B+ regulator with variable reference

Skip this original post and related fumbling completely. Here's the end result: http://groupdiy.com/index.php?topic=51232.msg651854#msg651854








If I want to switch B+ between let's say 200VDC and 150VDC, would this work?



Actual values don't matter. Could be 300VDC and 200VDC. I'm wondering if this concept is stable. Initial reference to the MOSFET is a zener shunt. This is further padded down using a simple voltage divider and the switch selects between the two.

I'm wondering if I've made a kill-switch, or something that might work.

Could this be expanded further by simply cascading more voltage dividers and using a rotary switch to select between the many steps of this "ladder" of references?

Thanks,
Mike

it is correct,

the much higher impedance source to the mosfet with the switch open could give less rejection to high frequency ripple noise (without C6) but the C6 attenuates a lot it the same.

Moreover, the high frequency ripple noise is attenuated by the other RC filters

Great! on to optimisation then.

How do I determine the optimal value for R4? I'm not quite sure what the "load" is. The zeners? What is the load of the MOSFET gate presented here? Knowing these would also help optimise C5/C6 and the voltage divider.

I've penciled in "1mA" for the initial calculations - also for the voltage divider - but might as well do this properly.


[edit]

there's a typo in the above schematic. The MOSFET is of course IRF840.

The values you have....

R4 flows say (220V-200V)/100K = 0.2mA
R7 R8 flow 200V/(22K+82K) = 1.9mA

There is NO place for that extra 1.7mA to come from.

Either R4 must be lower, R7 R8 must be higher, or a Miracle Must Occur.

Also you want a "decent" current in the Zeners. For good regulation and reliability, often 1/10th to 1/2 of the diode power rating. 0.2mA in 40V is 0.008 Watts, or say a 0.08W Zener, which is mighty small. 0.5W is more typical.

ZD1 ZD2 can't flow the same current at 40V and 160V and get similar operating points. If I assume "ZD2" is short-hand for four 40V Zeners, it does make sense.

0.5W diodes, 0.1W target dissipation, gives 2.5mA target current. 0.25W-0.05W range gives 6mA to 1mA range. So if R4 is dropping 20V to 50V (really too small!), R4 would be more like 8K or 20K, not 100K.

MOSFET Gate draws "no" current. So R7 R8 can flow quite small current, much smaller than diode current. When that wall-power sags you may have 1mA in diodes, so pencil 0.1mA in R7 R8. Total 200V/0.1mA = 2Meg, divided 470K and 1.5Meg to give your nominal voltages.

R7 R8 are not really needed if you have free choice of Zeners. Use four 50V Zeners stacked. You can switch the MOSFET Gate among them for 200V 150V 100V 50V, though as the switch changes the voltage can be undetermined. You could short-out one of the Zeners for 200V/150V, though that will increase the drop in R4 and thus the Zener current.

You can go back to the 470K+1.5Meg plan, and un-ground the 1.5Meg to get 200V. C6 will slow the change from one to the other, about 50mS (you might like more C6 for less thud).

With all the high impedances and low currents, I suspect R9 2K is not doing a thing.

PRR said:

Also you want a "decent" current in the Zeners. For good regulation and reliability, often 1/10th to 1/2 of the diode power rating. 0.2mA in 40V is 0.008 Watts, or say a 0.08W Zener, which is mighty small. 0.5W is more typical.

ZD1 ZD2 can't flow the same current at 40V and 160V and get similar operating points. If I assume "ZD2" is short-hand for four 40V Zeners, it does make sense.

0.5W diodes, 0.1W target dissipation, gives 2.5mA target current. 0.25W-0.05W range gives 6mA to 1mA range. So if R4 is dropping 20V to 50V (really too small!), R4 would be more like 8K or 20K, not 100K.

MOSFET Gate draws "no" current. So R7 R8 can flow quite small current, much smaller than diode current. When that wall-power sags you may have 1mA in diodes, so pencil 0.1mA in R7 R8. Total 200V/0.1mA = 2Meg, divided 470K and 1.5Meg to give your nominal voltages.

Click to expand...

Thank you. I didn't quite know what to do with the zeners this far. In my previous utilities they have been sitting "blind", not caring what current goes through them. For this reason I've had to way over-spec their watts, paying more than I needed - even if marginally. Great to know how to work them now. I'll redraw the switch area with another more flexible solution based on your tips.

PRR said:

R7 R8 are not really needed if you have free choice of Zeners. Use four 50V Zeners stacked. You can switch the MOSFET Gate among them for 200V 150V 100V 50V, though as the switch changes the voltage can be undetermined. You could short-out one of the Zeners for 200V/150V, though that will increase the drop in R4 and thus the Zener current.

Click to expand...

This project has a bad history with just zeners and transistors. Original attempt ended with a smoked zener (I know now it received way too much current) and a dead transistor (base-emitter voltage probably went way over maximum during the switch). http://www.groupdiy.com/index.php?topic=41371.0

Here's a solution I didn't understand at the time but now used as a starting point: http://www.groupdiy.com/index.php?topic=41509.msg539385#msg539385

I prefer the voltage divider due to the fact I can even set it up with a 1 meg pot/trimmer.

