Singxer SDA-2 DAC and Headphone Amplifier Review

[Wombat]

Audiophonics item is SDA-1, not SDA-2.

Thanks.

 

[amirm]

@amirm, ref comparison with DX3 pro, DX3 pro uses AK4493, SDA-2 uses AK4497. Not exactly the same chip (even if anecdotical).

I use the term generically. I have yet to find specifics of the chips mattering so I don't keep track of them.

 

[ayane]

It's hard to take a DAC seriously when it fails this spectacularly at filter performance, regardless of SINAD. Whatever happened to the sampling theorem? All of these DACs offer multiple filters and yet most of the time, most of those filters - all in this case - fail to satisfy what is necessary of them as per the Nyquist-Shannon sampling theorem.

Only Benchmark and Chord sell products that actually do their jobs as DACs, and of those two, only the former actually demonstrates a sound understanding of scientific theory. As far as I'm concerned, a DAC that offers multiple filter choices is really a toy, not a professional solution for digital to analog conversion.

It's hard to design an analog filter that attenuates sharply with a small passband, but this is digital filters we're talking about. The math has been beaten to death and now we have silicon so powerful and cheap that we can not only make filters that fulfill the Nyquist attenuation, but they can do it with thousands of taps...

 

[gvl]

What's going on with the sharp filters these days with -10dB attenuation at Nyquist? They are built-in on the DAC chip, so one would think the designers should know better? Perhaps the measurement was done at 48kHz sampling and labeled incorrectly?

 

[boXem]

now we have silicon so powerful and cheap that we can not only make filters that fulfill the Nyquist attenuation, but they can do it with thousands of taps...

As far as I remember, thousands of taps = thousand times 1/Fs delay.
While this is not a problem for music replay at home, this becomes one for other use cases like live music mix, multimedia replay, ...

 

[boXem]

I use the term generically. I have yet to find specifics of the chips mattering so I don't keep track of them.

I had a quick look for you.
AK4493 => AK4497
S/N +5 dB
THD+N -3 dB
Price/1000 +18 dB

 

[ayane]

As far as I remember, thousands of taps = thousand times 1/Fs delay.
While this is not a problem for music replay at home, this becomes one for other use cases like live music mix, multimedia replay, ...

What about minimum phase filters for real time applications? Sure, there are drawbacks to minimum phase filters, but they solve the problem of having an acausal system.

The real answer to the problem is higher sampling rates for recording and mixing with sane tap lengths and then downsampling during the final mastering and distribution.

 

[milosz]

There's what looks like an HDMI socket that's labelled I²S

I guess that's become a quasi standard starting with PS Audio....?

 

[Wombat]

There's what looks like an HDMI socket that's labelled I²S

I guess that's become a quasi standard starting with PS Audio....?

Philips spec. 1986, revised 1996: https://www.cypress.com/file/133906/download

 

[boXem]

What about minimum phase filters for real time applications? Sure, there are drawbacks to minimum phase filters, but they solve the problem of having an acausal system.

The real answer to the problem is higher sampling rates for recording and mixing with sane tap lengths and then downsampling during the final mastering and distribution.

Since you propose to down-sample for distribution, DAC problem is not solved
Anyhow, in sigma delta DACs, the filtering is already run on the over-sampled data, but it is usually only a 6 or 7 taps filter.

Keep in mind that a DAC is not a computer, the technologies used to build it are completely different (mixed signal stuff). Thus the internal algorithms have to stay relatively simple to have a business case standing on it's feet.

 

[milosz]

That Philips spec is the electronic / data spec for I²S. Yes, I know what I²S is. I've used in in some things I've built - DSP systems, and a USB interface to an ES9023-based DAC that I built.

Philips does not specify the use of the HDMI type connector. This is the first time I've seen an I²S input connector on a DAC/ Headphone amp.

 

[Veri]

That Philips spec is the electronic / data spec for I²S. Yes, I know what I²S is. I've used in in some things I've built - DSP systems, and a USB interface to an ES9023-based DAC that I built.

Philips does not specify the use of the HDMI type connector. This is the first time I've seen an I²S input connector on a DAC/ Headphone amp.

Amir has done a little article on it before: https://www.audiosciencereview.com/forum/index.php?threads/study-is-i²s-interface-better-for-dacs-than-s-pdif-or-usb.7105/

 

[Tatteredmidnight]

It's hard to take a DAC seriously when it fails this spectacularly at filter performance, regardless of SINAD. Whatever happened to the sampling theorem? All of these DACs offer multiple filters and yet most of the time, most of those filters - all in this case - fail to satisfy what is necessary of them as per the Nyquist-Shannon sampling theorem.

Only Benchmark and Chord sell products that actually do their jobs as DACs, and of those two, only the former actually demonstrates a sound understanding of scientific theory. As far as I'm concerned, a DAC that offers multiple filter choices is really a toy, not a professional solution for digital to analog conversion.

It's hard to design an analog filter that attenuates sharply with a small passband, but this is digital filters we're talking about. The math has been beaten to death and now we have silicon so powerful and cheap that we can not only make filters that fulfill the Nyquist attenuation, but they can do it with thousands of taps...

Is this an issue due to the higher bandpass limit causing aliasing?

 

[magicscreen]

Thanks.

For Europe:
Magna Hifi

 

[AudioSceptic]

This is a review and detailed measurements of the Singxer SDA-2 DAC and balanced headphone amplifier. It was kindly sent to me by shenzhenaudio online store. The SDA-2 costs US $580 from them.

The SDA-2 is a rather serious looking DAC with a professional feel:

View attachment 36755​

I like the yellow dot matrix display showing me everything at a glance. And the buttons to directly select what you want rather than going into menus. I do wish that if you are trying to change the volume for it to show that in large fonts so that you can see it from across the room. For desktop use it is not an issue.

I did not take a picture of the remote but it is a step above in what comes from competitors with an ergonomic back and properly labeled buttons.

The back panel sports nice connectors and usual set of inputs and outputs:

View attachment 36756​

Looks like an Audiolab but with an 80s Cambridge Audio volume knob. Good, in other words. <https://www.cambridgeaudio.com/gbr/en/about-us/our-history>

 

[steve29]

I included the verbiage from their website on my graph. It says "less than 10 ohm" for 1/4 inch and "less than 20 ohm for XLR" now.

I believe that the image in your graph regarding the impedance says: "(XLR 4pin BAL) headphone impedance is less than 2Ohm" and "(SE 1/4) headphone impedance is less than 1Ohm". With spaces, it should say "2 Ohm" and "1 Ohm" respectively.

 

[ayane]

Anyhow, in sigma delta DACs, the filtering is already run on the over-sampled data, but it is usually only a 6 or 7 taps filter.

Bingo. Running a small tap length filter on a data stream whose audio bandwidth is just a small fraction of the nyquist bandwidth is the way it's done, has been done, and will continue to be done. There are just too many drawbacks to other approaches. Noise shaping math is truly a beautiful thing.

 

[amirm]

Perhaps the measurement was done at 48kHz sampling and labeled incorrectly?

No, it is 44.1 kHz. If you look at the DAC chip specs, you see almost all of them target 24 kHz bandwidth for 44.1 kHz sampling. This allows them to use a more gentle filter.

Aliasing is not a huge issue here since music content doesn't have a lot of amplitude above 20 kHz so not much reflects back.

Still, as I noted in the review, I like to have a choice of a sharper filter.

 

[Veri]

It's so odd most all filters aim for 24 kHz ... I guess, not an audible issue? There remains the 'why', though.

 

[amirm]

Here is the AKM datasheet: