Hello Salas,
I have completed and tested the build of Simplistic Mosfet HV Shunt Regs
After the regulator is turned on and adjusting the VR, I got about 252 volts that slowly creeps to about 269 volts in an hour's time and settles down within +-0.5 volt. The voltage across the 33R is 1.803 and I assume that it's working perfectly. Am I right?
I also roughly measure the AC output ripple that's less than 0.5 mV using Fluke 189 but it's fluctating very rapidly (I know this is crude). Is the circuit oscillating? Should I use back IRFP9240 as the PMOS?
Thanks in advance for any response.
All the best,
housing
Shows 54mA constant current, the JFET comes to thermal stability and settles IDSS so the Vref comes up with time. Looks normal.
You can't tell its oscillating if there is no certain frequency caught. Is there a dummy load on its output?
Thanks Salas.
Yes, it's loaded with 5 pieces of 47K 20W resistors in parallel, 9.73K by measurement.
Why drift with the mains? It has its own voltage reference, plus its CCS ''firewall''. The drift is thermal, then stops. No.2 is an all time classic.
Yes, it's thermal drift. I'll try to improve it soon by adding mini heatsink to the 2SK170 and changing the VR to 15 ppm vishay trimpot. I've used 5 ppm S106K (?) 135K (2 x 200K S104K + 2 x 70K S102K) resistor as voltage reference.
Best is a low ppm resistor instead of trimpot if the application will be fixed after measuring the set trimpot value. The JFET will always slide a bit more IDSS with time until its equilibrium, its inherent to translate enough in a large voltage Norton source, but a mini sink on it may help it do it less. Shoot for the final voltage value as nominal. If it was a Zener it would be much more noisy and its filter cap would be a large electrolytic due to low impedance. Both detrimental to subjective tests. Don't use carbon composition resistors as reference under the JFET, even if they are nice in signal even if Riken and the like. They are large ppm and noisy relatively. You did well to use low ppm.
According to Broskie, shunt regulators will drift with the mains. In other words if you plug in a 1500W heater on the same circuit as the shunt reg the reg wil still kill ripple but the output voltage will drop.
If your shunt reg doesn't do this I'd be very surprised but enthused because I have a hybrid amp with a dc servo that is very touchy to any changes in B+.
Can you measure the output of your reg by turning on a heater and monitoring the B+ voltage to see if it moves? Now your mains may be such that it doesn't shift when turning on a high wattage device plugged into the same circuit so you would need to check that to.
Hi Regal,
This is defintiely thermal drift. I have tested it with a swinging fan blowing against the PCB and the voltage dropped 5 volts almost instantaneously.
Hi Regal,
This is defintiely thermal drift. I have tested it with a fan and the voltage dropped 5 volts almost instantaneously.
Big fan or little fan? Could be your mains dropped when the fan turned on?
Still interested in hearing reports on if the voltage changes with the mains.
Sounds to me that you need a bigger heatsink if blowing a little air on it causes that much change in voltage.
Last edited:
Mains dropped when the fan turned on? That's impossible. I have a 100Amp 3 phase 220V supply. Moreover the drop in the HV shunt regulator then remains at about 265 volts.
That doesn't concern the heatsink. It's about thermal equilibrium. It's the Jfet's Id that's sensitive to temperature change and it's inherent in the circuit.
I don't know where you live but my mains drop when I plug in about anything over 500 watts.
Measure your AC at the wall plug, you will be surprised how much your house voltage fluctuates, sometimes it is just the power company, during the day my outlets wander from 110V to 120V's . In California I heard it is even worse. Just interested in how this amp responds to mains fluctuations.
As far as the Jfet is there anyway to mitigate this ? Like glue it to a heatsink ?
I'm
Hong Kong. Yes, the main wanders by one or two volts but it won't sag even when I turn on the 3 hp air-conditioner during summer.
I've tried heatsinking the Jfet but to no avail. With heatsink and a low ppm VR, it just stabilize at about 2 volts lower than before, i.e. it stabilizes when thermal equilibrium is reached. So I think the VR controlling the Id of the jfet (and hence the output HV) should be retained (not to be replaced by a fixed resistor) as it may need to be further adjusted when the whole thing is placed in a closed working environment, i.e. inside a chassis.
Last edited:
When you say stabilized you mean by how much, i.e. less than 100mV B+ drift?
Thanks for all your info this is looking like a possibility for me, had planned on a whole new amp to use this reg but I may give it a shot with this touchy hybrid.
Also I'm wondering if quanghao is following this thread? I know Salas said it really wasn't his place to help troubleshoot quanghao's boards.
The Jfet will drift some idss, then it stops. A mini sink on it will set it somewhat differently for final Idss and will make it a little less sensitive to environment but that is all. When we translate Idss and magnify to 265V, its normal that the drift will be enough Volt until it settles. Then it does not wander. I stated that someone must set it after 15mins and in final environment again, very early in the thread. No practical problem with valves since they don't get upset starting 5% B+ lower, nor they will sing earlier anyway. Keep the trimmer for final assembly and substitute with a low ppm resistor in the end. Its more reliable in the long run, though not a sure failure is expected.
Do you have the link he discusses that? Is he referring to super regs (i.e. CCS feeding loop gain shunt) or just ballasted shunt elements like Zeners and VR tubes?
Actually there is. My Tube Hybrid headphone amp has a DC offset servo that cancels out the DC from the SS output buffer. Now this SS buffer is coupled to the tube stage via caps. The catch is that a very slow B+ wander (like .01 HZ) is basically a form of AC and passes thru the caps but the servo sees it as DC and can't keep up so I end up with mV's DC going to my headphones (I added a DC protection circuit.)
So I had to resort to a tightly SS regulated B+ (the Broskie PS-1), as you can imagine a tightly series reg B+ isn't the best for SQ.
My issue with other B+ powersupplies was the mains oscillating, the other shunt reg followed the oscillation and I had a very clean ripple free B+ that had about 100mV 0.1HZ wandering B+ supplying the tubes. Normally this would never be a problem. But that 100mV makes it thru the coupling cap since its technically AC, end result was DC protection would kick in on the output. Also to complicate issues the buffer has a huge input impedance (design feature to have as little load on the tube as possible.)
It sounds like this shunt regulator doesn't wander with the mains and the B+ drift is thermal. This may be random enough as to not pass thru the coupling caps as AC. I know "use smaller coupling caps", but this hurts SQ somehow.
I'm going to give it a try with this Salas Shunt before I build a new amp. If it works I'll be so thrilled cause this amp is amazing except for the B+ PS.
Thanks for bearing with me on this unusual amp that probably only affects a couple other members.
He recommended the PS-1 for me by personal email. But he touches on it here about halfway down:
Tubes and Religion; New Kits and PCBs
Read it. He basically said that his tube shunt is not self referenced, its like a shunt choke. So it will pass what play the rectified voltage gives, and that is defined by primary's AC. That will be slow, but can perplex an integrator as yours. Very rare to find an op amp servo in tube ccts though.
The other one looks like a depletion HV shielded fixed current Maida to me.
The other one looks like a depletion HV shielded fixed current Maida to me.