Here's an update. It's now continuously variable and drawn with values that most reflect my current situation and parts on shelf.



There might be an issue with the RC time constant of VR1/C6 that is now also continuously variable. It's short at the highest voltage end and a bit slow on the lower voltage end. Some thud maybe, probably not a big deal and can be prevented with an additional series resistor with expense of some max voltage.

I also have to make sure the pot/trimmer can deal with +100VDC through it. Or it could be a rotary switch.

I council you to connect the zd2 cathode on positive lead of C7 to have a current limiting prot., so I council you zd2=10V to set the max curr. from  125mA to 160.

What is that a regulator, or some kind of a voltage dropping gizmo.  I don't like it. The full burden is placed on R1, uugh!

I would use R9 but bigger and a cap from gate to ground to take out zenner noise (maybe not necesary but...)

Then, with C6 to gate and bigger R9 from gate to zenner you can jump from zenner to zenner with no problems. You can use 3 5W 50V zenners and 10 0.5W 5.1V zenners and they will be working at good point both with the same current, and you have 5.1V steps from 150V to 201V.

A lot of zenners but nice definition and neither working curent problems nor pots and time constant changes.

If you want a pot you could use let's say two 5W 50V zenner and one 100V 5W zenner with a pot in parallel with the 100V zenner taking the same current of it and will be working at the same point of 50V zenner and time constant variation will be low, biggest twice lowest if R9=(VR1)/2

Let's say 20K pot droping 100V would be 5mA and .5W, another .5W from the 100V zenner and .5W from each 50V zenner. Then 10K for R9 and C6 for your favorite time constant. I know, 2W just for zenners but with 2W zenners just 0.8W, harder to find but it isn't so bad lose 2W... I've never seen 0.5W 100V zenner but if you find them you just need 0.2W

C5 could be working better if not just in parallel with zenner, spliting R4 at half and putting C5 at the middle would make better filtering for the zenner and then another filtering stage between R9 and C6 would make a nice filtering for your regulator and if your time constant is slow and cap big your mosfet is a capacitance multiplier and add even more filtering to your complete PS.

If you are dropping 150V 100mA in your mosfet 15W is some heat to get a nice sink, not a problem and don't came to me saying my 2W from zenners is a lot! You will need a 40VA or 50VA transformer here.

Ah, as analog said, if I take it right, take R1 out.

IMHO most tube don't need so much precision to need a pot in the regulator, the 'a lot of zenner solutions' I think is a really elegant solution and you will know the voltage without measure it. You can do it with 10V jumps with 0.5W 10V and 5W 100V zenner or 15V jumps with 0.5W 15V and 5W 150V zenner. Or 12V or whatever if you find 0.5W and 5W 1:10 V zenners.

JS

I've now built the second revision and it works well. C6 doesn't even have to be 10uF. I used a 2.2uF film cap and no thud, but I suppose that depends on the use case.

analag said:

What is that a regulator, or some kind of a voltage dropping gizmo.  I don't like it. The full burden is placed on R1, uugh!

Click to expand...

That part was just to "pencil in" input to the regulator (ie. we want reasonably clean DC to the regulator stage) so a casual reader wouldn't be left guessing. This is not exactly how I'm using it, only the input voltage matters here. Also, your own poorman compressor regulator is an inferior version of this - only a voltage divider as a reference, completely tied to the input voltage and expecting it to be perfect. When there is ripple at the input, your version will happily reference that too.

joaquins said:

I would use R9 but bigger and a cap from gate to ground to take out zenner noise (maybe not necesary but...)

Then, with C6 to gate and bigger R9 from gate to zenner you can jump from zenner to zenner with no problems. You can use 3 5W 50V zenners and 10 0.5W 5.1V zenners and they will be working at good point both with the same current, and you have 5.1V steps from 150V to 201V.

A lot of zenners but nice definition and neither working curent problems nor pots and time constant changes.

If you want a pot you could use let's say two 5W 50V zenner and one 100V 5W zenner with a pot in parallel with the 100V zenner taking the same current of it and will be working at the same point of 50V zenner and time constant variation will be low, biggest twice lowest if R9=(VR1)/2

Let's say 20K pot droping 100V would be 5mA and .5W, another .5W from the 100V zenner and .5W from each 50V zenner. Then 10K for R9 and C6 for your favorite time constant. I know, 2W just for zenners but with 2W zenners just 0.8W, harder to find but it isn't so bad lose 2W... I've never seen 0.5W 100V zenner but if you find them you just need 0.2W

C5 could be working better if not just in parallel with zenner, spliting R4 at half and putting C5 at the middle would make better filtering for the zenner and then another filtering stage between R9 and C6 would make a nice filtering for your regulator and if your time constant is slow and cap big your mosfet is a capacitance multiplier and add even more filtering to your complete PS.

Click to expand...

The time constant turned out to be a non-issue, but thanks for the tips. These are all good tweaks to remove zener noise especially. But R5/C7 is already filtering so much the minor tweaks won't actually any effect. Noise from zener is measured in millivolts at the regulator output and I'm filtering some 2-5 volts there already - all noise gone. It depends on load of course, but as you said, tubes don't need so much precision.

bad news,

the above revision was making strange ~8hz latching/oscillation at the regulator output, at the higher voltage end. No good, lets see with the scope. Turns out there is about 5V oscillation right at the MOSFET gate - actually too fast for my scope which only saw noise.

PRR said:

With all the high impedances and low currents, I suspect R9 2K is not doing a thing.

Click to expand...

The basis of my drawings came here: http://www.solderingpoint.com/projects/mosfetps/mosfetps.php

They are calling the gate resistor current limiter. Looks like it has other uses. I put it back and the oscillation is gone. Hence, rev3:



All capacitance now reflect the exact build. I use film caps whenever I can. Live long and prosper. I took off the rectifying back-end to keep analag happy. It was not on my bench in the first place. SW1 is just a quick time constant and oscillation test point, not necessary at all. I tested one more thing: 1uF from MOSFET gate to ground. The oscillation was there again. Looks like the gate doesn't like capacitance and wants some ESR to go with it.

There is a lot of room for improvement.

C5 not needed.
C6 should be 10u or bigger. This will improve ripple rejection a lot.
R7 not needed, C6 will drain (almost) through ZD2.
Imagine a potentiometer (or resitors+switches) in parallel with ZD1. That is the best place for voltage adj. It will drain C6 too.
From the wiper go through 100k resistor to C6
And as you noted, it is elementary to keep gate impedance above zero, 220 ohms from C6 to gate is enough.
R5 is too big to my taste. C8 does nothing. Replace C7 and C6 with a proper poly cap above 10 uF.

Result: Better ripple rejection, less noise, smaller output impedance, no electrolytic cap shit, nicely ramped voltage switching.

Since it was you who pointed this direction in the first place, here's rev4, thanks!



I don't necessarily agree with using these ridiculously-sized film caps, especially if this was a +250VDC design which it could easily do as well. But I very much like these low parts count optimised designs.

[edit]

changed pot to 1meg, and slightly decreased R1. This is now on the bench and very stable.

Looking good. Just reduce the size of R1 or increase the pot size in order to get closer to raw input voltage.
1W zener would be more stable at these currents.

Do visit my place some day and get a Mundorf polyprop 30u/600V "tubecap", they are quite compact and not expensive when bought in quantities.

Jonte Knif said:

Just reduce the size of R1 or increase the pot size in order to get closer to raw input voltage.
1W zener would be more stable at these currents.

Click to expand...

I tested this in one application already and was wondering why the thing became interactive with bigger swings of the input voltage. Yeah that pot should be 1 meg instead. I built this as a 120V reference this time (3X 39V 2W zeners). Very flexible utility overall.

I'm thinking of using a very similar circuit for an adjustable 120V capsule polarization supply.

I'm wondering if there are any other tweaks possible, given that the capsule essentially draws nothing but leakage current (probably in the uA range).  Given there is no error feedback other than via the VGS mechanism I'm guessing it doesn't matter.

I was thinking perhaps a 1mA CCS (via a JFET or something similar) on the output would help a polarization supply remain stable as the output voltage was adjusted (again, since the only DC current on the output would be via R6).

Matador said:

I'm wondering if there are any other tweaks possible.

Click to expand...

You could just branch your polarisation from the very well filtered zener reference. It's the most stable voltage point of the design. No need for the transistor. Its only job here is to pass current and you need "none".

Jonte Knif said:

Do visit my place some day and get a Mundorf polyprop 30u/600V "tubecap", they are quite compact and not expensive when bought in quantities.

Click to expand...

I did a version with 4.7uF wima polyprops I had around. One for the zener filtering, another for output. The application required 20mA and even if there was so little capacitance, the result was perfect. I could not measure any ripple at the output. Looks like I will be gravitating from big electrolytics to "small" poly caps in all my future PSU's.

I had it in my mind that the MOSFET would work to keep the output impedance relatively constant as the pot was twiddled.  And perhaps it does, but it probably doesn't matter in my case.

I'll try a plain CRCRC filtered Zener string as a first shot.

Kingston said:

Jonte Knif said:

Just reduce the size of R1 or increase the pot size in order to get closer to raw input voltage.
1W zener would be more stable at these currents.

Click to expand...

I tested this in one application already and was wondering why the thing became interactive with bigger swings of the input voltage. Yeah that pot should be 1 meg instead. I built this as a 120V reference this time (3X 39V 2W zeners). Very flexible utility overall.

Click to expand...

I like your regulator, foil caps seems like a great solution. What values have input filters in ver.4 at 20mA? Have you also tried 1meg pot at 300v in ver.4? I'm just looking at typical solutions for input filters, would you explain a little how you prefer them?. It would be nice to have variable regulator like this up to 300V, i think elco caps for input filters improved and are not that problematic when well chosen. I calculated working hours at typical conditions for some Epcos elcos and it seems like they should last very long time. Btw, Fearn uses simple zener shunt regulator and it seems ok there.