ESS ES9039PRO and ES9027PRO announced

[pLudio]

smsl su9 already got this chip

full name is es9039mspro, what the hell is ms inserted there

That's the MQA tax. Buy a DAC without it.

 

[evan123]

The new one. SU-9 Pro.

Measurements of SMSL SU9 Pro ESS9039Pro DAC - L7Audiolab

台式设备 今年五月份（2022年5月）的时候ESS发布了9038Pro芯片的下一代（？）产品，ESS9039Pro。当时有过一阵子讨论，从编号来看9039Pro更像是9038Pro的"小改"，双目三林这次给我们抢先带来了 SMSL SU9 Pro 这款基于9039Pro芯片的产品

www.l7audiolab.com

In the L7 review, forum member nameless999 wrote some detail about the practical differences between 9038 and 9039. Quoting nameless999, using a machine translation to English:

The first 9039? Not for smsl, I really want to say a few words about this chip. The official website of the data manual has been released. How can I see that it is a shrinking version of 9038!? The biggest feature is the more power-saving hyperstream modulator. What channels does this power saving rely on? I looked at it at a glance and saw two points from the data manual 1. The worse reconstruction filter eliminates the brickwall, adds an additional minimum phase, and worse filter attenuation. 2. For lower-frequency hyperstream modulators, mclk (MasterClock) 9038 has an asynchronous mode of 100mhz (>2048fs) and 9039 is 50mhz (1024fs). You know, the data on the 9008 white paper is 40mhz. Isn't this equivalent to doing it back... This mclk is related to the final modulation process, which affects noise shaping and also directly affects the resolution.

 

[Matias]

It must be a trade off of some sort: lost a few things here to gain other things we don't know yet? Maybe this solves the low level artifacts and/or the hump?

REVIEW: Sabaj A20d 2022 Version DAC [Part II] - Resolution & Distortion. (Windows 11 22H2 available. Greetings from Peru!)

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

 

[KSTR]

9039PRO and 9027PRO also have revised "recommended output stages" in their datasheets (https://www.esstech.com/products-overview/digital-to-analog-converters/sabre-audiophile-dacs/) which use a back-to-back de-glitch capacitor plus no "angst" resistor for the I/V capacitor (sounds familiar, doesn't it ;-). They also use a common-mode control loop with an extra opamp to force the output voltages to be exactly 0V rather than using a bias voltage directly.
The 9027PRO circuit shown in the datasheet looks, let's say, somewhat special... probably a drawing error...

 

[BR52]

9039PRO and 9027PRO also have revised "recommended output stages" in their datasheets (https://www.esstech.com/products-overview/digital-to-analog-converters/sabre-audiophile-dacs/) which use a back-to-back de-glitch capacitor plus no "angst" resistor for the I/V capacitor (sounds familiar, doesn't it ;-). They also use a common-mode control loop with an extra opamp to force the output voltages to be exactly 0V rather than using a bias voltage directly.
The 9027PRO circuit shown in the datasheet looks, let's say, somewhat special... probably a drawing error...

this "auto bias" savs money no adjustment anymore + long time stability.

 

[BR52]

this "auto bias" savs money no adjustment anymore + long time stability.

interesting the capacitor between the outputs before the i/v circuits. Looks like the output stage is the same cr,,,,,,, like before. Did they read your posts? Why do they use the fast
expensive 1611 for this task??

 

[KSTR]

this "auto bias" savs money no adjustment anymore + long time stability.

interesting the capacitor between the outputs before the i/v circuits. Looks like the output stage is the same cr,,,,,,, like before. Did they read your posts? Why do they use the fast
expensive 1611 for this task??

Normally there isn't any adjustment pot, rather the proper resistor selection is made for the Vref divider ceating the bias voltage. But there still is always a residual error and yes, aging / temp changes etc can move the DC bias around.

But the most important point is that the dynamic (AC) common mode signal is removed by this servo loop (a known technique in instrumentation). By this, you can use one output leg single-ended with almost no degradation wrt distortion and noise, quite in contrary to the simple biased circuits.

The CM servo loop also is the lowest noise and distortion circuit to balance an output around the output reference. In single ended mode it's a bit less good but still better than the usual approach with a subtractor circuit after the I/V's.

The CM servo loop is forming a composite amplifier loop (the I/V's being the slave buffers, paralleled). This, and the previous remark, are two reasons why you want to use a nice fast precision OpAmp here. Further, you want to have controlled conditions for the master and slave OpAmps bandwidths and very similar bandwidths are a good operating point for such a composite. Finally, using the same OpAmp type throughout is convenient and cost-effective in mass production (the 1611 will actually be 1/2 1612 in most implementations, and in a true precision application the OPA1612's will be OPA2211's -- same chip but selected for better offset ratings).

Conceptually, the whole shebang is equivalent to a fully differential OpAmp and one can use a good one but probably specs would be a bit worse than with the discrete solution, plus you're restricting yourself to non-standard parts and footprints (bad idea these days) and there is no room for optimization (like the positive feedback seen in the 9039 circuit which creates an ever so slightly negative input impedance... whatever that will do for circuit perfomance).

