My friend, Chiaki, is assembling DAC board he re-designed for ES9038PRO. His original design was for dual mono ES9018S. Ten ES9038PRO chips, a minimum quantity for one order were delivered to a personal user like us about two months ago from a local distributor in Japan.
Within a few weeks, I will have a chance to try Chiaki's prototype DAC board.
I have been working with the 9038 chip.
We got initial samples which had a serious flaw - very disappointed with those - and we now have the fixed chips as well.
They are quite a departure from the 9018 in many ways - to the extent that our I/V stage is a completely new design.
One thing to note - the DS is wrong on several points. Dual mono in a production setting where you apply chip gain correction to adjust for chip to chip output differences (up to 28% difference from chip to chip) will not even come close to the datasheet numbers - because of the dynamic range lost because you must reduce the output swing in order to have headroom to apply sufficient correction.
Unfortunately - the 9018 also has the very same chip-to-chip output variance flaw - but it does not even have a way to apply correction automatically like the 9028/9038 do.
So unless you are willing to live with as much as 28% output level difference between channels - dual mono is not an option without losing dynamic range/SNR.
Cheers!
Russ
We got initial samples which had a serious flaw - very disappointed with those - and we now have the fixed chips as well.
They are quite a departure from the 9018 in many ways - to the extent that our I/V stage is a completely new design.
One thing to note - the DS is wrong on several points. Dual mono in a production setting where you apply chip gain correction to adjust for chip to chip output differences (up to 28% difference from chip to chip) will not even come close to the datasheet numbers - because of the dynamic range lost because you must reduce the output swing in order to have headroom to apply sufficient correction.
Unfortunately - the 9018 also has the very same chip-to-chip output variance flaw - but it does not even have a way to apply correction automatically like the 9028/9038 do.
So unless you are willing to live with as much as 28% output level difference between channels - dual mono is not an option without losing dynamic range/SNR.
Cheers!
Russ
I'm afraid he can't do that. All the chips at his hand now are already assigned to his project. Some users made reservation for his DAC board.
If I understood that correctly, you're saying the full scale output current varies as much as 28% from chip to chip?
Wow...
The low master clock speeds needed for synchronous mode are also surprising - lower than AKM, TI, etc. need for running the modulator at the highest rate. For ES9018 I guess you really could never turn off the ASRC even if the clocks were synchronized.
Last edited:
That is correct - the error (as documented) is +/- 14% from nominal - and it is explicitly stated in the datasheet in the newer chips - but just to check out a hunch I measured some of my own 9018s against each other and found two that were about 20% apart... I was very disappointed to say the least. Funny I never thought to check until I saw the new datasheets.
So the situation for dual mono is pretty grim - because you have to reduce the output level on both chips - and then adjust the master trim to get them identical - only now they are not operating at the same digital scale at all and each will have different dynamic range - and you are still left with two parts that have different output impedances... you actually can't adjust for that in the digital domain satisfactorily.
So my inclination is not to use more than one chip unless you went dual stereo (or n-channel) - basically just stacking chips.
So the situation for dual mono is pretty grim - because you have to reduce the output level on both chips - and then adjust the master trim to get them identical - only now they are not operating at the same digital scale at all and each will have different dynamic range - and you are still left with two parts that have different output impedances... you actually can't adjust for that in the digital domain satisfactorily.
So my inclination is not to use more than one chip unless you went dual stereo (or n-channel) - basically just stacking chips.
By dual stereo do you mean a situation where you're feeding each I/V converter with 8 total outputs (4 from each chip)?
Essentially, a stacked version of their quad-differential current mode example?
Essentially, a stacked version of their quad-differential current mode example?
Correct - but there is no reason it would have to be 2 channels - it could be 4 - or 6 or 8 etc... the idea is that you would never have a channel to channel difference in total relative output impedance. This approach would quickly reach the point of diminishing returns. 🙂
That said - it makes no sense to do any of that - at least with the 9038 which is basically 4 9028s stacked already - at least on the analog end (800R output-Z for 9028 200R for 9038 per channel - of course with +/- 14% error). There is really no need to stack - or to go dual mono. You won't really see any payoff in the real world.
