开云体育

Adjustable, yes. I would look at varying the voltage which is currently presented in the circuit by the zener...... Make this a variable voltage. Put a bypass cap on it too to suppress any potentiometer noise, which would be multiplied up at the output.

Yes, the HV Zeners do require sorting to match the voltage at a given current. I used a breadboarded cd V-700 style circuit to sort some 330V nominal Zeners into sets of 3 to give 900 to 930V. I thought it worked ok in a 700..... but if you tried t put that on a HV supply that can source much more current it would go closer to 330V nom x3 = 990Vnom..... not good.

rogerw

Roger:
?Yes.
But its my fault for not explicitly stating that my interest in the zener multiplier circuit( Codatron) is that I was thinking about using it in devices other than the CDV-700's.
1) I had poor luck in using surplus high voltage zeners. I bought a few bags of ~ 90V zeners but knee wasn't consistent. It was a bother to get the string to regulate at 900V.?
It also wasn't easily adjustable.
2) I had poor luck in using a resistive divider to make a properly regulated supply. I often got ripple due to the time constant of the feedback network, and also its susceptibility to noise as it was high impedance.
?I would add capacitance to cure the noise, but then I would get low frequency ripple.?

I was thinking that a shunt regulator while not efficient may not have ripple issues (2).
Or more likely? is it that I want something to tinker with.
I can also think of cases were I don't want to permanently change the voltage of my CDV-700 or some other device? and I could use an external shunt regulator.
I believe that this shunt regulator can be made easily adjustable so I could adjust the hv easily for temporary uses ?

P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

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

Roger:
?Yes.
But its my fault for not explicitly stating that my interest in the zener multiplier circuit( Codatron) is that I was thinking about using it in devices other than the CDV-700's.
1) I had poor luck in using surplus high voltage zeners. I bought a few bags of ~ 90V zeners but knee wasn't consistent. It was a bother to get the string to regulate at 900V.?
It also wasn't easily adjustable.
2) I had poor luck in using a resistive divider to make a properly regulated supply. I often got ripple due to the time constant of the feedback network, and also its susceptibility to noise as it was high impedance.
?I would add capacitance to cure the noise, but then I would get low frequency ripple.?

I was thinking that a shunt regulator while not efficient may not have ripple issues (2).
Or more likely? is it that I want something to tinker with.
I can also think of cases were I don't want to permanently change the voltage of my CDV-700 or some other device? and I could use an external shunt regulator.
I believe that this shunt regulator can be made easily adjustable so I could adjust the hv easily for temporary uses ?

P

开云体育

BTW, I have given some more thoughts to my comments in this thread about how this shunt regulator we are discussing "requires more current" than a simple Zener string:

1) The schematic I have shown and we have discussed shows a 1000V Unreg HV? with a 500K (or, say, 3M as discussed) load resistor shown. This resistor is setting the total current in the circuit and based on a 1000V unreg HV that supplies whatever is demanded. However, if you look at a Lionel CD V-700 schematic, there is a 1.8Meg resistor in series coming out of the HV transformer secondary. This resistor is actually perhaps all you need, AND,

2) The regulator circuit requires a current of Vz/1Meg in the lefthand path ( = 15V/1M = 15uA in my version, but only 5.6uA in your version). All of the extra current supplied thru the 500K to 3Meg load resistor is shunted to ground thru the right-hand leg and thru the Zener. This current is simply whatever current sink is required to regulate the output by feedback, and therefore similar to a simpler Zener string shunting whatever is required to regulate. The load resistor simply needs to be low enough value to provide enough current to the Zener, in excess of the current to the lefthand leg, so that the Zener has sufficient current to get decent Zener voltage.

Thus, in effect, I think the load resistor of 1.8Meg? in the Lionel, or other, is already there, and the rest of the contemplated Cascoded Shunt Regulator actually behaves pretty much like the Zener String, albeit with two current paths and still requiring somewhat more current perhaps.

Does that make sense?

rogerw

Good deal..... and btw after you simulate this circuit, I have a simple improvement to nearly totally alleviate the BJT base current error, but I suggest you try it as is (with components tweaked to your requirements) first.

rogerw

On 3/28/2022 10:32 AM, peter via groups.io wrote:

Roger:
?Thanks again. Your mind-sim is a newer version than my version. I could only think all BJT or all Mosfet. Hybrid didn't even cross my mind
I didn't think of the current mirror only as BJT and the cascoded transistors as Mosfets.
I was fixated on looking for some HV BJT's and couldn't remember where I put them.
After work today, I will try? to tinker with your mind-sim output.

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

Good deal..... and btw after you simulate this circuit, I have a simple improvement to nearly totally alleviate the BJT base current error, but I suggest you try it as is (with components tweaked to your requirements) first.

rogerw

Roger:
?Thanks again. Your mind-sim is a newer version than my version. I could only think all BJT or all Mosfet. Hybrid didn't even cross my mind
I didn't think of the current mirror only as BJT and the cascoded transistors as Mosfets.
I was fixated on looking for some HV BJT's and couldn't remember where I put them.
After work today, I will try? to tinker with your mind-sim output.

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 am re-posting a schematic of how I would use a junction transistor in this present scheme. This is "stream of consciousness" and un-simulated nor breadboarded.

We used to call this a "Mind-sim."? I think it will work pretty good :-)

rogerw

[Edited Message Follows]

Roger:
?earlier you had mentioned that using a junction transistor may be an alternative to a n-ch mosfet.
Is the HFE sufficient in 1000V rated transistors so that the base drive current won't be a significant issue when compared to use a MOSFET ?
I noticed that 1000V rated n-ch mosfets aren't cheap nor commonly available as surplus.
Would the configuration for a cascode junction transistor be similar in topology or better yet could you please post a circuit of how? npn junction transistors could be used in a comparable circuit ( cascode and not-cascode)?

P

ps:?
pps:?

-- 
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 am glad the cascode worked for you. The regulation will not be super great because, if you think in terms of the "opamp" explanation I posted, the single M1 gain stage does not make an ideal opamp and the feedback loop gain for regulation is kinda low.

I have enclosed what I THINK would be an improved regulator on the present scheme (not simulated, nor breadboarded). The bipolar will provide higher loop gain that a FET. One could also actually employ a real opamp, but that would be overkill complexity..... I would change the 500K to about 3Meg I think and bring the current way down, but then the Zener voltage will drop some and other values will need adjustment to compensate that. This version will have less error in the voltage appearing on the 1Meg to be multiplied up......

I think the problem you MAY have is that this regulator circuit demands more current from the unregulated HV supply. I don't think the CD V-700 HV generator will supply it. In the version I was discussing below it was something like 260uA. Of course by tweaking component values you can get that way way down, but in the end there are still two current-legs to divide your total current budget between.......

With a Zener string ALL of the Unreg HV current is crow-bar'ed to ground, but no more current than is available is required and the output is well regulated to the requirement.

A coupla years ago I found some 330V Zeners pretty cheap to use in a string to regulate 900+ volts.They were not sold as Zeners, per se, but under some other functional name like arc suppression, or such.? At such low currents like 10-20uA they are less than 330V and work pretty well for 900+ volts. Maybe I can look that up again.....

rogerw

开云体育

Roger:
?earlier you had mentioned that using a junction transistor may be an alternative to a n-ch mosfet.
Is the HFE sufficient in 1000V rated transistors so that the base drive current won't be a significant issue when compared to use a MOSFET ?
I noticed that 1000V rated n-ch mosfets aren't cheap nor commonly available as surplus.
Would the configuration for a cascode junction transistor be similar in topology or better yet could you please post a circuit of how? npn junction transistors could be used in a comparable circuit ( cascode and not-cascode)?

P

ps:?
pps:?https://www.tubecad.com/2006/11/15/simple%20ss%20series%20regulator.png

开云体育

Peter,

I re-read your question below and now I think I understand what you are asking..... No, it does not impose a large Vgs on the upper cascode M3. Instead, M3's gate is at a fixed dc voltage and a fixed current will be drawn out of the M3 source (by M2), resulting in the source node being a normal Vgs3 volts below the M3 gate voltage.