 

[BR52]

9039PRO and 9027PRO also have revised "recommended output stages" in their datasheets (https://www.esstech.com/products-overview/digital-to-analog-converters/sabre-audiophile-dacs/) which use a back-to-back de-glitch capacitor plus no "angst" resistor for the I/V capacitor (sounds familiar, doesn't it ;-). They also use a common-mode control loop with an extra opamp to force the output voltages to be exactly 0V rather than using a bias voltage directly.
The 9027PRO circuit shown in the datasheet looks, let's say, somewhat special... probably a drawing error...

Yesterday it was a little late.... Today I need a little help to understand the ESS schematic.
Mostly we can see here some tweaks of the I/V stage, from KSTR already shown here February 2022.
The offset problem is here solved with an active stage. So far, so good. Now I was expecting a slow (integrating)
offset compensation. Here we can see something very fast acting (51 pF!!) integrating (compensating) capacitor and a fast OP -amp. Some Ideas?

​

 

[KSTR]

Here we can see something very fast acting (51 pF!!) integrating (compensating) capacitor and a fast OP -amp. Some Ideas?

Yes, it's a very fast "servo", the idea is to remove all DC and AC common-mode voltage at the outputs, up to and beyond audio frequencies.

 

[BR52]

Normally there isn't any adjustment pot, rather the proper resistor selection is made for the Vref divider ceating the bias voltage. But there still is always a residual error and yes, aging / temp changes etc can move the DC bias around.

But the most important point is that the dynamic (AC) common mode signal is removed by this servo loop (a known technique in instrumentation). By this, you can use one output leg single-ended with almost no degradation wrt distortion and noise, quite in contrary to the simple biased circuits.

The CM servo loop also is the lowest noise and distortion circuit to balance an output around the output reference. In single ended mode it's a bit less good but still better than the usual approach with a subtractor circuit after the I/V's.

The CM servo loop is forming a composite amplifier loop (the I/V's being the slave buffers, paralleled). This, and the previous remark, are two reasons why you want to use a nice fast precision OpAmp here. Further, you want to have controlled conditions for the master and slave OpAmps bandwidths and very similar bandwidths are a good operating point for such a composite. Finally, using the same OpAmp type throughout is convenient and cost-effective in mass production (the 1611 will actually be 1/2 1612 in most implementations, and in a true precision application the OPA1612's will be OPA2211's -- same chip but selected for better offset ratings).

Conceptually, the whole shebang is equivalent to a fully differential OpAmp and one can use a good one but probably specs would be a bit worse than with the discrete solution, plus you're restricting yourself to non-standard parts and footprints (bad idea these days) and there is no room for optimization (like the positive feedback seen in the 9039 circuit which creates an ever so slightly negative input impedance... whatever that will do for circuit perfomance).

Your answer was faster then my question

 

[BR52]

Your answer fas faster then my question

Thank you was very helpfull I could learn.

 

[KSTR]

^ You're welcome

 

[777]

C18.1=2.2n is equall good for es9038pro ?

 

[evan123]

If the 9039 internally remodulates to exactly 1024fs (as nameless999 suggests), I wonder if that might imply one could feed the DAC DSD1024 with no internal remodulation (just expansion to wide DSD for volume control). It would be unusual but definitely possible today to feed at DSD1024.

Historically ESS has been quite opaque about what is going on under the hood so we will probably never know, but that might have something to do with the change in remodulation rate from 100MHz in the 9038?

 

[piaseczek]

It's interesting that in fist implementations of the es9039pro (smsl su9pro & d400es) probably synchronous mode is used and yet measurements are great.

 

[Herbert]

The 9039Pro can run in Hardware Mode - Does this mean this DAC could run without a controller?
So simply SPDIF or I2s input, proper supply power, proper output stage and the DAC would work?
See page 10 - 14:

https://www.esstech.com/wp-content/uploads/2022/10/ES9039MPRO_ES9039PRO_v0.2.1.pdf

 

[David_M]

Yesterday it was a little late.... Today I need a little help to understand the ESS schematic.
Mostly we can see here some tweaks of the I/V stage, from KSTR already shown here February 2022.
The offset problem is here solved with an active stage. So far, so good. Now I was expecting a slow (integrating)
offset compensation. Here we can see something very fast acting (51 pF!!) integrating (compensating) capacitor and a fast OP -amp. Some Ideas?
View attachment 239109

​

Old post here ... Can someone please explain to me why there's positive feedback on the U14 opamps? Also, why the fast servo with a 2,600Hz bandwidth? I'm used to seeing BW values under 100Hz or so.

 

[BR52]

Old post here ... Can someone please explain to me why there's positive feedback on the U14 opamps? Also, why the fast servo with a 2,600Hz bandwidth? I'm used to seeing BW values under 100Hz or so.

It’s not a dc servo its a full bandwidth servo to suppress the broader bandwidth common mode.

 

[David_M]

The I/V stages for the ES9069 and 9039Q2M (they are identical) are nuts to me. So unlike the one for the 9039PRO shown in above post. Can some explain to me how this IV stage works? Why positive feedback on U2.1? Why apply DAC input to both U2.1 filter netwirk and non-inverting input of U2.2? How is this circuit truly balanced?

 

[David_M]

It’s not a dc servo its a full bandwidth servo to suppress the broader bandwidth common mode.

Cool! Thank you for clearing that up. Is the assumption here that music signals within the filter bandwidth are differential and not common to both +/- inputs?