48V to 2000V DC-DC smallest design

Its been a while but once again looking at a HV pulser type application.

It will charge up about 10nF to 2kV with about 1mA.  Then keep it toped
off as that cap discharges into another capacitive much smaller
capacitive load.   The 10nF will loose 100V or so then need to be
recharged witn 1mA before the next burst.

It need to be a custom approach due to environment constraints. Cost is
not a strong consideration but must be reasonable I can use custom
transformers etc.

The voltage does need to be variable from 1K to 2K.

no input to output ground isolation required.  but sometime the load
shorts and the supply must go into current limit and not damage itself
when this happens.

looking for small topologies to do this.

1) boost inductor feeding a voltage multiplier (8-10 stage)
	I have had 1A diodes blow when a previous VM design was shorted from
1.5kV.   even with a 100K resistor in series with the output.  never
investigated jsut sured up the source of the arcing and moved on with
fingers crossed.
  
2) flyback to do most of the boost with a doubler or tripler on the
output.  This should keep the turns ratio reasonable.

3) straight pushpull making use of the primary voltage doubling action
to get soem volatge gain.  PWM the center tap.

any other physically small power stage topologies to look into?

thanks




  

How small is small?

I made this without any particular care of size,
https://www.seventransistorlabs.com/Images/HVPower2.jpg
https://www.seventransistorlabs.com/Images/HVPower.jpg
(What, a legible, hand drawn schematic?  On SED? :^) )

Easy enough to fix up the DC supply side for efficiency (buck reg), fixed 
output, and drop the sense amp that's not needed.

The chopper is kinda here-nor-there, but I do recommend a resonant type in 
any case as the secondary capacitance is unavoidable, and the Baxandall 
oscillator is fairly foolproof as oscillators go.

Also, that was 12V input, but 48 isn't any trouble with appropriate changes.

It seems there aren't many resonant controllers for smaller applications 
(like DC-DC modules), so you may have to make a discrete or digital 
controller if you want to go that route more formally.

Probably a CCFL transformer would do, no need for custom transformers 
surprisingly enough.

If you need it small as balls*, you can stick CSP package GaN transistors to 
the backside of a CCFL transformer, use a QFN or BGA controller (which 
again, might end up sorta custom, an FPGA or something), a few ceramic caps, 
probably a 4 or at most 6 layer board, and be done.

*Erm, huh... literally about right, I think.  You'd take up the other ball 
for the reservoir cap.  And some insulation which might just wrap around the 
whole thing like a scrotum.  I'll, uh, stop now.

If that's too extreme, any combination of Si components and leaded or 
no-lead parts will do, requiring only somewhat more board area.

Tim

-- 
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Design
Website: https://www.seventransistorlabs.com/

"mook Jonhon" <mook@mook.net> wrote in message 
news:4o17F.80540$OU6.70486@fx40.iad...
>
> Its been a while but once again looking at a HV pulser type application.
>
> It will charge up about 10nF to 2kV with about 1mA.  Then keep it toped
> off as that cap discharges into another capacitive much smaller
> capacitive load.   The 10nF will loose 100V or so then need to be
> recharged witn 1mA before the next burst.
>
> It need to be a custom approach due to environment constraints. Cost is
> not a strong consideration but must be reasonable I can use custom
> transformers etc.
>
> The voltage does need to be variable from 1K to 2K.
>
> no input to output ground isolation required.  but sometime the load
> shorts and the supply must go into current limit and not damage itself
> when this happens.
>
> looking for small topologies to do this.
>
> 1) boost inductor feeding a voltage multiplier (8-10 stage)
> I have had 1A diodes blow when a previous VM design was shorted from
> 1.5kV.   even with a 100K resistor in series with the output.  never
> investigated jsut sured up the source of the arcing and moved on with
> fingers crossed.
>
> 2) flyback to do most of the boost with a doubler or tripler on the
> output.  This should keep the turns ratio reasonable.
>
> 3) straight pushpull making use of the primary voltage doubling action
> to get soem volatge gain.  PWM the center tap.
>
> any other physically small power stage topologies to look into?
>
> thanks
>
>
>
>
>
>
> 

On Wed, 21 Aug 2019 01:46:40 GMT, "mook Jonhon" <mook@mook.net> wrote:

>
>Its been a while but once again looking at a HV pulser type application.
>
>It will charge up about 10nF to 2kV with about 1mA.  Then keep it toped
>off as that cap discharges into another capacitive much smaller
>capacitive load.   The 10nF will loose 100V or so then need to be
>recharged witn 1mA before the next burst.
>
>It need to be a custom approach due to environment constraints. Cost is
>not a strong consideration but must be reasonable I can use custom
>transformers etc.
>
>The voltage does need to be variable from 1K to 2K.
>
>no input to output ground isolation required.  but sometime the load
>shorts and the supply must go into current limit and not damage itself
>when this happens.
>
>looking for small topologies to do this.
>
>1) boost inductor feeding a voltage multiplier (8-10 stage)
>	I have had 1A diodes blow when a previous VM design was shorted from
>1.5kV.   even with a 100K resistor in series with the output.  never
>investigated jsut sured up the source of the arcing and moved on with
>fingers crossed.
>  
>2) flyback to do most of the boost with a doubler or tripler on the
>output.  This should keep the turns ratio reasonable.
>
>3) straight pushpull making use of the primary voltage doubling action
>to get soem volatge gain.  PWM the center tap.
>
>any other physically small power stage topologies to look into?
>
>thanks

I did this out of standard parts we had in stock. A couple more diode
stages would get it to 2KV.

https://www.dropbox.com/s/e3n5af9sw1a1flh/28S840A_3.pdf?dl=0

With a different, maybe custom, transformer and higher voltage diodes
it could be smaller. I think Coilcraft has some HV type transformers.

I have the LT Spice model around here somewhere.

On Wednesday, August 21, 2019 at 11:46:45 AM UTC+10, mook Jonhon wrote:
> Its been a while but once again looking at a HV pulser type application.
> 
> It will charge up about 10nF to 2kV with about 1mA.  Then keep it toped
> off as that cap discharges into another capacitive much smaller
> capacitive load.   The 10nF will loose 100V or so then need to be
> recharged witn 1mA before the next burst.
> 
> It need to be a custom approach due to environment constraints. Cost is
> not a strong consideration but must be reasonable I can use custom
> transformers etc.
> 
> The voltage does need to be variable from 1K to 2K.
> 
> no input to output ground isolation required.  but sometime the load
> shorts and the supply must go into current limit and not damage itself
> when this happens.
> 
> looking for small topologies to do this.
> 
> 1) boost inductor feeding a voltage multiplier (8-10 stage)
> 	I have had 1A diodes blow when a previous VM design was shorted from
> 1.5kV.   even with a 100K resistor in series with the output.  never
> investigated jsut sured up the source of the arcing and moved on with
> fingers crossed.
>   
> 2) flyback to do most of the boost with a doubler or tripler on the
> output.  This should keep the turns ratio reasonable.
> 
> 3) straight push-pull making use of the primary voltage doubling action
> to get some voltage gain.  PWM the center tap.
> 
> any other physically small power stage topologies to look into?

http://www.sophia-electronica.com/Baxandall_parallel-resonant_Class-D_oscillator1.htm

The circuit diagram at the bottom of the page shows a circuit which looks like option 3).

http://sophia-elektronica.com/PMT-transformer.html

talks about the transformer (and it's stray capacitance).

http://sophia-elektronica.com/website_pmt_psu1.htm

gives the .asc file (which you'd have to rename as a .asc file).

Starting from a +45V supply would make the transformer design a lot easier.

The MOSFETS at M1 and M2 would see a lot higher drain voltage when off - up to about 150V - and the scheme to mark-to-space the gate drives for M3 and M4 gets marginally trickier (not that I bothered working that out).

None of it would get hot, so keeping it small wouldn't be difficult.

IIRR my circuit ran at about 50kHz, limited by the stray capacitance in the secondary. The mark-to-space sequences had much the same frequency, which doesn't necessarily allow for particularly fine control of output voltage, if they are the same from one cycle to the next, but by alternating between two adjacent mark-to-space ratio's you can get finer control as long as the low pass filtering at the output averages out the alternation.

-- 
Bill Sloman, Sydney

"mook Jonhon" <mook@mook.net> writes:

> Its been a while but once again looking at a HV pulser type application.
>
> It will charge up about 10nF to 2kV with about 1mA.  Then keep it toped
> off as that cap discharges into another capacitive much smaller
> capacitive load.   The 10nF will loose 100V or so then need to be
> recharged witn 1mA before the next burst.

