High voltage constant (low) current load

[inse]

I got myself a Electrophoresis power supply capable of 3kV output voltage, which I want to use as a HV source for experiments as it allows control of voltage, current and power.
The unit has been made by the former Swedish brand Pharmacia in the early 80‘s.
No documentation available so far…
It is working fine but has a feature that is limiting its use: the output requires a minimal load of ca 1mA to activate.
I don’t know whether it’s a kind of safety feature to avoid high voltage on exposed electrodes or just a technical limitation.
The behavior could be fixed by adding a constant current load of 1mA to the output, with only 3W of max. power dissipation.
The voltage range is in principle from 10 to 3000V (I plan to use it above 100V), the circuit would have to cover a huge range.
As MOSFETs are only available to max 1.6kV breakdown voltage, my idea was to make a cascode circuit of four 1kV MOSFETs and a simple current regulator with an OPAMP and a sense resistor.
Unfortunately I don’t have experience with cascode arrangement and wonder if this idea is reasonable.
Can anyone give me advice on HV designs?

[moffy]

This might suffice: https://au.mouser.com/datasheet/2/240/media-3321478.pdf

[Vovk_Z]

3W and 1 mA? Why don't you want to test it with just a simple passive resistor load?
What do you want from it so you need active electronic load?
Why 6x 510k 2W resistors in series are not enough to load it?

[inse]

I need to sink 1mA also at lower voltages, the power supply is adjustable from ~10V upwards.
The power supply won’t start even at lower voltage settings unless a minimal current is drawn.

[inse]

@moffy:hefty price tag

[moffy]

@moffy:hefty price tag

Not for what you want to do.

[BrokenYugo]

That MOSFET is way cheaper than my first thought of doing it with tubes.

If you want cheap get reverse engineering and hack it to work how you want it to work.

[inse]

Thanks for the suggestion.
Above the price of the MOSFET, do you think I will be able to get the regulation stable over such a huge range?

[moffy]

Thanks for the suggestion.
Above the price of the MOSFET, do you think I will be able to get the regulation stable over such a huge range?

There is no reason not to as long as the current regulation loop is moderately fast, and for current that is quite easy.

[Slh]

Heads up, that MOSFET is only rated to around 2000V at DC . Look at the SOA curves. The curves don't go as low as 1mA so you might get away with it but you won't know until you test it.

[moffy]

Heads up, that MOSFET is only rated to around 2000V at DC . Look at the SOA curves. The curves don't go as low as 1mA so you might get away with it but you won't know until you test it.

Good point.

[moffy]

Maybe a cascoded pair of SiC, JFETs they are certainly cheaper: https://au.mouser.com/datasheet/2/412/UF3N170400B7S_Data_Sheet-3177149.pdf
but the biasing network would be interesting.

[inse]

„The Art of Electronics“ also mentions the cascode for high voltage trick, but no further details on the bias network.
Have a happy new year, everyone!

[inse]

Heads up, that MOSFET is only rated to around 2000V at DC . Look at the SOA curves. The curves don't go as low as 1mA so you might get away with it but you won't know until you test it.

The curve drops vertically, I don’t think it would recover below 10mA - there must another reason for that limit.

[moffy]

„The Art of Electronics“ also mentions the cascode for high voltage trick, but no further details on the bias network.
Have a happy new year, everyone!

I did a similar thing for a HeNe laser gyroscope a number of years ago. The final laser current was controlled by a 200V JFET in a constant current feedback loop and the voltage had to be trimmed by placing a series resistor with the JFET drain to limit the maximum voltage. To aid this I designed and built a test jig with about 4 of the JFETs cascoded so that the optimum current for the laser could be set, you wanted max intensity without multi moding. The trick is to share the voltage equally across the JFETs, even in transient conditions.
Included is a rough template, but should not be used as is, it is just an illustration to be improved upon. You would probably want to consider some sort of gate clamping.

P.S. A happy new year to you also.

[inse]

Interesting application- how did you design the D/S snubber network?

[moffy]

Interesting application- how did you design the D/S snubber network?

Not really a design just an indication, the capacitance is for sharing transients equally between the FETs, the resistance just to limit transient current.

[ArdWar]

Maybe a cascoded pair of SiC, JFETs they are certainly cheaper: https://au.mouser.com/datasheet/2/412/UF3N170400B7S_Data_Sheet-3177149.pdf
but the biasing network would be interesting.

You can control N-JFET with the same opamp-based N-MOS control scheme, just give the opamp more negative supply headroom. The control "sign" is still the same after all.

[inse]

I will keep you informed once the components arrive…

[David Hess]

„The Art of Electronics“ also mentions the cascode for high voltage trick, but no further details on the bias network.

Usually it comes down to series connected resistors to control the voltage at each gate or base.  A high voltage zener diode may be placed across each resistor to limit the maximum voltage of each stage under dynamic conditions.  Then the bottom stage is configured to control the current.

[PCB.Wiz]

Heads up, that MOSFET is only rated to around 2000V at DC . Look at the SOA curves. The curves don't go as low as 1mA so you might get away with it but you won't know until you test it.

The curve drops vertically, I don’t think it would recover below 10mA - there must another reason for that limit.

It also varies with HV mosfet type and vendor.
I see the Mouser  G2R1000MT33J MOSFET 3300V  has a SOAR slope, and allows ~ 1A @ 80V and 100mA @ 800V, so should be fine for 1mA at 3.3kV

It is working fine but has a feature that is limiting its use: the output requires a minimal load of ca 1mA to activate.

Did you measure that 1mA across voltages ?  Maybe you can modify that limit ? (possibly they have  poor regulation at very light loads, so added that to stay in spec ? )

[inse]

Maybe I will have to do a deeper analysis of the circuit.
So far I found out that the HV drive is shut down in case of open circuit.
A bias voltage of ca. 12V is present across the output and once a minimum current is drawn, the driver starts up.
I haven’t figured out what the actual converter topology is, the transformer is driven by two transistors, possibly in half bridge configuration.
But I don’t know what topologies the controller TDA1060 actually supports and whether it needs a minimum load for stability or if there is a safety feature to avoid high voltage on open connections.

[moffy]

It also varies with HV mosfet type and vendor.
I see the Mouser  G2R1000MT33J MOSFET 3300V  has a SOAR slope, and allows ~ 1A @ 80V and 100mA @ 800V, so should be fine for 1mA at 3.3kV

The data sheet I have shows those points as on the 100ms pulse width curve, is that considered the same as a DC load? I thought they were different.
https://genesicsemi.com/sic-mosfet/G2R1000MT33J/G2R1000MT33J.pdf

[inse]

Shouldn’t the difference between DC and pulsed operation only be thermal constraints?

[moffy]

Shouldn’t the difference between DC and pulsed operation only be thermal constraints?

If you look at the SOA curve for: https://au.mouser.com/datasheet/2/240/media-3321478.pdf
The max DC voltage is below the max voltage for the pulsed 100ms, it's probably avalanche related.