Some of this information is out there so I am free to talk about it.
The DAC always uses internal 32-bit data pipeline for PCM and DSD. The special part for DSD is that it also expanded to this same internal format, and importantly it does this without any sample rate conversion! (in the conventional sense - decimation)
For PCM inputs the final sample rate is the rate after the oversampling filter, while for DSD it is the native DSD rate. But each "0" or "1" is scaled to 32-bits - so that it can have volume control applied. Everything then goes to the HyperStream-II modulator.
While downsampling to DSD to PCM is indeed a harmful process, for example: one that goes from a 2.8MHz DSD signal to a 352.8kHz PCM signal. Inside the ESS DAC what is happening that the DSD data is basically converted to a 32-bit stream but while retaining the original 2.8MHz sample rate!!! This is really important - because it means that the data is the same - it just can now be scaled.
The ESS DSD process is not a downconversion where the sample rate is decimated - instead you should think of it as an upward conversion where each 1 bit value because a 32bit value
It's actually a particularly awesome way to handle DSD. I can't say I have heard anything better than the ES9028/38 at handling DSD with volume control.
I hope that helps!
Cheers!
Russ
Last edited:
Hi Russ/Brian,
I had posted earlier requesting for some preview into the multichannel solution.
How are the 4 IVs stacked with the DAC board? Or is it a single IV board with 4 stereo channels?
Do the IVs have on board regulators?
How many power supplies are required in the multi-channel solution?
I had posted earlier requesting for some preview into the multichannel solution.
How are the 4 IVs stacked with the DAC board? Or is it a single IV board with 4 stereo channels?
Do the IVs have on board regulators?
How many power supplies are required in the multi-channel solution?
Last edited:
Some of this information is out there so I am free to talk about it.
The DAC always uses internal 32-bit data pipeline for PCM and DSD. The special part for DSD is that it also expanded to this same internal format, and importantly it does this without any sample rate conversion! (in the conventional sense - decimation)
For PCM inputs the final sample rate is the rate after the oversampling filter, while for DSD it is the native DSD rate. But each "0" or "1" is scaled to 32-bits - so that it can have volume control applied. Everything then goes to the HyperStream-II modulator.
While downsampling to DSD to PCM is indeed a harmful process, for example: one that goes from a 2.8MHz DSD signal to a 352.8kHz PCM signal. Inside the ESS DAC what is happening that the DSD data is basically converted to a 32-bit stream but while retaining the original 2.8MHz sample rate!!! This is really important - because it means that the data is the same - it just can now be scaled.
The ESS DSD process is not a downconversion where the sample rate is decimated - instead you should think of it as an upward conversion where each 1 bit value because a 32bit value
It's actually a particularly awesome way to handle DSD. I can't say I have heard anything better than the ES9028/38 at handling DSD with volume control.
I hope that helps!
Cheers!
Russ
Russ, this is awesome!
There was a huge controvesy in major China forums that, due to its controllable-volume nature in DSD, ES9018 always convert DSD to PCM, it is not a true 'Direct DSD' DAC. While I believe it is misunderstood, due to the lack of decimation filter applied, I have a hard time to explain it. Your clarification nailed it. Thanks
Hi
Just tried out my 28 Pro SE. No lock, silence (and buzzing on output to amp when power off, suggesting no input connection). All regs etc fine, and D1 etc, from an Otto, connected exacted the same as to my previous Buffalo II. I assumed that with the stereo firmware the serial inputs would be connected as default, but I suspect this is not the case? How do I correct this, and select the spdif input if required? I'm using Trident SR regs.
Thanks
Paul N
The bit about data-3 in the manual meant nothing to me I'm afraid
Just tried out my 28 Pro SE. No lock, silence (and buzzing on output to amp when power off, suggesting no input connection). All regs etc fine, and D1 etc, from an Otto, connected exacted the same as to my previous Buffalo II. I assumed that with the stereo firmware the serial inputs would be connected as default, but I suspect this is not the case? How do I correct this, and select the spdif input if required? I'm using Trident SR regs.
Thanks
Paul N
The bit about data-3 in the manual meant nothing to me I'm afraid
You should not need OTTO any more... the only way to input SPDIF is through the SPDIF input (not D1 as in the old BII etc).
Last edited:
Some of this information is out there so I am free to talk about it.
The DAC always uses internal 32-bit data pipeline for PCM and DSD. The special part for DSD is that it also expanded to this same internal format, and importantly it does this without any sample rate conversion! (in the conventional sense - decimation)
For PCM inputs the final sample rate is the rate after the oversampling filter, while for DSD it is the native DSD rate. But each "0" or "1" is scaled to 32-bits - so that it can have volume control applied. Everything then goes to the HyperStream-II modulator.
While downsampling to DSD to PCM is indeed a harmful process, for example: one that goes from a 2.8MHz DSD signal to a 352.8kHz PCM signal. Inside the ESS DAC what is happening that the DSD data is basically converted to a 32-bit stream but while retaining the original 2.8MHz sample rate!!! This is really important - because it means that the data is the same - it just can now be scaled.
The ESS DSD process is not a downconversion where the sample rate is decimated - instead you should think of it as an upward conversion where each 1 bit value because a 32bit value
It's actually a particularly awesome way to handle DSD. I can't say I have heard anything better than the ES9028/38 at handling DSD with volume control.
Thanks for the detail Russ. So - if I understand correctly- there is no advantage in using upsampled DSD if everything is processed at native 2.8mhz bitrate.
Mark
Hmm - not really what I meant to convey - DSD is scaled at the rate it arrived (and ultimately upsampled by the ASRC unless you turn it off) - my example was for standard rate DSD
The main point was - this is not a destructive decimation down-sampling to PCM. It is an upsampling to a 32bit 'PCM like' stream that contains the DSD data faithfully reproduced.
The main point was - this is not a destructive decimation down-sampling to PCM. It is an upsampling to a 32bit 'PCM like' stream that contains the DSD data faithfully reproduced.
Last edited:
Goto: My present scenario is a B-IIIse/NTD-1 (v.4) combo. I oversample everything to DSD 128 using Audirvana+ (which gives great control over filter parameters). I find the ESS chips sound better feeding them DSD, and higher rates might be even better, but I am not sure I can tell the difference (different filter parameters in A+ probably make more difference than DSD 128 vs 256, etc). Also, running everything to the DAC at DSD only simplifies set up, and the DAC can be further optimized (filter settings, analog filter corner, etc) just for DSD, without having to change settings etc.
For the B-IIIPRO 9038, I plan a Mercury output stage, and Amanero/Hermes/Cronus running a single Pulsar 45.1584 clock and running DSD 128. It is going to be awesome.
I do not know the clock rate needs of the 9038, but I am pretty sure you need >80 MHz masterclock for DSD 512 if you want to go that high, but I do suspect that if you run the ESS chip in async mode, it will resample the DSD data to the new clock rate, right Russ? To run sync with the Cronus, and get DSD 512 you could do a 90.3168 clock...
I plan to run synchronously from the 45.1584 on the Cronus, as I am fine with DSD 128 oversampling, and the lower phase noise of the lower clock rate offers an advantage. Also having only one oscillator in the DAC may give benefits as well in terms of power supply noise modulation form beating clock frequencies...
For the B-IIIPRO 9038, I plan a Mercury output stage, and Amanero/Hermes/Cronus running a single Pulsar 45.1584 clock and running DSD 128. It is going to be awesome.
I do not know the clock rate needs of the 9038, but I am pretty sure you need >80 MHz masterclock for DSD 512 if you want to go that high, but I do suspect that if you run the ESS chip in async mode, it will resample the DSD data to the new clock rate, right Russ? To run sync with the Cronus, and get DSD 512 you could do a 90.3168 clock...
I plan to run synchronously from the 45.1584 on the Cronus, as I am fine with DSD 128 oversampling, and the lower phase noise of the lower clock rate offers an advantage. Also having only one oscillator in the DAC may give benefits as well in terms of power supply noise modulation form beating clock frequencies...
Yes - with an async master clock it is eventually asynchronously reclocked (unless you disable the jitter eliminator) - with a sync clock the dpll has no work to do - (and you can be explicit about this in a register) - but once again - even when present the ARSC happens in a lossless way. IMO there is no down-side to running async - and there is actually quite a lot of upside. The bottom line is - either approach yields excellent results. Sync/async do what gives you the most pleasure. BTW the DPLL and ASRC is *far* better in the ES9028/38 than it was on the 9018. It is also significantly faster (to lock) and more accurate in initial esitmates.
The bottom line is - either approach yields excellent results. Sync/async do what gives you the most pleasure. BTW the DPLL and ASRC is *far* better in the ES9028/38 than it was on the 9018. It is also significantly faster (to lock) and more accurate in initial esitmates.