Simplistic Mosfet HV Shunt Regs

[nigel pearson]

Originally Posted by Sergelisses

Hi Salas

Is it a good idea to replace a 5AR4 by a HV Shunt reg ?

Best regards
Serge

Let me know if you do and how you liked it .

[triode_al]

Hi,
I built a Salas shunt for my preamp. The preamp draws between 30 mA, design is for 250V.

However I get some strange effects with the SSHV1.
I get less and less current out of it.

The current source I use is with 2SK585, slightly higher knee (5V/about1A) than the stock IRF6910 (4V/about1A).
My LEDS: an array of 5 red LEDS on 68k giving 7,9 volt. The source voltage is around 3 volts (across the bias resistor) and this starts to drop as the current demand increases.
With a test on a lab supply set I got 58mA (rbias=47ohm) out of it, with 270V in, but once installed in the preamp the current source drops in value.
I get

1. 39 mA with 74 ohm raw 298V
2. 41 mA with 68 ohm; the input starts to drop. 297V
3. 40 mA with 51 ohm. in = 291V
4. 30 mA with 42 ohm. in=249V

Pretty counterintuitive!
At the same time the raw input starts to drop. This means the supply is loaded (but the current drops ).

I reduced my output from initial 218V to 200, now it is barely 185V when the regulation starts.

My power supply: 300V AC (rs=50 ohm) / 5R4GY / L=1,75, 200ohm / C= 10 uF / L= 7 H, 300 ohm / 30 uF.
So it is choke loaded. The output starts to sag from 300V DC down to lower than 250V. I tried a solid state drop-in rectifier, same effects, but drop is less. The LCLC regulates well.
There is no oscillation visible on my scope.

Is the LED array to small? Do these effects then happen?

I first had an OD3 and a EL84 shunt on 250V; this gave a higher raw voltage in: 325 volts DC could be obtained. (But the Salas SSHV1 is so much more silent, I can post the pictures).

Has anybody noted such strange behaviour (not enough current and dropping) and detoriation (sagging output over time)? Every on-off cycle reduces the current capability???? The voltage of the knee where the shunt starts seems to drop every time. [Maybe I should add the 12Vzener to guard the shunt IRFIB5N65A ...]

My guess: the LEDS are too weak. I already added one extra to go from 4 to 5 pieces . Add one more ?? or replace one red (1,58V@4mA) with yellow (2V@4mA).

Hope this is clear
albert

[AndrewT]

There are two parts to the supply.
The CCS and the Shunt regulator.

You can build and test the CCS without any of the shunt components in place.

Build the CCS.
Apply a low safe working voltage. Maybe 20Vdc.
Attach a 1r0 power resistor from the CCS output to Power Ground.
Measure the voltage across the 1r0 dummy load resistor.
Adjust your CCS current to the value you need.

Remove the temporary low voltage supply.
Add into the build all the shunt components.
Replace or parallel the big voltage dropper in the sense circuit to reduce the regulated voltage to ~20Vdc. Now apply a temporary low voltage to the CCS input and see what current is flowing in the CCS. Measure the regulated output voltage. Determine the Vdrop across the CCS that allows correct operation of the CCS and the shunt regulator. This is the absolute minimum DC voltage that can be allowed during high voltage operation for the CCS+Shunt to work properly.
Now take off the temporary parallel resistor jumper to reset the sensor circuit to the high voltage version.

Both parts of the regulator have been tested. It is now relatively safe to apply high voltage to the CCS input. Measure the CCS Vdrop. It must be bigger than that minimum value you determined during the LV testing.
It must remain higher than that minimum Vdrop even when mains voltage has dropped to the minimum supply voltage that your electricity supplier has given you.

[triode_al]

I added a LED, and this solved the issue.
Now I can get:

• 57 mA with 68 ohms, V in is 287V
• 64 mA with 56 ohms, V in is 287V
• 240V out, with 21 mA shunt current (sub Herz not nice)
• but 210 is better, with 30 mA shunt draw

looked at the calculator again and this confirms my findings: at a higher current the voltage over the bias resistor drops (of course ).

The sub Herz flutter remains a problem; the shunt needs quite some current to cope with it and get a very smooth output line.
With a solid state rectifier I see some residue of below 1 mV pp of rectifier peaks; with the tube rectifier I did not see it.
So my next step: revert to the tube.
albert

[triode_al]

Originally Posted by AndrewT

There are two parts to the supply.
The CCS and the Shunt regulator.

You can build and test the CCS without any of the shunt components in place with a low voltage supply & 1 ohm shunt & adapted LED resistor.

Yes this s good idea. Much better than my tweaking method (in situ, replacing parts . . .)
albert

[Sergelisses]

Originally Posted by nigel pearson

Let me know if you do and how you liked it .

I have never worked with tubes so i am not sure that it's a good idea

Serge

[nigel pearson]

Someone will say I am sure . I have a design to use near my computer . Keeping it neat and tidy helps . I probably will not get time to do it .

[Salas]

Originally Posted by triode_al

I added a LED, and this solved the issue.
Now I can get:

• 57 mA with 68 ohms, V in is 287V
• 64 mA with 56 ohms, V in is 287V
• 240V out, with 21 mA shunt current (sub Herz not nice)
• but 210 is better, with 30 mA shunt draw

looked at the calculator again and this confirms my findings: at a higher current the voltage over the bias resistor drops (of course ).

The sub Herz flutter remains a problem; the shunt needs quite some current to cope with it and get a very smooth output line.
With a solid state rectifier I see some residue of below 1 mV pp of rectifier peaks; with the tube rectifier I did not see it.
So my next step: revert to the tube.
albert

If not having an ID/VGS graph of an alternative CCS MOSFET component to readily refer to, it certainly takes some experimentation for the level of the LEDs reference. But you saw that. You can also filter the main voltage reference in the shunt part with more uF to depress 1/f better but it will take a zener to protect the Jfet current reference from sustained overvoltage during charge up (see SSHV2). Transcoductance and ID curve of chosen MOSFETS plays a role on how well they reject with given spare mA at 1/f also. In general, the hotter the better (within practical bounds).

[triode_al]

Yep. I will add either a TIP50 (Hfe 150, 400V max) or an IRFIB5N65A (650V mosfet) as gyrator: with a 10k resistor and a 10-47 uF capacitor. This regulates sub-Herz a bit better than a 100H eChoke, I noted.
And maybe add a 220 uF electrolytic.
The 2SJ585 has embedded zeners, that is why I choose this one over the 2SJ449 I also have, both being 250V. So increasing to say 4 uF would be possible without extra zener.

You can also filter the main voltage reference in the shunt part with more uF to depress 1/f better

You mean increasing C2? (I have 1,2 uF), I have a 250V type (Philips MKP).
SSHV1.png

Just an idea - Could it be that with this type running against its max that I get this noise? Because the shunt seems to have behaved better initially.

Thanks, albert

[Salas]

Not sure if stressing the dielectric can do such. I use 400V there. More inclined to that the different Mosfets can be wilder at 1/f and you need more C2 value. But strap a 12V Zener from C2&R9 node to +Vo rail (cathode up) along upping the capacitance not to zap the JFET.