[Odezra]

R26 linearity degrades rapidly above 18bits, so you might want to do 19bit depth processing to match the ladder accuracy and then avoid dithering? Does doing 19bits cost much in processing time compared to 16bits with dithering?

Just for R2R reference, the Holo May shows what good engineering can do (maybe about four times better). That is unlikely to be audible usually.

If you put the r26 sinad into the calculator you get 16 bits tho?

https://www.everythingrf.com/rf-calculators/sinad-to-enob-calculator

 

[Nada]

If you put the r26 sinad into the calculator you get 16 bits tho?

https://www.everythingrf.com/rf-calculators/sinad-to-enob-calculator

That calculator does not take into account up sampling for a R2R DAC that processes discrete bit depth on a resistor ladder with a known linearity. In other words it's a misleading tool to use if doing PCM upsampling to the R26. Whether that's audible is another question.

 

[Odezra]

That calculator does not take into account up sampling for a R2R DAC that processes discrete bit depth on a resistor ladder with a known linearity. In other words it's a misleading tool to use if doing PCM upsampling to the R26. Whether that's audible is another question.

Cool - and how did you work out that 19 bits would be the optimal level for the r26 and pggb?

In terms of processing time - I have just tried a 256 track at 32 bits to 705/768 and it takes about 15 mins for a single track on a M1 MacBook Pro. So should be less than that to go 19 bits

 

[Nada]

work out that 19 bits ... ?

This is a bit over the top for most of us. PGGB lets you dial in the specific bit depth you want. So why not do it and leave the useless LSBs quiescent:

1. Work out the R2R DAC bit depth it can actually render – ignore the specs, eg if a DAC says 24bit don't believe it.
a. find a linearity plot with actually measurements processing 24 bit PCM
b. Choose a deviation level eg +/- 1dB to be kind eg R26 about -110dBFS
c. Bit depth is about dB/6 so the R26 is about 18 bits ( the resistors processing LSB below level 18 are not doing much useful)
d. Add one bit on for good measure, so you don't need to dither.

2. Make sure the chosen bit depth is more than you need for dynamic range, eg 19bits gives 116dB range which is way more than needed

 

[Alenotta]

Man this all sounds crazy. I'm over here streaming Tidal. I don't think I'm ready for that rabbit hole.

 

[Nada]

Its a lot less crazy than listening to Tidal and MQA

 

[rockcity007]

Man this all sounds crazy. I'm over here streaming Tidal. I don't think I'm ready for that rabbit hole.

+1 Allenotta...Way over my head!

 

[camrector]

Cool - and how did you work out that 19 bits would be the optimal level for the r26 and pggb?

In terms of processing time - I have just tried a 256 track at 32 bits to 705/768 and it takes about 15 mins for a single track on a M1 MacBook Pro. So should be less than that to go 19 bits

How was the sound compared to 16fS/128?

 

[Beatles1966]

I use the PGGB realtime plugin for FOOBAR. Prefer HQPlayer.

 

[camrector]

This is a bit over the top for most of us. PGGB lets you dial in the specific bit depth you want. So why not do it and leave the useless LSBs quiescent:

1. Work out the R2R DAC bit depth it can actually render – ignore the specs, eg if a DAC says 24bit don't believe it.
a. find a linearity plot with actually measurements processing 24 bit PCM
b. Choose a deviation level eg +/- 1dB to be kind eg R26 about -110dBFS
c. Bit depth is about dB/6 so the R26 is about 18 bits ( the resistors processing LSB below level 18 are not doing much useful)
d. Add one bit on for good measure, so you don't need to dither.

2. Make sure the chosen bit depth is more than you need for dynamic range, eg 19bits gives 116dB range which is way more than needed

Is there any benefit to sending 32bit to the R26 even though 19bit is the output?

 

[Davidv100]

Deleted - It was a question for Nada, but I am not sure I would have understood the answer anyway.

 

[JerseyD]

can you stand the R26 on its side? Any reason why I couldn’t or shouldn’t?

 

[incredulousity]

Received ad email today. It seems that Drop is now selling R26. Apologies if this is old news.

 

[The Hawk]

$1485 on Drop

https://drop.com/buy/gustard-r26-discrete-decoder-dac?utm_placement=0&referer=8BUP6T&mode=guest_open&utm_campaign=Automated Daily Promotional 2023-02-12&utm_source=SparkPost&utm_medium=email&utm_term=Daily Promotional&utm_content=1676190450794.614344437555206293962436&defaultSelectionIds=973580

 

[labeelen]

Is there any benefit to sending 32bit to the R26 even though 19bit is the output?

No, the opposite is true, sending >16 (or 20) bits may degrade the sound because of added distortion.

It was explained to me as follows:
“

It is not possible to make an R2R DAC that is linear to more than about 20 bits. Beyond that, the bits just produce incorrect values (distortion). This distortion is particularly noticeable with low level signals and around zero-crossing region of high level waveforms. So you need to check linearity sweep measurement of the DAC to see at which point it begins to go off and only use the linear range of the converter. When you do this with a suitable noise-shaper, the distortion disappears while having no loss of dynamic range (limited by analog thermal noise floor). If you use even one bit too much, the effect disappears and the distortion pops back.

”