[/ATTACH]
I'm not sure what your setup is, but have you tried an output stage that converts differential to single end signal, so that any noise coming from «analog» side of supply circuit gets cancelled? My topping d10s is arranged that way, so in theory sabre analog noise is greatly reduced... similar to attached file.
I'm not sure what your setup is, but have you tried an output stage that converts differential to single end signal, so that any noise coming from «analog» side of supply circuit gets cancelled? My topping d10s is arranged that way, so in theory sabre analog noise is greatly reduced... similar to attached file.
Attachments
Last edited:
Yes. The version of that circuit we recommended in this thread is attached to post #3003.
Agree that common mode noise is attenuated by such an arrangement. How much it gets attenuated verses frequency might be another question. Precise gain matching throughout the output stage can help a lot. However consider that use of .01% resistors may not be practical given resistance values needed for setting filter characteristics of the output stage. Thus in most implementations common mode rejection may not be fully optimized.
The above having been said, experiments with AVCC power supplies show that audible feedthrough effects exist. They don't necessarily matter much for AP metrics, since they should consist of noise. So long as the noise is low enough level then an FFT should look pretty good. However, AVCC voltage including any noise is multiplied by the dac chip internal output processing to produce the analog output of the chip. That means any AVCC noise becomes intermodulated with the audio signal. That leads to a condition sometimes referred to as 'noise floor modulation,' so named due to its appearance in FFT measurements. According to none other than ESS, noise floor modulation in dacs is something that the human ear is exquisitely sensitive to. They say some people can hear it an extremely low levels.
I should also probably say that for my own dac I have done much more than improve the AVCC regulation. But, I would say working on that subsystem could be a good starting point for hobbyists who are interested in doing some of their own experimentation with dac design.
yes i tried simple opamp and salas reg for avcc, but it was hard for me to pickup a difference with my hd650 headphones on my previous d10s which is killed with a bad unstable supply i think.. But i know how sensitive to PSU performace dac circuits are. Default avcc reg for D10s is an LD33 which is i believe a 1117 series reg, which is not bad from experience...it is noisy a little but sounds very natural to me...very good for digital portion which is less sentive to noise, in my experience. Now with my new d10s i will try tps7a series for AVCC, some times, looks like a good «all around» psu with good reaction to step loading, very decent noise performance, low output impedance.
Regarding resistor matching, not sure why you say higher value resistors makes things difficult...
i any case, thanks for sharing your experience!
Regarding resistor matching, not sure why you say higher value resistors makes things difficult...
i any case, thanks for sharing your experience!
We have recommended trying opamp buffers for AVCC, a schematic is attached to post #3003. Some commercial dacs use that method, and ESS recommended it before they started making dac chips that required too much AVCC current. Not sure I would recommend a Salas shunt for AVCC. Also, it is important that the AVCC regulator outputs be right next to the dac chip. Any connecting wires should be a short as possible, maybe less than a couple of cm if possible. ESS emphasized the need for that in their old output stage and AVCC paper attached below.
One other thing to be aware of is that dac chips radiate some RF locally. For that reason the AVCC opamp buffer I used was mounted on the bottom side to the dac PCB so that the ground plane could help provide some shielding for it.
One other thing to be aware of is that dac chips radiate some RF locally. For that reason the AVCC opamp buffer I used was mounted on the bottom side to the dac PCB so that the ground plane could help provide some shielding for it.
Last edited:
yes i tried that opamp buffer...i sounded decent, but nothing spectacular....i can't remember exactly my impressions, but nothing mind blowing...i prefered salas shunt v1.0 ,which is UBER low noise, but lacked dynamics... but i think it killed the chip, due to instability so i will pass on it for that specific use....anyway lots of others regs to try! i'm not fan of opamp in general, so looking to change the output stage/filter for a discrete one, maybe salas DCG3.
Other than that i would recommend salas shunts and amb sigma PSU's for voltages 5v and above...my absolute favorites!
Other than that i would recommend salas shunts and amb sigma PSU's for voltages 5v and above...my absolute favorites!
Last edited:
On the subject of AVCC psu and opamp buffered, I currently have in my system a PFM flea for AVCC. Before that I used the single rail portion of one of these Chinese psus. I actually cut the whole pcb down to leave me the small single 3v3 section. The dual section could be used for an opamp IV obviously. Maybe not the best solution but very compact. I did use the dual rail on the IV and it played very nicely.
￡19.20 31%OFF | NE5534 HIFI DAC Power Supply Class A dual DC ± 12V ± 15V power supply multi-outputs T0612
NE5534 HIFI DAC Power Supply Class A dual DC +- 12V +- 15V power supply multi outputs T0612|Amplifier| - AliExpress
The Chinese psu actually has an OP177 and in the Flea I have an AD797
￡19.20 31%OFF | NE5534 HIFI DAC Power Supply Class A dual DC ± 12V ± 15V power supply multi-outputs T0612
NE5534 HIFI DAC Power Supply Class A dual DC +- 12V +- 15V power supply multi outputs T0612|Amplifier| - AliExpress
The Chinese psu actually has an OP177 and in the Flea I have an AD797
Last edited:
@pidesd,
IMHO a discrete regulator such as a shunt has more potential than an opamp buffer, but there are more variables to optimize the more complex a regulator is. Regarding series regulators verses shunts, the latter can sound better but they tend to be harder to optimize for best sound. Regarding the Salas version of a shunt in particular, I have been told by a professional audio designer friend that it would not be his choice for a shunt design. The symptom of flat dynamics you alluded to is IME due to mis-optimization of the regulator. If you are up to the task of redesigning a Salas as needed then maybe the result could be very satisfying.
IMHO a discrete regulator such as a shunt has more potential than an opamp buffer, but there are more variables to optimize the more complex a regulator is. Regarding series regulators verses shunts, the latter can sound better but they tend to be harder to optimize for best sound. Regarding the Salas version of a shunt in particular, I have been told by a professional audio designer friend that it would not be his choice for a shunt design. The symptom of flat dynamics you alluded to is IME due to mis-optimization of the regulator. If you are up to the task of redesigning a Salas as needed then maybe the result could be very satisfying.
i am not up to design myself yet, plus i don't have an scope, so i stick to proven ones. the one which i think cooked the dac avcc had a bjt instead of mosfet for shunt duty...maybe it was the cause, but for now i'll stick to «safer» regs. Nevertheless, Salas shunts, especially the original simple version, is unbeatable IMO, if you are willing to put some bias current to it, like 200mA on the shunt to give a decent dynamic sound (according to my standards). And i use irfp240 for the CCS and shunt part....good for around 5V output and above. Never had any problem with this one. One has to «see» it for himself to believe...
If i want lower consumption, i always go to AMB Sigma (slightly modified), i even use it for my power amp (low power)! This reg sounds quite «natural» and has STIFF regulation.
If i want lower consumption, i always go to AMB Sigma (slightly modified), i even use it for my power amp (low power)! This reg sounds quite «natural» and has STIFF regulation.
Sorry to change subject -- Can anyone suggest a suitable replacement for the NJM7915FA regulator that will match with the 7815FA -------- it seems the 7915FA is obsolete & I'm not sure if I can properly spec a suitable replacement -- and if that means the 7815 should be changed as well.
thanks - Bart
thanks - Bart
on first glance that looks no different from any other 79XX 3 terminal regulator. They are available in both + and - complements from many sources - ON, ST, TI, etc. Generic and cheap -- and not terribly great. You can also use the slightly newer nad more universal LM317/337 pairs that are adjustable. Various tricks lower noise.
The NJM versions of 78xx and 79xx series devices were reputed by some people to sound better than other brands (at least for circuits where regulators may have audible effects). For a replacement probably some of the other 79xx parts are reasonably close. https://www.mouser.com/c/semiconduc...controllers/linear-voltage-regulators/?q=7915
Also, some people feel 78/79 series parts sounded better than LM317/LM337. Of course that doesn't matter if we can't get the parts we prefered anymore.
In addition, if a good supply is needed for analog audio circuitry, IMHO the discrete power supply shown in the schematic attached to one of the posts in the following thread might be of interest: https://www.diyaudio.com/community/threads/pass-hpa-1-what-do-we-know.300060/
Also, some people feel 78/79 series parts sounded better than LM317/LM337. Of course that doesn't matter if we can't get the parts we prefered anymore.
In addition, if a good supply is needed for analog audio circuitry, IMHO the discrete power supply shown in the schematic attached to one of the posts in the following thread might be of interest: https://www.diyaudio.com/community/threads/pass-hpa-1-what-do-we-know.300060/
i would be surprised if meaninful differences existed between what are generic part numbers from different makes. They have to be nearly identical to meet 2nd source requirements. As noted, if you want a lot better use a totally different part, a discrete reg, or as I noted, think abotu how to get the most out of those 78-79 xx parts. Some LDOs are something like 3X better overall. (they claim more). Frankly i often wind up qwith both
I'm sure his "some people" are "trained" listeners who are in the business and therefore their words should be taken as gospel, IHHO.
let's not be snarky. I encourage critical evaluation. Whether i believe many results is another matter....
That said i;ve caught more things with my ears than with my lab full of equiupment. Once i hear it I often know to look harder or where to look.
But i strongly think that worrying about whether one mediocre regulator is better than an almost identical mediocre regulator is pointless. Change the design to use a good one or a cleverer design. Now, if the signal at this point in the circuit is sufficiently large, it may be ok either way. Also, if its class-A linear analog circuitry, you can consider non -regulator solutions (filters, post regulator filters) - assuming you have a truly constant current draw.
That said i;ve caught more things with my ears than with my lab full of equiupment. Once i hear it I often know to look harder or where to look.
But i strongly think that worrying about whether one mediocre regulator is better than an almost identical mediocre regulator is pointless. Change the design to use a good one or a cleverer design. Now, if the signal at this point in the circuit is sufficiently large, it may be ok either way. Also, if its class-A linear analog circuitry, you can consider non -regulator solutions (filters, post regulator filters) - assuming you have a truly constant current draw.
Good for you. Should I be concerned or otherwise interested in your ears capabilities?
You have to keep in mind that Heisenberg's uncertainty principle applies to audio as well. The more we know about the physical properties of the circuit (schematic, layout, measurements ...) the less we know about how good it sounds. And vice versa. So the best sounding circuits are always the ones with no schematics, layouts or measurements, only subjective anecdotes.
However in contrast to quantum world in audio the uncertainty principle has one exception: close-in phase noise. So the lower the close-in phase noise the better it sounds regardless of anything else.
However in contrast to quantum world in audio the uncertainty principle has one exception: close-in phase noise. So the lower the close-in phase noise the better it sounds regardless of anything else.