SMSL D-6s Balanced DAC Review

[Angelilo]

I read my music with foobar, maybe that's why it's not working?

 

[staticV3]

I read my music with foobar, maybe that's why it's not working?

What output mode are you using in foobar?

 

[Angelilo]

I use ASIO.

 

[staticV3]

Has SMSL even published an ASIO driver for the D-6s?

 

[PyramidElectric]

I'm not saying that the rule is valid for all DACs... I have a friend who has a Topping by deactivating the volume and he found that the sound quality was really better...

Share measurements please. Unless the dac has a seperate analogue volume control circuit that is being bypassed (unlikely to say the least) then this absolutley should not be the case.

 

[Toku]

I read my music with foobar, maybe that's why it's not working?

Is Foober2000's volume control set to Max 0.00dB?
If it drops even a little, it won't work.

 

[Angelilo]

Yes 0,00 dB

 

[Angelilo]

SMSL a-t-il même publié un pilote ASIO pour les D-6 ?
[JOINDRE]327144[/JOINDRE]

Non, les drivers des D-6 S n'existent apparemment pas... peut-être existe-t-il effectivement des drivers très spécifiques ! mais jusqu'à présent, je ne les ai pas trouvés.

 

[Angelilo]

I have this one

 

[kschac]

I have a question for the D-6S processors! I realized that when playing a DSD or MQA file normally I should have a small light that lights up, on my DAC it does not light up. Do you have the same problem please?

On mine, when playing DSD, the word DSD lights up in the panel. When playing MQA the word MQA lights up and a bright purple-ish light illuminates.

 

[nanook]

I'm not saying that the rule is valid for all DACs... I have a friend who has a Topping by deactivating the volume and he found that the sound quality was really better...

The usual way volume is adjusted in these DACs is a so called "digital volume control". This is a re-scaling of the digital input signal before it reaches the actual digital/analog conversion. With 0.0dB attenuation the bits are supposed to pass unmodified, such that 0.0dB attenuation is basically without volume control.
Using this digital volume control will always lead to the fact that the DAC chip uses a smaller number of bits. Attenuating by e.g. 12dB will just shift data by 2 bits and the DAC chip will use "only" 22 bits out of its 24 bits (in reality even less because the Least-Significant-Bits will be masked by the noise floor.
Because of this I try to use only moderate attenuation with the digital volume control.

Here is the signal path from the ES9039Q2M datasheet:

I was tempted to get the FiiO K7-BT because it has an analog input and an analog attenuator (NJU72315, so far the best analog attenuator chip I've seen. The NJU72315 can only attenuate by 62dB in 2dB steps, but in conjunction with a digital volume control in the DAC chip this would be SOTA). This way the DAC chip gets operated almost without attenuation in the digital domain and thus does not loose resolution. Unfortunately the K7-BT cannot be remote controlled by the FiiO remote control unit (This is at least what the german distribution told me). It's real pitty because this would have been a really nice All-In-One solution for me, but input selection and volume control via remote control are a must for me.

I can imagine, that DACs with headphone amplifier provide 2 operation modes.
DAC-only: Line-out gets not attenuated, but headphone-out gets attenuated (either Line-out or headphone active at a time because the usual 2-channel DAC chip can only handle one output and thus one attenuation setting)
PreAmp: Line-out and headphone are subject to attenuation such that you can use the digital volume control to adjust volume with amplifier and speakers connected

Hope this clarifies the topic "digital volume control" vs. "analog volume control" a bit.

 

[Toku]

Yes 0,00 dB

I have this oneView attachment 327161

There is no problem with the ASIO driver installed.
I suspect that the settings on the Foober2000 side are suspicious. This behavior occurs if the settings are incorrect.

 

[Esprit]

On mine, when playing DSD, the word DSD lights up in the panel. When playing MQA the word MQA lights up and a bright purple-ish light illuminates.

