开云体育

Peter,

1) Well, I don't have an enviro chamber either..... in fact I don't even have a breadboard!? :-)

2) I thought the 2sk3563's had finally worked ok in earlier versions?

3) Best I found:

They are available at Mouser or Digikey, $2.58 at Digikey. Since it is 1000V only one cascode is needed, but two might still be worthwhile to avoid constant operation near the breakdown voltage (reliability).

4) I could not find any HV MosFet not in a power package like at least a TO-220. This circuit is extremely low power in the MosFets, never exceeding about 1/8 W in the cascodes.

"I don't like it but I guess things happen that way." - J. Cash

rogerw

Roger:
?I don't have an environmental chamber at home, I only have a hair dryer!.
I didnt temperature test. I did cobble the circuit using three 5.1V zeners( I did pick up the 6.2V Rohm zeners but didn't use them yet)
I think the over size MOSFETs (2SK3563) are causing issues because the regulation is not stiff, it still rises about 100V over 900V as the Vcc raised past 900V and up to 1150V.? Again, if I tie the top end of the resistor divider to Vcc instead of the drain of the upper mosfet, regulation improves, and I see only about a 25V increase.?

I got to look for some smaller 100mA size 1000V mosfets. Unfortunately I haven't found anything that uses every remotely close in the e-recycling bins! Most items have mosfets in the 200~ 500V range at several to tens of amps.

P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

开云体育

I began some temperature testing simulations and the On Semi 6.2V Zeners I originally used only exhibited about half the expected tempco in simulation. I don't know why.....? I changed them over to Rohm 6.2V Zeners (as in the schematic below) and the temperature compensation became excellent over the -25degC to +75degC range. I also approximately halved the Zener current changing R6 to 600K and also put R7=600K to give Q2 about another 1uA minimum bias. R7 is prob not necessary though.

R5 is re-adjusted to give 900+V at the output (it can be a trimmer to vary voltage output).

Circuit performs very well over variation of input voltage = UnRegHV and temperature.

It is also possible to use a single Rohm 10.0V Zener (same tempco as the three 6.2V Zeners) and remain temperature compensated. Then R5 would be re-adjust to give 900+V output....... I like the higher reference voltage though with the three 6.2V Zeners.

rogerw

The Rohm 6.2V should be fine.

I never simulated the temp performance, that was MindSim. Generally, in what we usually call Zener Diodes, the are actually two breakdown mechanisms: Zener breakdown and avalanche breakdown. Zener breakdown dominates below about 5.1V and avalanche dominates above that voltage. Avalanche mechanism has a positive tempco and Zener has a negative tempco.

It is very common design practice to pair a 6.1V or 6.2V Zener with a silicon diode or base-emitter junction to achieve close to zero tempco.

If the Zener is above 5.1 volts some will call it an "avalanche diode" instead of Zener Diode..... Truth is, all Zener Diodes have both mechanisms going on to varying degrees, hence a nonlinear change of tempco vs voltage spec.

rogerw

On 4/1/2022 3:09 PM, peter via groups.io wrote:

[Edited Message Follows]

Roger:
?The closest mosfet in my Micro-Cap library was a 2SK311 to the 2SK3563 that I was proto-typing with.?
Since these mosfets are much higher current parts, than those you may have used in your sim,? they don't work as well in the simulation I ran using a 2SK311,?
I had to tweak R6 to 6.8M to get the circuit to work and as you simulated, temp stability is great.?
Now I have a few more things to cobble together tonight.
I looked up the part I will be using its a Rohm MTZJ-6.2? zener.
will try to make time this evening to melt some solder
P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

Virus-free.

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

开云体育

The Rohm 6.2V should be fine.

I never simulated the temp performance, that was MindSim. Generally, in what we usually call Zener Diodes, the are actually two breakdown mechanisms: Zener breakdown and avalanche breakdown. Zener breakdown dominates below about 5.1V and avalanche dominates above that voltage. Avalanche mechanism has a positive tempco and Zener has a negative tempco.

