Author Topic: 3kV FET module with 10ns rise time  (Read 2220 times)

0 Members and 1 Guest are viewing this topic.

Offline LukeWTopic starter

  • Frequent Contributor
  • **
  • Posts: 686
3kV FET module with 10ns rise time
« on: January 13, 2019, 08:03:10 am »
I'd like to switch a 3kV supply and generate a square pulse with a rise time as fast as reasonably achievable.

Please give this a bit of peer review - is this realistically achievable? What passive component values should be used?

A Behlke HTS61-02 module is used, and a few components will be added around it - SHV connectors for the HV signals, connectors for the low-voltage signals, couple of resistors and capacitors.
They quote 5nS rise time, but if I can get even 10-12 nS I'd be happy.
The turn-off time constant is not critical, and the overall duty cycle is like 1%.

A 10uF decoupling capacitor is installed at the +5V supply to the module.
50 \$\Omega\$ RG58 cables are used throughout the system.
The pulse generator's output impedance is unknown.
(I would have suspected 50 \$\Omega\$, but installing a 50 \$\Omega\$ resistor destroys the signal so apparently it isn't.)
The output level is ~5V. I've connected it to the Behlke module via RG58, and left a placeholder for a possible terminator resistor at the trigger input of the module - not populated yet.

A GP02-40 idiot-proofing diode is put at the SHV connector that brings in the +3kV line, because these supplies can be switched to negative.
A 10nF HV capacitor provides decoupling at the +HV side of the module.

Short ground connections are used for everything - direct connections from the Behlke ground terminals to the metal chassis, with the SHVs directly chassis mounted close to the module.

The load is an electrode in vacuum for an ion-physics application - essentially a DC open circuit, with a small stray capacitance.
A simple 10M:100k voltage divider provides a 100x tap for measurement connected to a 'scope via RG58 cable - I know this is a bit quick and dirty and does not provide frequency compensation.
(Note: 10M resistors on diagram are actually 2M, total resistance of the 5x string is 10M.)

I'm not sure if the R5 / C3 snubber across the switch is needed. Place left for it but not installed at the moment.
Same for the output resistor R4, and the "discharge" resistor R3.

A small DC bias (say +100V) is applied to the output when the switch is off - a series resistor and HV diode are used to protect this power supply.
 

Offline Zero999

  • Super Contributor
  • ***
  • Posts: 20355
  • Country: gb
  • 0999
Re: 3kV FET module with 10ns rise time
« Reply #1 on: January 13, 2019, 10:27:32 am »
Wow, that module is a beast. How much did it cost?

I don't think the oscilloscope probe will work very well. How about using a 10M resistor and the 50Ω input on the oscilloscope or add a 50R resistor to the co-axial connector or use a terminator and T-piece? That will give 5mV per kV, which should be enough for most modern oscilloscopes. There will be no need for frequency compensation, since the transmission line will be terminated with its characteristic impedance.


The signal generator not being 50Ω might be an issue. When you say adding a 50Ω resistor to the output destroys the signal do you mean it goes down to 0V? If so it sounds like it has a very high impedance which isn't good. If the cable isn't correctly terminated, there could be reflections, causing it to trigger more than once. A buffer could be added right next to the signal generator's output, something like a MOSFET driver IC, to reduce its output impedance to near zero. The other end of the cable could be terminated with 50Ω to eliminate reflections.
 

Offline DaJMasta

  • Super Contributor
  • ***
  • Posts: 2392
  • Country: us
    • medpants.com
Re: 3kV FET module with 10ns rise time
« Reply #2 on: January 13, 2019, 05:12:46 pm »
Quote
The outside diameter of RG-58 is around 0.2 inches (5 mm). RG-58 weighs around 0.025 lb/ft (37 g/m), exhibits approximately 25 pF/ft (82 pF/m) capacitance and can tolerate a maximum of 300 V potential (1800 W).

From wikipedia, but it doesn't sound like you're going to find any RG-58 that is going to tolerate 3kV.
 

Online iMo

  • Super Contributor
  • ***
  • Posts: 5570
  • Country: va
Re: 3kV FET module with 10ns rise time
« Reply #3 on: January 13, 2019, 05:55:43 pm »
I would use 10kOhm (10x1k in series) at High-Z probe's tip with 50ohm termination (3kV/200=15V).
Readers discretion is advised..
 

Offline Yansi

  • Super Contributor
  • ***
  • Posts: 3930
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: 3kV FET module with 10ns rise time
« Reply #4 on: January 13, 2019, 05:59:56 pm »
Quote
The outside diameter of RG-58 is around 0.2 inches (5 mm). RG-58 weighs around 0.025 lb/ft (37 g/m), exhibits approximately 25 pF/ft (82 pF/m) capacitance and can tolerate a maximum of 300 V potential (1800 W).

From wikipedia, but it doesn't sound like you're going to find any RG-58 that is going to tolerate 3kV.

Not sure if much over 3kV, but RG58 can tolerate quite some voltage. Look harder.  ;)

Operating voltage 2.5kVrms:
http://www.rosnol.com/upload/product/RG-58C%EF%BC%8FU.pdf
or 1900Vrms: http://www.farnell.com/datasheets/2095749.pdf

etc..
 

Offline ahbushnell

  • Frequent Contributor
  • **
  • Posts: 751
  • Country: us
Re: 3kV FET module with 10ns rise time
« Reply #5 on: January 13, 2019, 06:12:41 pm »
I'd like to switch a 3kV supply and generate a square pulse with a rise time as fast as reasonably achievable.

Please give this a bit of peer review - is this realistically achievable? What passive component values should be used?

A Behlke HTS61-02 module is used, and a few components will be added around it - SHV connectors for the HV signals, connectors for the low-voltage signals, couple of resistors and capacitors.
They quote 5nS rise time, but if I can get even 10-12 nS I'd be happy.
The turn-off time constant is not critical, and the overall duty cycle is like 1%.

A 10uF decoupling capacitor is installed at the +5V supply to the module.
50 \$\Omega\$ RG58 cables are used throughout the system.
The pulse generator's output impedance is unknown.
(I would have suspected 50 \$\Omega\$, but installing a 50 \$\Omega\$ resistor destroys the signal so apparently it isn't.)
The output level is ~5V. I've connected it to the Behlke module via RG58, and left a placeholder for a possible terminator resistor at the trigger input of the module - not populated yet.

A GP02-40 idiot-proofing diode is put at the SHV connector that brings in the +3kV line, because these supplies can be switched to negative.
A 10nF HV capacitor provides decoupling at the +HV side of the module.

Short ground connections are used for everything - direct connections from the Behlke ground terminals to the metal chassis, with the SHVs directly chassis mounted close to the module.

The load is an electrode in vacuum for an ion-physics application - essentially a DC open circuit, with a small stray capacitance.
A simple 10M:100k voltage divider provides a 100x tap for measurement connected to a 'scope via RG58 cable - I know this is a bit quick and dirty and does not provide frequency compensation.
(Note: 10M resistors on diagram are actually 2M, total resistance of the 5x string is 10M.)

I'm not sure if the R5 / C3 snubber across the switch is needed. Place left for it but not installed at the moment.
Same for the output resistor R4, and the "discharge" resistor R3.

A small DC bias (say +100V) is applied to the output when the switch is off - a series resistor and HV diode are used to protect this power supply.
Your resistive voltage probe will not have the bandwidth to measure the rise time.  Rule of thumb.  RC/4 is the probe rise time where C is the stray capacitance on the resistors to ground.  10M*10 pF/4=25 us.  It's possible to put resistors and capacitors in parallel to over come the stray capacitance.  But there will be a capacitive load on your experiment.  Also at these time scales the capacitance of your experiment may cause ringing with the inductance in the switch.

 

Offline PartialDischarge

  • Super Contributor
  • ***
  • Posts: 1625
  • Country: 00
Re: 3kV FET module with 10ns rise time
« Reply #6 on: January 13, 2019, 06:13:35 pm »
Quote
The outside diameter of RG-58 is around 0.2 inches (5 mm). RG-58 weighs around 0.025 lb/ft (37 g/m), exhibits approximately 25 pF/ft (82 pF/m) capacitance and can tolerate a maximum of 300 V potential (1800 W).

