ESS Sabre Reference DAC (8-channel)

[tritosine]

...really, how about the self noise, all we know it has 4th order noise shaping. Can't find no measurements.

Is AKM still the best when it comes to out of band noise ? (not counting r2r ofc) In the AKM patent they say it has 8bit modulator, maybe it has even lower order noise shaping than ESS?

[needsp]

Hi Dustin

The circuit I'd probably try is shown here (Reply #29)-

monica - where are we now and moving forward

Paul

[qusp]

i'm not saying there is significant self noise, I dont have any specific numbers there either, but it only stands to reason that since its not a perfect world that there would be some noise created that would be common mode, thus mostly cancelled in the CMMR

[dusfor99]

Originally Posted by steve jones

Hi Dustin,

I'm using transformer output with my Buffalo II. Wired across + and -. I like it alot.

Was wondering if the ESS Sabre chip likes to see the output's + and - current returned to the DAC's AC signal ground @ AVCC/2 (by using the transformer's primary center tap), or is simply wiring the transformer from + to - (and ignoring the center tap as I have done) complete the shortest current path ?

Great job on the Sabre chip, BTW.

-Steve

Hi Steve,

To get the best out of the DAC, what you want to do is cause the current going into the power supply pins to be constant, that way you wont get power supply modulator causing harmonic distortion. This can be accomplished by returning the current back to some DC potential. The value of the DC the current gets returned to doesn't seem to matter a lot. Some have tied it to ground, some to DVcc/2 and some DVcc. Puting the DAC into a mode where its pins are at a virtual ground also acomplished this. The extra benefit of have the virtual ground is that the internal voltage coefficients of the DAC circuits are no longer able to cause THD. I have been wanting to try building up a discrete I/V for some time, but im just too busy with other things at the moment.


Dustin

[dusfor99]

Originally Posted by needsp

Hi Dustin, Calvin

I too am extremely interested in using the ESS9018 in a single ended way, with a discrete component I/V stage. I intended to use "half" of the differential output- i.e. to feed the I/V converter from either the + or the - output, and ground. The I/V circuit can be set up to provide 1/2 AVCC on its input.

But you mention grounding "DACb". What did you mean by this please?

Many thanks

Paul

I Paul,

So you plan to use only DAC and then tie the DACb off to Ground? This is OK, but you will get better performance using both sides. I am interested to see what people come up with.

Dustin

[Calvin]

Hi,

my aim would be to have as low a parts number count as possible in the signal path. At the moment I favour Owen´s idea as in here: A New Take on the Classic Pass Labs D1 with an ESS Dac
The Q remaining is how to generate a singleended signal from this I/V-stage. Possibilities are either with a dedicated differential-to-singleended converter, be it a OPamp or a discrete stage, or to simply take just one I/V-output and to ground the second DAC-output. One could ground via a relais, or an opto-switch.
THD would probabely be higher in the second case, but I´m happy to see numbers and to hear that the THD-increase is not due to the DAC-outputs but due to the capabilities of the analog stages.
Thanks for this info Dustin.

jauu
Calvin

[steve jones]

Originally Posted by dusfor99

Hi Steve,

To get the best out of the DAC, what you want to do is cause the current going into the power supply pins to be constant, that way you wont get power supply modulator causing harmonic distortion. This can be accomplished by returning the current back to some DC potential. The value of the DC the current gets returned to doesn't seem to matter a lot. Some have tied it to ground, some to DVcc/2 and some DVcc. Puting the DAC into a mode where its pins are at a virtual ground also acomplished this. The extra benefit of have the virtual ground is that the internal voltage coefficients of the DAC circuits are no longer able to cause THD. I have been wanting to try building up a discrete I/V for some time, but im just too busy with other things at the moment.


Dustin

Hi Dustin,

I'm using a 3.3v regulated constant current source to feed the Sabre's output supply.

I guess what info I am after is are the Sabre's outputs (+, -) exclusively current sourcing, current sinking, or mixed. If they are either both sourcing or both sinking, I'll need to use the center tap to a ground reference for current return. If they are mixed (can either source or sink), no center tap connection is required as each (+) output will sink what the other (-) output sources.

Thanks,
-Steve

[dusfor99]

Hi Steve,

The outputs can both sink and source current.

Hope this helps.


