ESS Sabre and DSD Volume Control

[Hiro]

Either way, the ESS chip does volume control on DSD.

 

Can one bypass the volume control inside the ESS chip and let the DAC do what it is supposed to do, that is D/A conversion?

[elcorso]

Can one bypass the volume control inside the ESS chip and let the DAC do what it is supposed to do, that is D/A conversion?

 

On exaSound DACs (ESS based) you can set volume control to 0.00 dB, according to the DAC user manual this way you bypass volume control.

 

Roch

[Hiro]

On exaSound DACs (ESS based) you can set volume control to 0.00 dB, according to the DAC user manual this way you bypass volume control.

 

Roch

 

Thanks for the info, Roch.

[magnum innominandum]

Hi,

 

I was talking purely about digital domain, what comes out of the software.

 

Wow. That is useful.

 

DAC's noise floor is typically at least 20 and frequently even 40 dB lower than typical power amp's.

 

Really? I am not sure how you come to such a conclusion. Before my current Amp's I owned a Goldmund Mimesis 9, the very similar (just slightly lower power) 8 was tested by Stereophile:

 

Goldmund Mimesis 8 power amplifier Measurements | Stereophile.com

 

109dB unweighted noise re 1W, output 23dB above 1W, 2.1V will give full output so referred to 2.1V input that is 132dB unweighted SNR. Please show me a DAC with > 132dB UNWEIGHTED SNR referred to 2V.

 

This Amp is > 20 Years old! Though noisier my current Pass Aleph's still do 115dB re. 2V Input (which incidentally gives full power).

 

When I measure power amp output, I can see only power amp's self noise.

 

Looks like you need better power amp's.

 

Which DAC puts out 10V? I don't have such DAC here. Most of those put out around 2 Vrms full level.

 

Based on their literature, Mytek, if you want the full SNR of 128dB....

 

What is most stupid thing to do, is to first output excessive voltage, then attenuate it a lot in analog domain just to amplify it again a lot.

 

This we both agree on.

 

However, interesting detail is that many DACs have lowest THD between -10 dBFS and -20 dBFS output levels.

 

So we should use that as a reason to throw away 10 - 20dB of Dynamic range?

 

And has horrible distortion characteristics. At least all the Pass gear I've seen measurements of.

 

You should fact-check before yapping senseless garbage:

 

https://www.passdiy.com/pdf/B1%20Buffer%20Preamp.pdf

 

That would be still very good, because most amps barely manage 108 dB SNR when analog volume is turned to max (0 ohm series resistance) and input voltage is tuned to be just barely below clipping.

 

That is garbage. I have no idea what poor quality amplifiers you use as reference, but it is not true for most. You should check 3rd party measurements. Many Amplifier measured by Stereophile do MUCH better than 108dB

 

DAC's output's noise floor is some 30 dB below best case amplifier performance.

 

I showed above an example of a > 20 Year old amplifier with -132dB unweighted noisefloor below 2V input. If what you claim were remotely true you should show a DAC with -162dB unweighted. Please do, orretract yourincorrect claims.

 

If you have turned potentiometer down so that it has 1k resistance, noise is at least -124 dBV, turn it down to 10k ohm and it's at least -114 dBV. 100k pot has at least -104 dBV noise at minimum setting.

 

Noise added for any Pot/Attenuator is lowest at minimum and maximum settings (where added noise is basically zero) and worst a setting where attenuation is 6dB, presuming the source impedance is low compared to the Pot, as in this case two resistors of equal value are connected in parallel to what is in effect AC ground, so the Thevenin equivalent impedance will be 1/4 of the nominal value of the Pot/Attenuator.

 

For a 100K Pot the worst case noise is -111dBV at -6dB progressively lowering for lower settings so that at -20dB setting noise is -115dBV and at -40dB setting it is -125dBV.

 

Please make sure you get your basics right.

