and replace the ES9822 for the new chip, if you'll change that 2 times you'll get grade-A quite possible.
I find it very hard to believe that I've ended up with two ES9822s landing in the tiny% that is C grade. Both of the one's I have currently soldered are within a couple of 0.1dB of one another.
If IVX's experience of 40% giving 128-129dbA with the front ended connected, and then 60% giving 127dB, are accurate then statistics alone says that the performance I've got is very unlikely related to random chance.
Technically, I suppose, I am getting datasheet performance, but I am far more inclined to believe that it's something systemic causing the reduction in noise floor rather than the chip itself.
This is why I wanted IVX to see what his noise floor is with the front end disconnected vs having it connected. As my front end is on a separate PCB it's easy to unplug and see. I am guessing IVX would have to remove the 4 resistors on the inputs to the ADC to accomplish the same thing.
It's possible his input stage is just stupidly quiet or mine isn't performing how it should.
The other possibility is that the way I've got the ADC wired up is having an impact on the performance. First of all I'm using it in slave mode. I would not expect this to have any impact on it's level of performance but it might. Secondly the ADC can accept its master clock input from a variety of different pins and it's possible that some are better than others for optimum performance. Then there's the jitter on the master clock itself, which I wouldn't expect to be a problem, but it could be. Anything is possible.
If IVX's experience of 40% giving 128-129dbA with the front ended connected, and then 60% giving 127dB, are accurate then statistics alone says that the performance I've got is very unlikely related to random chance.
Technically, I suppose, I am getting datasheet performance, but I am far more inclined to believe that it's something systemic causing the reduction in noise floor rather than the chip itself.
This is why I wanted IVX to see what his noise floor is with the front end disconnected vs having it connected. As my front end is on a separate PCB it's easy to unplug and see. I am guessing IVX would have to remove the 4 resistors on the inputs to the ADC to accomplish the same thing.
It's possible his input stage is just stupidly quiet or mine isn't performing how it should.
The other possibility is that the way I've got the ADC wired up is having an impact on the performance. First of all I'm using it in slave mode. I would not expect this to have any impact on it's level of performance but it might. Secondly the ADC can accept its master clock input from a variety of different pins and it's possible that some are better than others for optimum performance. Then there's the jitter on the master clock itself, which I wouldn't expect to be a problem, but it could be. Anything is possible.
I suppose the problem here is that 3dB of spread seems typical from ADC manufacturers. If you're building a dual mono system specifically to get 128dBA but actually get 125dBA you're going to be, and rightly so, annoyed. You've doubled up your silicon just to get the typical performance of a single chip.
Yes, I would think so. That's why I don't understand what's at stake here, all are in the data sheet spec, any outstanding result is pure chance that has nothing to do with good/bad engineering.
And at these level, comparing implementations has very slim chances to lead to any general and practical conclusions. Think the DAC/ADC in the instrument may have about the same contribution. There are about a million of other possible error sources.
I just ordered one of the Cosmos ADC's. I'm looking forward to exploring its performance. Clearly better than ESS's first effort. I'm not completely clear on the input impedance vs sensitivity switches. it seems the sensitivity goes up as the impedance goes up. It seems backwards so maybe an explanation could help. Also is the input AC or DC coupled?
hey, 1audio, did you see that? Cosmos ADC | index
Input ranges/impedance:
1.7Vrms 640Ω
2.7Vrms 1kΩ
3.5Vrms 1.3kΩ
4.5Vrms 1.66kΩ
6.7Vrms 2.46kΩ
7.6Vrms 2.82kΩ
8.5Vrms 3.12kΩ
10Vrms 3.48kΩ
43Vrms 13.6kΩ
Note: unbalanced impedance about 30% less.
Using the COSMOS ADC in balanced mono mode, ie parallelled inputs, would mean the input resistance is half that:
320 Ohms
500 Ohms
1250 Ohms
etc.
Correct?
320 Ohms
500 Ohms
1250 Ohms
etc.
Correct?
ESS adc Datasheets released without NDA
The NDA is no longer needed to get the datasheets.
Analog to Digital Converters – ESS Technology, Inc.
The NDA is no longer needed to get the datasheets.
Analog to Digital Converters – ESS Technology, Inc.
Nice to get access to the datasheets without an NDA. And it seems like a lot of information has been added compared to the one I got under NDA.
Coming back to the DAC for a moment.
Long time ago I made some measurements and experiments on a low cost DAC board with ES9038Q2M, see ES9038Q2M Board
As the measurements show, I was not able to get better THD at 1 kHz than around -103dB at 0dBFS. I have just measured again at -1dBFS, where I get around -105dB.
I would like to hear from members here, who have experience with the ES9038Q2M: Is this what can be expected without THD compensation or is there something that must be changed to improve it?
I would like to get a reasonable performance before making my own board, with improved low noise and low impedance supply etc.
Coming back to the DAC for a moment.
Long time ago I made some measurements and experiments on a low cost DAC board with ES9038Q2M, see ES9038Q2M Board
As the measurements show, I was not able to get better THD at 1 kHz than around -103dB at 0dBFS. I have just measured again at -1dBFS, where I get around -105dB.
I would like to hear from members here, who have experience with the ES9038Q2M: Is this what can be expected without THD compensation or is there something that must be changed to improve it?
I would like to get a reasonable performance before making my own board, with improved low noise and low impedance supply etc.
@IVX: Hi ... looks very interesting your ADC, however, I hope you may be able to help with a couple of clarifying questions ...
- Will the driver work with Windows 7?
- What is the difference between the various "bundles" in your Aliexpress shop?
- Will you be setting up within-EU distribution?
- and, lastly, are you considering an integrated ADC & DAC combination running off of the same clock (I understand it should be preferable for loudspeaker loop-back measurements)?
Have a good evening,
Jesper
- Will the driver work with Windows 7?
- What is the difference between the various "bundles" in your Aliexpress shop?
- Will you be setting up within-EU distribution?
- and, lastly, are you considering an integrated ADC & DAC combination running off of the same clock (I understand it should be preferable for loudspeaker loop-back measurements)?
Have a good evening,
Jesper
Jens, I was myself not able to bring a ES9038Q2M chip under -110dB THD+N at -1dB (with compensation), that was an heroic effort that included selecting the chip(s) (and destroying a couple of boards in the process). The -110dB i also the number that I see for the best commercial DAC implementations spec, although I don't think they are selecting the chips, so my boards (4 layers) could be 1-2dB less than optimal.
I believe -105dB @-1dB (without compensation) is a fair stating point to tweak the THD compensation, but you need to find out if it is the noise or the THD dominating the THD+N. I suspect it is the THD, so compensating may bring you to about the same numbers as above. I am using OPA1612 in the IV.
Side note, to me, the disappointment of the year is the CS43198 which is slightly worse than a non-compensated ES9038Q2M, and the noise dominates the output. I am not surprised the "high end" market penetration of this DAC chip is not very good. Or perhaps I am doing something very wrong with it, dunno...
Last edited:
sure
different ADC grades there are as bundles
better to ask in our Aliexpress shop, I have no idea what does mean setting up within-EU distribution
the physical delay from Cosmos ADC input to the 9038SG3 output is 10mS at Fs=192k ASIO4ALL.
Just take a look to the ESS 9038Q2M demoboard and make the same, with compensation you'll see THD+N a bit less than -120db(depends on the chip noise performance), without compensation probably -117db(depends on the chip distortions performance).