I went hunting for the problem and managed to sort it out. Seems to be the wall wart for the provider modem that's causing that. Getting close to it I can hear some whining.
I'll try and replace or at least turn it off when I take measurements.
I'll try and replace or at least turn it off when I take measurements.
The horror that many SMPS adapters exhibit this mains polluting phenomenon and then realising that you have to replace them all for linear ones and also need mains filters of good quality 😀
I made the first tests with LM337 + dienoiser on the diy power supply but then when I tried it on the Magni it just didn't work. I didn't have room to use the same output cap I used in testing.
In the end I did manage to make it work on the Magni as well, just that it was a pain of trial and error. I had to experimentally find the correct values for the comp network. In the end it worked with 40nF + 1R. That 1R seems to be critical. 3.3R/5R/10R wasn't stable. 0.33R seemed to worsen the performance. 0.5R was ok but touchy. I test for oscillations by touching Vout. If it's on the brink that usually gets it off into oscillation. 0.5R worked, it went crazy but would come back. With 1R it seemed to not care when I touch it. I am not sure what R is related to. I did use 4x10nF caps in parallel which would lower the ESR.
Also the output cap is kind of tricky. I tested some and it seems that anything over 40mOhm doesn't work. Most under 40mOhm worked, irrespective of capacitance.
So the LM337 + dienoiser can be made stable just that you really need a LNA to check your results. Osciallation isn't the only issue with it. While tweaking the compensation network + output cap I had it stable in some configurations but the performance was lacking. It would either get the ADC in protection for way too high noise, or it would be sub denoiser performance. To get it to perform as it should it took some time of trial and error. So don't go adding it to a circuit without seeing what's happening. You're safer with the LM337+denoiser. But if you have the tools and are willing to spend the time tweaking it then sure.
Here's the LM337+denoiser and LM337+dienoiser as installed in the Magni, measurement taken at output cap. Some 10dB improvement.
In the end I did manage to make it work on the Magni as well, just that it was a pain of trial and error. I had to experimentally find the correct values for the comp network. In the end it worked with 40nF + 1R. That 1R seems to be critical. 3.3R/5R/10R wasn't stable. 0.33R seemed to worsen the performance. 0.5R was ok but touchy. I test for oscillations by touching Vout. If it's on the brink that usually gets it off into oscillation. 0.5R worked, it went crazy but would come back. With 1R it seemed to not care when I touch it. I am not sure what R is related to. I did use 4x10nF caps in parallel which would lower the ESR.
Also the output cap is kind of tricky. I tested some and it seems that anything over 40mOhm doesn't work. Most under 40mOhm worked, irrespective of capacitance.
So the LM337 + dienoiser can be made stable just that you really need a LNA to check your results. Osciallation isn't the only issue with it. While tweaking the compensation network + output cap I had it stable in some configurations but the performance was lacking. It would either get the ADC in protection for way too high noise, or it would be sub denoiser performance. To get it to perform as it should it took some time of trial and error. So don't go adding it to a circuit without seeing what's happening. You're safer with the LM337+denoiser. But if you have the tools and are willing to spend the time tweaking it then sure.
Here's the LM337+denoiser and LM337+dienoiser as installed in the Magni, measurement taken at output cap. Some 10dB improvement.
Last edited:
Hi and thanks for all the measurents.
Is it possible for you to try with iron plates in your test setup too separate the noise sources from the electronics.
It could also be a learning task?
Here's the schematic and layout (C1/C4 are on the backside along with the jumpers):
My LNA and ADC are far from something that could truly characterize the performance of these circuits. They're rather tools that help me see if anything is obviously wrong. It's not worth the time for me improving the testing setup.
edit: the HF noise from that switchmode wall wart was coming through the 230VAC line. I think a power line filter would have helped take that out.
Last edited:
THANK YOU VERY MUCH Trileru i will try today can i use 7812/7912 voltage rgulator ic
warm regards
Andrew
warm regards
Andrew
Here's a comparison between LM317 + dienoiser and LT1085 + dienoiser. I tweaked the dienoiser resistor value, seems LT1085 gives better performance at around 1K vs the 560R recommended for LM317. The tests were done on the diy power supply, so the signal of interest is 100Hz and its harmonics. 50Hz is environmental pickup, and you can see it and its harmonics are not attenuated between the regulators. There's around 11.5dB extra attenuation at 100Hz from the LT1085 + dienoiser (orange LM317 and blue LT1085):
edit: 100uF/63V Panasonic FC on the output of LM317 and 100uF/63V Panasonic FR on the output of LT1085
edit: 100uF/63V Panasonic FC on the output of LM317 and 100uF/63V Panasonic FR on the output of LT1085
Last edited:
Is it ok to use a trimmer instead of a resistor from LM317 adjust pin to ground to adjust output voltage ?
