If i go ceramic, i go c0g if there is a value needed, to me it's as close to perfect cap as possible. Class 2 ain't good for r2r dac chip decouple. Rogic doesn't use acrylic. I like acrylic ones as well, and have ordered 0.1uf to give it a go in my tda dac. We'll see. I'll be making both dem modded and without.
Has anyone tried adding 0.1uf acrylic/film/c0g to these ukl 100uf? Measured results?
Has anyone tried adding 0.1uf acrylic/film/c0g to these ukl 100uf? Measured results?
Yes class 2 has nothing to do here. Though class I gives not as good results for the 14 DEM than plastic based smd caps.
Pedja was using Pan PPS but they were swapped this year by acrylic at Pan catalog iirc. Objectivly sigthly better. Pedja as Loesh as John in that discovering order were not happy with ceramic class I for this aera. Nothing bad for the crystal decoupling when pF are needed. But for the main decoupling clock power supply when neededoften acrylic gives better results...
Tda1541 is not really a r2r dac though.
See back in this thread for ukl and acrylic there.
Pedja was using Pan PPS but they were swapped this year by acrylic at Pan catalog iirc. Objectivly sigthly better. Pedja as Loesh as John in that discovering order were not happy with ceramic class I for this aera. Nothing bad for the crystal decoupling when pF are needed. But for the main decoupling clock power supply when neededoften acrylic gives better results...
Tda1541 is not really a r2r dac though.
See back in this thread for ukl and acrylic there.
Last edited:
There is no need of extra cap for the 50 Hz DEM , and the 100 uf UKL works great , I have made some FFT test today , and there are no noise in the DEM region at all , john was right for it's choice
.
I finally got 0.1 mv DC at both output , here is the best distortion figure I can get :
I tried various frequency and various level , result are all ok , got to find a pair of trafo to get any futher
I found that the way ( means where ) grounds between board are connected play a significant role in finding best numbers
.
I tried various frequency and various level , result are all ok , got to find a pair of trafo to get any futher
I found that the way ( means where ) grounds between board are connected play a significant role in finding best numbers
.
Last edited:
refweing to post #7279
at
https://www.diyaudio.com/community/...ltimate-nos-dac-using-tda1541a.79452/page-364
It would be much better to have separate Current injections of 2mA at each output. With JFET as CCS. More clear picture to set, measure and evaluate?
Then tie together.
Also probably it would be much better to "blend" 2 outputs of same chanell, mono mode DAC IC, to separate primary coils at the signal transformer.
So if You still does not require transformers, please consider types with 2 x prmary : 2 x secondary.
That could be used as multipurpose ways covering many topologies? Among them is one from the picture.
It is pretty clear but I can sketch a schematic?
.
If You have a time please try to lower Riv value to 25ohms and measure distortion in same surroundings.
Probably will be significantly lower and with less noise at the HF?
(Read the datasheet about the max voltage offset...)
Cheers
.
at
https://www.diyaudio.com/community/...ltimate-nos-dac-using-tda1541a.79452/page-364
It would be much better to have separate Current injections of 2mA at each output. With JFET as CCS. More clear picture to set, measure and evaluate?
Then tie together.
Also probably it would be much better to "blend" 2 outputs of same chanell, mono mode DAC IC, to separate primary coils at the signal transformer.
So if You still does not require transformers, please consider types with 2 x prmary : 2 x secondary.
That could be used as multipurpose ways covering many topologies? Among them is one from the picture.
It is pretty clear but I can sketch a schematic?
.
If You have a time please try to lower Riv value to 25ohms and measure distortion in same surroundings.
Probably will be significantly lower and with less noise at the HF?
(Read the datasheet about the max voltage offset...)
Cheers
.
Thanks for your input Zoran , fact is that I do require transformer to stepup the output because I want to stay passive all the way out , at the moment distortion figure are pretty good for an old pair of 16 bit TDA1541A , I am not looking for much better , I dont even known what better is or mean when the distortion is that low , dont known if these chips are able to get much lower , I dont have exemple of that
as said before I want to stick at the John's idea as much as possible
it's an setup with no digital filter , so HF noise will be lowered/canceled at the analog side
I will try 25 ohms instead of 100 ohms to see what's happens , I'll let you know
.
as said before I want to stick at the John's idea as much as possible
it's an setup with no digital filter , so HF noise will be lowered/canceled at the analog side
I will try 25 ohms instead of 100 ohms to see what's happens , I'll let you know
.
Last edited:
I received some parts today
first of all I change the AZ1117CH 3.3 reg of the CM6631A board for an AP7361C , and I gain a little in the noise floor
then I put some 33 ohm resistor in parralele of the 100 ohms , result is an lower output as expected from 70 mv rms, to 18 mv rms , it's about 50 mv peak to peak , not sure I can get any lower than that , here is the FFT :
for those who like to see the FFT all conected but without signals here it is :
wee can still see the fs at 48 khz
.
first of all I change the AZ1117CH 3.3 reg of the CM6631A board for an AP7361C , and I gain a little in the noise floor
then I put some 33 ohm resistor in parralele of the 100 ohms , result is an lower output as expected from 70 mv rms, to 18 mv rms , it's about 50 mv peak to peak , not sure I can get any lower than that , here is the FFT :
for those who like to see the FFT all conected but without signals here it is :
wee can still see the fs at 48 khz
.
Last edited:
FFT with the QA401 is a great and precise measurement but it is bandwidth limited , so Today I decided to use the oscilloscope at the outputs , and result is not that good , plenty of HF stuff from fs to over 240 Khz , I rearange the grounds and I got better results but still noisy , here is an 1Khz sine on both output
I dont know if it's the normal HF figure as it's not filtered , probe ( x10 ) pickup is directly at the I/V résistors , any thought welcome
I have to say that , bcko , leo , and data are clean at the DAC imput
.
I dont know if it's the normal HF figure as it's not filtered , probe ( x10 ) pickup is directly at the I/V résistors , any thought welcome
I have to say that , bcko , leo , and data are clean at the DAC imput
.
Last edited:
Speaking about grounds , it seems that decoupling the +5 to DGND is better ( noise wise ) than to AGND , -5 v and -15 v to AGND
an other good advice gave here : https://www.mvaudiolabs.com/digital/tda1541a-reference-dac/ , is the decoupling of the -5 to -15 v , less noise too
still learning
.
an other good advice gave here : https://www.mvaudiolabs.com/digital/tda1541a-reference-dac/ , is the decoupling of the -5 to -15 v , less noise too
still learning
.
Upgrade of the day , I have seen along the tread a film cap to ground at the exit of the biais trimmer , value was 1uf , I tried 0.47 and 1 to seen whats happen
huge gain in distorsion with 1uf , but lost in voltage output from 5Khz and beyond , less but still gain in distortion with 0.47 uf and less lost in voltage
here is the FFT , yellow graph with 1.0 uf and red graph with 0.47 uf
interesting is at the oscilloscope the sine wave in way less scaled with this cap , I guess there will be some compromise here :
.
huge gain in distorsion with 1uf , but lost in voltage output from 5Khz and beyond , less but still gain in distortion with 0.47 uf and less lost in voltage
here is the FFT , yellow graph with 1.0 uf and red graph with 0.47 uf
interesting is at the oscilloscope the sine wave in way less scaled with this cap , I guess there will be some compromise here :
.
do you mean those two grounds do not share a same reference ? Because DGND do be connected to AGND, where is the subtility about the routing and layout though... including the 14 DEM caps. Those 3 volatges are related together and have nothing to see with a said digital or analog manicheist ground... although there is a subtility yet about on how should be treated each other 5 V, but it is more about decoupling of the power supply (if pcb layout made rigth first !)Speaking about grounds , it seems that decoupling the +5 to DGND is better ( noise wise ) than to AGND , -5 v and -15 v to AGND
an other good advice gave here : https://www.mvaudiolabs.com/digital/tda1541a-reference-dac/ , is the decoupling of the -5 to -15 v , less noise too
still learning
.
Oups I said I was not imputing anymore... so my mistake and my last !