MQA here, it only works for me with Audirvana Studio and not with the Tidal app for Mac (I don't want to argue with Tidal).
It usually lights up blue, sometimes green, and once purple.
Obviously I'm not convinced that everything is working correctly.
I think MQA works really badly.
I would like to know from you, in order to compare and verify, which MQA song you are listening to, which frequency is shown on the DAC display and which color is illuminated. Thank you

 

[dsnyder0cnn]

MQA here, it only works for me with Audirvana Studio and not with the Tidal app for Mac (I don't want to argue with Tidal).
It usually lights up blue, sometimes green, and once purple.
Obviously I'm not convinced that everything is working correctly.
I think MQA works really badly.
I would like to know from you, in order to compare and verify, which MQA song you are listening to, which frequency is shown on the DAC display and which color is illuminated. Thank you

I created this TIDAL playlist for testing MQA functionality:

TIDAL - High Fidelity Music Streaming

TIDAL is the first global music streaming service with high fidelity sound, hi-def video quality, along with expertly curated playlists and original content — making it a trusted source for music and culture.

listen.tidal.com

I'm streaming the playlist from Roon to the D-6s via COAX S/PDIF connection from a WiiM Pro. With DSP and volume leveling disabled, here's what I see on the display for each of the seven MQA Studio tracks:

However, I normally have a 65k tap room or headphones correction filter and volume leveling enabled (target is -18 LUFS). In this case, Roon preserves the MQA signaling, applies the DSP settings, and then restores the MQA signaling so that the downstream DAC can complete rendering (hardware unfolding). With these settings enabled, I see basically the same thing on the display of the D-6s while playing the same seven tracks; however, the the color of the MQA LED changes from blue (MQA Studio) to purple.

I don't have a similar playlist for MQA tracks on TIDAL that are not "MQA Studio", but if I find an album that is just MQA on TIDAL, like Leave It Beautiful from Astrid S, I see the MQA symbol with a green LED when Roon's DSP and volume leveling is switched off:

I see the same results with the TOSLINK S/PDIF input. The D-6s is the only DAC for under $200 that I'm aware of that supports both MQA Decoding and Rendering on the S/PDIF inputs. From my testing, everything related to MQA works as advertised.

For completeness, here's the display when playing DSD content. In this case, I had to switch to the USB input, fed by a Raspberry Pi since the WiiM Pro has no support for DSD over DoP.

I won't bother showing standard PCM formats, but it's as you'd expect. You get the sampling rate with no MQA or DSD label. The display is limited to three digits, so, as above, rates like 352.8 kHz and 705.6 kHz show up as "352" and "705", respectively.

 

[kschac]

I created this TIDAL playlist for testing MQA functionality:

TIDAL - High Fidelity Music Streaming

TIDAL is the first global music streaming service with high fidelity sound, hi-def video quality, along with expertly curated playlists and original content — making it a trusted source for music and culture.

listen.tidal.com

I'm streaming the playlist from Roon to the D-6s via COAX S/PDIF connection from a WiiM Pro. With DSP and volume leveling disabled, here's what I see on the display for each of the seven MQA Studio tracks:

View attachment 327234

However, I normally have a 65k tap room or headphones correction filter and volume leveling enabled (target is -18 LUFS). In this case, Roon preserves the MQA signaling, applies the DSP settings, and then restores the MQA signaling so that the downstream DAC can complete rendering (hardware unfolding). With these settings enabled, I see basically the same thing on the display of the D-6s while playing the same seven tracks; however, the the color of the MQA LED changes from blue (MQA Studio) to purple.

View attachment 327235

I don't have a similar playlist for MQA tracks on TIDAL that are not "MQA Studio", but if I find an album that is just MQA on TIDAL, like Leave It Beautiful from Astrid S, I see the MQA symbol with a green LED when Roon's DSP and volume leveling is switched off:

View attachment 327242

I see the same results with the TOSLINK S/PDIF input. The D-6s is the only DAC for under $200 that I'm aware of that supports both MQA Decoding and Rendering on the S/PDIF inputs. From my testing, everything related to MQA works as advertised.

For completeness, here's the display when playing DSD content. In this case, I had to switch to the USB input, fed by a Raspberry Pi since the WiiM Pro has no support for DSD over DoP.

View attachment 327387

I won't bother showing standard PCM formats, but it's as you'd expect. You get the sampling rate with no MQA or DSD label. The display is limited to three digits, so, as above, rates like 352.8 kHz and 705.6 kHz show up as "352" and "705", respectively.

Great post!

And, yes, if I remove convolution and other DSP (I am using Roon), I get either a blue light for MQA Studio tracks or a green one for plain vanilla MQA tracks.

Also - the Kobayashi Chopin/Liszt recording is terrific!

 

[Jimster480]

The usual way volume is adjusted in these DACs is a so called "digital volume control". This is a re-scaling of the digital input signal before it reaches the actual digital/analog conversion. With 0.0dB attenuation the bits are supposed to pass unmodified, such that 0.0dB attenuation is basically without volume control.
Using this digital volume control will always lead to the fact that the DAC chip uses a smaller number of bits. Attenuating by e.g. 12dB will just shift data by 2 bits and the DAC chip will use "only" 22 bits out of its 24 bits (in reality even less because the Least-Significant-Bits will be masked by the noise floor.
Because of this I try to use only moderate attenuation with the digital volume control.

