Hello. For 5 years i have soekris 1021 dam dac. I have modded absolutly everythink. Dac is powerer by lifepo4 bateries with lots off ultra capacitors. Plays good, neutral sound. But im think what next? Mayby soekris 1121 , have little better clock. I want to change dac pcb only, not my PSU. But on allie There are lots of r2r dacs. Even 64bits. Have any one try chineese r2r diy pcb dacs? They looks great. For example.
https://m.pl.aliexpress.com/item/1005002703232363.html?gatewayAdapt=Pc2Msite
Or
https://m.pl.aliexpress.com/item/10...216754730608374124e5d08!12000028801405640!rec
Can any body compared soekris dac to something else from diy dacs? Very thanks for help.
https://m.pl.aliexpress.com/item/1005002703232363.html?gatewayAdapt=Pc2Msite
Or
https://m.pl.aliexpress.com/item/10...216754730608374124e5d08!12000028801405640!rec
Can any body compared soekris dac to something else from diy dacs? Very thanks for help.
Your first link looks to be a DSD only DAC, calling it 'R2R' is misleading. The second one seems to handle both PCM and DSD.
The Soekris is one of the better implementations of R2R DACs that I'm aware of.
Anybody who promises 64-bit performance from an R2R DAC is lying. There's no way you'll get to that level of precision even with ±0.01% tolerance resistors. Soekris gets to 17-18 bit resolution with ±0.05% resistors: https://www.audiosciencereview.com/...ds/soekris-dam1021-r2r-dac-measurements.2324/. You'd get 18-19 bits with ±0.01% tolerance resistors.
Tom
Anybody who promises 64-bit performance from an R2R DAC is lying. There's no way you'll get to that level of precision even with ±0.01% tolerance resistors. Soekris gets to 17-18 bit resolution with ±0.05% resistors: https://www.audiosciencereview.com/...ds/soekris-dam1021-r2r-dac-measurements.2324/. You'd get 18-19 bits with ±0.01% tolerance resistors.
Tom
I have tried this: https://www.aliexpress.us/item/3256804799332720.html
And this one: https://www.aliexpress.us/item/3256804799290351.html
Both need IIS input - which of course can come from many sources... But an easy USB to IIS board is this one: https://www.ebay.com/itm/353802762966
You can of course also use a HDMI input board like this: https://www.aliexpress.us/item/2251832845605595.html
In fact, I am listening to the balanced version of the DAC with the HDMI to IIS board now. Sounds great driven by my laptop playing ripped .WAV files...
All you have to add is regulated +5V DC and +/- 15VDC.
Good luck.
D.
And this one: https://www.aliexpress.us/item/3256804799290351.html
Both need IIS input - which of course can come from many sources... But an easy USB to IIS board is this one: https://www.ebay.com/itm/353802762966
You can of course also use a HDMI input board like this: https://www.aliexpress.us/item/2251832845605595.html
In fact, I am listening to the balanced version of the DAC with the HDMI to IIS board now. Sounds great driven by my laptop playing ripped .WAV files...
All you have to add is regulated +5V DC and +/- 15VDC.
Good luck.
D.
All these DACs with FPGAs had so much digital noise that was adding huge digital signature in whole sound outcome, For me...
I find more acceptable "glue logic", without processors, only digital ICs and R2R... Still has to be with analog filter at the output but more relaxing and natural sound. But that is my opinion only...
I find more acceptable "glue logic", without processors, only digital ICs and R2R... Still has to be with analog filter at the output but more relaxing and natural sound. But that is my opinion only...
Yes
They probably advertised this data from I2S LE line, booth channels (2x32 BCK) length... 🙁
For each channel With standard digital bus input, max lenght is 32bit per channel and ofcourse only when data word is "full" with 32bit. In all other cases rest of the bits are zeroes, and there are no point to put r2r ladder in these bits. I didt it in one discrete dac i made and it of course worked out with 32bits music. But I think that it adding a noise in other cases...
...
But this feature is configurable even in discrete design it deserves just to disconnect lower bits (driver IC and R2R part of ladder) and set that point to GND as finished word of LSB.
You seem incessant on playing with the tech for what reason? In all honesty DACs are really a solved problem and the least of your concerns in terms of audio performance. They are good for tinkering and getting overly involved in for no real benefit, though.
When You by some finished boards, You dont know how is the % tolerance and PPM of Rs... 🙁
The small tolerance and very low PPM This is the from crucial importance for R2R lader dacs. Because You cant trim it due to change of temperture...
So the firs thing is to find reliable manufactirer.
It is not a big deal to exchange all of the r2r ladder net, but still, it is bigger handy job than cost of the resistors (also not cheep)...
...
I will give a tip to try first some cheeper diskrete "glue logic" DAC without FPGAs and programming.
And try to obtain some FPGAa R2R to compare for Yourself?
I have found something colled Total r2r dac. Dont know how to buy pcb for this. It dont have resistors and i dont se clock on board. The good think about soekris dac is that we have pure raw output from r2r ladder,with no opa and capacitor .
AND guaranteed % tollerance value of the resistors...
Still it is "processor" based digital core dac...
@Johnsykes
The relatively simple, but only digital logic ICs + ladder R2R net was present decades ago...
this is one of the first examples (Japanese DIYers acomplishes more good "discrete" desings even before 2000 year...)
Most of that working experiments are not in the net anymore...
.
the Monica DAC from Sonic illusions. 2006. Year
http://www.sonicillusions.co.uk/discrete_dac.htm
more links on the page
Schematic for 18 bit with 595 IC
http://www.sonicillusions.co.uk/images/dac18.png
.