It is very common design practice to pair a 6.1V or 6.2V Zener with a silicon diode or base-emitter junction to achieve close to zero tempco.

If the Zener is above 5.1 volts some will call it an "avalanche diode" instead of Zener Diode..... Truth is, all Zener Diodes have both mechanisms going on to varying degrees, hence a nonlinear change of tempco vs voltage spec.

rogerw

[Edited Message Follows]

Roger:
?The closest mosfet in my Micro-Cap library was a 2SK311 to the 2SK3563 that I was proto-typing with.?
Since these mosfets are much higher current parts, than those you may have used in your sim,? they don't work as well in the simulation I ran using a 2SK311,?
I had to tweak R6 to 6.8M to get the circuit to work and as you simulated, temp stability is great.?
Now I have a few more things to cobble together tonight.
I looked up the part I will be using its a Rohm MTZJ-6.2? zener.
will try to make time this evening to melt some solder
P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

开云体育

Peter,

That last "Way-Improved" circuit is working extremely well, I encourage you to try it. You do not have to have all three zeners to get functionality, you merely adjust the bottom trim resistor for a 900+ output and the only thing you lose more or less is temperature compensation (3Vbe + 3Vz of 6.3V).

btw, I experimented with increasing the "dropper resistor" all the way up to 10Meg and the circuit functioned fine with the output dropping from about 918V with 1Meg to about 904V with 10Meg.

Pretty danged good!

That corresponds to what would happen if the the UnRegHV generator current limits at about 30uA at 1200V and will not put out anymore current. That 30uA figure is 22uA in the Zeners (plus 6.7uA in the resistor string, plus 1.3uA in the BJT) which can be reduced, but I have not tried it. It can probably be made to be about 15uA in the zeners without much change in performance.......

All this is simulation, no breadboarding.

rogerw

Actually, the original values are working quite well, with only the 2.85Meg resistor being trimmed for nominal RegHV value.

On 4/1/2022 12:06 PM, Roger Whatley via groups.io wrote:

Peter,

If you decide to try this Way-Improved Cascoded Shunt Regulator, I will suggest the following, same circuit as before but with resistor-value changes.

BTW, by taking the voltage to the 600K/Zeners from the Source of M3, an additional help is occurring. As UnRegHV goes up and the current drawn by Q1 increases through M3, Vgs3 gets larger so that node drops in voltage somewhat. Thus the Zener current drops somewhat and this helps hold the overall reference voltage (= 3*Vz + 3Vbe) more constant as Vbe1 increases with current.

Voila!? (wallah!)

On 4/1/2022 11:01 AM, Roger Whatley via groups.io wrote:

OK, found it. Interesting.......

Mindsim tells me that doing that will reduce the feedback regulation of the voltage by essentially reducing the forward gain of the "opamp." By the "opamp," I mean the BJT whose collector current drives the load resistor which you are calling the "dropper resistor." That voltage gain = gm*Rload is the forward gain, and the feedback gain = (R1)/(R1+ R2 + R3 + R5). By taking the feedback point at less than the full forward gain, the loop gain is reduced, hence regulation is reduced.

It may well be that while you are introducing more loop gain error you are also compensating some other circuit error.....

That's my mindsim guess...

In the meantime check out my last posted circuit..... ("Even I was amazed" was a recurring joke-line in my old circuit design group in the 1980s. We were Designers once, and Young.....)

rogerw

On 4/1/2022 10:34 AM, peter via groups.io wrote:

Roger
Please see the attachment in message #299
After my boo-boo, which caused the voltage to sag, it was feedback. So I changed the amount of feedback until the voltage would not rise or fall much.

I will check spec of the zeners I have access to.
I try to find something that is close to -2.2mV/C at low current.
P


P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

Virus-free.

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

开云体育

Actually, the original values are working quite well, with only the 2.85Meg resistor being trimmed for nominal RegHV value.

Peter,

If you decide to try this Way-Improved Cascoded Shunt Regulator, I will suggest the following, same circuit as before but with resistor-value changes.

BTW, by taking the voltage to the 600K/Zeners from the Source of M3, an additional help is occurring. As UnRegHV goes up and the current drawn by Q1 increases through M3, Vgs3 gets larger so that node drops in voltage somewhat. Thus the Zener current drops somewhat and this helps hold the overall reference voltage (= 3*Vz + 3Vbe) more constant as Vbe1 increases with current.

Voila!? (wallah!)

On 4/1/2022 11:01 AM, Roger Whatley via groups.io wrote:

OK, found it. Interesting.......

Mindsim tells me that doing that will reduce the feedback regulation of the voltage by essentially reducing the forward gain of the "opamp." By the "opamp," I mean the BJT whose collector current drives the load resistor which you are calling the "dropper resistor." That voltage gain = gm*Rload is the forward gain, and the feedback gain = (R1)/(R1+ R2 + R3 + R5). By taking the feedback point at less than the full forward gain, the loop gain is reduced, hence regulation is reduced.

It may well be that while you are introducing more loop gain error you are also compensating some other circuit error.....

That's my mindsim guess...

In the meantime check out my last posted circuit..... ("Even I was amazed" was a recurring joke-line in my old circuit design group in the 1980s. We were Designers once, and Young.....)

rogerw

On 4/1/2022 10:34 AM, peter via groups.io wrote:

Roger
Please see the attachment in message #299
After my boo-boo, which caused the voltage to sag, it was feedback. So I changed the amount of feedback until the voltage would not rise or fall much.

I will check spec of the zeners I have access to.
I try to find something that is close to -2.2mV/C at low current.
P


P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

Virus-free.

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

开云体育

Peter,

If you decide to try this Way-Improved Cascoded Shunt Regulator, I will suggest the following, same circuit as before but with resistor-value changes.

BTW, by taking the voltage to the 600K/Zeners from the Source of M3, an additional help is occurring. As UnRegHV goes up and the current drawn by Q1 increases through M3, Vgs3 gets larger so that node drops in voltage somewhat. Thus the Zener current drops somewhat and this helps hold the overall reference voltage (= 3*Vz + 3Vbe) more constant as Vbe1 increases with current.

Voila!? (wallah!)

OK, found it. Interesting.......

Mindsim tells me that doing that will reduce the feedback regulation of the voltage by essentially reducing the forward gain of the "opamp." By the "opamp," I mean the BJT whose collector current drives the load resistor which you are calling the "dropper resistor." That voltage gain = gm*Rload is the forward gain, and the feedback gain = (R1)/(R1+ R2 + R3 + R5). By taking the feedback point at less than the full forward gain, the loop gain is reduced, hence regulation is reduced.

It may well be that while you are introducing more loop gain error you are also compensating some other circuit error.....

That's my mindsim guess...

In the meantime check out my last posted circuit..... ("Even I was amazed" was a recurring joke-line in my old circuit design group in the 1980s. We were Designers once, and Young.....)

rogerw

On 4/1/2022 10:34 AM, peter via groups.io wrote:

Roger
Please see the attachment in message #299
After my boo-boo, which caused the voltage to sag, it was feedback. So I changed the amount of feedback until the voltage would not rise or fall much.

I will check spec of the zeners I have access to.
I try to find something that is close to -2.2mV/C at low current.
P


P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

Virus-free.

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

开云体育

OK, found it. Interesting.......

Mindsim tells me that doing that will reduce the feedback regulation of the voltage by essentially reducing the forward gain of the "opamp." By the "opamp," I mean the BJT whose collector current drives the load resistor which you are calling the "dropper resistor." That voltage gain = gm*Rload is the forward gain, and the feedback gain = (R1)/(R1+ R2 + R3 + R5). By taking the feedback point at less than the full forward gain, the loop gain is reduced, hence regulation is reduced.

It may well be that while you are introducing more loop gain error you are also compensating some other circuit error.....

That's my mindsim guess...

In the meantime check out my last posted circuit..... ("Even I was amazed" was a recurring joke-line in my old circuit design group in the 1980s. We were Designers once, and Young.....)

rogerw

Roger
Please see the attachment in message #299
After my boo-boo, which caused the voltage to sag, it was feedback. So I changed the amount of feedback until the voltage would not rise or fall much.