Think of the source voltage as "hanging" a normal gate-source voltage below the fixed gate voltage, and due to a fixed current draw. The approximate Vgs for a given Id current will be given in the transistor datasheet.

rogerw

oops dammit. had another error left on the schematic I forgot to fix..... try this enclosure.

rogerw

On 3/27/2022 7:12 PM, Roger Whatley via groups.io wrote:

Peter,

Well, for starters I made another mindless math error: " split the 56M resistor into two 23M resistors" should be "two 28M resistors."

I have enclosed a schematic drawing of the cascode connection I intended. Does that fix it?

rogerw

On 3/27/2022 6:31 PM, peter via groups.io wrote:

Roger:
?Huh ? Doesn't that resistor configuration impose a large Vgs on? the upper Mosfet ?

Also when turning the circuit on, depending upon the turn on speed etc, won't one Mosfet be exposed to the full potential of the HV instead of being evenly divided across the two mosfets ?

?

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

oops dammit. had another error left on the schematic I forgot to fix..... try this enclosure.

rogerw

Peter,

Well, for starters I made another mindless math error: " split the 56M resistor into two 23M resistors" should be "two 28M resistors."

I have enclosed a schematic drawing of the cascode connection I intended. Does that fix it?

rogerw

On 3/27/2022 6:31 PM, peter via groups.io wrote:

Roger:
?Huh ? Doesn't that resistor configuration impose a large Vgs on? the upper Mosfet ?

Also when turning the circuit on, depending upon the turn on speed etc, won't one Mosfet be exposed to the full potential of the HV instead of being evenly divided across the two mosfets ?

?

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

Peter,

Well, for starters I made another mindless math error: " split the 56M resistor into two 23M resistors" should be "two 28M resistors."

I have enclosed a schematic drawing of the cascode connection I intended. Does that fix it?

rogerw

Roger:
?Huh ? Doesn't that resistor configuration impose a large Vgs on? the upper Mosfet ?

Also when turning the circuit on, depending upon the turn on speed etc, won't one Mosfet be exposed to the full potential of the HV instead of being evenly divided across the two mosfets ?

?

P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

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

Roger:
?Huh ? Doesn't that resistor configuration impose a large Vgs on? the upper Mosfet ?

Also when turning the circuit on, depending upon the turn on speed etc, won't one Mosfet be exposed to the full potential of the HV instead of being evenly divided across the two mosfets ?

?

P

开云体育

See enclosed schematic. If you are used to simple op amp circuits the explanation will jump out at you.

M1 is basically connected in a non-inverting gain op amp configuration with M2 used solely to correct for M1's horrible input offset voltage (= Vgs1). Then, the input voltage to the non-inverting gain stage is Vz and the gain is (56M+1M)/1M = 57. Thus a 15V zener produces a regulated 855V output voltage, approximately.

That is the simplest way I know to view the theory of operation of this circuit.

There are significant simplifying assumptions in that calculation, the biggest being that the output will also sum in Vgs2, and also (Vgs2-Vgs1)*57.So, since M1 and M2 might be matched, they are operating at different currents and will necessarily have different Vgs voltages. Since Vgs1 is reliably greater than Vgs2 by component values chosen, this difference results in a larger output voltage.

rogerw

correction:

"I would suggest that the two Mosfets should be run at much closer currents for better matching and Vgs2=Vgs1, and also to avoid as much loading on the input Vhv= 1000V supply. I would suggest a 3Meg resistor for Ids1 = 45uA and Ids2=15uA."

After re-reading my note, it may not read as clear as I had hoped. Questions are encouraged. Here are some clarifying edits:

M1 and the 500K load resistor forms a amplifier where the gate is the negative input and the source is the positive input. You could think of this as a crude opamp with both positive and negative inputs, but a huge offset voltage = Vgs1.

The output of this M1 amplifier connects back to the negative input (in a unity gain opamp connection) thru M2= Vgs2 voltage drop. And thus whatever input voltage the source (positive input) "sees" is applied to the 1Meg load resistor. But whatever voltage drop appears on the 1Meg load resistor must also appear 56X on the 56Meg feedback resistor, thus multiplying the input voltage from the Zener by (56+1)/1 = 57X times.