Here is the signal path from the ES9039Q2M datasheet:
View attachment 327163
I was tempted to get the FiiO K7-BT because it has an analog input and an analog attenuator (NJU72315, so far the best analog attenuator chip I've seen. The NJU72315 can only attenuate by 62dB in 2dB steps, but in conjunction with a digital volume control in the DAC chip this would be SOTA). This way the DAC chip gets operated almost without attenuation in the digital domain and thus does not loose resolution. Unfortunately the K7-BT cannot be remote controlled by the FiiO remote control unit (This is at least what the german distribution told me). It's real pitty because this would have been a really nice All-In-One solution for me, but input selection and volume control via remote control are a must for me.

I can imagine, that DACs with headphone amplifier provide 2 operation modes.
DAC-only: Line-out gets not attenuated, but headphone-out gets attenuated (either Line-out or headphone active at a time because the usual 2-channel DAC chip can only handle one output and thus one attenuation setting)
PreAmp: Line-out and headphone are subject to attenuation such that you can use the digital volume control to adjust volume with amplifier and speakers connected

Hope this clarifies the topic "digital volume control" vs. "analog volume control" a bit.

This is really cool, thanks for the Diagram. I've always thought about how this "volume control" worked on the DACs' themselves.
Personally I always thought they were analog as you mentioned later, I didn't realize they were digital in this way.

 

[PyramidElectric]

The usual way volume is adjusted in these DACs is a so called "digital volume control". This is a re-scaling of the digital input signal before it reaches the actual digital/analog conversion. With 0.0dB attenuation the bits are supposed to pass unmodified, such that 0.0dB attenuation is basically without volume control.
Using this digital volume control will always lead to the fact that the DAC chip uses a smaller number of bits. Attenuating by e.g. 12dB will just shift data by 2 bits and the DAC chip will use "only" 22 bits out of its 24 bits (in reality even less because the Least-Significant-Bits will be masked by the noise floor.
Because of this I try to use only moderate attenuation with the digital volume control.

Here is the signal path from the ES9039Q2M datasheet:
View attachment 327163
I was tempted to get the FiiO K7-BT because it has an analog input and an analog attenuator (NJU72315, so far the best analog attenuator chip I've seen. The NJU72315 can only attenuate by 62dB in 2dB steps, but in conjunction with a digital volume control in the DAC chip this would be SOTA). This way the DAC chip gets operated almost without attenuation in the digital domain and thus does not loose resolution. Unfortunately the K7-BT cannot be remote controlled by the FiiO remote control unit (This is at least what the german distribution told me). It's real pitty because this would have been a really nice All-In-One solution for me, but input selection and volume control via remote control are a must for me.

I can imagine, that DACs with headphone amplifier provide 2 operation modes.
DAC-only: Line-out gets not attenuated, but headphone-out gets attenuated (either Line-out or headphone active at a time because the usual 2-channel DAC chip can only handle one output and thus one attenuation setting)
PreAmp: Line-out and headphone are subject to attenuation such that you can use the digital volume control to adjust volume with amplifier and speakers connected

Hope this clarifies the topic "digital volume control" vs. "analog volume control" a bit.

Yep, you're right, but I think losing bits is not really a problem these days as ESS converts everything to 32 bit float (see attached 2011 pdf), so you don't run into any meaningful resolution loss until you reduce the signal into the noise floor, which on this DAC is basically SOTA.

 

[nanook]

Yep, you're right, but I think losing bits is not really a problem these days as ESS converts everything to 32 bit float (see attached 2011 pdf), so you don't run into any meaningful resolution loss until you reduce the signal into the noise floor, which on this DAC is basically SOTA.

Thanks a lot for the presentation! This really nails this topic analog vs. digital volume control down.
It's really quite a while ago since DAC chips started to have 32bit data path.

I grew up with 14bit 4x oversampling and 16bit 4x oversampling and when you listened to a 400 Hz sinewave that was attenuated by ca. 40 - 50dB (not fully unrealistic for a voice in a pianissimo part of a classical concert, where headroom was required as well) you could easily hear the 400Hz sinewave getting a staircase. Attenuating e.g. 48dB leaves just ca. 8bits for a native 16bit DAC - no wonder that this got audible.
Having a 24bit DAC with >20bit real resolution is way more comfortable, this is why I wrote "only 22 bits". With this SOTA DAC you can easily afford to "loose" a couple of bits.