 

You seem to have no idea about the noise levels in modern amplifiers (easily > 130dB below full power) or of the noise levels of a Pot/Attenuator (with 20dB attenuation a 10K pot will show -125dBV).

[Miska]

Goldmund Mimesis 8 power amplifier Measurements | Stereophile.com

 

109dB unweighted noise re 1W, output 23dB above 1W, 2.1V will give full output so referred to 2.1V input that is 132dB unweighted SNR. Please show me a DAC with > 132dB UNWEIGHTED SNR referred to 2V.

 

Seems to have extremely poor THD+N. Almost 0.2% distortion at 20 kHz.

 

Looks like you need better power amp's.

 

Would you like to suggest good options for THD & IMD below 0.001% and capable of putting out at least 1.5 kW to 1 ohm without protections kicking in? Just to at least match my current one.

 

Something that would have SNR > 135 dB and THD+N < -120 dB from 0 - 20 kHz, then it would match ES9018. I think I could easily parallel eight ES9018's per channel to improve figures further.

 

Price tag safely below 5k€ and preferably no more than 2x price of the Mytek. I'm not interested on gear that costs over 10k€ because majority of audiophiles don't have such.

 

Based on their literature, Mytek, if you want the full SNR of 128dB....

 

Hmmh, mine doesn't. It can output around 4.24 Vrms. But now I have TEAC UD-501 connected to the amp.

 

This Amp is > 20 Years old! Though noisier my current Pass Aleph's still do 115dB re. 2V Input (which incidentally gives full power).

 

If we look for example at

Pass Labs XA60.5 monoblock power amplifier Measurements | Stereophile.com

 

We can see it is even worse and has HUGE IMD, components peaking at -52 dB!

 

So we should use that as a reason to throw away 10 - 20dB of Dynamic range?

 

Because when you drop the level from 0 dBFS to -10 dBFS, THD+N improves.

 

https://www.passdiy.com/pdf/B1%20Buffer%20Preamp.pdf

 

I already see from the schematic that it's not the type I would touch. It lacks cable compensations and has at least 1k output impedance.

 

That is garbage. I have no idea what poor quality amplifiers you use as reference, but it is not true for most. You should check 3rd party measurements. Many Amplifier measured by Stereophile do MUCH better than 108dB

 

I don't remember seeing many with THD+N exceeding 108 dB.

 

This looks promising:

Benchmark Media Systems, Inc. - New! Benchmark AHB2 Power Amplifier - Shipping Soon

with 130 dB SNR and -113 dB THD+N. It could get a bit closer to what a DAC output is capable of. Especially promising is that it's a low gain design, so no need to do lot of attenuation before amplification.

 

I showed above an example of a > 20 Year old amplifier with -132dB unweighted noisefloor below 2V input. If what you claim were remotely true you should show a DAC with -162dB unweighted. Please do, orretract yourincorrect claims.

 

I can get pretty impressive SNR figures for my DAC too because output relay switch the output to ground for silence.

 

But that amp has completely horrible THD+N figures.

 

You seem to have no idea about the noise levels in modern amplifiers (easily > 130dB below full power)

 

I'd be interested to see measurements of such with THD+N < -130 dB.

[sandyk]

You seem to have no idea about the noise levels in modern amplifiers (easily > 130dB below full power)

 

I'd be interested to see measurements of such with THD+N < -130 dB.

 

I have seen very few recent published designs, even those using the distortion reducing techniques as described by Audio Designer and Author Douglas Self, that do better than 122dB S/N at rated power,(e.g. 135W in 8 ohms) and with THD with 2 or 3 zeroes from 20 to 20kHz.Most readily affordable PAs aren't that good. Even at such low distortion levels, a preponderance of odd harmonic distortion is quite audible. Some Halcro amplifiers designed by Bruce Candy can even reach 4 zeroes !!!