I will check spec of the zeners I have access to.
I try to find something that is close to -2.2mV/C at low current.
P


P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

开云体育

So, I eluded below to the fact that changing current in the Zener(s) is the remaining dominant error in regulation over voltage. So check out the enclosed circuit and regulation curve.....

Probably overkill, but it sure is nice!

From 1000V to 1200V the regulated output varies by....... drum roll please...... ONE volt!

Of course this is simulation, but still..... Be still my heart!

rogerw

1) I don't understand, can you post a schematic?

2) In the earlier current mirror configuration, the only purpose of the diode-connected BJT is to drop one Vbe so that the voltage on the resistor would be nominally equal to the Zener voltage. There are several sources of error in that occurs, and a significant one is BJT base current. The base current in the right-most BJT with the Zener can have its collector current go from less than 10uA to well over 100uA or 200uA depending on the value of UnRegHV, etc.So, yes, you could employ Darlingtons in the current mirror configuration and it would greatly alleviate the base current error.

But it will not alleviate the mismatch in the two Darlington current mirror BJTs.

(The error quantity, delta Vbe, is extremely well-defined math-wise, and used in other circuits to good effect. In a Band Gap Reference the negative tempco of a Vbe is compensated by the positive tempco of delta Vbe, and when this is done the physics of it is you have just achieved the physical bandgap voltage of the semiconductor. For silicon ~1.2V.)

By eliminating the diode-connected device, the Vbe with its negative tempco becomes part of the reference voltage seen by the base resistor. But base current when the collector current hits above 100uA is a significant error in the 66M+66M, etc, so going to a Darlington alleviates that.

Then, with two Vbe's giving -2mV/degC each of tempco, using two 6.3V zeners giving +2mV/degC each does temperature conpensation.

It works well, I think...... try it.

BTW, in the real application where the "dropper resistor" might already be in the HV generator output circuit, AND the HV generator itself is not capable of putting out 100's of uA even though it has an open circuit output of, say 1200V...... Well in reality what you have is a 1200V source with a nonlinear output resistance and in series with an external "dropper resistor." As the UnRegHV approaches higher voltages, the shunt regulator (either the present one OR a zener string) pulls more and more current until the current limits of the HV Generator are being approached.... and the effective "dropper resistance" is increasing to limit the UnRegHV...... So, just like with a series zener string, shunt regulation!

rogerw

1) any comments of using two resistors for the dropper and taking the mid point connections to feed the bias resistor string ? I found that it greatly reduces any voltage rise as the Vcc is raised.
2) in the earlier current mirror configuration with two bjt's, is there any advantage on using two darlingtons?

P
PS I will have to remember to get the 6.2V zeners today!

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

Virus-free.

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

开云体育

1) I don't understand, can you post a schematic?

2) In the earlier current mirror configuration, the only purpose of the diode-connected BJT is to drop one Vbe so that the voltage on the resistor would be nominally equal to the Zener voltage. There are several sources of error in that occurs, and a significant one is BJT base current. The base current in the right-most BJT with the Zener can have its collector current go from less than 10uA to well over 100uA or 200uA depending on the value of UnRegHV, etc.So, yes, you could employ Darlingtons in the current mirror configuration and it would greatly alleviate the base current error.

But it will not alleviate the mismatch in the two Darlington current mirror BJTs.

(The error quantity, delta Vbe, is extremely well-defined math-wise, and used in other circuits to good effect. In a Band Gap Reference the negative tempco of a Vbe is compensated by the positive tempco of delta Vbe, and when this is done the physics of it is you have just achieved the physical bandgap voltage of the semiconductor. For silicon ~1.2V.)

By eliminating the diode-connected device, the Vbe with its negative tempco becomes part of the reference voltage seen by the base resistor. But base current when the collector current hits above 100uA is a significant error in the 66M+66M, etc, so going to a Darlington alleviates that.

Then, with two Vbe's giving -2mV/degC each of tempco, using two 6.3V zeners giving +2mV/degC each does temperature conpensation.

It works well, I think...... try it.

BTW, in the real application where the "dropper resistor" might already be in the HV generator output circuit, AND the HV generator itself is not capable of putting out 100's of uA even though it has an open circuit output of, say 1200V...... Well in reality what you have is a 1200V source with a nonlinear output resistance and in series with an external "dropper resistor." As the UnRegHV approaches higher voltages, the shunt regulator (either the present one OR a zener string) pulls more and more current until the current limits of the HV Generator are being approached.... and the effective "dropper resistance" is increasing to limit the UnRegHV...... So, just like with a series zener string, shunt regulation!

rogerw

1) any comments of using two resistors for the dropper and taking the mid point connections to feed the bias resistor string ? I found that it greatly reduces any voltage rise as the Vcc is raised.
2) in the earlier current mirror configuration with two bjt's, is there any advantage on using two darlingtons?

P
PS I will have to remember to get the 6.2V zeners today!

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

Roger:
Couple morning questions:
1) any comments of using two resistors for the dropper and taking the mid point connections to feed the bias resistor string ? I found that it greatly reduces any voltage rise as the Vcc is raised.
2) in the earlier current mirror configuration with two bjt's, is there any advantage on using two darlingtons?

P
PS I will have to remember to get the 6.2V zeners today!

开云体育

Attached is a simulated performance curve for the Darlington BJT Cascoded HV Regulator. As you can see, when UnRegHV goes from 1000V to 1200V, the RegHV output goes from 904V to 918V. I think that is about comparable to Zener strings I have used to regulate the same.

In fact if you think about it, this circuit derives its reference in the same way but with a zener at low voltage whose current will also increase proportionately as UnRegHV is increased.... hence the regulated voltage goes up even with perfect circuit feedback regulation because the zener voltage goes up.

Still, this is very good performance, very adequate to the task, and affords easy oppty for adjustment.

Also, if there is any reliability issues with HV Zeners in a string, this circuit may alleviate that. But a circuit failure here might have the same effect as a circuit failure in the zener string.

I ramped the simulated circuit from UnRegHV = 0V to 1200V and did not see any HV on the BJTs that might cause breakdown during power-on transient, which if it occurs can shorten the life of the BJT. So that seems safe......

rogerw

of course you can use any zener to cobble it up for functionality, and adjust the bottom resistor for correct output. 6.3V zener is just for the tempco.

rogerw

On 4/1/2022 12:34 AM, peter via groups.io wrote:

Roger:

Thank you again. Of course i will give this latest circuit a try.?
Hopefully, nothing exciting will happen this weekend, so I can tinker away.
I don't have any 6.3V zeners handy at home. I will swing by the parts cabinet at work tomorrow and get couple so I can cobble this circuit up.
P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

Virus-free.

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

开云体育

of course you can use any zener to cobble it up for functionality, and adjust the bottom resistor for correct output. 6.3V zener is just for the tempco.

rogerw

Roger:

Thank you again. Of course i will give this latest circuit a try.?
Hopefully, nothing exciting will happen this weekend, so I can tinker away.
I don't have any 6.3V zeners handy at home. I will swing by the parts cabinet at work tomorrow and get couple so I can cobble this circuit up.
P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.

开云体育

Before you quit try the enclosed version. In simulation the single BJT base current was biggest circuit error left. Darlington pair simulated and regulated quite well.

Two 6.3V zeners at +2mV/degC each compensates the two Vbe's -2mV/degC? over temp.

rogerw

Roger:

Micro-cap sim doesn't look that good but on the proto board it works well.
Put two 5.1V zeners in series.
Used two resistors for the dropper to get some negative feedback.
Seems to regulate well. Set to 900V and then crank the Vcc to 1150V, and the regulation increases about 30V.

I think I got my fun out of this circuit.
May be I will try the BJT version posted.
Again thanks for all the suggestions.

P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

The PanDemic is over, but the DemPanic goes on.
MisDisMal is a dictator's tool of repression.