I've spent the last few weeks looking into how to create a digitally controlled high voltage variable resistor and I came to the conclusion of MCU>DAC>common emitter amplifier>source follower, and that works for some of my intended designs but now I need to create a high voltage power supply and I want it to have really good transient response.
The voltage range will be 0-600v and the current range will be 0-100ma.
I want to create as close to an ideal voltage source as I can. This particular power supply will be used for audio purposes. I will probably use a CRC filter for the initial filtering before the regulator.

I wanted to use a 32bit ADC/DAC feedback loop to feed the pass transistor for this design but I read that an analog feedback loop will have way better transient response than a digital feedback loop. I don't know if this includes a potential 32bit feedback resolution but such a system will be expensive to create in either case so it's worth looking into the analog loop.

So now I'm faced with the possibility of using an analog feedback loop with a digitally controlled voltage reference.
I need advice on what the best approach is for best transient response.

 

It's impossible to resolve 32 bits in an analog system.
That is 4nV out of 1V, well below the thermal noise level for any normal circuit.

What transient response time do you need?

 

What transient response time do you need?

I'm going to assume that since it will be used for audio the transients should not be more than 20khz which is a response time of 0.05ms.

 

Nevermind, I just realized the output bypass capacitor of the regulator should take care of any transients missed by the regulator. Right?

 

The voltage range will be 0-600v and the current range will be 0-100ma.
I want to create as close to an ideal voltage source as I can. This particular power supply will be used for audio purposes

You probably didn't mean this since you say it is for an audio amp. But even at 100 ma if you wanted 5 volts, 55 or more watts would be dissipated in your transistor.
It can be really hard to close the loop in software. How fast is your mico.
As the transients go.... It depends on how big the filter cap is. 600 volt caps don't come cheap.
What kind of amplifier are you building that wants 600 volts at .1 amps?

 

You probably didn't mean this since you say it is for an audio amp. But even at 100 ma if you wanted 5 volts, 55 or more watts would be dissipated in your transistor.

Dissipation is pretty irrelevant to me. I can just use a bigger heat sink or watercooling and I have SIC FETs that can handle ridiculous amounts of power. Not that I would often expect to use 5v@100ma.

It can be really hard to close the loop in software. How fast is your mico.

The micro is not yet purchased, I'm waiting to see what I need first, the plan is evolving as I learn more.

As the transients go.... It depends on how big the filter cap is. 600 volt caps don't come cheap.

I'm going to stack caps in series to increase the voltage ratings or use film caps. By filter cap I assume you mean the bypass cap on the output of the regulator?

What kind of amplifier are you building that wants 600 volts at .1 amps?

Not so much an amp as much as a prototyping power supply. I doing 600v for headroom and not many tubes use more than 100ma.


So I was looking into series regulator vs shunt regulator and I noticed that shunt regulators have super low output impedances which get even lower as you shunt more current. There are many claims that when people increase the current of their CCS fed shunt regs that the sound quality is improved due to this. Plus I read that shunt regulators work well with constant current loads which is also another plus for me.

So I think instead of a series reg I will go for a shunt reg.
So this is the basic design of a shunt reg

Or one of the heralded "super regulators" that get a lot of praise for their "sound" but they are way above what I current can hope to design.

I'm not sure what is going on in the above picture, I'm not quite there yet. I'm gunna take a jab and say J2 changes its resistance based on the output voltage of the regulator which then changes the resistance of the voltage divider below it which biases M1 for compensation.
The designer claims the M2 is a current "limiter" I'm not sure if that means it limits maximum current and doesn't maintain a constant current. I'm not sure how you are supposed to get a constant voltage if you don't have a constant current to shunt.

I have some of these http://www.mouser.com/ds/2/196/Infineon-IJW120R100T1-DS-v02_00-en-776538.pdf SIC Jfets lying around. Jfets are considered the most quiet transistor and since my SIC versions are high power/voltage rated I would like to see if I can implement them into the shunt design.

I'll probably have the CCS run at some redundantly high current and then water-cool it just so I can get a really low output impedance.

So I want to digitally control the voltage of this regulator and get a Jfet to act as a CCS and a shunt element. I'm not sure if or how I can implement a jfet as a CCS in series with an RC filter and the load, I've learned a lot but I still suck at transistor design and Jfets are even more of a mystery to me.

 

I though about ditching the Jfet idea until I realized SIC mosfets don't have a P variant. I can't seem to get an analog jfet CCS to work so I guess I will just use a digital control loop since I only need to control DC level current unless someone can think of an analog solution.

This is what I have so far except for the fact I want the mosfets to be jfets.


As for the shunting device, the feedback will need to react to AC level voltages so I might have to use an analog loop but I read that they are noisy, is this a function of the analog circuity or something else? Would a digital control loop introduce noise?
If I did use a digital control loop then I would need it to have reaction time of 50 microseconds in order to handle the AC voltages, is this even possible?