[Miska]

Some, IMO, well performing amps from Stereophile:

 

One integrated:

NAD Master Series M3 integrated amplifier Measurements | Stereophile.com

 

And one power amp:

Marantz SM-11S1 Reference power amplifier Measurements | Stereophile.com

 

...as examples. Both have S/N slightly over 100 dB and good THD+N and IMD performance.

[Skeptic]

This looks promising:

Benchmark Media Systems, Inc. - New! Benchmark AHB2 Power Amplifier - Shipping Soon

with 130 dB SNR and -113 dB THD+N. It could get a bit closer to what a DAC output is capable of. Especially promising is that it's a low gain design, so no need to do lot of attenuation before amplification.

 

 

 

I'm quite interested in the amplifier myself, as I am thinking about putting together a new speaker system soon, but the thing which concerns me is that using a stereo DAC and amplifier seems sub-optimal for audio performance - even if you have state-of-the-art components.

 

This is because passive crossovers seem so primitive compared to what you can do with a well designed active crossover (e.g. Genelecs) or simply taking a multichannel amplifier, ripping out the crossovers in your speakers, and doing it in software.

 

 

Towards the end, even John Siau remarks that he prefers active designs (on a theoretical level, I suppose) even though active speakers often have much worse amplification.

[Miska]

This is because passive crossovers seem so primitive compared to what you can do with a well designed active crossover (e.g. Genelecs) or simply taking a multichannel amplifier, ripping out the crossovers in your speakers, and doing it in software.

 

Yes, that's another way too Genelec says their amps have THD, SMPTE-IM, CCIF-IM and DIM-100 <= 0.05% and SNR >= 100 dB.

 

ATC is not very clear, but since their power amps/modules spec from 105 dB to 110 dB SNR the speaker units are probably around similar figures.

 

Bryston's multichannel (6B SST2) power amps say IM and THD+N <0.005% and SNR:

29 dB gain: >110 dB

23 dB gain: >113 dB

17 dB gain: >116 dB

...which makes it good combination for DACs and such in order to avoid excessive attenuations before PA and running software cross-overs with something like exaSound E28.

 

But otherwise, 28B SST2 is beast to my taste.

[InfernoSTi]

Miska,

 

If a pre-amp had the follow specs, that would be considered a competitive match for DAC volume control with the ESS 9018:

 

THD+N

<0.001%, 20Hz-20KHz

 

Dynamic Range

>130dB, 20Hz-20KHz, A-weighted

 

The pre-amp is the Auralic Taurus Pre and the corresponding DAC is the Auralic Vega DAC. It is well reviewed subjectively at low listening levels but this seemed to be a good place to get a "specifications" response.

 

Thanks,

John

[Boris75]

Can one bypass the volume control inside the ESS chip and let the DAC do what it is supposed to do, that is D/A conversion?

 

The Benchmark DAC2 allows bypassing volume control with the "home theater bypass" function. I personally find the name confusing (what are home theaters doing here? This DAC is stereo anyway), but this functionality does the job.

[Andrew Allen Ballew]

Sorry, haven't been through this thread lately! Lots of interesting comments.

 

As far as Thorsten's diagrams, Miska, they were oversimplified on purpose. I think he would tell you the same thing.

 

I respect Thorsten's thoughts on the matter as much as anyone, and have conversed with him quite a bit about the subject at hand, that is, volume control on the ESS as it relates to DSD.

 

Bottom line?

 

As Miska has said a couple times, no one really knows. ESS isn't in the telling mood.

 

One of two things are possible.

 

DSD is decimated (like Wolfson does) and follows the exact same signal path (with the only variant being the nature of the oversampling filter) as PCM. Personally, I think this is likely, especially considering the ASRC that as far as I know cannot be bypassed with DSD.

 

 

or

 

 

DSD is coverted to a multi-bit representation in the FIR filter. Not decimated. Rather, is like 'DSD-wide'. (like the Cirrus chip) In this state, DSP can be performed. This intermediate representation is re-modulated in the ESS 'hyperstream' modulator. This is how Mytek would tell you it is done.