Dustin

[Spartacus]

Hi Dustin,

Nice to have you around these parts. A couple of Q's while you're here .... first, what circuit design was used for the 108dB measurement for voltage mode output? I'm guessing an opamp as balanced to SE converter?
Second question, I understand there are ways to optimise the Sabre's modulators for best DNR. Russ White mentioned randomzing the modulators ... do you have information on this? I'm currently putting together my own 'C' firmware for the Sabre.

Thanks in advance for any help you can give,
Dan

[OzOnE_2k3]

Hi, all,

Here's another of my famously long posts, but I've tried all sorts of things to fix this final issue I'm having...

I'm STILL messing with my amp build - it has been sitting on the shelf for the past seven months but I really want to finish it now.

I bought a small prototyping adapter so I can now fit the CPU chip (from a Denon 1905) next to the Denon DSP board into my amp chassis.

For a bit of background info - my previous ramblings and a close-up photo of the DSP board + Sabre DAC are here...

ESS Sabre Reference DAC (8-channel)

I've attached another photo of the current layout of the amp.

I'm now testing some THAT1200 opamps as an output stage for the Sabre (2 channel only atm). They appear to work great, and provide very good ground isolation which stops all the "ground loop" type noise I was getting before.

The THAT1200 outputs go through some 1uF M-Caps and then directly into the Sure amp module.

Firstly, the good news is that I've now confirmed that the cause of the digital noise when the Sabre first locked onto a signal was indeed caused by the default "low" DPLL bandwidth value of the Sabre. When I set it to medium bandwidth (via I2C using a buspirate), the noise disappears completely!

The problem I'm having now is that the Sabre locks onto the I2S audio from the DSP perfectly fine when using an optical input from a DVD player, but...

...If I use the optical output from the PC (tried two different sound cards), the Sabre locks on and plays audio for three seconds, then it looses lock for one second and the process repeats?

The sound cards work perfectly through optical to a commercial amp (even through a very cheap 5 metre optical cable). I tried three different cables with the Sabre DAC and I still have this "lock" problem?

The decoded I2S signals from the DSP board look identical on the scope whether I'm using the DVD player optical output (audio fine), or PC optical output (audio intermittent)?

Of course, on my amp build, the Denon DSP is responsible for decoding SPDIF and converting to I2S for the Sabre, so it's probably the way I have it all connected (as usual )...

The I2S signals from the DSP board are at 3.3v levels, but I'm getting a huge amount of overshoot? Is this likely due to incorrect impedance matching, or the use of a "1X" scope probe?

I've attached o'scope shots of the I2S bitclock and wordclock going into the Sabre board (please note the different time / division).

You can see that the overshoot is pretty huge. The flat peaks are at 0v and 3.3v (0V at same position when scope input set to "GND"), so the overshoot actually goes BELOW 0v and ABOVE 3.3v?

Apart from the overshoot, the I2S signals stay solid at ALL times even when the Sabre looses lock? Assuming the overshoot is not an artifact of the scope impedance, could it be the cause the Sabre loosing lock?

The CPU / DSP / Sabre ground is now isolated from the Sure amp / Tracopower SMPS / mains ground. I do appear to have around 130mV of noise on the CPU / DSP / Sabre ground though (around 295KHz, so very likely to be SMPS switching noise).

It's important to know that the I2S signals pass through some buffer chips and series resistors on the DSP board before they reach the Sabre.

The Sabre PCB also has a 47K pull-down resistor on each digital input line, and then has a 22 ohm series resistor between each input and the ESS chip.

The buffer chips on the DSP board are all powered from 3.3V. So, the signals follow this path...

(Note: Bit clock and Word clock generated by Sanyo LC89057 SPDIF receiver. So, clocks recovered from SPDIF PLL.)
LC89057 -> 47ohm -> 74LCX244 -> 47ohm -> (47K pull-down) -> 22ohm -> Sabre chip.


(I2S audio data generated by SHARC DSP. SHARC is also clocked by the Sanyo SPDIF receiver.)
SHARC -> 74LVX157 mux -> Sabre chip.


Note: The CPU uses the mux chip to "mute" the I2S audio signal (tie output to ground) when it detects a decoding error. The mux is not activated while the Sabre looses lock, so it's basically just a buffer.

Should I bypass some of the series resistors on the Bit clock and Word clock to try to reduce the apparent overshoot?

EDIT: The overshoot on the actual I2S audio signal is only around 0.48v above / below the flat peaks (ie. around 4.26v between lowest and highest overshoot).

Thanks in advance,
OzOnE (Ash).