Thus the feedback loop is regulating the voltage on the 1Meg resistor to equal the Zener voltage, Vz. M2 is a diode-connected device to generate a matching and cancelling voltage to the input characteristic of M1. Since Ids2 flowing in 1Meg resistor also flows thru the 56Meg resistor, the voltage that M1 must maintain at the output is approximately (56M + 1M)/1M = 57 times the Zener voltage (and plus one Vgs voltage, a small error).

I hope that helps..... :-)

btw, I believe that the reason for choosing Mosfets is probably that they are most easily found as HV transistors. However, I think the circuit would be generally improved accuracy with bipolar HV transistors and if they cannot be found with high enough Vcemax, then two or more may easily be cascoded (ie, vertically stacked) to achieve the HV output. An advantage of the Mosfets is no input current, Ib, which will give a small error term.I would actually suggest that a simulation would help answer which is best.

rogerw

On 3/27/2022 1:48 PM, Roger Whatley via groups.io wrote:

OK, here in words is how it works: (reference enclosed schematic)

M1 and the 500K resistor forms a negative feedback loop regulating the voltage on the 1Meg resistor to equal the Zener voltage, Vz. M2 is a diode-connected device to generate a matching and cancelling voltage to yhe input characteristic of M1. Since Ids2 flowing in 1Meg resistor also flows thru the 56Meg resistor, the voltage that M1 must maintain at the output is approximately (56M + 1M)/1M = 57 times the Zener voltage (and plus one Vgs voltage, a small error).

Analysis:

Starting at ground and summing voltages (Kirchoff's Voltage Law) around the loop we go up Vz volts, and up Vgs1 volts, and then down Vgs2 volts, and down by the IR drop on the 1Meg resistor, or,

Vz + Vgs1 = Vgs2 + (Ids2)* 1Meg, and re-arranging,

Vz = (Vgs2-Vgs1) + (Ids2)*1Meg. Now if we can assume M1 and M2 are matching Mosfets and operating at the same current (or some fixed ratio of current will give a relatively fixed error), then Vgs2-Vgs1 = 0, and

Vz = Ids2 *1Meg

Then, for Vz = 15V,

Ids2 = 15V/1Meg = 15uA

Vhvreg = Vz *(56M+1M)/1M = 15V * 57 = 855V

Ids1 + Ids2 = (Vhv - Vhvreg)/500K = 290uA

Ids1 = 290uA - 15uA = 275uA

I would suggest that the two Mosfets should be run at much closer currents for better matching and Vgs2=Vgs1, and also to avoid as much loading on the input Vhv= 1000V supply. I would suggest a 3Meg resistor for Ids1 = 30uA and Ids2=15uA.

rogerw

On 3/26/2022 10:18 PM, peter via groups.io wrote:

Geo:
What was posted is not a patent. It is Provisional Patent Application.
It is not reviewed by the PTO. It is typically used when someone is in a rush to publicly announce something and wants to get some IP protection.?
I searched for the same inventor name and didn't come up with a granted patent. ( it could have been filed differently, ie different primary inventor ?)
my crude simulation:

?Have a series resistance of 500k to 1M before the shunt
The upper resistor is 56M, lower resister is 1M
Zener is a 16V part, and Microcap says shunt voltage varies about 15 V over 0~ 100C range. Voltage goes up with temperature going down.
If I use a lower voltage "zener" diode the shunt voltage is not as stable.
So a 5.1V which is close to neutral is no good? and a lower value which is actually a zener is worse.
I used the same IRFAF22 n-ch MOSFET for the Microcap simulation. I think? the Mosfet that doesn't get the full applied voltage have can be different lower voltage rated part but I didn't try it yet.

I am still a newbie at using a simulator. May be if we are lucky Roger W will do an analysis.
Still the regulation was ok for a 100C temperature swing. May be with some tweaking it would get even better ?