All oter designs are just itteration of the same way...
another one:
https://www.diyaudio.com/community/threads/my-humble-discrete-dac.333898/page-2
.
One of the implementations Bartola Valves made BUT with very low tolerance and from the best quality Rs. Expensive type.
I dont remember which one, but it could be traced by the photos...
https://www.bartola.co.uk/valves/tag/discrete-dac/
.
I didnt use 595 because it is sort of "combo" and more complicated.
I used DIP, bigger versions. Old school 🙂 164 for the shift register and 564 / 574 as buffer-driver for R2R. For ballanced mode.
.
There are many diskrete designs of DACs
BUT nobody wrote clear instructions about the input format, before the shift register.
The input format is LEFT justified. AND MSB must be inverted. Clock is the same as in I2S structure.
Left means that LE should be delayed for one cycle of BCK.
I compare to I2S because this is dominant format from the USB digiral interfaces (like Amanero)
.
Discrete dac can not play with I2S input.
Last edited:
That is brand made DAC and probably You cant find the sch or PCB. Main thing is that Total DAC uses very expensive R. low tolerance from TI. I think Z foil...
Hi Zoran: you are right that these FPGA R2R DACs have digital background noise. I overcome this by using a volume control after the DAC and before any downstream gear. I have had great success with https://www.aliexpress.us/item/3256804344033842.html. Then let the digital source (PC in my case using HDMI out) run at max volume. The digital noise is still there, but the signal is now so strong that the signal to noise ratio drowns out the noise. I no longer hear it - because low volume at using the volume control after the DAC also dampens the background noise.
Good luck.
Good luck.
Maybe... Too bad they blew the layout. Can you guarantee that it's indeed R2R or is it R1.999R on one step and R2.0001R on the other?
And what on earth is going on in the lower right of that picture?
I would have mounted those all the way to the PCB to limit variation in lead length. Too bad they made the footprint for the resistors wider than the pin spacing on the resistors.
My point is that it takes more than good resistors to design a good R2R DAC. You don't necessarily need good absolute accuracy in the resistors, but you do need very, very tight matching between them. 1/(2^24) = 60 ppb. That's b for billion. 1/(2^20) is just shy of 1 ppm.
Tom
Hi i just share some examples, already published on the sites. I didnt made these PCBs.
One for sure I would not put these very expensive (but very good) types at first. AND certanly NOT in direct (mosfet devices) drive from Digital ICs...
I think that i will rather go to BJT or JFET segment prior to R2R network.
And I will change the format configuration. Talking of simple ICs discrete NOS dac with R2R ladder.
In all designs without processors, the format is Left justified, MSB inverted, based on the I2S L-R Word select Fs line.
There are no triggering the loaded data, the LE is just as it is Latch Enable.
I will made sort of time simultaneous type with booth channels digital data are present in the same time. And LE must be present for booth channels. Also in Left justified format, as in the I2S too, BCK is running all of the time. Even when no data at all. That inject more than double amount of digital noise and ground bouces...
This is not complicated operation just one ch datas should be delayed. and like before LE from I2S should be delayed by 1 BCK.
BCK has to be stopped when word is finished.
In all designs without processors, the format is Left justified, MSB inverted, based on the I2S L-R Word select Fs line.
There are no triggering the loaded data, the LE is just as it is Latch Enable.
I will made sort of time simultaneous type with booth channels digital data are present in the same time. And LE must be present for booth channels. Also in Left justified format, as in the I2S too, BCK is running all of the time. Even when no data at all. That inject more than double amount of digital noise and ground bouces...
This is not complicated operation just one ch datas should be delayed. and like before LE from I2S should be delayed by 1 BCK.
BCK has to be stopped when word is finished.
As you probably know already, the resistance of the switches matter. You want to ensure that you get the same low resistance in every switch through out the entire signal swing.
Last time I built a DAC I made a sub-ranging ladder DAC. I only needed 8-10 bits of precision so I made it a 12-bit DAC to have some margin. To my surprise I actually got 11+ bits of precision out of it. This was in a BiCMOS process where the absolute tolerance of the resistors was ± several percent and the matching between them (which is the important spec) not guaranteed. It makes me wonder if you could make a sub-ranging DAC with good performance for audio. The biggest drawback of the sub-ranging architecture is that the DNL can be pretty bad when the DAC switches from one range to the next.
Tom
Last time I built a DAC I made a sub-ranging ladder DAC. I only needed 8-10 bits of precision so I made it a 12-bit DAC to have some margin. To my surprise I actually got 11+ bits of precision out of it. This was in a BiCMOS process where the absolute tolerance of the resistors was ± several percent and the matching between them (which is the important spec) not guaranteed. It makes me wonder if you could make a sub-ranging DAC with good performance for audio. The biggest drawback of the sub-ranging architecture is that the DNL can be pretty bad when the DAC switches from one range to the next.
Tom
Hi Tom, Yes but we end up with what are inside of the digiral ICs end? If I am not wrong most of digital ICs has MOSFET complementary pair at the output? These devices has some different Rdson, so different Ri from the source is when "0" and "1"?
So that issue questioning the % tolerance of the Rs. That delta in internal source resistance is small maybe few ohms
Plus the Vcc is typically +5V or +3.3V or even smaller... In these terms maybe it would be better to go with the higher power supply and use SE switching method?
There is one old (gogle translated) part of the Sukema discreet DAC module, I could not find it in the net anymore. 🙁
.
Unfortunatley I am not in that level of expertise 🙁. But there are on the forum from my knowledge few members that are experts in that field.
But this switching concept may be of interest. I found in the article https://www.researchgate.net/figure...voltage-mode-with-extra-toggle_fig7_259580170