Adventures in Upsampling to DSD

[Tsarnik]

Thank you barrows. :-)

2-ch 5,6448 MHz x 32 bit is a nice throughput.

IIRC, ESS developers back then were talking about going with 29 bits into the internal pipeline (9008?), I think that was at the DIY forum. Again: IIRC, someone mentioned that everything is being modulated to ~20 MHz in the ES9016 (~40 MHz in the ES9018), but I could be wrong here.

[T]he ESS volume control does absolutely nothing to the data when set to full scale, whether it is "on" or "off"

Yes.

My amp's inputs need the Stereo192-DSD's FS output to be reduced by a little less 1 dB.

Attenuate 2 dB more and most ISPs wont cause problems either.

At the same time I don't want the ES9016's internal modulator to overload from ISP while playing back CDs over S/PDIF, so I decided against the -6 dB jumpers.

So I end up using the ESS volume control, because I attenuate by 3 dB via the 'Volume Trim' menu, the 'Volume Control' menu being set to 'BYPASS'.

When I upsample in Foobar I set its 'replay gain' to -3 dB (or less) in order to reduce the chance of overloads in the SDM modulator (SACD plugin). This way the music signal is attenuated twice: 1) in Foobar and 2) in the ESS volume control.

This way also the shaped noise contained in the DSD128 signal sent to the DAC is attenuated by 3 dB in the ESS volume control. (Or is it??)

Alternatively, I could set Foobar's 'replay gain' to -6 dB and the Stereo192-DSD's 'Volume Trim' to DISABLED.

That way I would bypass the second of the two digital gain stages, but at the same time increase the DSD128 noise relative to the music signal.

So I went with the first option and my question was if this reasoning was reasonable. :-)

And hope just the same that "@Miska will correct me if understand this incorrectly."

Or anyone else, of course, for that matter. :-)

[barrows]

@Tsarnik, yes, it is a complex issue!  I do not know the final rate at which the ESS chips run, it is not included even in the (proprietary) data sheets, although the massive thread on diyaudio might have the answer somewhere, as Dustin Foreman (of ESS) did answer a lot of questions there.  It is some massively high rate.

I think in your situation you have to trust your ears.  Of course the other option (DIY-analog) would be to add a resistor pad at the input of your amplifier, this is not hard to do, and for the small level of attenuation you need would be harmless to SQ as long as good high precision resistors were used, since you just need a few I would probably use Z-foils (TX-2575), and by putting this attenuation at the amplifier input it would not negatively affect amplifier drive in any way.

 

In any case, your approach sounds totally reasonable to me, and the amount of attenuation you are now using in the ESS chip would not bother me one bit (or even 8 bits, hahaha!).

 

I think you need to ask Mytek about the volume trim feature, can they answer that question for you?  If it is happening in the analog VC chip, then it is not doing what you desire for the SPDIF input...

 

As far as the DSD noise goes, does the Mytek allow you to adjust the parameters of the IIR filter for DSD?  If you are running everything at DSD 128 you can set this to 50 kHz, or 60 kHz (I would listen to both).  Of course there is also going to be an analog filter after the ESS chip, and who knows what cutoff frequency and slope that is operating at, again have to ask Mytek that one.

[Tsarnik]

Yes, with the Stereo192-DSD one can choose from three IIR DSD filters: 50, 60 and 70 kHz with -18 dB/oct.

I am using the 60 kHz filter with DSD128 and am happy with the result. 

(I have to admit I never scratched my head about the final analogue filter... uh-oh...) 

Thank you for the tip with the resistor pad! I wouldn't feel comfortable to chance a circuit myself, though.

Cambridge Audio technicians were so kind to mail me the max. input level of the amp, my own experiments confirmed this and I also checked with Fs/4 PCM signals that ISPs clearly survive by the same amount the 'Volume Trim" is being reduced, so I believe it's not a VC after the chip. :-)

But I do not fully grasp how the ESS volume control attenuates DSD signals.

Therefore it's not clear to me whether the shaped noise contained in the DSD stream at the ESS' input is attenuated together with the "music signal" or some  DSD remodulation occurs which might even introduce more noise... 

[barrows]

9 minutes ago, Tsarnik said:

Therefore it's not clear to me whether the shaped noise contained in the DSD stream at the ESS' input is attenuated together with the "music signal" or some  DSD remodulation occurs which might even introduce more noise... 

I am almost certain that there are no worries with this.  No matter if you attenuate the signal or not the ESS chip re-modulates the incoming DSD to what is now basically termed "DSD wide", so it is always doing this regardless of volume setting:

 

DSD to DSD at 32 bits.  

 

Of course a single pass of SDM might be better overall...  But we also might be picking nits here.

[UELong]

First post on this site.  So, please forgive if a bit obtuce.  After reading this thread, I'd like to try upsampling DSD64/128 up to 256 using FB2K.  I use a program that permits me to move my mouse with keyboard commands, which also allows me to read what's on the screen with speech.  So, my simple question is:  How is it done?

 

[UELong]

On ‎9‎/‎2‎/‎2019 at 2:39 PM, UELong said:

First post on this site.  So, please forgive if a bit obtuce.  After reading this thread, I'd like to try upsampling DSD64/128 up to 256 using FB2K.  I use a program that permits me to move my mouse with keyboard commands, which also allows me to read what's on the screen with speech.  So, my simple question is:  How is it done?

 

Thanks guys.  May you lose your sight.