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

Break the 56M resistor into two 23M resistors in series. A new mosfet M3 will have its gate connected at the new mid-resistor node. M1's drain will connect to M3's source, and M3's drain will connect to the junction of the 500K load resistor and the top 23M resistor. In this new cascoded configuration M1 and M3 only see half as much Vds as M1 in the original circuit.

This idea can? be extended to more cascoded devices.

rogerw

Roger:
?thanks.
BTW how would you cascode the MOSFET? in this application ?


? ? they are still being sold and expensive!
P

-- 
rogerw
On the Catclaw
ΜΟΛΩΝ ΛΑΒΕ

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

Roger:
?thanks.
BTW how would you cascode the MOSFET? in this application ?


? ? they are still being sold and expensive!
P

开云体育

correction:

"I would suggest that the two Mosfets should be run at much closer currents for better matching and Vgs2=Vgs1, and also to avoid as much loading on the input Vhv= 1000V supply. I would suggest a 3Meg resistor for Ids1 = 45uA and Ids2=15uA."

After re-reading my note, it may not read as clear as I had hoped. Questions are encouraged. Here are some clarifying edits:

M1 and the 500K load resistor forms a amplifier where the gate is the negative input and the source is the positive input. You could think of this as a crude opamp with both positive and negative inputs, but a huge offset voltage = Vgs1.

The output of this M1 amplifier connects back to the negative input (in a unity gain opamp connection) thru M2= Vgs2 voltage drop. And thus whatever input voltage the source (positive input) "sees" is applied to the 1Meg load resistor. But whatever voltage drop appears on the 1Meg load resistor must also appear 56X on the 56Meg feedback resistor, thus multiplying the input voltage from the Zener by (56+1)/1 = 57X times.

Thus the feedback loop is regulating the voltage on the 1Meg resistor to equal the Zener voltage, Vz. M2 is a diode-connected device to generate a matching and cancelling voltage to the input characteristic of M1. Since Ids2 flowing in 1Meg resistor also flows thru the 56Meg resistor, the voltage that M1 must maintain at the output is approximately (56M + 1M)/1M = 57 times the Zener voltage (and plus one Vgs voltage, a small error).

I hope that helps..... :-)

btw, I believe that the reason for choosing Mosfets is probably that they are most easily found as HV transistors. However, I think the circuit would be generally improved accuracy with bipolar HV transistors and if they cannot be found with high enough Vcemax, then two or more may easily be cascoded (ie, vertically stacked) to achieve the HV output. An advantage of the Mosfets is no input current, Ib, which will give a small error term.I would actually suggest that a simulation would help answer which is best.

rogerw

OK, here in words is how it works: (reference enclosed schematic)

M1 and the 500K resistor forms a negative feedback loop regulating the voltage on the 1Meg resistor to equal the Zener voltage, Vz. M2 is a diode-connected device to generate a matching and cancelling voltage to yhe input characteristic of M1. Since Ids2 flowing in 1Meg resistor also flows thru the 56Meg resistor, the voltage that M1 must maintain at the output is approximately (56M + 1M)/1M = 57 times the Zener voltage (and plus one Vgs voltage, a small error).

Analysis:

Starting at ground and summing voltages (Kirchoff's Voltage Law) around the loop we go up Vz volts, and up Vgs1 volts, and then down Vgs2 volts, and down by the IR drop on the 1Meg resistor, or,

Vz + Vgs1 = Vgs2 + (Ids2)* 1Meg, and re-arranging,

Vz = (Vgs2-Vgs1) + (Ids2)*1Meg. Now if we can assume M1 and M2 are matching Mosfets and operating at the same current (or some fixed ratio of current will give a relatively fixed error), then Vgs2-Vgs1 = 0, and

Vz = Ids2 *1Meg

Then, for Vz = 15V,

Ids2 = 15V/1Meg = 15uA

Vhvreg = Vz *(56M+1M)/1M = 15V * 57 = 855V

Ids1 + Ids2 = (Vhv - Vhvreg)/500K = 290uA

Ids1 = 290uA - 15uA = 275uA

I would suggest that the two Mosfets should be run at much closer currents for better matching and Vgs2=Vgs1, and also to avoid as much loading on the input Vhv= 1000V supply. I would suggest a 3Meg resistor for Ids1 = 30uA and Ids2=15uA.

rogerw

On 3/26/2022 10:18 PM, peter via groups.io wrote:

Geo:
What was posted is not a patent. It is Provisional Patent Application.
It is not reviewed by the PTO. It is typically used when someone is in a rush to publicly announce something and wants to get some IP protection.?
I searched for the same inventor name and didn't come up with a granted patent. ( it could have been filed differently, ie different primary inventor ?)
my crude simulation:

?Have a series resistance of 500k to 1M before the shunt
The upper resistor is 56M, lower resister is 1M
Zener is a 16V part, and Microcap says shunt voltage varies about 15 V over 0~ 100C range. Voltage goes up with temperature going down.
If I use a lower voltage "zener" diode the shunt voltage is not as stable.
So a 5.1V which is close to neutral is no good? and a lower value which is actually a zener is worse.
I used the same IRFAF22 n-ch MOSFET for the Microcap simulation. I think? the Mosfet that doesn't get the full applied voltage have can be different lower voltage rated part but I didn't try it yet.

I am still a newbie at using a simulator. May be if we are lucky Roger W will do an analysis.
Still the regulation was ok for a 100C temperature swing. May be with some tweaking it would get even better ?

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, here in words is how it works: (reference enclosed schematic)

M1 and the 500K resistor forms a negative feedback loop regulating the voltage on the 1Meg resistor to equal the Zener voltage, Vz. M2 is a diode-connected device to generate a matching and cancelling voltage to yhe input characteristic of M1. Since Ids2 flowing in 1Meg resistor also flows thru the 56Meg resistor, the voltage that M1 must maintain at the output is approximately (56M + 1M)/1M = 57 times the Zener voltage (and plus one Vgs voltage, a small error).

Analysis:

Starting at ground and summing voltages (Kirchoff's Voltage Law) around the loop we go up Vz volts, and up Vgs1 volts, and then down Vgs2 volts, and down by the IR drop on the 1Meg resistor, or,

Vz + Vgs1 = Vgs2 + (Ids2)* 1Meg, and re-arranging,

Vz = (Vgs2-Vgs1) + (Ids2)*1Meg. Now if we can assume M1 and M2 are matching Mosfets and operating at the same current (or some fixed ratio of current will give a relatively fixed error), then Vgs2-Vgs1 = 0, and

Vz = Ids2 *1Meg

Then, for Vz = 15V,

Ids2 = 15V/1Meg = 15uA

Vhvreg = Vz *(56M+1M)/1M = 15V * 57 = 855V

Ids1 + Ids2 = (Vhv - Vhvreg)/500K = 290uA

Ids1 = 290uA - 15uA = 275uA

I would suggest that the two Mosfets should be run at much closer currents for better matching and Vgs2=Vgs1, and also to avoid as much loading on the input Vhv= 1000V supply. I would suggest a 3Meg resistor for Ids1 = 30uA and Ids2=15uA.

rogerw

Geo:
What was posted is not a patent. It is Provisional Patent Application.
It is not reviewed by the PTO. It is typically used when someone is in a rush to publicly announce something and wants to get some IP protection.?
I searched for the same inventor name and didn't come up with a granted patent. ( it could have been filed differently, ie different primary inventor ?)
my crude simulation:

?Have a series resistance of 500k to 1M before the shunt
The upper resistor is 56M, lower resister is 1M
Zener is a 16V part, and Microcap says shunt voltage varies about 15 V over 0~ 100C range. Voltage goes up with temperature going down.
If I use a lower voltage "zener" diode the shunt voltage is not as stable.
So a 5.1V which is close to neutral is no good? and a lower value which is actually a zener is worse.
I used the same IRFAF22 n-ch MOSFET for the Microcap simulation. I think? the Mosfet that doesn't get the full applied voltage have can be different lower voltage rated part but I didn't try it yet.

I am still a newbie at using a simulator. May be if we are lucky Roger W will do an analysis.
Still the regulation was ok for a 100C temperature swing. May be with some tweaking it would get even better ?

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