That is truly excellent news Phuong, has put an extra lift into my steps this morning 
P.S. 12.6V is a bit on the high side, the maximum supply voltage is set by the voltage limitation of the AD8017 whose absolute max is 13.2V. My latest DAC design doesn't have this limitation though - I've moved to using AD815.
P.S. 12.6V is a bit on the high side, the maximum supply voltage is set by the voltage limitation of the AD8017 whose absolute max is 13.2V. My latest DAC design doesn't have this limitation though - I've moved to using AD815.
Last edited:
Abraxalito , please edit the first post (you can) and add the last versions of schematics, to be updated as necessary . It is difficult to follow the whole view, as you have truncated the circuit into several functions and updating them individually . For example, you are using now op amp ad8017 as power supply in the filter stage . What happened to the power supply of the dac stage ? I am unable to sort out the correct version.
hi abraxalito,
according to post#3, where do i find the latest set of Gerbers for all 4 boards?
BR
Günni
I can edit the first post sure, when I try to add attachments I find that I'm unable to. I mention this inability at the end of the first post. Did you read that post?
Clearly there is some misunderstanding here. 'lingDAC' design has not been updated, it has remained unchanged. Rather there is a newer design 'PhiDAC' which is a single board and easier to DIY whereas lingDAC needs 3 or 4 boards.
Does my earlier paragraph help you sort things out more clearly?
I had a quick scroll back along the thread, posts #58 and #74 have gerbers for the 3 boards. It turns out I've yet to post up the gerbers for the fourth (power supply) board, this is because there was an error created by the gerber generation software which caused a short-circuit. If you're willing to correct that by hand then I'll be happy to share the gerbers for the PSU, subject to that caveat that it'll not work without modification.
Now add the posted numbers on your first post to direct a new or casual reader like me, access directly to to the last version that you edit the posted number at the next modification. You could have started a new thread for the phiDAC , now the discussion of both circuits will be mixed up.
Last edited:
Do you (anyone, not just @kokoriantz) feel mixing up the discussions of the two is a major drawback - even though there are commonalities between them? If so I'll create a new thread for the next design.
@kokoriantz - you've not included the lingDAC PSU schematic in post #346, was this intentional? (Its post #124).
@kokoriantz - you've not included the lingDAC PSU schematic in post #346, was this intentional? (Its post #124).
Last edited:
I have not seen it .
I have great esteem in designers like you as Mark Levinson who use ears rather than measuring instruments . This why your designs are valuable experiences .
Last edited:
Personally I think it's the thread-starter's call; it's likely about personal preference for other thread participants (i.e. different people will have different opinions).
I started the tda1387 front-end thread which morphed into being mostly about the RPI DAC HATs. In my mind they were close enough to not to warrant a second thread. I edited the first post to be an index of sorts, hoping it would help folks quickly navigate the thread. I think a decent index (which could link to other threads just as easily as self-linking) reduces the difference between multi-threads or one big thread.
Hi abraxalito,
ok i will give it a try. Maybe i also can modify the gerbers, that i can have a errorfree PSU design
Günni
When people around the net mention ''R2R'' they usually mean multibit** and/or NOS, which the TDA1387 will get you for pennies, at this point ''R2R'' seems like mostly a marketing term in consumer world but Im assuming since R2R still has a following on this site they must have some advantage, what is that advantage? or if its believed to have subjective advantages, are there theories around why it does?
**(multibit PCM, if thats the correct term , considering multibit DS exists) edit:correction, NOS already implies multibit PCM
Also @abraxalito in this post
you mention going from using stacked chips to single chip, was it just down to optimizing cost and/or circuit layout?
From reading that it sounded like a amplification stage was used with stacked chips, but I thought higher current with lower I/V would achieve same levels.
Im noticing aswell for the noDAC filters and NOS reconstruction filters, inductors are being used instead of resistors, what is the reason for that?
advantage of LC is less voltage loss, does that mean there is a disadvantage to low value R in RC aside from bigger caps needed?
**(multibit PCM, if thats the correct term , considering multibit DS exists) edit:correction, NOS already implies multibit PCM
Also @abraxalito in this post
you mention going from using stacked chips to single chip, was it just down to optimizing cost and/or circuit layout?
From reading that it sounded like a amplification stage was used with stacked chips, but I thought higher current with lower I/V would achieve same levels.
Im noticing aswell for the noDAC filters and NOS reconstruction filters, inductors are being used instead of resistors, what is the reason for that?
advantage of LC is less voltage loss, does that mean there is a disadvantage to low value R in RC aside from bigger caps needed?
Last edited:
@lasercrape - I'm wondering if it has something to do with the noise shaping that sigma delta DACs use & the possibility that this noise shaping has an unanticipated effect in the form of noise floor modulation?
I remember Mallinson, the chief design engineer at ESS, had an RMAF presentation (there's a video on Youtube) in which he talks about this noise floor modulation (which I believe he said was an issue in all SD DACs?) that they found in the ESS DAC when developing it & set about alleviating the problem - don't know if they fully succeeded based on some measurements that I've seen posted on this forum & others?
Is the lack of this noise shaping part of the advantage of multibit DACs?
I remember Mallinson, the chief design engineer at ESS, had an RMAF presentation (there's a video on Youtube) in which he talks about this noise floor modulation (which I believe he said was an issue in all SD DACs?) that they found in the ESS DAC when developing it & set about alleviating the problem - don't know if they fully succeeded based on some measurements that I've seen posted on this forum & others?
Is the lack of this noise shaping part of the advantage of multibit DACs?
Possibly, I was only considering the issue with interpolation, and that if did you want oversampling a computer has potential to do a better job than the DAC.
Based on what I heard with 9038Pro DS DACs, software oversampling to 705.6/768kHz did make a difference to the sound but it could not be considered much of an improvement in any aspect.
it could mean a few things, that internal oversampling was still taking place somehow, that the DAC oversampling is as good as PC or that noise shaping is biggest deciding factor of sound quality. I think its a mixture of the last 2
It was not very clear but In my previous I was actually asking specifically about expensive true R2R NOS chips vs the TDA1387 continuous calbration NOS , which appear to achieve the same result.
After a quick look at the datasheet compared to R2R chips, I see that R2R chips have better SNR and THD, and more bits, but that doesnt explicitly mean that these specs wouldnt be possible with continuous calibration, just that the best NOS DAC chips produced were R2R before DS tookover
Based on what I heard with 9038Pro DS DACs, software oversampling to 705.6/768kHz did make a difference to the sound but it could not be considered much of an improvement in any aspect.
it could mean a few things, that internal oversampling was still taking place somehow, that the DAC oversampling is as good as PC or that noise shaping is biggest deciding factor of sound quality. I think its a mixture of the last 2
It was not very clear but In my previous I was actually asking specifically about expensive true R2R NOS chips vs the TDA1387 continuous calbration NOS , which appear to achieve the same result.
After a quick look at the datasheet compared to R2R chips, I see that R2R chips have better SNR and THD, and more bits, but that doesnt explicitly mean that these specs wouldnt be possible with continuous calibration, just that the best NOS DAC chips produced were R2R before DS tookover
Last edited:
On the topic SNR, in many cases (e.g Sabre DACs) paralleling devices will lower noise floor and that is no expection with tda1387.
I guess the reasoning is that the SNR is still fine, and only looks 'issue' when you stack it against the ridiculous SNR levels of modern DACs, which are rarely/never achieved in real world implementations, and almost certainly wont be preserved through preceding amplification.
THD would be the bigger concern
I guess the reasoning is that the SNR is still fine, and only looks 'issue' when you stack it against the ridiculous SNR levels of modern DACs, which are rarely/never achieved in real world implementations, and almost certainly wont be preserved through preceding amplification.
THD would be the bigger concern
Last edited:
The advantages are purely in subjective terms as measurement-wise S-D DACs have a clear advantage (both in terms of ultimate SNR and also in THD). The hypotheses I have for why multibit is subjectively preferred are a) that multibit DACs generally don't come with an integrated digital filter so its possible to optimize this or run without ;b) the same applies to integrated opamps in the output stage; c) the DEM is done in hardware (in the cases where multibit DACs do in fact have DEM (TDA1541 and 1387)) whereas for S-D the DEM is entirely software. Software DEM is a distortion hiding process whereas hardware DEM is a distortion elimination process.
That was a couple of factors - stacking chips means that they share the same power supply and I figured this places more demands on the PSU integrity than does a single chip. Also stacked chips tends to suit passive I/V as resistors don't mind much what current they're asked to handle but active circuits tend to do better with lower currents (particularly opamps with their classAB output stages).
You're ultimately limited by the compliance range of the DAC chip but yeah stacking chips is a good way to get lower output impedance when passive I/V is in use.
Inductors allow much steeper filters to be built than can be done (passively) with R and C alone.
R2R is a DAC topology which uses resistors of two values (one twice the other) in a ladder configuration. The resistors are driven by switches and one of the weaknesses of this kind of DAC arises from the timing of the switches - if they don't all switch at the same instant there can be significant output glitches generated. A true R2R DAC I reckon can't have DEM ('Continuous Calibration' is Philips' implementation of DEM in CMOS DACs) so its dependent on good thermal and time tracking of its internal Rs.
If you build an RC filter with more than one pole you find you can't achieve a 'tight' corner at the cut-off frequency, the response droops rather in the passband. Another way of putting this is by saying that the Q of an RC-based passive low-pass filter can't get higher than 0.7.
If you build an RC filter with more than one pole you find you can't achieve a 'tight' corner at the cut-off frequency, the response droops rather in the passband. Another way of putting this is by saying that the Q of an RC-based passive low-pass filter can't get higher than 0.7.