 

 

I wish ESS would come out and tell us. But the fact that they haven't cleared up this matter also makes me lean to the first option, rightly or wrongly.

 

In any case, if I wanted to be sure I was getting native DSD processing, (and by native, I include both single and multi-bit delta sigma. I don't care if it is one or the other) I would not have an ESS based DAC.

 

I currently use an iFi iDSD nano, and will be upgrading to the iDSD micro when it is available. These use the BB DSD1793, which does conversion of DSD natively, in a way very similar to Miska's own DSD dac.

 

 

And yes, as far as listening is concerned, I do prefer DSD. That is pure DSD. Like the kind we get from Channel Classics, with minimal microseconds worth of PCM editing (Thank you tailspn!!)

 

There is a certain purity, lack of digital edge there. I do agree with Thorsten that DSD lacks resolution and is colored by the ultrasonic noise. But these deficiencies don't matter so much to me. I know they are there, and I recognize them in listening. But, they to my ear aren't as evil as the PCM decimation filter. Therefore the question of DSD processing, and whether a chipset decimates or not, is very important to me.

 

Andrew

[firedog]

Hi-

 

I've personally corresponded with Michal, designer of Mytek DACs on these issues. He has written to me:

Most of inner details are not in the data sheet....Sabre DAC works in purely DSD mode, the signal path in this mode is 1 bit input >32 bit fader> DSD filter performed in DSD domain > 6 bit DSD (DS PWM) DAC > analog

 

The fact that the DAC is 6 bit DOES NOT mean it's not DSD. Basic DSD is 1 bit but it can be any number of bits, in the case 6 if it helps the performance.

 

Number 1 is a subset of 6 - a 6 bit DSD offers better D/A performance than 1 bit , that's why it's there. It's still DSD though. Sabre in DSD mode DOES NOT use interpolation filters or any other PCM processing ...In fact Sabre DAC sounds better in DSD mode because signal path is simpler.

 

Sounds like native DSD processing to me.

 

Not to insult you, but I'm willing to bet that Michal KNOWS a lot more about the subject than your speculations.

[Andrew Allen Ballew]

I am sure he does.

 

But he doesn't know anything more about the ESS chipset than say, Miska, or Thorsten Loesch. Both of them will tell you that ESS keeps the real inner working of their chips as a black box. They don't tell. And that includes Michal at Mytek, sorry.

 

I am sorry to inform you that it just isn't that cut and dry with the ESS chipset. And you are right. I can speculate all I want. I don't know one way or the other. But until ESS comes out and tell us what is what, everything is a bit of speculation.

 

If you want to believe Mytek, who has a vested financial interest in their product being DSD native, go ahead. As for me, the jury is still out.

 

 

 

EDIT.

 

Better yet, I am going to find out for myself. To confirm if decimation occurs or not, all you need are some DSD test tones like the ones Miska used with the Schitt Loki. 200 khz, right? I am going to put together a test setup sometime in the next few months and see for myself exactly what the ESS chips can do. Think a Herus is good enough? That is advertised as native DSD right?

 

Will get back to this thread when I have some results...

[AudiophileWanabe]

I have a Resonessence Labs Invicta DAC\HeadAmp, which is based on ESS Sabre DAC chips.

When in early 2013 a firmware update delivered the possibility of listening to DSD via DoP, I contacted Ressonessence Labs with this question:

 

I have updated the firmware of my INVICTA, and now I am able to play DSD files.

As I listen via headphones, I always need to use the volume control, and as far as I understand, the volume control in the INVICTA is digital, correct?

In that case, will this mean that the DSD stream will have to be converted to PCM in the Sabre DAC chip?

If so, what about if we keep the volume at 0 Dbfs? Will the DSD stream be directly converted to Analog?

 

Mark Mallinson asked for a few days to see if he could get an answer and later replied with this:

"Inside the Sabre DAC chip the DSD stream is processed at the DSD rate, but it is extending to have a 32 bit representation so that it can be scaled. This is NOT the same as converting to PCM as that implies decimation filtering. Inside the Sabre, all DSD data is processed in a PCM domain, however its PCM at 64*44.1kHz = 2.8224MHz. This data is then applied to the modulator."

 

I assume this came from ESS, since there are close ties between both companies.

 

Hope this helps.

[matthias]

"Inside the Sabre DAC chip the DSD stream is processed at the DSD rate, but it is extending to have a 32 bit representation so that it can be scaled. This is NOT the same as converting to PCM as that implies decimation filtering. Inside the Sabre, all DSD data is processed in a PCM domain, however its PCM at 64*44.1kHz = 2.8224MHz. This data is then applied to the modulator."

 

Can someone, perhaps Miska explain this? For me it appears to be some sort of contradiction.

 

Matt

[Miska]

"Inside the Sabre DAC chip the DSD stream is processed at the DSD rate, but it is extending to have a 32 bit representation so that it can be scaled. This is NOT the same as converting to PCM as that implies decimation filtering. Inside the Sabre, all DSD data is processed in a PCM domain, however its PCM at 64*44.1kHz = 2.8224MHz. This data is then applied to the modulator."

 

Can someone, perhaps Miska explain this? For me it appears to be some sort of contradiction.

 

That description is pretty much the same what Martin Mallinson (CTO of ESS) gave in his RMAF presentation (no available in YouTube anymore).

 

Sabre uses internal 32-bit data pipeline and DSD is also expanded to this same internal format, without sampling rate conversion. For PCM inputs it's the rate after the oversampling filter, and for DSD it is the native DSD rate. But it is always 32-bit. It then goes to their HyperStream modulator to output 6 bits (not parallel, but in time).

[Boris75]

"Inside the Sabre DAC chip the DSD stream is processed at the DSD rate, but it is extending to have a 32 bit representation so that it can be scaled. This is NOT the same as converting to PCM as that implies decimation filtering. Inside the Sabre, all DSD data is processed in a PCM domain, however its PCM at 64*44.1kHz = 2.8224MHz. This data is then applied to the modulator."

 

Can someone, perhaps Miska explain this? For me it appears to be some sort of contradiction.

 

Matt

 

Many thanks. It is good to know that no downsampling is happening inside the ESS DAC chip, especially after all the comments according to which the ESS chip was converting DSD streams to lower-rate PCM.

[wisnon]

Processing DSD in PCM domain????

[Boris75]

Processing DSD in PCM domain????

 

What's wrong with a 32-bit 2.8Mhz signal? Is there some magic attached to one-bit 2.8MHz encoding?

[wisnon]

Boris, here is a better question to address, the one Matthias pointed out...the obvious contradiction in the email answer above...

 

"Inside the Sabre DAC chip the DSD stream is processed at the DSD rate, but it is extending to have a 32 bit representation so that it can be scaled. This is NOT the same as converting to PCM as that implies decimation filtering. Inside the Sabre, all DSD data is processed in a PCM domain, however its PCM at 64*44.1kHz = 2.8224MHz. This data is then applied to the modulator."

[wisnon]

What's wrong with a 32-bit 2.8Mhz signal? Is there some magic attached to one-bit 2.8MHz encoding?

 

Simple answer is YES!!!!!!

[firedog]

Simple answer is YES!!!!!!

 

and what's magic about it?

[Bunpei]

I think their explanation has no contradiction in it.

It reminded me of a DSD-wide approach implemented in Sonoma(?) by Sony Professinal Audio.

AES E-Library » DSD-Wide. A Practical Implementation for Professional Audio.

In the case of the conventional DSD-wide, its bit depth is merely 8 and sampling rate is 2.8 MHz while Sabre32 adopts 32 and 24 MHz(?) at maximum!!

It's really amazing!

[wisnon]

and what's magic about it?

 

 

 

The sound quality.