Affordable 100vDC Reference for Met Experiments

[enut11]

I needed a stable 100vDC supply for calibrating various meters and general Met experiments.

What follows is not a construction project but rather a proof-of-concept design.

Good HV supplies are very expensive, even used ones. I wanted to see what was possible using cheap readily available parts.

Initially, I did not set any performance goals, instead just going along on a discovery journey.

The first candidate was the TL783 high-voltage adjustable regulator. In TO220 form it is capable of outputting 1.25v-125v
My requirement of 100v at less than 10mA would be easy going for this chip that can handle up to 700mA within its thermal rating limits.

https://www.ti.com/lit/ds/symlink/tl783.pdf?ts=1600909697623&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FTL783

All I had to do was breadboard the basic circuit - application 9.2 in the TI spec sheet - and provide a 150v DC unregulated from 110vAC. The only thing to watch is the output voltage setting resistor R2 can get quite hot so a 5 watt unit is recommended.

I must stress here that the circuit is to be run from an isolated AC source. This means a proper transformer with distinct primary and secondary windings. I made the mistake of assuming a 240vAC to 110vAC step-down 'transformer' to be suitable. Most of these available online are actually auto-transformers - ie not electrically isolated.
More to come...
enut11

[enut11]

Well, the mock-up turned out to be 110vDC but as I was just testing for stability, the output voltage was not critical and could be made adjustable later.

The plot shows voltage from a cold start with the supply running at a couple of milliamps. Ambient temperature was stable around 25-27C.

It took about 40 min to stabilise, thereafter dropping about 15mV over the next 40min (0.014%).

A pleasing result for such a simple circuit but was it good enough? Probably not.

Concerns: the large regulator voltage drop (>40 volts) and the presence of the on-chip TL783 reference meant that it could have a significant temperature co-efficient.

[enut11]

The high input-output voltage drop was solved with a cascaded regulator setup.

Instead of using another TL783 as a pre-regulator, a friend put me onto a TI application sheet LB-47 which surprisingly uses an LM317 as a precision HV regulator.

https://www.ti.com/lit/an/snoa648/snoa648.pdf

The secret to improved performance comes from an external compensated precision zener (LM329) to stabilise the LM317 output.

According to TI
"improves temperature stability since the LM329B has a guaranteed TC of +-20PPM/degC and improves regulation because more loop gain is available from the LM317"

[enut11]

While I was waiting for the LM329 chip to arrive I tested an ordinary 6.8v zener for stability at 1mA. The plot shows a fairly bumpy output with variations of up to 1 mV.

Ambient temperature was relatively stable at around 22C.

[enut11]

As expected, the LM329 compensated zener was much better with higher frequency 'noise' showing up in the mico-volt region.

More to come...
enut11

[EmmanuelFaure]

Look at that AN from TI, "Stacking the REF50xx for High-Voltage References" :
https://www.ti.com/lit/an/sbaa203/sbaa203.pdf?ts=1600954791619

[enut11]

Thank you @EmmanuelFaure. I have already looked at series Vrefs to increase voltage but have temporarily put the idea aside due to cost. 

However, the HV regulator described above would make a great power source for stacked Vrefs.
enut11

[enut11]

The next step in this journey was to build the LM317 based HV regulator circuit. This needed more parts so it ended up on veroboard.
The interesting part of this design is how it allows a 40v device to operate at hundreds of volts. LB-47 describes how this is done.

Refer Reply #2 for circuit LB-47 page 2.
Again, the voltage setting resistor (R7) has to be high wattage as it is across the 100V output.  Initially I used three 3.3K 3 watt resistors in series plus a 1K 2 watt WW pot for trimming. The tempco of this resistor string has to be low for a stable output.

My requirement was for 100v at less than 10mA, mostly less than 1mA, so a moderate heatsink was used. In fact, the 240/110v transformer created most of the heat and a decision was made to mount this in a separate box.

For a 100vDC output from an unregulated 150vDC I set up each regulator to drop about 25vDC.

The XL snapshot shows the performance of the cascaded regulators (TL783+LM317) at idle current. Ambient temperature was stable at 22-23C.
Again, it takes about 40 min for the voltage to stabilise but this time the variation, once the curve flattens, is around 5mV over the next hour, an improvement over the single regulator version.

For my requirements, this is now a viable design as short term stability is likely to be much better. The second capture shows a blow-up of the last 40 min of the test.
enut11

[enut11]

I have now checked the short term stability of the cascaded regulators 100vDC source and it is approx 100uV drift per minute (0.0001%) after a couple of hours warmup at ambient 22C.

I believe I know the likely cause of this drift. There are two output-setting resistors that dissipate approx 2 watts.

If I relocate these outside the box this may allow even better short-term stability.
enut11