ZTX can bring improvement only in the nonoiser or dienoiser designs, where noise falls below 1 nV/rtHz. Your numbers are as expected. LM337, as pointed out by bohrok2610, has slightly higher noise. 10 nF compensation capacitor is enough in my experience. Higher values negatively affect high frequency PSRR.
As you said (should put that in taller letters) Just follow Mark's previous guidance.
So, Trileru
don’t you think that you owe an apology to the OP. How did your story about “bad design”, “shafted users”, “immature behavior” …. turn out?
don’t you think that you owe an apology to the OP. How did your story about “bad design”, “shafted users”, “immature behavior” …. turn out?
I don't have to apologize for anything, I haven't said anything wrong.
The pcb design hasn't been updated to follow the denoiser recommendations. The design itself is not optimal for the denoiser. And it's rather wasteful with 4layers if you ask me.
The whole issue was that the VRDN creator chose to blame the TI LM337 instead of trying more to make it work. I invested some time into the denoiser project and I want to be clear that the denoiser works fine as long as the implementation recommendations are followed.
Also I don't think you realize the optics of the VRDN creator having you carrying his battles. He hasn't addressed anything and he chose to let other users try to sort it out for him.
Also, we need to make sure the VRDN is working with the values that other people used, based on rev.A2 which hasn't been measured yet.
The pcb design hasn't been updated to follow the denoiser recommendations. The design itself is not optimal for the denoiser. And it's rather wasteful with 4layers if you ask me.
The whole issue was that the VRDN creator chose to blame the TI LM337 instead of trying more to make it work. I invested some time into the denoiser project and I want to be clear that the denoiser works fine as long as the implementation recommendations are followed.
Also I don't think you realize the optics of the VRDN creator having you carrying his battles. He hasn't addressed anything and he chose to let other users try to sort it out for him.
Also, we need to make sure the VRDN is working with the values that other people used, based on rev.A2 which hasn't been measured yet.
I simply can’ believe the number of wrong statements in your reply.
Pointing out to you, what is wrong is … well, pointless.
Pointing out to you, what is wrong is … well, pointless.
Using some true RMS voltmeter or oscilloscope, measure ripple voltage at the input of LM chip. You need some load at the output of regulator to have substantial ripple voltage. Let’s say it is measured to be 100 mV RMS. Using some online voltage to dBV calculator determine corresponding dBV value. For the 100 mV it is -20 dBV.
After calibrating REW using V/rtHz scale, switch to dBV scale. Measure where is top of the 120 Hz “hump” at the regulator output. Let’s say it is at -120 dBV.
PSRR is 120 -20 = 100 dB.
His Instrument input on his ADC has a high imput impedance. Wouldn't a 4.7uF capacitor in series with the ADC input be enough to allow him to see the ripple in ARTA/REW? Then subtract that from the output ripple measurement in dB.
I tried but if you check his previous comments you'd understand my answer.
He accused me of manipulating my measurements to fit member's expectations and said I used a "rectal extraction method" for my measurements.
If you're going to suggest to anyone to show more respect I think it should be to him.
He accused me of manipulating my measurements to fit member's expectations and said I used a "rectal extraction method" for my measurements.
If you're going to suggest to anyone to show more respect I think it should be to him.
I have a second VRDN board that was in service powering my preamp. I took it apart and checked it also.
The second board has an On Semi LM317 and a TI LM337, the comp network on both rails is the original 6.8nF +15 ohms.
The negative rail denoiser was not working after measuring it.
The positive rail was working perfectly.
I removed the ESR correction resistor on the negative rail output resistor and the negative rail is now working perfectly.
BTW, both rails in this configuration had a little less noise 3-4nF/rtHz for both rails.
I am going to use Mark's recommended compensation network values even if it slightly increases noise, it will increase stability.
The second board has an On Semi LM317 and a TI LM337, the comp network on both rails is the original 6.8nF +15 ohms.
The negative rail denoiser was not working after measuring it.
The positive rail was working perfectly.
I removed the ESR correction resistor on the negative rail output resistor and the negative rail is now working perfectly.
BTW, both rails in this configuration had a little less noise 3-4nF/rtHz for both rails.
I am going to use Mark's recommended compensation network values even if it slightly increases noise, it will increase stability.
I am going to hold off posting any more measurements until I get the board configured as Mark recommended. After I get the boards fixed, I will post the final measurements. I'm sure they are end up being between 3-9nF/rtHz.
At that time I will also measure the PSRR.
At that time I will also measure the PSRR.
Dumb question, just to make sure I do these measurements correctly. How do I calibrate the V/rtHz scale? Convert noise of 1k resistor to dB? I think most of the calculators do that, from memory -124.9dB.
I should also add the second VRDN board was measured with voltage at +-21.5V (as high as I could get it and maintain balance between rails) and 120 mA load. I'm using it to power Wayne's preamp.
Yes, just follow Mark’s recommended values. It was bad advice on my part.
That On Semi LM regulators have lower noise, confirmed by properly conducted measurement so no doubt, is very useful information fort those planning VRDN build.
No need for conversion math. It is done automatically. Using V/rtHz scale calibrate reading and just switch to dBV scale/view using drop down selection list at left top corner of RTA window. It will immediately display correct dBV reading.
You can use REW to measure input ripple voltage directly only if ripple voltage is below level that would bring LNA to the clipping, and that is around 10 mV for 60 dB gain and +-15V LNA supply.
This is not correct. I'll again share the findings of Elvee on noise for the denoiser (not dienoiser or nonoiser):
D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based V.Reg.
Where it's clearly stated that the best noise results were with the ZTX transistors. There should be a clear difference between ZTX851 and BC337 (0.16uV vs around 0.3uV).
Yes, and since the BW was 10kHz, it would roughly translate into 1.6nV/rtHz and 3nV/rtHz at 1 kHz respectively (somewhat less in fact, because of the excess LF noise).
It is important to note that these value only take into account the transistor and regulator noise, since the input supply was a clean lab supply, and the load was purely static (resistor).
In a real setting, the input ripple and the load-induced voltages would also contribute, as would parasitic inductions and ambient electro-smog.
IIRC, the regulators tested were from ST and ONsemi, and gave very similar performances
It is important to note that these value only take into account the transistor and regulator noise, since the input supply was a clean lab supply, and the load was purely static (resistor).
In a real setting, the input ripple and the load-induced voltages would also contribute, as would parasitic inductions and ambient electro-smog.
IIRC, the regulators tested were from ST and ONsemi, and gave very similar performances
