[Odezra]

@labeelen I did extensive a/b testing between pcm upsampling and dsd via Roon and HQplayer. I don’t believe I am getting any noise - nothing that’s audible to my ears when playing tracks.

If anything - I found the opposite of your results with the test. I found pcm to be ever so slightly hazier / veiled and dsd to have a clearer sound with less / no noise.

Not disputing your results. Happy to try and repeat the test on my machine if you want to DM the file?

My machine was set to nos on for pcm and dsd, with a ock-1 external clock in place, via LAN (connected via an external second google nest router with LAN isolators). My hypothesis is that it could be noise / jitter in your line, and it’s there on both pcm / dsd but only audible when upsampled via dsd. Pure guess. It could also obviously be a defective unit but would be good to rule out the former?

 

[Jake2]

Jussi (HQPlayer) supplied me with a test file (with recorded silence -120dB). When I upsample this to DSD (using Roon or HQPlayer) I hear a soft noise on R26; when I upsample the same file to pcm there is no noise. I have asked the seller (shenzhenaudio) to contact Gustard and verify if this is a known issue and if it can be fixed.

I’ve been using Roon > HQP > R26 for a while now, Mac Mini M1 upscaling to DSD512 with great results. Sounds excellent.

I’m still on Gustarender v1.3 - will upgrade to 1.4 when I return home after the Christmas hols.

I got bad popping at the start of DSD tracks initially when I first got the R26 but in the course of changing a few HQP settings it went away. Wasn’t entirely sure which one(s) did it but was happy to have it gone for the past 6 weeks.

Revisiting my HQP settings this week it seems that the popping is only there when I have Adaptive rate output checked (selects upscaled frequency from same rate family as source frequency. Ie. an integer multiple ) vs greyed (adaptive with interpolation to selected maximum up-sample frequency). I only recently changed back from the latter to the former this week as it is supposed to be better sonically. The pops at start of DSD tracks are now back alas.

Edit: this is with DSD direct ON, which is theoretically better/purer than off as the R26 would presumably attempt to do further up sampling, doubling up. Or at least if not actually upsampling as DSD512 is already at the R26’s max, ‘some’ processing by the K2 PLL, as OFF certainly sounds different, a little softer.

 

[labeelen]

Is the R26 in DSD direct mode?
Is NOS on or off?

DSD Direct was ON, when I switch it OFF the noise is gone. (thanks !)
PCM NOS was ON but switching it to OFF has no effect (I assume because it only affects PCM)

 

[camrector]

DSD Direct was ON, when I switch it OFF the noise is gone. (thanks !)
PCM NOS was ON but switching it to OFF has no effect (I assume because it only affects PCM)

Nice!!!
I’ve always wanted to know what was happening with the DSD signal path
In Roon I use “native” for DSD playback.
Does having the R26 with DSD Direct ON, bypass the clock of the R26?
Maybe @Jake2 and @Odezra were not having this issue because they used an external clock?

I do not upsample DSD but prefer the Direct setting off.

 

[Jake2]

Nice!!!
I’ve always wanted to know what was happening with the DSD signal path
In Roon I use “native” for DSD playback.
Does having the R26 with DSD Direct ON, bypass the clock of the R26?
Maybe @Jake2 and @Odezra were not having this issue because they used an external clock?

I do not upsample DSD but prefer the Direct setting off.

To clarify my recent observations:
- I have an external clock and I do have the DSD popping issue with v1.3, upsampling adaptive rate ticked when using DSD Direct ON. I can't recall if it happened with DSD Direct OFF as I don't generally use it.

 

[camrector]

To clarify my recent observations:
- I have an external clock and I do have the DSD popping issue with v1.3, upsampling adaptive rate ticked when using DSD Direct ON. I can't recall if it happened with DSD Direct OFF as I don't generally use it.

Gotcha. The popping issues I was very aware about (never experienced though), so I was more addressing this new “issue” @labeelen was experiencing.
Glad it’s solved.
It appears that all the issues with DSD come from the multiple settings and different combinations of those settings.
Kind of crazy to think about all a company must take into account when designing a DAC lol.

 

[labeelen]

Gotcha. The popping issues I was very aware about (never experienced though), so I was more addressing this new “issue” @labeelen was experiencing.
Glad it’s solved.
It appears that all the issues with DSD come from the multiple settings and different combinations of those settings.
Kind of crazy to think about all a company must take into account when designing a DAC lol.

Unfortunately the search is not over yet.
Jussi (HQPlayer) explained to me that the noise is gone because when the DSD Direct is off in R26, it is converting DSD to PCM and playing it as PCM through the R2R ladder instead of feeding it straight to the separate DSD D/A converter.

