It should be quite straightforward to implement a 24-bit parallel dac interface with my STM32 based USB-I2S board as there are plenty of available GPIOs. But I have no interest in this.
No current PC has a PCI bus, they are all PCIE, and have been for years.
So you're want to get a really old PC, running some ancient version of windows for this?
Have you actually done it, or is this all theoretical?
Switching is common to the one ppoint when the outputs from logic IC going to R2R passive ladder network.
Other points before are just digtal lines and there are no same effects as final output point say D to A
In the concept schematic that is marked as multiplied switches.
The best way to keep other digital noise small is to migrate to simoultaneous format
MSB anyway has to be inverted. That is actualy stopped BCK opperation. BCK is present only with DATA. And the same is for LE that has to be ENABLE all within tha DATA length.
So when no data word ALL lines are "0", and LE is common for booth channels.
There are no glitches allover the lines from BCK and other lines mixing in the PCB and PS.
That and simple putting R between ICs to improve signal integrity, with good decoupling of every IC for better ground bounces will significantly decrease the amount of HF noise.
(Glitches will stay because they are product of format and happening every time when MSB change state from 0 to 1. Almost random...)
.
And maybe use some separate IC with +ENABLE input for LE line with own dedicated power supply.
.
Or even tri state logic for the "output" stage prior to R2R ladder net. bits are foung to EN inputs of IC and LE foing to real input. I done it for 1bit DSD just to check the concept and it was much better sounding than standard DSD to analog.
MSB anyway has to be inverted. That is actualy stopped BCK opperation. BCK is present only with DATA. And the same is for LE that has to be ENABLE all within tha DATA length.
So when no data word ALL lines are "0", and LE is common for booth channels.
There are no glitches allover the lines from BCK and other lines mixing in the PCB and PS.
That and simple putting R between ICs to improve signal integrity, with good decoupling of every IC for better ground bounces will significantly decrease the amount of HF noise.
(Glitches will stay because they are product of format and happening every time when MSB change state from 0 to 1. Almost random...)
.
And maybe use some separate IC with +ENABLE input for LE line with own dedicated power supply.
.
Or even tri state logic for the "output" stage prior to R2R ladder net. bits are foung to EN inputs of IC and LE foing to real input. I done it for 1bit DSD just to check the concept and it was much better sounding than standard DSD to analog.
Stefano,
How important are the 12v supplies? (accuracy, noise etc)
I made a design error by using LM317/337 regulators fed from the Lipo battery but the drop out is 3v so when the voltage had fallen t0o low the regulation failed.
There was intermittent and horrible digital noise and distortion when the dac was receiving 10.5v per rail.
I have replaced the regulators for LDO versions of the 7812/7912 type but annoying, the Rohm 12v regulator is giving 12.6v (so much for the 0.1% accuracy).
I might replace them with discrete regulators if they are mission critical.
I am using separate AMS1117-5 regulators for the 5v supplies to the dac and Dir9001.
How important are the 12v supplies? (accuracy, noise etc)
I made a design error by using LM317/337 regulators fed from the Lipo battery but the drop out is 3v so when the voltage had fallen t0o low the regulation failed.
There was intermittent and horrible digital noise and distortion when the dac was receiving 10.5v per rail.
I have replaced the regulators for LDO versions of the 7812/7912 type but annoying, the Rohm 12v regulator is giving 12.6v (so much for the 0.1% accuracy).
I might replace them with discrete regulators if they are mission critical.
I am using separate AMS1117-5 regulators for the 5v supplies to the dac and Dir9001.
Hi, the analog circuit take the reverence voltage from +12V. Any shift from here will shift the mean of output signal. It must be inside +/- 0.5V. Concerning the noise coming from +/-12V you will find on output directly. I use standard 7812 and it's noise suppression from 50Hz is ok.
Have care to fix the unic GND point on SJ3 and SJ1. If not you can have loops Digital <-> Analog noise problems.
Regards
Stefano
Hello members,
attached a new version of R2R DAC doubling D/A section with opposite bitperfect output.. coming soon. 🙂
Advantages:
Regards, Stefano
attached a new version of R2R DAC doubling D/A section with opposite bitperfect output.. coming soon. 🙂
Advantages:
- Keep out DC and (possible) common noise from digital section without reference voltage -> you can power supply analog section with any voltage.
- Balanced Outputs (you can use only + signal for standard single output)
Regards, Stefano
Good news. I feel balanced is best as common mode noise cancels.
