I think I asked this question here, but I didn't get an useful answer, or was disregarded as not that important.
On his latest regulator version, where he made a few changes, including the pre-regulator, the type he suggested (recommended?) for the 120uF caps were HQ and FC types, which if I'm not wrong are not made anymore.
https://refsnregs.waltjung.org/Improved_PN_Regs.pdf
The BOM parts list on the Linear Audio pdf file does not specify any, just the 120uF value. But am I wrong to guess some specific model, or at least having specific specs, are to be preferred, particularly ESR and others that might apply or have proven better in stability or maybe audio quality?
I'm not expecting an "official" recommendation, just some suggestions.
On his latest regulator version, where he made a few changes, including the pre-regulator, the type he suggested (recommended?) for the 120uF caps were HQ and FC types, which if I'm not wrong are not made anymore.
https://refsnregs.waltjung.org/Improved_PN_Regs.pdf
The BOM parts list on the Linear Audio pdf file does not specify any, just the 120uF value. But am I wrong to guess some specific model, or at least having specific specs, are to be preferred, particularly ESR and others that might apply or have proven better in stability or maybe audio quality?
I'm not expecting an "official" recommendation, just some suggestions.
Or, when preparing the PCB layout, arrange four 0.25W resistors in parallel, and connect this parallel assemblage in series between the output capacitor and ground. Now you can perform experiments with ESR to find out exactly how much or how little of it, your board needs.
At the end of this procedure you will know with the certainty of measured data, data which you actually trust, just exactly how much ESR gives the test results that make you happy. And also: you are happy! A delightful outcome.
- Install the lowest ESR capacitor you can find
- Stuff and solder only one of the four parallel resistors. Stuff and solder 400 milliohms {closest standard value: 390} of ESR. Test the regulator. Are you happy?
- Stuff and solder a second parallel resistor, this one 400 milliohms. Now the parallel combination is 200 milliohms of ESR. Test the regulator. Are you happy?
- Stuff and solder a third parallel resistor, this one 200 milliohms. Now the parallel combination is 100 milliohms of ESR. Test the regulator. Are you happy?
- Stuff and solder a fourth (the final) parallel resistor, this one 100 milliohms. Now the parallel combination is 50 milliohms of ESR. Test the regulator. Are you happy?
I've tried super regs with Allo DAC's amongst others and always found them to be better than any currently available LDO's, even the LT3042/5's. I was using the Andrew Weekes version which has the added LM317 tracking pre reg which I keep. I've just got the Khadas board today and am impressed even sharing the RPi4 supply via USB. Over this weekend I'll be trying an ALWSR set to just below 5.5 volts as the 3.3volt regs on the Khadas have a max rating of 5.5v.
John
In the absence of a vector network analyzer, or a scope with FFT, a rough way you can judge happiness by the current drawn by the regulator -- when they oscillate they tend to draw a lot more current than would be expected from the overhead and load.
So we're all on the same page, load the regulator for 25mA @13.7V = ~ 560 Ohms and go with Marks procedure.
Then when you're done, place w cheap, noisy 5V switching wall-wart near the regulator and see if it is excited into oscillation.
Interesting, I just bought a Singxer SU-1 (to send USB via AES to my DAX Discrete) and a mod is to swap in a LT3045 after the linear PSU in the SU-1.
How about instead of a 317, use a LT3045 as the pre-reg to a Jung-Didden?
Interesting. So it's local RF as much as anything then that sends this super reg into apoplexy?
Nice. I read that Walt used the 120uF's as he happened to have some in a drawer. I happen to have some Nichicon KZ 220uF in a drawer with no use for them, now I have a use, in my next JD super reg, probably in the Singxer.
And I have a 120uF low cost for the output.Have you fellas had a look at, reverse engineered, measured, one of those Sparkos Labs discrete regs, crafted from tiny tiny parts onto a three pin PCB? Sparkos Labs, Inc. audio power supply discrete voltage regulators
The Singxer I bought has a 3.3 Volt one. Should I swap it out for a JDSR? Please remind me, will the JDSR supply 3.3V.
Thanks.
Last edited:
@ianas
"I read that Walt used the 120uF's as he happened to have some in a drawer."
Hmmm... in the TAA articles the narrative was that the circuit and the parts had been exhaustively fine-tuned and builders shouldn't deviate from the parts list provided.
Not sure which way to jump at this point.
"I read that Walt used the 120uF's as he happened to have some in a drawer."
Hmmm... in the TAA articles the narrative was that the circuit and the parts had been exhaustively fine-tuned and builders shouldn't deviate from the parts list provided.
Not sure which way to jump at this point.
The tests have been done with 120uF, so we don't know how it would be with 47uF or 200uF.
It's not that I try to evade questions, but there is an infinite number of permutations between values of parts and normally you base your design on calculations and experience and the majority of the infinite permutations are not explored.
We've only a non-infinite life ;-)
That said, I have used 100uF if that was what I happened to have, with no issues.
Jan
It's not that I try to evade questions, but there is an infinite number of permutations between values of parts and normally you base your design on calculations and experience and the majority of the infinite permutations are not explored.
We've only a non-infinite life ;-)
That said, I have used 100uF if that was what I happened to have, with no issues.
Jan
Last edited:
With the AD797 error amplifier -- it doesn't have this problem with AD825 or LME49710.
Way back in this thread I mentioned that I could get it to oscillate at 770kHz -- the frequency of the 50kW clear-channel station in NYC.
Pity the LME49710 is obsolete and unavailable.
Any other you tried with good results and no oscillation?
@jan
PS to your edit:
I hear you and understand. I wasn't implying that it was your responsibility to test every permutation (except, perhaps, when you report that you kinda did). That would be nuts.
But if you reached into the junk drawer and used the first unit that came to hand and didn't see any reason to change it -- even though it was an odd-ish value -- then that would have been helpful info to share. (I apologize if you have; the thread is... epic)
And if you (or anyone) later tried a different junk drawer part -- and it worked too -- that would have been helpful to know to enrich the data set as actual experience deepens -- to add the next data point. Incrementalism.
Make sense? Reasonable?
Incidentally, after years of dithering I actually have boards on the way (with solder in part as a way to support diyaudio) so my comments/questions are a bit less academic that they've been in the past. Thanks for your patience with me over the years.
PS to your edit:
I hear you and understand. I wasn't implying that it was your responsibility to test every permutation (except, perhaps, when you report that you kinda did). That would be nuts.
But if you reached into the junk drawer and used the first unit that came to hand and didn't see any reason to change it -- even though it was an odd-ish value -- then that would have been helpful info to share. (I apologize if you have; the thread is... epic)
And if you (or anyone) later tried a different junk drawer part -- and it worked too -- that would have been helpful to know to enrich the data set as actual experience deepens -- to add the next data point. Incrementalism.
Make sense? Reasonable?
Incidentally, after years of dithering I actually have boards on the way (with solder in part as a way to support diyaudio) so my comments/questions are a bit less academic that they've been in the past. Thanks for your patience with me over the years.
A while back in the thread, Jackinnj was kind enough to answer my question about opamp choices. My paraphrase of his answer is that most, even almost all, opamps seemed to work. The exception is the stability of the AD797, which is a historically finicky opamp.
I just built my super regulator using OPA1611 because they were in the parts bin and they worked fine. No measurements, though.
Jac
We have some TI lurkers and participants here -- I've tested some new NE5534A and their noise and THD% are better than spec. I haven't tried OPA1611 but would look good on paper.
I had trouble with the LT1028, ymmv.