The noise is gone and DSD files sound fine, but they are still not being processed via the pure DSD DAC.

 

[camrector]

Unfortunately the search is not over yet.
Jussi (HQPlayer) explained to me that the noise is gone because when the DSD Direct is off in R26, it is converting DSD to PCM and playing it as PCM through the R2R ladder instead of feeding it straight to the separate DSD D/A converter.

The noise is gone and DSD files sound fine, but they are still not being processed via the pure DSD DAC.

How does he know that this is the case?
The R26 is still seeing and displaying that it is DSD. There is no information from Gustard saying that a change to PCM is occurring.

 

[camrector]

Unfortunately the search is not over yet.
Jussi (HQPlayer) explained to me that the noise is gone because when the DSD Direct is off in R26, it is converting DSD to PCM and playing it as PCM through the R2R ladder instead of feeding it straight to the separate DSD D/A converter.

The noise is gone and DSD files sound fine, but they are still not being processed via the pure DSD DAC.

The R26 still sends the DSD signal to the R2R ladder, bypassing the interpolating FIR filter. That’s not the same as converting to PCM.

 

[cobrabucket]

I think that if there is any volume attenuation, the signal must be converted to PCM as native DSD can't have any DSP. So in this case, my understanding is that with "DSD Direct" set to OFF and giving you the ability to adjust the volume via the Gustard that there is a conversion happening internally. Please let me know if I am mistaken about that.

 

[Odezra]

Yep - i think there are two things here.

In roon - selecting 'enable native DSD processing' sends a DSD file to the 1 bit DSD chip. I use this mostly when upsampling and it sounds better to my tastes, than PCM oversampling.

I also think Cam is right - as according to the user manual - the final D/A conversion is done via the ladder without any filters on (when NOS is selected).

I'd love to know more about the precise differences here at the technical level.

In either case - with Roon upsampling and DSD direct on, I get no noise. I do get a small pop at the start of any playlist kicking off, but it's just once and doesn't reoccur. I am on Gustarender v1.3.

 

[Odezra]

Nice!!!
I’ve always wanted to know what was happening with the DSD signal path
In Roon I use “native” for DSD playback.
Does having the R26 with DSD Direct ON, bypass the clock of the R26?
Maybe @Jake2 and @Odezra were not having this issue because they used an external clock?

I do not upsample DSD but prefer the Direct setting off.

I think the clock choice (internal v external) is purely driven by selecing of 'int' or 'ext' in the clock menu setting on the Gustard.

With 'int' - the internal clock dictates things, with 'ext' the internal clock synthesiser works to the rate as dictated by the external clock (whether directly connected or connected via DDC).

 

[camrector]

Yep - i think there are two things here.

In roon - selecting DSD native sends a DSD file to the 1 bit DSD chip. I use this mostly when upsampling.

I also think Cam is right - as according to the user manual - the final D/A conversion is done via the ladder without any filters on (when NOS is selected).

I'd love to know more about the precise differences here at the technical level.

In either case - with Roon upsampling and DSD direct on, I get no noise. I do get a small pop at the start of any playlist kicking off, but it's just once and doesn't reoccur. I am on Gustarender v1.3.

This can help on the technical side of things. https://www.analog.com/media/en/training-seminars/tutorials/mt-017.pdf


@cobrabucket the attenuation happening here is with the interpolation filter.

 

[Odezra]

This can help on the technical side of things. https://www.analog.com/media/en/training-seminars/tutorials/mt-017.pdf


@cobrabucket the attenuation happening here is with the interpolation filter.

Thanks - that file is really good. I'll need to read it again to fully take it all in.

The one thing it doesn't really answer though is the Gustards 1bit architecture and the R2R ladder. None of the diagrams have that example in line with the Gustard's architecture, where the Gustard uses it's 1 bit DAC and then uses the R2R ladder to do the final d/a work..

It seems a bit of a contradiction really. I would of thought the 1 bit DAC would of done the full conversion, without the need for the R2R ladder? That's why I'd love to know more about the technical details behind why the Gustard manual says with 'DSD Direct on', the file still goes through the R2R ladder. I always thought the ladder was for PCM conversion and wasn't needed for DSD direct?

My knowledge here is not great so would love to know more.

 

[camrector]

I think the clock choice (internal v external) is purely driven by selecing of 'int' or 'ext' in the clock menu setting on the Gustard.

With 'int' - the internal clock dictates things, with 'ext' the internal clock synthesiser works to the rate as dictated by the external clock (whether directly connected or connected via DDC).

Yup finding this about the internal fir filter being bypassed answered my question.

I think that bypassing this oversampling filter with DSD direct on doesn’t spread out the quantization noise enough and increases the signal-to-noise ratio.