How to get a negative 1000V at 0.1mA?

[razberik]

Flyback topology is definitely not suitable for PMT aplication.

Use Jim Williams aproach described in AN118: http://www.analog.com/media/en/technical-documentation/application-notes/AN118fb.pdf
Use Coiltronic/Eaton transformer like CTX21xxx. There are obsolete, but if you want to built one-piece, it is possible to source them.

[kony]

I just happend to have our AAS service manual on table, so here is solution coutesy of Varian, cca 1985.

So just cascoded shunt regulator. The HV source does not necessarily have to be auxilitary power transformer winding with a doubler, resonant converter with CCFL transformer is viable as well.

[Lee Leduc]

Lots of circuits on the net. Google "photomultiplier tube power supply circuit" and search images.
Here's one that I've used. "An adjustable 0 to 1500V+, regulated, low ripple HV power supply
for powering PMT's and geiger tubes".

It outputs a positive voltage but you may modify the circuit to change the polarity.
http://charliethompson.50megs.com/LowRippleHVsupply.html

Typical CCFL inverter sold by Digikey.
https://www.jkllamps.com/pdfs/BXA-24529.pdf

[tecman]

To answer a comment made earlier, negative voltage for a PMT is quite common.  The negative voltage is applied to the cathode, which makes the anode signal output relative to ground. 

For the power supply yoou will likely need a milliamp or two.  There should be a resistor divider for the dynodes.  If you make the value too high, then the tube will suffer non-linearity over the signal range.  20K to 1 meg are used, typical is 50K to 100K, which requires 1-2 ma just for the divider string.  As for the power supply, a LCD backlight supply (for CCFL) is ideal.  Since they are typically AC out, you can rectify to the polarity you want.  A simple op-amp circuit and transistor can be used to regulate the output by modulating the DC input to the inverter.  I have built several.  They are cheap (ebay) and effective.

paul
 

[voltsandjolts]

I have my doubts about HV noise performance of this opensource design, as part of a pulse height analyser project, but posting here since it is quite an interesting application:

https://www.theremino.com/en/blog/gamma-spectrometry

[ocset]

i would use a flyback.....you can actually use the reflected output voltage on the primary to sense when your up to your required voltage.
The flyback can be isolated as such.
Anyway, just use it the "wrong way round" for a negative output.

This flyback way is how they do it for the detonator power supplies.....they use high voltage like you.

There are other ways  of doing it including combining an smps with a cockcroft walton multiplier......but a straight flyback is good.

You can  even buy high voltage bobbins so you can get the voltage separation for your primary and secondary

[David Hess]

A voltage multiplier can be attached to the output of a boost, inverting, or flyback converter however it will have much higher noise than a Royer or inverter based design.

[Wolfgang]

True. Flybacks are simpler, but have steep transistions and create a lot of wideband noise.
Royers are quieter and can be slew-rate limited to make them even more quiet.
And in extreme cases you could use sine-wave inverters.

my question:

- What is the spec for your PMT tube application ? How much ripple and noise can it tolerate in what frequency ranges ?

[voltsandjolts]

I suppose ripple translates as instantaneous PMT gain change, so what percentage pulse height noise is acceptable in the application? Maybe for simple gamma count rate PMT applications higher ripple of flybacks would be tolerable, just set discriminator a bit higher. For spectral applications psu noise would smear the spectrum to some degree (higher FWHM).

The Hamamatsu range mostly have ripple specified in the 10 to 100mV range but they don't say much about it's bandwidth.
https://www.hamamatsu.com/resources/pdf/etd/High_Voltage_Power_Supply_TACC0005E.pdf

OrTec is <15mV 20Hz to 20MHz
https://www.ortec-online.com/products/electronics/power-supplies-and-nuclear-instrument-module-nim-bins/556-and-556h

CAEN NIM PSU modules seem to be 10mV typ. at 3Kv/200uA. Some quote <5mV 100Hz - 100MHz.
http://www.caen.it/csite/Function.jsp?parent=21&idfun=15

ISEG 10mV
https://iseg-hv.com/en/products/dc-dc

Matsusuda claim <1mV for this
https://www.matsusada.com/pdf/hpms.pdf
https://www.matsusada.com/product/psel/hvps2/

And a CERN flyback design that claims a 'low' 700mV ripple:
https://indico.cern.ch/event/83060/contributions/2101687/attachments/1069924/1525757/Presentation_TWEPP.pdf
http://iopscience.iop.org/article/10.1088/1748-0221/5/12/C12049/pdf

So, 10mVpp ripple on a 1000v psu, 20Hz-20MHz, seems feasible to aim for, 0.001% reg.

[voltsandjolts]

Cockroft-Walton design. Doesn't discuss output noise.

[David Hess]

Check out figure 17 on page 11 and figure 27 on page 18 of Linear Technology application note 45.  The first shows a flyback converter driving a voltage tripler and the second is the same but with just an inductor in boost configuration and lower output voltage.

I think the transformer shown was intended for a Royer CCFL inverter but since you are not running on 1.5 volts, a photographic flash flyback transformer would be more suitable and easier to get.

For the inductor version, a cascode transistor will need to be added to the common emitter or source switching transistor to handle higher voltages.

I would not necessarily use the LT1073 shown.

[james_s]

Building a Royer converter to get anywhere up to several kV is very easy, I have done so on numerous occasions. If you want a regulated output you can monitor the output through a voltage divider and use that as feedback to control the input voltage to the converter. At only 0.1mA you might even get away using a linear shunt regulator across the output for regulation.

[voltsandjolts]

Does anyone have any hints and tips for measuring millivolt ripple on a 1500VDC psu?

How about using a 1uF PP film cap int (HV rated) into a resistor load to ground (with bidirectional zener across it for some scope protection) - measure noise across R.
Should get down below 100Hz with that.
Or is that a dumb idea?

[voltsandjolts]

...and a nice answer comes once again from the late great Jim Williams in Linear App Note 118. RIP.

[Ian.M]

I'm a *tiny* bit dubious of the Jim Williams circuit - if the HT supply you are characterising suffers a flashover,  with nothing to limit the transient current out of the coupling cap except its ESR and the cable impedance, the transient current through one of the MUR-110s may be rather higher than you'd like, so that 50R input its direct coupled to had better be fairly robust.   Of course, in his day, the testgear he used typically was that robust.  Modern low voltage highly integrated gear typically isn't, unless the designer put a *LOT* of effort into input protection.

Does anyone have any hints and tips for measuring millivolt ripple on a 1500VDC psu?

How about using a 1uF PP film cap int (HV rated) into a resistor load to ground (with bidirectional zener across it for some scope protection) - measure noise across R.
Should get down below 100Hz with that.
Or is that a dumb idea?

Replace the resistor load with a 2:1 divider, with the clamping at the tap for the scope, compensated for the loading of the scope probe,  or maybe with the tap buffered by a video OPAMP driving a 50R cable terminated at the scope and I think you'd be onto something.   The top resistor in the divider, with only a few pF of compensation across it limits the max clamping current.    Also, there was a topic that got bumped today:
https://www.eevblog.com/forum/beginners/multimeter-input-protection-what-are-these-bjts-doing/msg719790/#msg719790 that's worth a look as the transistor clamping described can have much lower leakage and junction capacitance than a pair of Zeners.  Alternative very low capacitance and leakage clamping if the scope input can stand >100V would be a NE-2 neon bulb.

  Its behaviour could be characterised with a cheap chinese coin cell stack in a plastic pipe^* to get a low noise approx 1KV DC bias voltage, and driving the bottom end of that from a signal generator. 

* A 1KV CR2016 stack is only 334 cells and just over half a meter tall.  Split it in two and it can easily be fitted in two 30cm tubes bonded side to side, with a conducting crossbar to link them and retain the cells at one end and two contact springs retained by pins at the other.   Insertion is easy if you stack cells 16 at a time and wrap the edges with just under a single turn of 25mm sellotape. then insert them with a slotted plastic plug that can be removed (and replaced again) after complete assembly preventing contact with the conductive crossbar, so you don't have to work live on a 1KV DC source capable of over 200mA short circuit current.

[Wolfgang]

Hi,

I think the comment about the maximum flashover current is valid. The pulse could fry your protection diodes (35A max.) and the next thing to suffer then is your scope input.

What I would do is a 50Ohm series resistor (HV type) at the input, a neon bulb or spark arrester to ground,
another series 50Ohm resistor (normal one is OK) and the antiparallel diodes to ground afterwards.

That should limit the voltage to a 50Ohm scope to something survivable.

[David Hess]

I think you are underestimating how tough a silicon diode is.  If you do manage to destroy the junction short of disintegration, it turns into a short fulfilling it protection function anyway.  The 35 amp maximum specification of the MUR-110 is for a half cycle of 60 Hz and reflects the maximum rise in junction temperature.

If you are still worried, then the same protection circuit can be used with a 1 megohm input but 50 ohm input instruments have lower noise.

[Wolfgang]

... I might be pessemistic, but the 35A for the diodes are the only pulse power spec that is available in the datasheet. I used this scheme once (with different diodes and a larger cap to get a lower frequency limit) and they did not survive (fortunately the failed short so the scope was undamaged).

If the diodes would fail open and you fail to take notice of this, your next dead thing is your scope.

I dont know what exactly killed them, it could also be that due to a very high dV/dt the conducting area was just a small portion of the die, making this one punch thru.

[vindoline]

I got a photomultiplier tube that takes a negative 1000V to drive and takes about 80 uA of current.
I need to provide this from a 12V supply.

So far i've been doing that with a push-pull converter out of another project, fed over an isolated 12V->12V DCDC brick, which is an extreme overkill, bulky and pricey.
I've looked around for pre-made DCDC power modules that can produce -1KV, and while there are some from XP Power  and Traco, dear god are they expensive.

The question now is - how to make one?
It has to be regulated, with low ripple, and i have a strong preference for something that won't require winding inductors/transformers.

I'm not sure even what topology to go for, and google was surprisingly unhelpful. There are some variants of a boost converter that can do negative voltages (buck-boost?), and it appears to be possible to attach a voltage multiplier to one.
Would that work in practice?
Or is there a more straightforward solution?

I've had exactly this problem. By far the simplest and cheapest option is to use a CCFL inverter intended for automotive headlamp use. On eBay search for "angel eyes inverter." They go for about $5 each. Use a simple LM317 circuit to provide a variable input from 0-12v and add a voltage double/rectifier on the output. This works great for PMT supplies.

User G8RPI over at the gamma spectroscopy forums (https://groups.io/g/GammaSpectroscopy) has a great article on exactly how to do this. Posted with his permission.