Damien
Damien
For the denoiser I don't see a problem, but for the dienoiser ideally you'd adjust the dienoiser resistor value for the output voltage. But if you're using it for trimming the voltage close to target then you would be ok without adjusting that resistor as well.
Good find. Did you sim it in ltspice?
These are Nazar's versions:
S-Audio.Systems // "Nazar`s regulator", variants
Last edited:
You will need a very good (and expensive) opamp to match the noise performance of a humble BC337 or even a S9013.
The 1.5K input resistor alone will generate 20 to 30 times the noise of a transistor.
The advantage of an opamp is that the 100Hz gain can be pushed to arbitrarily high levels, to reject the ripple.
Other characteristics will be mediocre.
Elvee,
I remember that Nazar_lv was introduced this circuit as "Capacitorless regulator" but it didn't take too much attention. Probably he was referring to same AC principles as your Denoiser which makes no need for big electrolytic filter capacitors. What do you think?
If I remember correctly, he used this dual rail design on his commercial projects too. Which means it has no stability issues with 2 transistor design on negative rail, right?
S-Audio.Systems // Питание High-end апаратуры без применения электролитов
http://s-audio.systems/public/uploads/PSU/capacitorless-630x547.gif
I remember that Nazar_lv was introduced this circuit as "Capacitorless regulator" but it didn't take too much attention. Probably he was referring to same AC principles as your Denoiser which makes no need for big electrolytic filter capacitors. What do you think?
If I remember correctly, he used this dual rail design on his commercial projects too. Which means it has no stability issues with 2 transistor design on negative rail, right?
S-Audio.Systems // Питание High-end апаратуры без применения электролитов
http://s-audio.systems/public/uploads/PSU/capacitorless-630x547.gif
Attachments
L78xx/79xx are very good at high frequencies by themselves , they are specd at 100khz. So you only need to improve their lower frequency behavior which is what Technics denoiser does as you mentioned.
The best advantage of an op-amp is that it supports with minimum space on PCB both regulators when using a dual op-amp. I am using MC7805/7905 with zenner ADJ lifter for higher voltages, as mc7805 is the fastest and lowest noise of the series and here you can see two double regulators +-20v, +-10v regulators and you see much higher capacitors than needed...
Attachments
They are just "Finesse regulators" by Charles Wenzel. They were mentionned at the start of the thread:
D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg.
They work, but they require uncomfortable tradeoffs like impedance vs. noise reduction.
The denoiser not only maintains the key parameters, but it betters them and the DC capabilities of the initial regulator need not be derated.
The 78/79 might be relatively good at HF, but their noise cannot match a BJT, and the very high theoretical PSRR attained by an opamp denoiser is going to be meaningless if it isn't matched by the LF noise from the opamp and its input resistance.
Managed to cascade two LT1085+dienoiser. First one is the +12V rail and the second one is VACC/Vref (in my DAC).
I tried it with LM317+dienoiser for +12V and LT1085+dienoiser for 5V (Vref) but the LM317 didn't like it at all. Then I replaced the LM317 with LT1085. Had to tweak the comp network but got it stable.
Sadly the DAC environment is kinda crap and measurements seem similar to the previous ones. Ideally I'd have to make a better pcb but at the moment I'll use it as it is. Sounds absolutely great!
There was something off about that ams1117+dienoiser for Vref. There was lots of detail but bass was a bit off, kind of lacking. I don't trust those unknown source ams1117.
I tried it with LM317+dienoiser for +12V and LT1085+dienoiser for 5V (Vref) but the LM317 didn't like it at all. Then I replaced the LM317 with LT1085. Had to tweak the comp network but got it stable.
Sadly the DAC environment is kinda crap and measurements seem similar to the previous ones. Ideally I'd have to make a better pcb but at the moment I'll use it as it is. Sounds absolutely great!
There was something off about that ams1117+dienoiser for Vref. There was lots of detail but bass was a bit off, kind of lacking. I don't trust those unknown source ams1117.
Since I'm almost done with the dienoiser mods in my gear I figured I might test the nonoiser as well. I made a simple tht version that is also diy-able. It's 27mm X 21mm. The inductor is actually the ferrite bead, I didn't have a 3D model for it. You could bypass with a wire and install the bead on the bjt's leg. I attached the Kicad project with pdf for diy and gerbers for fab.
Elvee, if it isn't stable what could be tweaked? I had C/Rcomp for the dienoiser.
This design has not been tested, you make it at your own risk.
Elvee, if it isn't stable what could be tweaked? I had C/Rcomp for the dienoiser.
This design has not been tested, you make it at your own risk.
I made a comparison between the dienoiser that is onboard the DIY supply I made, from post #1396 and the add-on dienoiser from post #1590 that I attached to the DIY supply, after removing the onboard 220uF from the ADJ pin. I wanted to see if there's a significant difference of performance between the one that is onboard and the dangling add-on. The result at 12V and 80mA is that they both behaved almost identically. The add-on had around 4-5cm of wire between.
Last edited: