** ALL SOLD . Thank you guys ** A few of local audiophiles requested for 16 bit NOS discrete R2R DAC for Raspberry pi , I designed and built 8pcs for them, and I have 2 pcs left.
Spec:
16 Bit NOS Discrete R2R DAC
support " R-pi DAC" as driver for Volumio,Pcp, Moode audio etc.
Hand picked and matched thin film resistors , matched down to 0.005 %
discrete output stage , without negative feedback
tested up to 192K sampling rate. Works up to 24 bit 192K , ( truncated )
not support software volume control
International shipping
Please PM me for more detail .
Spec:
16 Bit NOS Discrete R2R DAC
support " R-pi DAC" as driver for Volumio,Pcp, Moode audio etc.
Hand picked and matched thin film resistors , matched down to 0.005 %
discrete output stage , without negative feedback
tested up to 192K sampling rate. Works up to 24 bit 192K , ( truncated )
not support software volume control
International shipping
Please PM me for more detail .
Last edited:
Hi all,
I am currently building a new batch of 24 bit version ( version V2.4) for group purchase by audiophile members out there. I will build a few extra as usual, and will post it here once done.
I start to design and built R2R DAC since 2018 , and I have more than 20 versions built. I try hard to make it cheaper but with hand-match thin film resistors down to 0.005%, and using good parts seems not really possible to cut the cost too much.
As for shipping cost, I'm using FedEx Priority International, estimate around USD35 to USA and some Europe countries. Cheaper shipping option (EMS ) at around USD22.90 but not as good as FedEx service.
Thanks
Ew
I am currently building a new batch of 24 bit version ( version V2.4) for group purchase by audiophile members out there. I will build a few extra as usual, and will post it here once done.
I start to design and built R2R DAC since 2018 , and I have more than 20 versions built. I try hard to make it cheaper but with hand-match thin film resistors down to 0.005%, and using good parts seems not really possible to cut the cost too much.
As for shipping cost, I'm using FedEx Priority International, estimate around USD35 to USA and some Europe countries. Cheaper shipping option (EMS ) at around USD22.90 but not as good as FedEx service.
Thanks
Ew
@skickaskinka nope, south germany, far away...
How much would be the DAC himself ?
How much would be the DAC himself ?
A little while ago I worked out that you can build an R2R DAC without bothering to match the resistors (I expect this is well known but it was new to me when I was thinking about it).
You can just select the values such that the sum of bits 0-N is guaranteed to be slightly larger than the value of N for any possible values of resistors within the tolerance band (1% resistors for example).
Then you add a couple of extra bits. So, you might perhaps implement 32 hardware bits and end up with 30 or so effective bits.
After that, for each device you construct, you have to measure the values of the bits you happen to have ended up with given the actual resistors used for that device. Once you know the actual voltage contributed by each bit, you need to preprocess the digital audio to select the bits required to get the correct output.
This can be done with a greedy algorithm that just recursively selects the highest value bit that is less than or equal to the remaining value required. This is a very efficient algorithm which doesn’t require introduction of any audio delay so can be done in real time.
It would also be possible to measure the resulting output and make ongoing adjustments for temperature changes to keep things accurate.
You don’t actually need to know the absolute value of the bits, you just need to arrange them on a linear scale which I think can be done offline using a comparator between the left and right channels. So, whilst an ADC might make things simpler, it could be done more cheaply without. The trick here is that the bits 0-N are guaranteed to be greater than bit N+1 so it’s possible to find a combination of the lower bits that’s equal to the upper bit within the error bounds of the extra least significant bits. I think if you do this for each bit it is possible to solve for the relative values of the bits (within the error bounds).
If you double up the R2R hardware then it would be possible to continually calibrate offline using the comparator and regularly rotate calibrated R2R networks into use to keep the audio side calibrated across temperature changes (which would still probably be cheaper than using a good ADC to do online calibration).
The limits on system performance from this approach are the sensitivity of the comparator, the time required for making a comparison, the number of comparisons required to calibrate and the drift rate of the resistor values.
There’s no need to stop at 32 bits. You could do more for extra dynamic range for a volume control.
You could go farther and use the digital output to switch higher voltage on the R2R network to eliminate the requirement for a gain stage and integrate it directly with a power stage to get a digital amp.
Anyway, I just thought I’d throw this up on your thread in case you were interested in going beyond 16 bits when it eventually becomes impossible to proceed by matching resistors alone.
You can just select the values such that the sum of bits 0-N is guaranteed to be slightly larger than the value of N for any possible values of resistors within the tolerance band (1% resistors for example).
Then you add a couple of extra bits. So, you might perhaps implement 32 hardware bits and end up with 30 or so effective bits.
After that, for each device you construct, you have to measure the values of the bits you happen to have ended up with given the actual resistors used for that device. Once you know the actual voltage contributed by each bit, you need to preprocess the digital audio to select the bits required to get the correct output.
This can be done with a greedy algorithm that just recursively selects the highest value bit that is less than or equal to the remaining value required. This is a very efficient algorithm which doesn’t require introduction of any audio delay so can be done in real time.
It would also be possible to measure the resulting output and make ongoing adjustments for temperature changes to keep things accurate.
You don’t actually need to know the absolute value of the bits, you just need to arrange them on a linear scale which I think can be done offline using a comparator between the left and right channels. So, whilst an ADC might make things simpler, it could be done more cheaply without. The trick here is that the bits 0-N are guaranteed to be greater than bit N+1 so it’s possible to find a combination of the lower bits that’s equal to the upper bit within the error bounds of the extra least significant bits. I think if you do this for each bit it is possible to solve for the relative values of the bits (within the error bounds).
If you double up the R2R hardware then it would be possible to continually calibrate offline using the comparator and regularly rotate calibrated R2R networks into use to keep the audio side calibrated across temperature changes (which would still probably be cheaper than using a good ADC to do online calibration).
The limits on system performance from this approach are the sensitivity of the comparator, the time required for making a comparison, the number of comparisons required to calibrate and the drift rate of the resistor values.
There’s no need to stop at 32 bits. You could do more for extra dynamic range for a volume control.
You could go farther and use the digital output to switch higher voltage on the R2R network to eliminate the requirement for a gain stage and integrate it directly with a power stage to get a digital amp.
Anyway, I just thought I’d throw this up on your thread in case you were interested in going beyond 16 bits when it eventually becomes impossible to proceed by matching resistors alone.
That would be great, but I live far from Linköping unfortunately. Thanks for the offer though 🙂