Simplistic MosFET HV Shunt Regs

diyAudio Chief Moderator
Joined 2002
Paid Member
Stixx said:
The heatsink on the IRF840 remained stone cold during the 15 mins of my testing (which again surprised me since I expected more heat on the second Mosfet...).

That shows that your circuit was getting it almost all and did not leave any excess reserve for the 840 to drive to earth. The more the reserve, the hotter the 840. Maybe that is why you could not get more voltage easily.
 

iko

Ex-Moderator
Joined 2008
OK, salas, please don't get upset, but IMHO the HV regulator design should be rethought. I put a good number of hours and I have a cemetery of parts. Some things I observed: the load current imposes some changes in the rest of the regulator; the first resistor value (in series with the source of the irfp9240) needs to be recalculated. In that part of the circuit the biasing of the irfp9240 needs to be quite exact, when a larger load current is in place. Three 2V LEDs were not quiet right, and I didn't have any 1.7V. Another thing I noticed was the sensitivity of the circuit to the value of the resistor on the emitter of BC560B. The circuit oscillates easily and the parts fry quickly. Even when in simulation it worked well I ended up with fried parts. As far as I am concerned I will take a step back now.

I don't mean to be negative, these are just some things I observed.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
No you are not being negative. Those things are endemic in to this HV shunt practice. The LEDs and their Vdrop are directly proportional to the IRFP VGS and R1. I have written about calculations right from the start. That is why I have given 2 examples for a good range and certain values and showing mA range setting. The VGS will change with different Vin and current settings or load. Live probing can kill small parts due to inductance, also lead inductance from resistors can do it. The thing is that it plays for months now, also for Cygnus, and even Kofi did it recently. Main thing for making it simpler and less dangerous, is having a PCB with good tight layout I guess. Also a 47R base stopper for the MPSA or equivalent HV BJT can help. I had no particular problems so I don't know why change it, given the sound it delivers. But its not a straightforward thing to make and test for sure.
 

iko

Ex-Moderator
Joined 2008
I have one more 9240 left, and one irfpc60. I will try one more time later tonight. I still have a couple of mpsa92s. I got no bc560b, but can use 2n4402s there. I also don't have any 1.7V LEDs.

salas, perhaps you could help out a bit with an opinion. Considering these parts, what values would you recommend for R1 and anything else that would need to change, for a load of 125mA, 255V in, 245V out?
 
diyAudio Chief Moderator
Joined 2002
Paid Member
P.S. If in your shoes, I would base stop all small BJTs with 47R, use a 22R 2W R1, 1k shunt gate stopper, put water in my wine for B+ temporarily given my trafo so to keep at least 50V Vin-Vout diff, connect, power on, and take good distance with my eyes covered. I do such stuff, I don't know about you.
 

iko

Ex-Moderator
Joined 2008
Salas said:
P.S. If in your shoes, I would base stop all small BJTs with 47R, use a 22R 2W R1, 1k shunt gate stopper, put water in my wine for B+ temporarily given my trafo so to keep at least 50V Vin-Vout diff, connect, power on, and take good distance with my eyes covered. I do such stuff, I don't know about you.

Yes, done this for the last couple of days, how do you think I've a cemetery of parts now. I cannot say it's totally void of fun :D
 
diyAudio Chief Moderator
Joined 2002
Paid Member
It is compensated at the output with the Zobel and we must avoid further capacitors later. Which is one of the main benefits for listening better. They will degrade the transient performance and if there is enough parasitic inductance in them, or in the cabling leading to them, oscillation is possible. The procedure is, we use a dummy load for about half the CCS set consumption, already the trimmer is at half, from there, power on, we set the Vout after 15 minutes. Power down, connection with short twisted leads to audio circuit. Power on, listen. A base stopper of 47R for the MPSA92 is helping it, making it less sensitive to mishaps.
 

iko

Ex-Moderator
Joined 2008
Here's my update, perhaps it will help someone. Found 1.8V LEDs and used them, R1=22R, shunt mosfet tried two types, irfpc60 and stw20nm60. Both high voltage, the stw20nm60 is a quality item, with lower gate charge, low Rdson (not that it matters). Instead of bc560b I tried bd140, and 2n4402. As gate resistor I used 1k for the shunt mosfet, 50R for the mpsa92 (mosfet driver), and 560R for the 9240. This actually behaved better on the bench supply. Got the input voltage up to about 190V, and it would regulate, nice oscilloscope trace, steady, for about 20 seconds. Then it would start oscillating. The layout was carefully done.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
OK, that's progress. Now don't use the oscilloscope probes again. Just the dummy load. Leave it on for 15 mins. Set the Vout then. If the output voltage does not jump around, its probably OK. Power it down. Wait for 2 minutes. Make sure that there are no capacitors other than a possible small EMI bypass on the audio circuit's B+. Power on. See if it works as a whole.
 

iko

Ex-Moderator
Joined 2008
Unfortunately, the last experiment I did last night was to put the shunt in the real circuit (amp), to see if it would oscillate there. The scope probes were connected, as were the DMM probes. After a few seconds of oscillation the output went to 0.6V and stayed there. I think it had another heart attack. It was too late already and I was too sick of it so I just turned the power off. There was no hammer handy to give it a few smacks! :smash: