More discrete R2R stuff, this time for the diy'er

[georgehifi]

And it’s cheap!!
Looks like it’s designed in Maracaibo, Venezuela and made and shipped from China

R-2R (R2R) Ladder Multibit DAC discrete components NOS/non-filtering COAX/optic | eBay

R-2R R2R 24 bit Ladder +-0.5% Multibit DAC discrete components NOS/non-filtering | eBay

Cheers George

 

[soekris]

And it's crappy, I notice the not very precise resistors, and it's not sign magnitude, which mean distortion with be quite high.... It also have high output impedance, so noise will also be high....

Also missing a lot of features, like I2S input, digital filters, digital volume control and DSD support, no reclocking or FIFO.

So price/performance is not really that impressive.... Compared to my dam1021-05 I would call them expensive....

 

[abraxalito]

The guy does not market himself very well at the first link - his bullet point list says 1% resistors but if you investigate the part number he gives two lines beneath it turns out to be a Yageo 0.1% type. Still not great for such a DAC but an order of magnitude better than his 1% claim.

At the 2nd link, he claims 0.5% and the Susumu resistors he cites are indeed 0.5%. But why use a looser tolerance resistor for a higher bit-depth DAC?

 

[lcsaszar]

I experimented with discrete R2R DAC some 10+ years ago:

DDAC project

Please use Google translator. Here is the circuit diagram attached. I built only one channel with 0.1% resistors. It worked 😀 but I did not listen to music on it, so I can't comment on its performance.

 

[adason]

...and i thought you guys were talking about reel to reel 🙂

 

[georgehifi]

I experimented with discrete R2R DAC some 10+ years ago:

DDAC project

Please use Google translator. Here is the circuit diagram attached. I built only one channel with 0.1% resistors. It worked 😀 but I did not listen to music on it, so I can't comment on its performance.

So the DDAC was you, bit of praise back then for it, did you ever do any more with it?

Cheers George

 

[miro1360]

I experienced with a current R-2R architecture + deglitcher based on a hold capacitor.
The resistors has to be matched at least to a 0.003% accuracy for 16 Bits DAC.

ltspice simulation for an 8-Bit architecture worked fine:

https://ctrlv.cz/shots/2020/05/03/b0qM.png
https://ctrlv.cz/shots/2020/05/03/OaG6.png

before detlitcher:
https://ctrlv.cz/shots/2020/05/03/CyH5.png

after deglitcher:
https://ctrlv.cz/shots/2020/05/03/sCr0.png

but it escalated into crazy diagram with a crazy PCB 😀
https://ctrlv.cz/shots/2020/05/03/EzVo.png

It works a bit, but not as I wanted, because these switches (1 ohm TS5A4624) are not able to handle the speed of higher current switching through 1k and 2k resistors. I may try to replace these resistors with a higher value, like 20k for lower current, but I haven't tried it yet. I discovered DAC8811 😀 Still no big motivation for that. AD1862 sounds superbly and costs a funny fraction of the price 😀

 

[olsond3]

You really can't buy resistors with sufficient precision or switches with matched pull up/down resistance. I believe the modern approach is to just use the 0.1% resistors, make a 24 or 32 bit network and then use an FPGA to calibrate it using a lookup table.

 

[lcsaszar]

So the DDAC was you, bit of praise back then for it, did you ever do any more with it?

Cheers George

No, I lost motivation
BTW I applied the variable resistor calibration approach, by using a switchable digital staircase generator for each of the top 6 bits, to be viewed on an oscilloscope. For the lowest 10 bits, I found 0.1% resistors sufficient. Also the switching MOS-FETs' RDSS of 0.1 ohm is negligible compared to the lattice resistors of 200+100 kohms. I never realized the deglitcher, neither the second channel, neither 2s complement to offset binary converter, neither...