Help with HV adjustable power supplies for radiation detectors.

[Cai Arcos]

Hi:

I'm in the process of building a variable (0-600V), relatively silent (400mV peak to peak noise maximum) and able to sustain it's output voltage under a 1Mega omh load power supply for work in development of radiation detectors. The schematic I post have achieved that, with a noise of around 360mV peak to peak. It is based on a design that appears at the 3rd edition of The Art of Electronics (also included), substituing the HV module with a CCFL inverter and a rectifier/filter. The idea to use a resistor in the output filter (as well as it's value) is from App Note 118 of LT "High Voltage, Low Noise, DC/DC converters" by Jim Williams. Along the prototyping part of the project I have developed some doubts that I hope this board can help me with:

1) If I understand correctly, the Cc and Rc in the original schematic serve to create a PI controller. When using them and adjusting the output voltage, the output comes to the needed voltage more slowly and also "rings" a little bit, which I believe proves their function. However, they also significantly increase the noise creating a sinusoidal wave (whose frequency is controlled by Rc and Cc) that with the values suggested reach around 1,5 V p-p (reducing the value of the capacitor reduces a little this value). This is the reason they are not in my schematic. Why this increase in noise? Do they serve other purposes?

2) What is the function of the 162 kilo ohm resistor present in the original schematic? Since the op amp practically draws no current, it seems useless.

3) This is the most puzzling. I originally used a 1.2k resistor and 6.1V zener to generate the 6V required at one leg of the pot that controls the feedback. This reasulted in a sinusoidal-ish noise output with around 1,6V p-p. The issue was completely resolved when this voltage was derived using only the divider. Furthermore, when the feedback pot is turned maximum (and thus the zener conducts to clamp the feedback voltage) the same effect (although slightly reduced) appears again. What is happening with the zener diode to create that noise at the output?

Thanks!

[John Futter]

Why did you leave the high frequency gain reduction out??

Also a small cap across the 1 meg resistor would help inspeeding up the feedback loop

As it is the circuit is very short on stability ie it might oscillate rather than regulate

[Rich Feldman]

Are CCFL inverters OK with having one output terminal grounded?

If yes, then who needs bridge rectifier?

If no, then driving both AC terminals of a bridge rectifier won't make it OK to ground one of the DC terminals.

CCFL inverter does seem like compact and inexpensive way to avoid generating high frequency AC yourself.
Maybe the inverter outputs can be AC coupled into some kind of rectifier or multiplier circuit.

THere was some discussion on fusor.net about 10 years ago, about combining the power from many CCFL inverters using something like that.

[Cai Arcos]

John:

I wrote this late at night and made an important mistake. The Rc and Cc components don't introduce oscillation. I thought so because I tested the whole circuit when using the zener diode (the source of noise). I have now included them back and the circuit works the same way (it didn't oscillate before at all, but better be safe than sorry). Sorry for the error.

I include two captures of the output (AC coupled). The one that is "silent" is using the circuit attached below with the high frequency roll of components included. The other is the same: except that the divider at one leg of the pot is replaced by a 2.2K and 6.1V zener. This is what is causing most of my headaches.

What value should the speed-up capacitor be?

[Cai Arcos]

Rich:

I had found two posts regarding CCFL inverters:

viewtopic.php?f=13&t=12097&p=78757&hilit=CCFL#p78757
viewtopic.php?f=11&t=4775&p=31184&hilit=CCFL#p31184

In Doug Coulter's Forums there is also extensive discussion on noise and such: the problem is that no feedback is used in that case (which makes sense for what they're using it. As they say, you have problem if your He3 tube draws enough current to cause the output voltage to drop).

I think I've not expressed myself well (as per usual). The CCFL converter is a Royer Oscillator, whose output is a 120Khz sinusoidal voltage isolated from the input. I rectify/filter it, and then connect the ground of the bridge to the ground of the primary circuit because I don't need any isolation. I now see the schematic was not well made, and thus not clear.

[Cai Arcos]

John:

I problem I've noticed when using Rc and Cc: after leaving the circuit for around 2 or 3 minutes, the noise increases dramatically. I post the noise measurement with different values of Rc and Cc used and the result. It seems as this is the phenomena I was first describing. It seems as if the circuit is oscillating: which did not happened without them.

[John Futter]

Try a speed up cap across R1 47pf to 220pf but it should be ca 1kV type

[Rich Feldman]

When using scope to measure ripple on HV output, pay attention to the voltage rating of oscilloscope in AC mode.
It might be safer to use your own DC blocking capacitor, with well known voltage rating, and leave scope channel in DC mode.

[Cai Arcos]

John:

In the end, I actually recreated one of the schematics posted in the attached posts and buffered the feedback. Probably due to the very limited slew rate of the LM324, no oscillations happened, an it fact the noise is now under control. Adding the speed capacitor actually reduces a little bit more the noise, and so I have added it. Now the problem is that the CCFL inverter I took from an emergency light does not have enough power, and now I'm looking for something like this:

https://www.pollin.at/p/ccfl-inverter-c ... -kv-121244

Rich:

I'm considering recreating what you have made in here:

viewtopic.php?f=13&t=12649&start=20

What roll-off frequency have you chosen?

[Cai Arcos]

Hi:

I've been using some time of the quarantine to work on the high voltage supplies. Unfortunately, due to somo environmental regulations I was unable to buy the inverters showed in a previous post, so I tried other ways.

I *must* recommend to everyone interested in this type of work the App Note 118 of Linear Technologies by Jim Williams "High Voltage, Low Noise, DC/DC Converters". What I did find was an old LCD monitor. I teared it up and discovered that the HV section was a center tapped transformed being push-pulled by two N-Channel MOSFETs. By operating the transformer at it's resonant frequency, and using the filter shown in the figure 13 of the App Note, I am able to get around 250mV of noise at 400 V. One can also add a small capacitor (100pF in my case) between gate and drain in order to slow down the transitions (in a manner similar to that shown in the App note but without the specialized IC), but the noise was worse than that obtained with the sinuisoidal voltage at resonance. In both cases, the supply is a lot more stiff that the CCFL inverter, not buckling under a 1M load and going down to around 350V with a 100K load (I can increase the power supply of the primary and sustain the 400V no problems). Salvaging these screens seems the way to go, and most come with two transformers (because of the number of cold cathode lamps in the display). In the future I probably will try to make it adjustable, but for my current quenching circuit experiments in GM tubes a fixed voltage should be good enough.

I also tried to use the always popular "Joule Thief" (in the version used by HV enthusiasts using a secondary). While undoubtedly a very ingenious and interesting circuit, it is most suited for applications needed a relatively low output impedance and low component count, since its discontinuous power delivery and high frequency content make it hard to obtain reliable low noise. One can, however, make a compromise on the output voltage and reduce the noise by reducing rise times by inserting a small capacitance between colector and base of the transistor.