[FredA]

in here he says

Accuracy of the ladder resistors (tolerance):

...
The solution is to correct the ladder and not only depend on the tolerance of resistors. It’s a combination of both: Ultra-low tolerance resistors controlled by a correction technology using very high speed FPGA are applicable in in our design.​

but if you scroll down to these sections ..

Importunacy of the FPGA in the R2R 7:
FPGA tasks in the R2R 7​

... you'll see no mention of correcting resistors


does anyone know ...
1) what the physical resistor precision is
2) if correction is being applied

Two fpgas are used on each of the da-7 dac boards. Compensation is done, that's been mentioned. I think 0.01% resistors are used. Good enough given compensation.

Two fpgas are used on the digital board to handle the inputs and fifo reclocking. So the r2r 7 is a six-fpga dac, and the da-7 modules are even two-sided (there are approximately 800 precison resitors used counting both boards). So you are talking high-end implementation. You know your money is not going to marketing, distributor and resellers when popping up the hood. Check the nad m51 on the other end.

 

[DACLadder]

The R2R 11 DAC boards appear uncompensated as the description is different from the other models. Before I step in it further... Are the R2R 1, 2, and 7 DAC boards similar designs (with compensation)? Or are we getting feature creep or cost cutting with the newer models?

The digital board alone has three programmable parts. Two Xilinx CPLDs that handle the I/Os plus the Altera FPGA for digital filtering and driving the analog boards.

 

[DACLadder]

Deleted duplicate post...

 

[FredA]

Yeah, R2R2 is done
and looking at the R2R1 vs R2R7 graphs, I honestly don't see a difference. So DA7/DA8 isn't such a big difference after all?
I'm guessing there will be a freeze on orders for the R2R7 until the verdict is out on the R2R1

Look closely: the noise floor is lower with the r2r 7. The distorsion peaks are 10dB lower, and it!s reflected in overall disto: 0.0005% to 0.003% in favor of the r2r 7. The r2r 7 will provide better details and imaging, according to the graphs. It even has an edge over the pcm1704, which is to say it's top notch. It's great that Kingwa produces 3 versions for 3 different price backet and performance levels.

 

[Wynnytsky]

Two Xilinx CPLDs that handle the I/Os plus the Altera FPGA for digital filtering and driving the analog boards.

and in the case of the R2R7 there are four DA7 boards, with each carrying two Xilinx CPLDs (need to type it in hopes of committing to memory). 8 CPLDs in total doing resistor correction and a single Altera Cyclone 4 FPGA chip exposed in the middle.

Perhaps the R2R7 is starting with better measuring resistors (ie: so less correction applied)
If the resistors didn't measure as well for the R2R1 then it would need to correct more aggressively, and that would help justify the price diff.

 

[FredA]

The R2R 11 DAC boards appear uncompensated as the description is different from the other models. Are the R2R 1, 2, and 7 DAC boards similar designs with compensation? Or are we getting feature creep or cost cutting with the newer models?

The R2R 7 digital board alone has three programmable parts. Two Xilinx CPLDs that handle the I/Os plus the Altera FPGA for digital filtering and driving the analog boards.

I assume they all are. The difference is in the number of resistors (overall complexity) and possibly the fpga's specific model, would have to look closely. There is no reason not using compensation in the cheaper offerings, in my understanding, as far as production cost is concerned. I wonder how this works because to compensate resistors that are not accurate enough, you would have to do a calibration on each board to have the proper values outputted.

 

[FredA]

and in the case of the R2R7 there are four DA7 boards, with each carrying two Xilinx CPLDs (need to type it in hopes of committing to memory). 8 CPLDs in total doing resistor correction and a single Altera Cyclone 4 FPGA chip exposed in the middle.

Perhaps the R2R7 is starting with better measuring resistors (ie: so less correction applied)
If the resistors didn't measure as well for the R2R1 then it would need to correct more aggressively, and that would help justify the price diff.

Hard to say but i think there are just two da-7 boards. I know at least that they are two-sided. And i have never seen the other side, anybody did? The da-7 is a bit like the moon!

 

[Wynnytsky]

to compensate resistors that are not accurate enough, you would have to do a calibration on each board to have the proper values outputted

I would have thought the correction coefficients unique to my DAC were programmed into the Cyclone4 chip somewhere, and the 8 Xilinx chips crunch those parameters to apply the correction. But if that's the case, how would those parameters be safe from overwrite when I replaced accurate with smooth?

Speculating in a knowledge vacuum is so much fun

 

[FredA]

I would have thought the correction coefficients unique to my DAC were programmed into the Cyclone4 chip somewhere, and the 8 Xilinx chips crunch those parameters to apply the correction. But if that's the case, how would those parameters be safe from overwrite when I replaced accurate with smooth?

Speculating in a knowledge vacuum is so much fun

Each board has to be individually calibrated so they can be replaced in the future. Plus otherwise the Altera would need a different firmware for each dac. Therefore a dac board has to be a black box, each one within tolerance from one another.

 

[Wynnytsky]

And i have never seen the other side, anybody did? The da-7 is a bit like the moon!

Those brass threaded binding posts are used to stack them puppies 2 per side.
The R2R1's DA8s are smaller, but you can easily see them stacked.

Therefore a dac board has to be a black box, each one within tolerance from one another.

agreed. The correction coefficients must be somewhere directly on each DA board. Perhaps that Xilinx has some write-once registers.

Both the R2R 1and7 use 4 DA boards.
I asked this before, but is that because each output channel requires a dedicated board? (2 rcas, 2 xlrs)

 

[DACLadder]

Both the R2R 1and7 use 4 DA boards.
I asked this before, but is that because each output channel requires a dedicated board? (2 rcas, 2 xlrs)

I would assume 2 boards per channel in a balance configuration. Each balanced XLR output is buffered and single ended RCAs are buffered and summed from the balanced DACs. That is typical but could be something different.

 

[FredA]

I would assume 2 boards per channel in a balance configuration. Each balanced XLR output is buffered and single ended RCAs are buffered and summed from the balanced DACs. That is typical but could be something different.

From Currawong's picture, i only see one board per channel. @Currawong, can you confirm? And Kingwa told me the boards are double-sided. Plus there is the aluminum cover on top. I don't see that there is room for a 4th level (at the bottom, you have the output stage). You have a sign magnitude ladder on each half, each one implementing one phase of the balanced dac.

 

[Currawong]

You're correct. They are boards with pairs of 0.05% Vishay resistors (equivalent to 0.0025), as Kingwa told me that he couldn't get 0.01 and the Chinese 0.01 resistors he didn't think sounded as good.

 

[FredA]

You're correct. They are boards with pairs of 0.05% Vishay resistors (equivalent to 0.0025), as Kingwa told me that he couldn't get 0.01 and the Chinese 0.01 resistors he didn't think sounded as good.

Kingwa has done his homeworks, it seems and the response is very good so far. It's rare to see such a buzz around an entry level like the r2r 11. I am still really tempted by one for the office. I am only worried it will sound so good, as it was the case with the nfb-5 i had been using for a few weeks at some point, that i won't be able to work and listen to it at the same time. And there is perhaps something historical about this unit. Best dac under 500$ ever probably and a discrete r2r one, with a capable headphone amp as a bonus. It's almost irresistible.

 

[DACLadder]

@FredA Thanks. Currawong’s pictures looks like one board in a stack with a metal shield on top. Double sided circuit board with parts on both sides makes sense.

Dumb question.,, How is DSD implemented in R2R 7?