Still waiting for my unit to be delivered - I'm really curious.

 

[bogi]

Yep, you're right, but I think losing bits is not really a problem these days as ESS converts everything to 32 bit float (see attached 2011 pdf), so you don't run into any meaningful resolution loss until you reduce the signal into the noise floor, which on this DAC is basically SOTA.

It's 32 bit fixed point, not 32 bit floating point. But that's not crucially important for the topic.

The document you are referring to is related to ESS DAC chips and not to complete DAC devices. So the mentioned noise floor at -135dB is related to ESS chips only. ESS chip needs always to be followed by analog output stage behind the chip because analog output of ESS chips is not sufficiently filtered and is not suitable as direct device output. You don't get full scale -135 dB noise floor from analog output of a complete DAC device. And of course DAC devices of different price levels and quality exist. DAC resolution may be reduced from different implementation related reasons.

Let DAC resolution is 20bit and your input is 24bit recording. Every lowering of digital volume causes more valid audio data bits to be moved under DAC noise floor.
Now let consider 16bit recording. You get 20 - 16 = 4 bits = 24dB headroom for digital volume control.

That is also a bit simplified. Volume control in ESS chip is performed on ovesampled digital data. Oversampling increases dynamic range (that's the reason why intersample overflows may happen). So 16 bit recording may expand to say 18 bits of dynamic range because of oversampling. Then lowering volume by more than 2 bits (12 dB) already moves valid audio data under DAC noise floor. And of course many current DACs and most of older devices don't reach 20 bits of resolution.

Then, there si topic measurement vs. real conditions. For example influence of other connected devices, which may cause ground loop or other type of noise spreading into analog output. You could see for example noise floor pictures before and after applying a galvanic isolator. That are real pictures which don't show 20 bits of resolution.

Assume also that DAC device measurements are performed under artificially isolated conditions. The intention is that only the device of interest is measured. From this reason measurement gear is calibrated. So these isolated DAC measurements don't show how DAC performs under real conditions, when digital data source (for example computer) and amplifier are connected. Furthermore, you are not listening to DAC output. You are listening to amplifier output. You don't get the measured DAC values on amplifier output.

What is audible and what not - I will not speculate. But your statement "you don't run into any meaningful resolution loss until you reduce the signal into the noise floor, which on this DAC is basically SOTA" is too simplified and too optimistic.

 

[dsnyder0cnn]

I always like to spend a few weeks listening to a new DAC before taking measurements so that my subjective impressions are not swayed by objective data. After playing with the seven filter settings, I settled on FL3 as sounding the most "right" to my ears. I finally got around to taking some measurements today to see what's what.

TL; DR - avoid filters FL1 and FL6 because of distortion at ~14 kHz and excessive attenuation in the highs. FL3 and FL7 are the most useful. Choose FL3 for most listening, but FL7 may be beneficial with some music, depending on your tastes.


F2 is the default filter. Here's what I see for Amplitude and Impulse response, and Distortion.

You'll have to mentally compensate for the inverted horseshoe shared amplitude response. I don't have a great way to compensate for the ADC response of my Focusrite Scarlett 2i2 audio interface, and the vertical scale is only 1 dB, so pretend that this is flat.

Impulse response shows extended pre-ringing ripples and exaggerated post-ringing with this filter. No obvious issues with distortion.

With the FL1 and FL6 filters, I'm seeing early attenuation in the high frequencies and a sharp increase in 2nd order distortion at around 14 kHz:

I'm not sure how audible this is, but I would avoid FL1 and FL6 for this reason.

After looking at response of all seven filters, FL3 and FL7 are the most useful. FL3 has a little more pre-ringing than I would like, but it's under 5% from -500 to -140 microseconds and zero before -500 microseconds.

In contrast, FL7 has more extended pre-ringing ripples, although very low in level. It also has a much longer and more exaggerated post-ringing, which can be a useful effect with some music.

The default FL2 and FL7 have virtually identical amplitude and impulse response, but FL7 had lower distortion in my measurements. YMMV.

I may do some sampling rate comparisons to see if there's a rate that works best with the D-6s. So far, I've not seen much difference, though. In particular, impulse response for 44.1 and 48 kHz look virtually identical. That's not been the case with all ESS DACs I've measured.

Edit: there's a slight time-domain benefit to upsampling 44.1 kHz to 88.2 and 48 kHz to 96 kHz.

If you're using Roon, you can use Sample rate conversion to do this easily enough. No need to bother with frequencies above 96 kHz.