From wikipedia, but it doesn't sound like you're going to find any RG-58 that is going to tolerate 3kV.
3kV is fine for a rg58, but it mostly depends on the energy of the source (low in this case) and the frequency, clearly energy will be lost capacitively by those fast pulses anyway.
 

Online ejeffrey

  • Super Contributor
  • ***
  • Posts: 4027
  • Country: us
Re: 3kV FET module with 10ns rise time
« Reply #7 on: January 13, 2019, 06:16:26 pm »
The pulse generator's output impedance is unknown.
(I would have suspected 50 \$\Omega\$, but installing a 50 \$\Omega\$ resistor destroys the signal so apparently it isn't.)

These types of switches are not meant to drive impedance matched loads.  Think about it, 3 kV into 50 ohms would be 200 kW.  They are designed to deliver very high peak current into a capacitive load.  Usually you need to keep the leads to the load as absolutely short as possible.  I have used similar drivers for driving Pockels cells, and the switch and modulator are built into a module with short flying leads (3-4 inch) to the electrode terminals to minimize extra capacitance.  SHV coax was used for the DC power input, not for the high speed pulse.

Since your load is in a vacuum chamber, your life is a lot more difficult.  You still want to get your pulse generator as close as possible to the load.  You will get the best risetime performance by using separate unshielded + and - leads going through their own individual vacuum feedthroughs and keeping them as short as possible.  Unfortunately this will be terrible for EMI, especially for the part outside the vacuum.  The vacuum can itself provides shielding, but of course if you have any other electronics inside the vacuum they will see it.  If you can literally bolt the switch enclosure to the metal vacuum chamber with the feedthroughs inside the enclosure, that would be best.
 


Offline Zero999

  • Super Contributor
  • ***
  • Posts: 20355
  • Country: gb
  • 0999
Re: 3kV FET module with 10ns rise time
« Reply #9 on: January 13, 2019, 06:55:09 pm »
Quote
The outside diameter of RG-58 is around 0.2 inches (5 mm). RG-58 weighs around 0.025 lb/ft (37 g/m), exhibits approximately 25 pF/ft (82 pF/m) capacitance and can tolerate a maximum of 300 V potential (1800 W).

From wikipedia, but it doesn't sound like you're going to find any RG-58 that is going to tolerate 3kV.
But he's not using the co-axial cable or HV, so that doesn't matter.

I would use 10kOhm (10x1k in series) at High-Z probe's tip with 50ohm termination (3kV/200=15V).
The power dissipated in the 10k resistor would be enormous at 3kV. P = 30002/10×103 = 900W. I know it's a low duty cycle, but it's still a lot.

The pulse generator's output impedance is unknown.
(I would have suspected 50 \$\Omega\$, but installing a 50 \$\Omega\$ resistor destroys the signal so apparently it isn't.)

These types of switches are not meant to drive impedance matched loads.  Think about it, 3 kV into 50 ohms would be 200 kW.  They are designed to deliver very high peak current into a capacitive load.  Usually you need to keep the leads to the load as absolutely short as possible.  I have used similar drivers for driving Pockels cells, and the switch and modulator are built into a module with short flying leads (3-4 inch) to the electrode terminals to minimize extra capacitance.  SHV coax was used for the DC power input, not for the high speed pulse.

Since your load is in a vacuum chamber, your life is a lot more difficult.  You still want to get your pulse generator as close as possible to the load.  You will get the best risetime performance by using separate unshielded + and - leads going through their own individual vacuum feedthroughs and keeping them as short as possible.  Unfortunately this will be terrible for EMI, especially for the part outside the vacuum.  The vacuum can itself provides shielding, but of course if you have any other electronics inside the vacuum they will see it.  If you can literally bolt the switch enclosure to the metal vacuum chamber with the feedthroughs inside the enclosure, that would be best.
I think he's talking about the pulse generator used to drive the switch, not the switch itself.

If the load is capacitive, a series resistor with the same value as the transmission line's characteristic impedance should absorb the reflection when it gets back to the source.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf