Looking back at the output measurements I see a worst case difference on low gain of 26dB and best case on high gain 40dB. Since I'm measuring the 50Hz which also is parasitic around the test area it could easily have muddied some measurements, but only in the favor of the denoiser. Whenever there's extra 50hz it takes away from the measurement of the denoiser's performance. On higher levels it maybe is not such a problem but on lower level measurements it might cover some of the denoiser performance.
-116dB total improvement is around 40dB extra on the around -75dB or so total that the lm3x7+adj cap offers. So this seems to be somewhat consistent with the output measurements I made yesterday. At least on the best measuring channel on high gain. Unless I did something wrong in the measurements.
-116dB total improvement is around 40dB extra on the around -75dB or so total that the lm3x7+adj cap offers. So this seems to be somewhat consistent with the output measurements I made yesterday. At least on the best measuring channel on high gain. Unless I did something wrong in the measurements.
I looked in the alc892 datasheet, page 74 shows an input impedance of 64k. I used a 6.8uF cap just to be sure I don't filter anything at 50Hz. I connected it directly on the input of each regulator. Then I directly connected the cable into the line-in in PC.
ARTA has 3672mVpp (1.298Vrms) for the 0dB reference.
This is the result for the positive input:
And this is the result for the negative input:
The negative output was -130.14dB and the negative input is -18.35dB. The difference is 111.79dB. That is almost 37dB improvement over an average -75dB of a typical lm3x7+adj cap. Crazy.
For the positive rail I measured -76.92dB on the output so total -136.92dB. The input is -18.96dB, so the difference is almost 118dB. That's around 43dB improvement vs a lm3x7+adj cap. I find this hard to believe. Did I use the wrong scale or something?
edit: yes, used dBFS instead of dBu. Have to start from scratch

ARTA has 3672mVpp (1.298Vrms) for the 0dB reference.

This is the result for the positive input:

And this is the result for the negative input:

The negative output was -130.14dB and the negative input is -18.35dB. The difference is 111.79dB. That is almost 37dB improvement over an average -75dB of a typical lm3x7+adj cap. Crazy.
For the positive rail I measured -76.92dB on the output so total -136.92dB. The input is -18.96dB, so the difference is almost 118dB. That's around 43dB improvement vs a lm3x7+adj cap. I find this hard to believe. Did I use the wrong scale or something?
edit: yes, used dBFS instead of dBu. Have to start from scratch

Last edited:
I measured again using dBu. Measured input of regs, 6.8uF cap to ADC. Measured the output of the reg, directly into the LNA, 60dB gain.
Negative input:
Negative output:
Positive input:
Positive output:
That works out for a total of -111.43dB for the negative reg+denoiser and -119.57dB for the positive reg+denoiser. Subtracting about 75dB for the lm3x7+adj cap that means denoiser offered extra 36.43dB for negative and 44.57dB for positive.
I don't know if I am doing something wrong with measurements. Seems a bit on the high side, especially for the positive one. It nears dienoiser territory.
This is the first time I used ARTA. Is there anything else I should have done?
Headamp sounds fantastic!
Negative input:

Negative output:

Positive input:

Positive output:

That works out for a total of -111.43dB for the negative reg+denoiser and -119.57dB for the positive reg+denoiser. Subtracting about 75dB for the lm3x7+adj cap that means denoiser offered extra 36.43dB for negative and 44.57dB for positive.
I don't know if I am doing something wrong with measurements. Seems a bit on the high side, especially for the positive one. It nears dienoiser territory.
This is the first time I used ARTA. Is there anything else I should have done?
Headamp sounds fantastic!
I needed to make a sanity check. The only thing I wasn't sure of was the gain of the LNA, as I initially tested it with the crappy oscilloscope.
LNA on 60dB gain has an input sensitivity of 26mVpp max for 26Vpp max output. I used my phone as a signal generator, for 50Hz. I had to fiddle as the ADC would overload fast and I had to tweak the min volume of the phone for some signal to be seen with phone directly connected to ADC. I did two tests at slightly different levels, and as you can see from the following screenshots the LNA is pretty close to the 60dB spec. I also didn't change the volume between switching phone from ADC to LNA.
Phone output directly to ADC:
Phone output to 60dB LNA, and LNA to ADC:
And second test at slightly different output level:
So this test makes me believe the relative measurement of the denoisers.
LNA on 60dB gain has an input sensitivity of 26mVpp max for 26Vpp max output. I used my phone as a signal generator, for 50Hz. I had to fiddle as the ADC would overload fast and I had to tweak the min volume of the phone for some signal to be seen with phone directly connected to ADC. I did two tests at slightly different levels, and as you can see from the following screenshots the LNA is pretty close to the 60dB spec. I also didn't change the volume between switching phone from ADC to LNA.
Phone output directly to ADC:

Phone output to 60dB LNA, and LNA to ADC:

And second test at slightly different output level:


So this test makes me believe the relative measurement of the denoisers.
What is the measured voltage AC gain of the denoiser at 100 Hz ?
Vadj/Vadj
How could I measure this?
Ah, you mean probing the adj pins? I'll try it and report back. There's the 100Hz harmonic so I should see it.
Last edited:
This is the positive adj pin into 60dB LNA:
And the negative adj pin:
I made it for 50Hz as that was what I measured on the output. But 100Hz seems to scale the same.
Seems around 30.27dB for the negative and 34.86dB for the positive.
Might have busted the lm317 during these last measurements. Quick turn on/turn off and no protection diodes
Gonna have fun filing down the tab from a to220. I don't have any d2pak lm317.

And the negative adj pin:

I made it for 50Hz as that was what I measured on the output. But 100Hz seems to scale the same.
Seems around 30.27dB for the negative and 34.86dB for the positive.
Might have busted the lm317 during these last measurements. Quick turn on/turn off and no protection diodes

Gonna have fun filing down the tab from a to220. I don't have any d2pak lm317.
Managed to replace the broken LM317. Filing it down wasn't that hard. And looks like a proper d2pak. I also measured this one's output and it's similar to the original one, about 1dB less. The whole setup is a bit touchy, I get a 3dB range depending on cable layout on the desk, my presence near the whole thing etc.
Since I had to search for the lm317 stash I also found lm338, lt1084 and lt1085. I might test them as well in the diy supply I firstly made with the denoiser/dienoiser. Since I have the LNA now I'm curious of how these would perform on both denoiser circuits.
Since I had to search for the lm317 stash I also found lm338, lt1084 and lt1085. I might test them as well in the diy supply I firstly made with the denoiser/dienoiser. Since I have the LNA now I'm curious of how these would perform on both denoiser circuits.




Typo....What is the measured voltage AC gain of the denoiser at 100 Hz ?
Vadj/Vadj
Read ..AC gain...Vadj/Vout.
Isn't it 34.86dB? When I measure ADJ doesn't it have the AC from the output just that 34.86dB amplified? If you look at the positive output it's -68.83dB (first lm317 that died). Positive ADJ pin shows -33.97dB. The difference is 34.86dB so isn't this the gain of the denoiser?
Yes, if I understand correctly.
35dB makes sense.
Denoiser is nothing but an amplifier, its
gain and bandwidth fully characterizes its effects on the regulator.
35dB makes sense.
Denoiser is nothing but an amplifier, its
gain and bandwidth fully characterizes its effects on the regulator.
So for 35dB gain of the denioser it's a max of 35dB attenuation of lm317 noise/power supply ripple? Or it can go higher? I try to understand why the positive worked out to around -44dB for the 50Hz ripple. Negative would be in spec at around -35dB psrr but measured -30dB gain at ADJ pin.
Ah no, I got confused. It's around 30-35dB over the cap on adj pin version. So the total attenuation is more than the gain of the denoiser.
Trileru, I am confused too. I have simulation results that don't fit. I have not found yet what's going on, some simulation artifact or wrong reasoning ?
So forget my last posts for the moment, do not waste your time on those.
I am investigating.
So forget my last posts for the moment, do not waste your time on those.
I am investigating.
I tested LM338 and it works with the denoiser but is worse by 3dB or so. LT1084 and LT1085 doesn't work. I mean I think it works, but it's 60dB total. Better use adj cap than denoiser, for LT1084/5
Trileru, I am confused too. I have simulation results that don't fit. I have not found yet what's going on, some simulation artifact or wrong reasoning ?
So forget my last posts for the moment, do not waste your time on those.
I am investigating.
Check this post by Elvee:
D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg.
Details how to simulate psrr, noise of lm317 and also output impedance. Maybe it is of help.
I am well aware of these techniques.
I have trouble how these fit with loop gain and the gain of the denoiser.
I have trouble how these fit with loop gain and the gain of the denoiser.
I tested the denoiser and dienoiser (lm317) on my diy supply. The single rail one from few pages back.
I noticed that I have to give it some time. The noise lowers in 3-5 minutes or so. I had it covered with a kitchen metal bowl, and I used a 150ohm resistor for around 100mA current draw. I chose a high input ripple of around 1.4Vpp (simulted in LTSpice). I used a single 680uF cap on the input, C filter, nothing else. Crappy oscilloscope shows around 1Vpp, ARTA shows around 2.2Vpp. Somewhere around there. Doesn't matter, it's relative anyway, I'm interested of the difference between denoiser and dienoiser.
This is the input:
This is the output of the denoiser:
And this is the output of the dienoiser:
The ripple is 100Hz in this test as I used a bridge.
So around -109dB for the denoiser and around -128dB for the dienoiser. About 19dB difference between them.
edit:
also ADJ pin, which shows the same for both denoiser and dienoiser:
I noticed that I have to give it some time. The noise lowers in 3-5 minutes or so. I had it covered with a kitchen metal bowl, and I used a 150ohm resistor for around 100mA current draw. I chose a high input ripple of around 1.4Vpp (simulted in LTSpice). I used a single 680uF cap on the input, C filter, nothing else. Crappy oscilloscope shows around 1Vpp, ARTA shows around 2.2Vpp. Somewhere around there. Doesn't matter, it's relative anyway, I'm interested of the difference between denoiser and dienoiser.
This is the input:

This is the output of the denoiser:

And this is the output of the dienoiser:

The ripple is 100Hz in this test as I used a bridge.
So around -109dB for the denoiser and around -128dB for the dienoiser. About 19dB difference between them.
edit:
also ADJ pin, which shows the same for both denoiser and dienoiser:

Last edited:
- Home
- Amplifiers
- Power Supplies
- D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based VReg