That said - it makes no sense to do any of that - at least with the 9038 which is basically 4 9028s stacked already - at least on the analog end (800R output-Z for 9028 200R for 9038 per channel - of course with +/- 14% error). There is really no need to stack - or to go dual mono. You won't really see any payoff in the real world.
I also post in the deleted thread...I get the point rather than disappointing from the information.Similar mechanism(I think) DACs, NPC's old SM5865 and SM5866 is stated 10% and 20% absolute tolerance (V-I resistance) but they include relatively matched feedback resistors(I-V resistance), thus their fullscale level are less dependent on their absolute tolerance.ESS's detecting resistors also must be relatively matched to output resistance.
Otherwise we can adjust it in analog stage or reference voltage, which is analog supply in these DACs.The trimming job is not much problem for amateur.
Otherwise we can adjust it in analog stage or reference voltage, which is analog supply in these DACs.The trimming job is not much problem for amateur.
Trimming does not address root problem - which is once you decimate the output you can't achieve the datasheet numbers for dynamic range - and you will have unequal performance between the chips.
One problem was faulty DoP detection which would cause random unlocks with PCM input when using certain data pins. That made the DAC very hard to prototype. I have yet to determine if it is actually fixed. 🙂 We shall see.
I am not sure there are others yet (other than the output impedance matching) - it is too soon to say. Hopefully none! 🙂
Please drop me an usual internet e-mail (not a PM function on this forum).
You will find my address on this post
http://www.diyaudio.com/forums/digital-source/142562-microsd-memory-card-transport-project-56.html#post3300744
Dear Russ,
Information you have brought to us here are really great!
I think we all owe to you.
Best regards,
Bunpei
Information you have brought to us here are really great!
I think we all owe to you.
Best regards,
Bunpei
My pleasure - it's the kind of thing any of us would have found eventually.
Cheers!
Russ
In this web page in Japanese language, we can read Dustin's interview by Mr. Shima of OPPO Japan.
ESS Technology‚Ì‹Z�pŠ²•”‚É•·‚*�A�Vƒtƒ‰ƒOƒVƒbƒvDAC�uES9038PRO�v�i‰»‚̃|ƒCƒ“ƒg (1/3) - Phile-web
( Google translated https://translate.google.co.jp/translate?sl=ja&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=http%3A%2F%2Fwww.phileweb.com%2Finterview%2Farticle%2F201611%2F02%2F407.html&edit-text= )
He talked about some new features on ES9038PRO.
After reading the page, I understood that in the case of ES9038PRO, each channel has four DA units, four times of DA unit in ES9018S. Consequently, ES9038PRO outputs more analog current than that of ES9018S. Some of my friends told me that there would be no single op-amp that could accept such a large current when multiple channel analog outputs are tied in parallel. How can we deal with this definitely? Dustin's answer in the interview is "connecting to op-amp by each channel first and summing up outputs of op-amps secondly". Has anyone tried any alternative approach?
ESS Technology‚Ì‹Z�pŠ²•”‚É•·‚*�A�Vƒtƒ‰ƒOƒVƒbƒvDAC�uES9038PRO�v�i‰»‚̃|ƒCƒ“ƒg (1/3) - Phile-web
( Google translated https://translate.google.co.jp/translate?sl=ja&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=http%3A%2F%2Fwww.phileweb.com%2Finterview%2Farticle%2F201611%2F02%2F407.html&edit-text= )
He talked about some new features on ES9038PRO.
After reading the page, I understood that in the case of ES9038PRO, each channel has four DA units, four times of DA unit in ES9018S. Consequently, ES9038PRO outputs more analog current than that of ES9018S. Some of my friends told me that there would be no single op-amp that could accept such a large current when multiple channel analog outputs are tied in parallel. How can we deal with this definitely? Dustin's answer in the interview is "connecting to op-amp by each channel first and summing up outputs of op-amps secondly". Has anyone tried any alternative approach?