Author Topic: OP Supply Bootstrapping: Substituting the p-Mosfet against an n-Mosfet possible?  (Read 1738 times)

0 Members and 1 Guest are viewing this topic.

Online Echo88Topic starter

  • Frequent Contributor
  • **
  • Posts: 866
  • Country: de
Its possible to design a +-1kVout OP-Amp-bootstrapped supplies with series connected BJTs/MOSFETs.
See here for a usual design: https://www.analog.com/media/en/technical-documentation/application-notes/an67f.pdf Page 58 Figure 79 and a series connected HV-design: http://bbs.38hot.net/thread-11133-1-1.html or https://www.analog.com/media/en/technical-documentation/application-notes/an87f.pdf Page 58 Figure 84
Since n-MOSFETs are easily available with more than 1kV Breakdownvoltage, while p-MOSFETs are only available up to 600V, theres no chance to achieve +-1kV without using series connected p-MOSFETs or BJTs.
IGBTs and SiC-MOSFETs might also be interesting, but they are more difficult to get and p-Channel-variants are rare and therefore thats not a really an alternative.

But im interested if theres a solution by using only two high voltage n-MOSFETs, instead of a n- and p-MOSFET.
I just cant wrap my head around how to control the n-MOSFET, which regulates the negative supply by using the OP-output.
Maybe someone can give a suggestion how that may be doable?
Simulation-file from AN67 is attached.

Thanks.
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7453
  • Country: pl
You can't easily control a gate which is anywhere between 1V and 1000V below your bootstrapped VEE. Maybe some kind of PNP/P-ch cascode to send a control current down there, but then you need an HV device again. Perhaps an optocoupler.

Alternatively, don't control it at all. Set up an N-ch current sink pulling your VEE down to ground and bootstrap yourself by sinking current from HV through a bootstrapped N-ch cascode and dumping it onto VEE to saturate that CCS.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Use the n-channel source follower as usual on the positive side but use an n-channel common source current sink on the bottom with a shunt regulator across the operational amplifier circuits to protect them.  AC coupling to the current sink may be used to improve performance.

Update:

I might use optocoupled feedback from the shunt regulator to the n-channel low side current sink to adjust the current for constant current through the shunt regulator even with load changes.
« Last Edit: December 09, 2019, 02:49:42 am by David Hess »
 

Online Marco

  • Super Contributor
  • ***
  • Posts: 7043
  • Country: nl
Do you want a two quadrant DC power supply or a high voltage amplifier?
« Last Edit: December 09, 2019, 01:49:53 am by Marco »
 

Online Echo88Topic starter

  • Frequent Contributor
  • **
  • Posts: 866
  • Country: de
I want a voltage buffer, which enables me to measure +-1kVmax without loading the source (using a pA-bias current OP-Amp) and being able to source (sinking not necessary) a few mA to attach a 8.5 digit DMM for example.
Since the DMM would load the source with its 10M input resistance on 100V-range and 1kV-range, it cant be used to directly measure voltages across capacitors for example.
At the moment its just a simulation and i was pondering wether the bootstrap-circuit could be build like asked in my first post.
Seems the necessity for an isolated mosfet-gate controller/optocoupler + current sink will lead to more necessary parts, than the standard series connected mosfet solution.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
At the moment its just a simulation and i was pondering wether the bootstrap-circuit could be build like asked in my first post.
Seems the necessity for an isolated mosfet-gate controller/optocoupler + current sink will lead to more necessary parts, than the standard series connected mosfet solution.

Active drive with the low side current sink is only required to support higher power levels with higher efficiency.  For a buffer application, the low side current sink can be fixed and a zener diode used for the shunt regulator.

The tricky part then becomes how to level shift the output positive to drive the high side source follower if a high side current source is not available.  I remember one design which just used batteries in series since the MOSFET gate current is essentially zero.  Another way is to use a DC restorer circuit like you would find in a CRT circuit; pump charge through capacitors and diodes to make a floating bias supply.  A high voltage depletion mode MOSFET could be useful to make things very simple.

The best way may be to use an isolated low voltage supply to power both sides of the bootstrapped buffer.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
N-only amplifiers?  Look deep into history, back when "N" type was all we had -- vacuum tubes.  There are various totem-pole designs out there.  You will find performance limits are driven by the necessarily high-impedance gain node driving the high-side device.

The "mu follower" is probably a likely suspect.  Normally configured as a common-cathode amplifier with gyrator load, which doubles as a cathode follower.  (Gyrator meaning, the network converts a capacitance into an inductance -- that is, the high side load has a low impedance at DC, but a high impedance at high frequencies where its gain acts to bootstrap its resistance.  A rising impedance versus frequency, has inductive phase.)  You might change things around to vary the high side DC bias as well (either by just doing it directly, or with a parallel amp stage), and also think about improving pull-down slew rate by bypassing the follower with a diode (so the first amp can pull down the output directly).

Also may be worth researching classic logic and analog circuits: 7400 TTL is all NPN, and used an arrangement of transistors turning an internal voltage or current signal into alternate pull-up and -down actions.  Very early op-amps were mostly or all NPN as well (i.e., before lateral PNP was developed; 1960s).  I don't have any example part numbers, perhaps others can hint?

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7453
  • Country: pl
I just realized you want ±1kV. Out of curiosity I looked at the prices of 2kV FETs and I must say that this ladder circuit by lymex doesn't even look that bad anymore :P
 

Offline Micke

  • Regular Contributor
  • *
  • Posts: 101
  • Country: se
I bought a defective calibrator capable of +/- 1200V and 1000V 1kHz. Output current 10mA
The HV part was faulty, wrong HW transformer so I had to special make a 2x1270VAC 70VA transformer, not super cheap   :(
To make it even more funny, the HV amplifier was also broken and potted in epoxi, grrrr
But luckily I got the service manual from a friendly forum member, so I did a clone of the HV amplifier.
In the end I could repair the original potted version, they did not pot secondary side of PCB, there were some almost invisible cracked solder joints.
Anyway, I ordered 100pcs transistors MJE3439 from Mouser rather cheap and build the clone, worked perfectly!

Output voltage is controlled with current to the optocoupler, on my prototype:
0mA= -1479V
1,13mA = -1000V
1,71mA = -500V
2,25mA = 0V
2,79mA = 500V
3,32mA = 1000V
4,05mA = 1573V

I have Gerber files if you want, or PCB.


 
The following users thanked this post: thm_w

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
When I consider the difficulties and performance compromises in an all n-channel design, it makes a stacked complementary circuit seem reasonable and especially so if a single n-channel transistor will only achieve marginal derating.

Still, I think the all n-channel circuit I described will work.
 

Offline Micke

  • Regular Contributor
  • *
  • Posts: 101
  • Country: se
To return to the subject using all N-MOSFET, newer versions of the calibrator had MOSFET HV amplifier instead. Apparently it was prone to errors, the guy that helped me has repaired his HV amplifier several times. He did some reverse engineering, I have just drawn it from his handwritten notes to CAD. Less component count rather than the bipolar version with 48 NPN transistors! (But bipolar is perhaps more robust after all?) Anyhow, might be useful to get inspiration from!  ;)
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
The schematic is not quite right but I can see the general idea.  It would not surprise me if it had dynamic stability problems with all of those high value unbypassed resistor chains.  High voltage amplifiers usually include either series (transconductance) or shunt (transimpedance) local feedback.  Cascode amplifiers like that can have surprisingly high bandwidth and clean response.

Bipolar transistors have a big cost advantage at high voltages so I am surprised anybody would use that many MOSFETs.  With a 3600 volt total supply, even p-channel MOSFETs could have been used which gives the possibility of a higher performance design.

 

Offline Micke

  • Regular Contributor
  • *
  • Posts: 101
  • Country: se
Thanks for the analysis David!
Hmm, might have done some errors redrawing or the original had some errors...
Anyhow, I stick with my 48pcs NPN bipolar version!   :)
Having a proper PCB, not that difficult with that many components anyway.
But with a quiescent current of 7mA at 3600V that is 25W just in idle power  :scared:
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Oscilloscope z-axis amplifiers which drive the grid to control CRT brightness run at about 100 volts with a similar configuration with up to two transistors on the bottom and two transistors on the top to provide a delay in the 10s of nanoseconds range and rise/fall times in the range of several nanoseconds.  The big difference is that they use local shunt feedback so operate as a current in and voltage out transimpedance amplifier.

Two transistors are required because the high frequency transistors are not high voltage and the high voltage transistors are not high frequency.

I have seen at least one 50 MHz class-AB output amplifier for a function generator which also used a similar configuration.  Very high power audio amplifiers may also use cascode output stages.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf