The LM317 was always stable with the dienoiser. I had issues with LM337 + dienoiser but that was down to the high ESR output cap I used for the LM317. Once I used low ESR I manged to got it stable.
The nonoiser board is not stable. I've tried with LM317/LM338/LT1084 with different output capacitors ESR.
The output voltage rises while the overall noisefloor falls but at some point it goes wild.
I'll not try to make it work, I'm happy with the dienosers at the moment.
edit: also just to make things clear both small single smd board (from this post D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg. ) and the double tht one (from this post D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg. ) work with LM3x7 and LM78xx/LM79xx both for denoiser and dienoiser configurations. You just need to adjust the comp network for dienoisers.
The nonoiser board is not stable. I've tried with LM317/LM338/LT1084 with different output capacitors ESR.
The output voltage rises while the overall noisefloor falls but at some point it goes wild.
I'll not try to make it work, I'm happy with the dienosers at the moment.
edit: also just to make things clear both small single smd board (from this post D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg. ) and the double tht one (from this post D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg. ) work with LM3x7 and LM78xx/LM79xx both for denoiser and dienoiser configurations. You just need to adjust the comp network for dienoisers.
Last edited:
There is about no use of stability considerations with no knowledge of stability margin as used in servo systems.
Phase margin, Black abacus are the right tools from servo system theory.
Spice simulation with a Tian probe is the way to get at the open loop gain that will reveal the stability margin.
Fiddling on breadboards, with trial and errors, can only run into circles to end up at unpredictable results.
Let us, not forget that stability is in the most important parameters for a power supply. Who cares about noise and regulation if it can blow up what is powered.
Phase margin, Black abacus are the right tools from servo system theory.
Spice simulation with a Tian probe is the way to get at the open loop gain that will reveal the stability margin.
Fiddling on breadboards, with trial and errors, can only run into circles to end up at unpredictable results.
Let us, not forget that stability is in the most important parameters for a power supply. Who cares about noise and regulation if it can blow up what is powered.
I found that I can usually make a marginal stable dienoiser go into oscillation if I tap Vout or the emitter of the PNP (for positive). When it's stable the noise jumps a bit but instantly settles back. When it's marginally stable it would either go wild for a few seconds and recover either stay oscillating. That's how I chose the compensation values, and for the ones that passed this test and that I monitored for longer while working I had no issues.
I had no denoiser that went into oscillation. Just the dienoisers are a bit finicky.
I had no denoiser that went into oscillation. Just the dienoisers are a bit finicky.
Do not confuse " marginally stable" with "margin" as used in servo system theory.
The former is a vague so to speak wording, the latter is well defined to engineers.
Marginally stable, means to me: Not so stable, near to burst in oscillation.
Phase margin, dumping, Q, have accurate meanings to me.
No wonder the denoiser is always stable.
I understand you are speaking of this circuit standing alone.
It has no loop, it is an open circuit, a typical common emitter stage. It can only be stable.
The issue is unconditional stability, against components variations, unknown parasitics and various loads, a pretty tough requirement.
The former is a vague so to speak wording, the latter is well defined to engineers.
Marginally stable, means to me: Not so stable, near to burst in oscillation.
Phase margin, dumping, Q, have accurate meanings to me.
No wonder the denoiser is always stable.
I understand you are speaking of this circuit standing alone.
It has no loop, it is an open circuit, a typical common emitter stage. It can only be stable.
The issue is unconditional stability, against components variations, unknown parasitics and various loads, a pretty tough requirement.
I think you are referring to peace of mind implementing it in commercial products that would account for component/batch variations. My dealings were only about diy use of the circuit. If anyone wants peace of mind in using this for commercial applications they should spend the money and time on testing
I understand DIY is a different world and I am too much influenced by engineering.
Personally, I do not like to play Russian roulette and I only trust bullet proof designs.
Personally, I do not like to play Russian roulette and I only trust bullet proof designs.
Advances occur when people take risks and try new, unproven ideas.
"If I had asked people what they wanted, they would have said faster horses." -- Henry Ford
"If I had asked people what they wanted, they would have said faster horses." -- Henry Ford
You think that using the denoiser is the equivalent of a Russian roulette? That seems a bit of a stretch even for people under the influence of engineering.
I was asking if you think that, I am not saying you did.
From my experience neither denoiser or dienoiser went over the set voltage. So far I have added them to +- of the output stage of the DAC, to the Vref/VACC, to the analog line for spdif clock recovery on the digital receiver chip (3.3VDC), to the +- rails of the Schiit Magni 3 headamp and to the +- rails of a modified Behringer CX2300 active crossover. All are working, never went over the set voltage even while oscillating in testing. It wobbles between set voltage and lower voltage never went over set voltage.
The only time it went over was on the DIY supply because I shorted IN and OUT. But that was my mistake.
From my experience neither denoiser or dienoiser went over the set voltage. So far I have added them to +- of the output stage of the DAC, to the Vref/VACC, to the analog line for spdif clock recovery on the digital receiver chip (3.3VDC), to the +- rails of the Schiit Magni 3 headamp and to the +- rails of a modified Behringer CX2300 active crossover. All are working, never went over the set voltage even while oscillating in testing. It wobbles between set voltage and lower voltage never went over set voltage.
The only time it went over was on the DIY supply because I shorted IN and OUT. But that was my mistake.
Throwing additional loop gain into a system designed to be stable with no external gain could be a recipe for disaster.
A proper stability analysis could tell whether it is reasonable or not, but there are no real DS data or sim models accurate enough to determine with 100% confidence what is really going to happen.
Experience shows that a properly implemented 317 denoiser works at least 99.9% of the time; for the dienoiser and the nonoiser, things are more complicated, but it is possible to have them working most of the time.
Note that most of the "vetted" applications of the 317 and other regulators break some of the rules: PNP-boosted regs, Maida's, etc.
In general, they seem to work. Sometimes they don't -but even the fully legit applications sometimes fail-
A proper stability analysis could tell whether it is reasonable or not, but there are no real DS data or sim models accurate enough to determine with 100% confidence what is really going to happen.
Experience shows that a properly implemented 317 denoiser works at least 99.9% of the time; for the dienoiser and the nonoiser, things are more complicated, but it is possible to have them working most of the time.
Note that most of the "vetted" applications of the 317 and other regulators break some of the rules: PNP-boosted regs, Maida's, etc.
In general, they seem to work. Sometimes they don't -but even the fully legit applications sometimes fail-
Can this new version of the Dienoiser work free of oscillations?
If you look closely, the 180K resistor connection has changed slightly. The 22 nF compensation capacitor has been removed.
I confess that I have not yet been able to test it.
If you look closely, the 180K resistor connection has changed slightly. The 22 nF compensation capacitor has been removed.
I confess that I have not yet been able to test it.
Attachments
Last edited:
I could give it a try when I have some time to test. My diy testing supply is set up for positive regulators at the moment but I presume it should work with LM317 as well?
However, on Multisim 14 it would appear to be working.
Possibly, a real test has the last word.
I estimate it could be implemented with the LM317 as well.
Best regards
Possibly, a real test has the last word.
I estimate it could be implemented with the LM317 as well.
Best regards
In LTSpice your last version seems to be having similar performance to the adj cap version. So that should be pretty simple to tell in testing if it works or not.
So apparently just removing the 10nF comp cap seems to make the dienoiser to behave like the adj+cap version. So your proposed mod might work. Hope I find some time to test it.
You have shorted the 10n comp cap so no wonder it behaves like adj+cap. Instead of shorting you should remove it.