SMSL D-6s Balanced DAC Review

[Lyricman]

Unless you can hear the DAC's noise floor through your speakers (hiss with the D-6S plugged into the Amp, but gone with it plugged out), you are not losing any performance or resolution by using the D-6S' volume control, as in that case any potentially lost resolution was inaudible anyway.


Such a poll would have to be accompanied by blind test results of each participant to have any meaning or value, which would be exceedingly unlikely.

Thanks static for your reply. I don't have any hiss whatsoever so all good. so what was the discussion about losing 3-4 bits of resolution when volume is not set to 99 talking about, why would a volume control reduce the bits of resolution Just education for me.

I appreciate your answer on the poll but I wasn't asking which is the best using AB tests just interested what people were using without qualification.

 

[somebodyelse]

Thanks static for your reply. I don't have any hiss whatsoever so all good. so what was the discussion about losing 3-4 bits of resolution when volume is not set to 99 talking about, why would a volume control reduce the bits of resolution Just education for me.

It's the difference between what's measurable at the DAC terminals and what's audible in your room after the signal has been through amp and speakers. If you can't hear the DAC's own noise then you can't hear what's being lost. Technically some resolution will be lost if you are playing a 24 bit source, but in a real room with an amp you probably won't hear it. Background noise in a 'quiet' room during the day is around 25-30dB while 120dB is painful and damaging. This is the basis of 16 bits being 'enough' to cover the practical range of what we can hear. Closed back headphones or IEMs push it a little further because of the isolation.

 

[Lyricman]

It's the difference between what's measurable at the DAC terminals and what's audible in your room after the signal has been through amp and speakers. If you can't hear the DAC's own noise then you can't hear what's being lost. Technically some resolution will be lost if you are playing a 24 bit source, but in a real room with an amp you probably won't hear it. Background noise in a 'quiet' room during the day is around 25-30dB while 120dB is painful and damaging. This is the basis of 16 bits being 'enough' to cover the practical range of what we can hear. Closed back headphones or IEMs push it a little further because of the isolation.

Thank you very clear. I have an Audiolab CQ8200 CD/DAC/Preamp which is over 10 years old and whilst it has served me well i felt it was lacking so I thought I would buy D 6s to see how technology has advanced and wow what a difference for a modern £160 DAC to £1000 Audiolab an absolute bargain. Great soundstage extra depth and excellent detail stereo separation. Oh and throw in bluetooth which is a bonus

 

[JIW]

The ESS DAC used 32 bits for the signal regardless of what comes in. For 24 bits, it can attenuate by 8 bits, i.e. 48 dB, before you lose resolution from the source and for 16 bits, it can attenuate by 16 bits, i.e. 96 dB.

 

[nanook]

Having read through all the threads I realise that there are many highly technical wizz kids on here.
My question is about the volume control on the outputs. I have the D6s connected to a power amp and therefore use the volume control between -26 and -30 on the d6s,from the threads it would appear that I will be losing some performance and resolution from not this or I have read it wrongly. If so what would be the solution to correct this.

Here's the graph for output level[dBu] = f(volume-as-displayed) for the D-6s:

You are talking about "99 - 26" or about "26"?
With volume = 26 ... 30 you are at round about 36dB attenuation. That's somewhat more than 5 bits - nothing to really worry about.

 

[Lyricman]

Here's the graph for output level[dB] = f(volume-as-displayed) for the D-6s:

View attachment 395634

You are talking about "99 - 26" or about "26"?
With volume = 26 ... 30 you are at round about 36dB attenuation. That's somewhat more than 5 bits - nothing to really worry about.

So if I understand correctly it is pointless playing back any 24 bit music as I could only get 19 bits at volume level of 30 on the D 6s

 

[audiomaestro]

Here's the graph for output level[dB] = f(volume-as-displayed) for the D-6s:

View attachment 395634

You are talking about "99 - 26" or about "26"?
With volume = 26 ... 30 you are at round about 36dB attenuation. That's somewhat more than 5 bits - nothing to really worry about.

Excuse me if this question is silly, but does this mean that if the song is a 16 bit recording, the bits would only be 11 bit at 36db attenuation? So this would be another benefit of 24 bit audio, it would give more leeway to attenuation without getting into too low in bit depth at higher attenuation levels?

 

[somebodyelse]

Excuse me if this question is silly, but does this mean that if the song is a 16 bit recording, the bits would only be 11 bit at 36db attenuation? So this would be another benefit of 24 bit audio, it would give more leeway to attenuation without getting into too low in bit depth at higher attenuation levels?

NO! Your 16 bit recordings would be padded internally to get to 24 bit by adding zeros in the least significant 8 bits. If the attenuation removed the 5 least significant bits you still have 3 bits worth of zero padding left. No actual data has been lost yet.

That's the simplified version. Some volume controls operate at 32 bit, or convert to floating point to do the calculation, then back again.

 

[staticV3]

So if I understand correctly it is pointless playing back any 24 bit music as I could only get 19 bits at volume level of 30 on the D 6s

Excuse me if this question is silly, but does this mean that if the song is a 16 bit recording, the bits would only be 11 bit at 36db attenuation? So this would be another benefit of 24 bit audio, it would give more leeway to attenuation without getting into too low in bit depth at higher attenuation levels?

Your playback environment is unlikely to offer much more than 12-14 bits of SNR anyway, so all this discourse around audio bit depth and potential bit depth loss from digital attenuation is largely meaningless anyway.

Unless you can hear your DAC's noise floor in the form of loudspeaker hiss at your MLP, you're getting the maximum playback dynamic range that your environment permits, regardless of what bit depth your audio tracks have or by how much you're attenuating digitally.

 

[nanook]

That's the simplified version. Some volume controls operate at 32 bit, or convert to floating point to do the calculation, then back again.

The modern chips are working with 32 bits internally, so you can indeed attenuate significantly using the digital volume control of the DAC. I agree with staticV3 that the practical limit is the noise floor getting audible.

 

[somebodyelse]

So if I understand correctly it is pointless playing back any 24 bit music as I could only get 19 bits at volume level of 30 on the D 6s

It's easy to look at the numbers and talk about theoretical performance while forgetting about the real world. I do it on a regular basis, and a reality check is worthwhile from time to time.

Background noise in a 'quiet' room is somewhere in the 20dB to 30dB range. Add the 96dB of 16 bit and you're around 120dB - that's not good for you, and your speakers probably can't reach that level anyway. Noise cancelling headphones, or those with good passive isolation like the etymotics, can make the background quieter while being able to reach ear-damaging levels, so might give you more than 16 bits of dynamic range.

We've not thought about the recordings yet. If it was actually recorded rather than synthesised then the noise level in the recording will be higher than 19 bits down, probably higher than 16 bits down. You will also have to find a recording that makes use of the potential dynamic range, which may be tricky given the modern tendency to compress music and push it to the edge of full scale, or even beyond.

I really ought to have saved the references to papers looking at whether people could actually hear a difference between CD and HD formats. One thought they had, but only on certain source material. It turned out the sources where people could hear a difference had a different mix on CD and HD. When the HD was downsampled to CD rates and compared, they were back to tossing a coin.

 

[Lyricman]

What an education we get from you guys I now understand this well, bloody clever these chips are!!
With reference to quality I am always seeking out good recordings to enjoy the quality of my hifi and can say that I think all of my favourite recordings are in fact 16 bit CD quality which for my mature ears sound fantastic. Many of the 24 bit are taken from Vinyl which are not the best.

 

[audiomaestro]

NO! Your 16 bit recordings would be padded internally to get to 24 bit by adding zeros in the least significant 8 bits. If the attenuation removed the 5 least significant bits you still have 3 bits worth of zero padding left. No actual data has been lost yet.

That's the simplified version. Some volume controls operate at 32 bit, or convert to floating point to do the calculation, then back again.

Well explained. Thanks.

 

[antcollinet]

So if I understand correctly it is pointless playing back any 24 bit music as I could only get 19 bits at volume level of 30 on the D 6s

That’s not how it works. More bits gets you a lower noise floor - that is all.

If the noise floor is low enough that you don’t hear it during silent sections of the music, (and with 24 bit - or more - DSP you will not) then you can reduce the volume as much as you like - you still won’t hear the noise.

There is no penalty in using digital volume control.

 

[antcollinet]

Excuse me if this question is silly, but does this mean that if the song is a 16 bit recording, the bits would only be 11 bit at 36db attenuation? So this would be another benefit of 24 bit audio, it would give more leeway to attenuation without getting into too low in bit depth at higher attenuation levels?

No, a 16 bit song will be processed at 24bits or 32 bits in the volume control. So you can attenuate by at least 8 bits (48dB) and lose nothing of the original 16 bits.

Then see my previous post about more bits = lower noise floor. If you are not hearing the quantisation noise (and you will not be) then there is no penalty for digital volume control.

EDIT : Ninjad by earlier posts.

 

[miero]

Can someone check/measure if the PE on this DAC is connected with GND on RCA audio outputs ot SPDIF and USB inputs please?

If yes, then what resistance is there? 0 Ohm or more?

Thanks.

 

[nanook]

Can someone check/measure if the PE on this DAC is connected with GND on RCA audio outputs ot SPDIF and USB inputs please?

If yes, then what resistance is there? 0 Ohm or more?

Thanks.

I'm out this week so I can't double check.
As far as I remember the internal Gnd is connected to PE.
On two of the units I'm using RCA and I'm not having problems with Gnd- loop related hum.

 

[trungdtmc]

Can someone check/measure if the PE on this DAC is connected with GND on RCA audio outputs ot SPDIF and USB inputs please?

If yes, then what resistance is there? 0 Ohm or more?

Thanks.

Yup, i checked the PE is connect together with GND RCA, SPDIF

 

[miero]

OK, thanks. So ground loops can be easily created using this DAC.

So it might be reasonable to use an USB isolator if used with a source in electrical class I.
And also using a preamplifier/amplifier in electrical class I could be also a problem with unbalanced interconnection.

 

[antcollinet]

OK, thanks. So ground loops can be easily created using this DAC.

Ground loops are not created by a component but by a system and how it is interconnected. All DACS can form part of a ground loop (as can any other component in the chain) unless they include galvanic isolation. I am not aware of any that do.

You don't need to use a isolator, or other mitigation unless you experience noise - hiss or hum, which is audible at your normal listening volume.