Would be good to have direct access to the output for external I/V.
I am going to compare my balanced TDA1541 dac with yours this weekend and I use transformer I/V - no opamps or tubes, so will try them if I buy your new dac.
Thanks for the update.
Not had any more issues with noise since using low dropout 78/7912 regs but I am not sure of the cause.
I might swap the 072 op amps for TLE2072 versions - lower noise, higher slew rate, however I am thinking that a higher slew rate is not an advantage.
I'm interested in this as well. I haven't heard many discrete R2R dacs that I preferred over a good dual balanced 1541a, but since you're taking the time and effort to make this one better than the other diy implementations out there, I think yours could be significantly better and win me over. The hitch is always going to be the resistor accuracy and board grounding / PS layout and scheme. I'll be watching and waiting optimistically...
Stefano is producing a balanced version - at the prototype stage currently and if it's better than the existing one, it will be truly impressive.
I see you now have the balanced dac for sale. Good luck with it.
I have been enjoying the original one for several weeks now.
PS might I suggest that you use Google Translate to add the English version of the description to your ebay listings?
I have not quite finished rebuilding my 1541 dac.
I have been sidetracked this week with a pair of 35year old Naim SBL speakers which I hope to audition with the R2R dac from tomorrow.
Still not completed rebuilding my dual Tda1541 dac so no comparison yet.
I see that chiurutu is now selling his dual R2R dac on ebay so this might be worth a listen.
https://www.ebay.co.uk/itm/325410066392
I see that chiurutu is now selling his dual R2R dac on ebay so this might be worth a listen.
https://www.ebay.co.uk/itm/325410066392
Now rebuilt my dual balanced TDA1541 and initial comparisons are: (both powered from a +- 20v Lion battery pack) Source is optical spdif from a LG smart tv, 24bit 48khz)
The 1541 gives a smoother sound quality but no less detailed.
The ladder dac has a slightly more strident and edgy sound with strings and brass giving them more 'bite' and therefore is not as polite as the 1541s. However it's not harsh nor tiring.
Neither dacs produce any glare or grain and no unwanted aliasing noises.
(rest of my system: Yamaha C-45 preamp with remote control motorised pot added - a great 1980s NE5532 preamp using Nichicon Muse and Black Gate caps), Quad 306 with capacitor upgrades and sensitivity reduced to 1v RMS, Naim SBL speakers, also from the mid eighties but with new Scanspeak tweeters.)
More listening needed and also to find which is better for CD audio.
The 1541 gives a smoother sound quality but no less detailed.
The ladder dac has a slightly more strident and edgy sound with strings and brass giving them more 'bite' and therefore is not as polite as the 1541s. However it's not harsh nor tiring.
Neither dacs produce any glare or grain and no unwanted aliasing noises.
(rest of my system: Yamaha C-45 preamp with remote control motorised pot added - a great 1980s NE5532 preamp using Nichicon Muse and Black Gate caps), Quad 306 with capacitor upgrades and sensitivity reduced to 1v RMS, Naim SBL speakers, also from the mid eighties but with new Scanspeak tweeters.)
More listening needed and also to find which is better for CD audio.
Last edited:
Batteryman, pls elaborate on your dual balance 1541. Is it a finished kit just like Chiurutu's?
Thanks
Thanks
No, not like Stefan's. His is about 6 years too late for me or I wouldn't have designed mine! (not commercially available and no pcbs left)
I use IanCanada's Fifo II, Dual clock board and I2S to PCM to drive the 1541s in simultaneous mode with clocks stopped after data latched. Not sure if Stefan's does this.
I am thinking of buying Stefan's because I like the idea of using discrete logic and this works very well on his R2R ladder dac.
Stefan's is not bit perfect like iancanada's because he just inverts the data which gives a 1lsb error.
I tried doing this with an Ebay dual 1541 dac rewired into balanced mode (one chip for l/r and one for -l/r) and couldn't hear any side effects from this method.
I use IanCanada's Fifo II, Dual clock board and I2S to PCM to drive the 1541s in simultaneous mode with clocks stopped after data latched. Not sure if Stefan's does this.
I am thinking of buying Stefan's because I like the idea of using discrete logic and this works very well on his R2R ladder dac.
Stefan's is not bit perfect like iancanada's because he just inverts the data which gives a 1lsb error.
I tried doing this with an Ebay dual 1541 dac rewired into balanced mode (one chip for l/r and one for -l/r) and couldn't hear any side effects from this method.