Sounds a bit familiar - one our design discharged 47nF from 1000V to 4V
at 300Hz. We use a flyback step up from 12-16V to 4*250VDC, which are
then put in series. The transformer is a custom design. Starting from
48VDC should be easier, but I think 8*250V windings gets complicated, so
you'd have to take care of higher secondary voltages with this approach,
which increase the size.

There are also provisions for short circuit protection, low voltages,
flyback behaviour at low voltages etc.

--
mikko

On a sunny day (Wed, 21 Aug 2019 01:46:40 GMT) it happened "mook Jonhon"
<mook@mook.net> wrote in <4o17F.80540$OU6.70486@fx40.iad>:

>
>Its been a while but once again looking at a HV pulser type application.
>
>It will charge up about 10nF to 2kV with about 1mA.  Then keep it toped
>off as that cap discharges into another capacitive much smaller
>capacitive load.   The 10nF will loose 100V or so then need to be
>recharged witn 1mA before the next burst.
>
>It need to be a custom approach due to environment constraints. Cost is
>not a strong consideration but must be reasonable I can use custom
>transformers etc.
>
>The voltage does need to be variable from 1K to 2K.
>
>no input to output ground isolation required.  but sometime the load
>shorts and the supply must go into current limit and not damage itself
>when this happens.
>
>looking for small topologies to do this.
>
>1) boost inductor feeding a voltage multiplier (8-10 stage)
>        I have had 1A diodes blow when a previous VM design was shorted from
>1.5kV.   even with a 100K resistor in series with the output.  never
>investigated jsut sured up the source of the arcing and moved on with
>fingers crossed.
>  
>2) flyback to do most of the boost with a doubler or tripler on the
>output.  This should keep the turns ratio reasonable.
>
>3) straight pushpull making use of the primary voltage doubling action
>to get soem volatge gain.  PWM the center tap.
>
>any other physically small power stage topologies to look into?
>
>thanks

Not sure maybe I did not read it right what your power requirements are.

9 V to 500 V for GM tube
 http://panteltje.com/panteltje/pic/gm_pic/

5 V to 1250 V for PMT
 http://panteltje.com/panteltje/pic/sc_pic/

5 V to 1750 V for PMT
 http://panteltje.com/pub/PMT_regulated_power_supply_diagram_img_3182.jpg

etc etc

Note the last one is a SINE oscillator, just a few turns on an E core
Stabilizing is via supply of the oscillator.
 http://panteltje.com/pub/PMT_HV_supply_with_regulator_img_3175.jpg
it all depends, takes half an hour to wind a core like that.

Not always is flyback the way to go, harmonics...

mook Jonhon wrote...
>
> 2) flyback to do most of the boost with a
> doubler or tripler on the output.  This
> should keep the turns ratio reasonable.

 My flyback approach, previously detailed here
 on s.e.d., is different.  Three flyback stages,
 each with its own MOSFET, all running from one
 controller and gate driver.  The DC input-V of
 each stage is the previous stage's DC output.
 Starting with say 12V, you need a 167x stepup.
 Three 5.5x stages gets you to 2kV, and that's
 a pretty mild step-up ratio.  66V, 363V, 2kV.
 Stage currents and inductor values scale, since
 they all run with the same time parameters.

 Three feedback taps, each with a diode to the
 controller's FB pin.  The highest one controls,
 to prevent any one stage from going excessively
 over its voltage limit, but the last stage gets
 the controlling vote.  Very simple. Except for
 HV winding technique of the 3rd inductor.  :-)


-- 
 Thanks,
    - Win

Tim Williams wrote:

> How small is small?
> 
> I made this without any particular care of size,
> https://www.seventransistorlabs.com/Images/HVPower2.jpg
> https://www.seventransistorlabs.com/Images/HVPower.jpg
> (What, a legible, hand drawn schematic?  On SED? :^) )
> 
> Easy enough to fix up the DC supply side for efficiency (buck reg),
> fixed output, and drop the sense amp that's not needed.
> 
> The chopper is kinda here-nor-there, but I do recommend a resonant
> type in any case as the secondary capacitance is unavoidable, and the
> Baxandall oscillator is fairly foolproof as oscillators go.
> 
> Also, that was 12V input, but 48 isn't any trouble with appropriate
> changes.
> 
> It seems there aren't many resonant controllers for smaller
> applications (like DC-DC modules), so you may have to make a discrete
> or digital controller if you want to go that route more formally.
> 
> Probably a CCFL transformer would do, no need for custom transformers
> surprisingly enough.
> 
> If you need it small as balls*, you can stick CSP package GaN
> transistors to the backside of a CCFL transformer, use a QFN or BGA
> controller (which again, might end up sorta custom, an FPGA or
> something), a few ceramic caps, probably a 4 or at most 6 layer
> board, and be done.
> 
> *Erm, huh... literally about right, I think.  You'd take up the other
> ball for the reservoir cap.  And some insulation which might just
> wrap around the whole thing like a scrotum.  I'll, uh, stop now.
> 
> If that's too extreme, any combination of Si components and leaded or
> no-lead parts will do, requiring only somewhat more board area.
> 
> Tim


Thanks Tim,

I see you used a royer (push pull) with a voltage doubler on the back
side.

by small I need something to fit in a 1" v 2" board with the tallest
components about 1/4" above or below the board.   It could be 1/2" tall
as the board sits on 1/4" standoffs.   think pot core bolted to teh
chassis on one end of the board.   maybe some wiggle room to 2.5" in
length.


I will need ot run at highish frequencies (300+Khz) to get the
magnetics down but the continuous power levels will be low.


what you are suggesting is where I was going.   Just making sure there
was not trick or "old magic" that I was not aware of to simply make
high votlage from low.  :)

thanks

mook Jonhon wrote:

> 
> Its been a while but once again looking at a HV pulser type
> application.
> 
> It will charge up about 10nF to 2kV with about 1mA.  Then keep it
> toped off as that cap discharges into another capacitive much smaller
> capacitive load.   The 10nF will loose 100V or so then need to be
> recharged witn 1mA before the next burst.
> 
> It need to be a custom approach due to environment constraints. Cost
> is not a strong consideration but must be reasonable I can use custom
> transformers etc.
> 
> The voltage does need to be variable from 1K to 2K.
> 
> no input to output ground isolation required.  but sometime the load
> shorts and the supply must go into current limit and not damage itself
> when this happens.
> 
> looking for small topologies to do this.
> 
> 1) boost inductor feeding a voltage multiplier (8-10 stage)
> 	I have had 1A diodes blow when a previous VM design was shorted from
> 1.5kV.   even with a 100K resistor in series with the output.  never
> investigated jsut sured up the source of the arcing and moved on with
> fingers crossed.
>   
> 2) flyback to do most of the boost with a doubler or tripler on the
> output.  This should keep the turns ratio reasonable.
> 
> 3) straight pushpull making use of the primary voltage doubling action
> to get soem volatge gain.  PWM the center tap.
> 
> any other physically small power stage topologies to look into?
> 
> thanks
> 
> 
> 
> 
>   

All great ideas and food for thought.   great to see this group is
still a great resource.   I tell my EE friends about it and that look
at me and ask "USEnet... whats that"   :)

On 21/08/2019 9:34 am, Winfield Hill wrote:
> mook Jonhon wrote...
>>
>> 2) flyback to do most of the boost with a
>> doubler or tripler on the output.  This
>> should keep the turns ratio reasonable.
> 
>   My flyback approach, previously detailed here
>   on s.e.d., is different.  Three flyback stages,
>   each with its own MOSFET, all running from one
>   controller and gate driver.  The DC input-V of
>   each stage is the previous stage's DC output.
>   Starting with say 12V, you need a 167x stepup.
>   Three 5.5x stages gets you to 2kV, and that's
>   a pretty mild step-up ratio.  66V, 363V, 2kV.
>   Stage currents and inductor values scale, since
>   they all run with the same time parameters.
> 
>   Three feedback taps, each with a diode to the
>   controller's FB pin.  The highest one controls,
>   to prevent any one stage from going excessively
>   over its voltage limit, but the last stage gets
>   the controlling vote.  Very simple. Except for
>   HV winding technique of the 3rd inductor.  :-)
> 
> 

But surely that triple cascaded flyback topology you suggest would need 
a MOSFET with 2000V Vds rating?

piglet