?