High-voltage, high-speed pulse generator

[mbless]

I have a project requiring the pulse-echo ultrasound technique, so I've been looking around for fast, high-voltage pulse generators. I came across this paper "A Low-Cost, High-Performance Pulse Generator for Ultrasound Imaging" that looks perfect for my use, so I want to start out by reproducing the circuit and then work on a more modern version. I assume I'm not allowed to attach the paper here, but I'll post the relevant images.

I want to achieve pulses widths ranging from 20-500ns with voltages of -200 to -500V. The figures in the paper shows it can achieve a pulse width of 10ns with -110V peak (bottom left graph), or 40ns width with -350V (top right graph). It reaches -350V in ~20ns. These numbers look great to me, so that's why I want to start with replicating this design. The schematic is also quite simple.

The schematic has 3 parts: a high impedance input, a triple push-pull gate driver, and the output stage. The paper is from 2002, so only 2 of the 4 active components are still made, which gets to why I am posting.

I can't buy the Supertex (now Microchip) VP1304. There is the VP2106, which would complement the VN2106, but you can see the input capacitance increased from 20pF to 45pF which rise & fall times increasing by 2-3ns. I imagine this is close enough but wanted to get your thoughts since the original circuit was reaching -350V in 20ns.

The other part I can't buy is the 501N04 mosfet that was made by Directed Energy, who was bought by IXYS, who was then bought by Littlefuse. Well, Littlefuse has discontinued the IXYS RF line, so I have no chance of buying that mosfet. The closest one I can find is IXYS IXZH10N50L2A. Its typical gate threshold is 2V higher (5.4V vs. 3.4V) than the 501N04, and the input capacitance increases from 570pF to 611pF. Again, I'm worried the slightly worse specs will slow down the output slew rate. Also, it's a THT and not SMD, so I imagine the increased lead inductance won't help.

Thank you for any feedback! I also welcome suggestions of more modern designs that use different drive architecture or different mosfets like GaN FETs, etc.

[KE5FX]

I'd suggest taking a look around page 246 in the Art of Electronics x-Chapters volume.  It should give you some good pointers.  That particular page describes a 500V 20A 20 ns pulser made with currently-available parts.

The one in your paper will be more robust simply because the gate drivers won't tend to explode like Hill & Horowitz's initial attempt did.  They recommend TC4422 drivers for use at repetition rates near 10 MHz.

[mbless]

I'd suggest taking a look around page 246 in the Art of Electronics x-Chapters volume.  It should give you some good pointers.  That particular page describes a 500V 20A 20 ns pulser made with currently-available parts.

The one in your paper will be more robust simply because the gate drivers won't tend to explode like Hill & Horowitz's initial attempt did.  They recommend TC4422 drivers for use at repetition rates near 10 MHz.

I didn't know there was an x-chapters book, so I'll look into that.

Thanks for the reminder about repetition rate. Mine need is <10kHz.

[tggzzz]

These have been discussed on comp.arch EDIT: I meant sci.electronics.design (doh!); you might be able to pick up some hints there, but no complete circuits.

Look for posts by Win Hill and John Larkin.

[mbless]

I'd suggest taking a look around page 246 in the Art of Electronics x-Chapters volume.  It should give you some good pointers.  That particular page describes a 500V 20A 20 ns pulser made with currently-available parts.

It turns out that section is a sample they've posted on their website! Section 3x.15.3 is an negative high-voltage pulser, so I have an alternative to the one in my original post.

These have been discussed on comp.arch; you might be able to pick up some hints there, but no complete circuits.

I didn't know about that group. It's hard to search but I've found some interesting topics.

[tggzzz]

I'd suggest taking a look around page 246 in the Art of Electronics x-Chapters volume.  It should give you some good pointers.  That particular page describes a 500V 20A 20 ns pulser made with currently-available parts.

It turns out that section is a sample they've posted on their website! Section 3x.15.3 is an negative high-voltage pulser, so I have an alternative to the one in my original post.

These have been discussed on comp.arch; you might be able to pick up some hints there, but no complete circuits.

I didn't know about that group. It's hard to search but I've found some interesting topics.

I'm an idiot. I meant sci.electronics.design, not comp.arch. There is some high quality stuff in s.e.d, but there is an awful lot of crap too. Filtering on author "Win Hill" would be a good starting point.

[Marco]

The Hill designs are all push pull, if the load can just be relied upon to pull itself up again after the pulse such as in the circuit you posted in the original post push pull is needlessly complex.

Just replicate it, but with a modern fast driver instead of that whole chain of totem poles and a better MOSFET. There's no reason to find components close to what they used if there's simply better available.

[KE5FX]

I didn't know there was an x-chapters book, so I'll look into that.

It's a dangerous book.  Costs me several hours every time I open it, including yesterday when I looked up the chapter in question.   
 

[mbless]

Just replicate it, but with a modern fast driver instead of that whole chain of totem poles and a better MOSFET. There's no reason to find components close to what they used if there's simply better available.

It's a requirement to replicate a proven design to serve as a baseline, and this one was chosen for that... I was able to replicate it in LTspice. I tried different mosfets, but they aren't responding as I would expect based on their gate charge (smaller charges taking longer to turn on). I'm not sure if it's real or the models just aren't tuned for nanosecond response. I also tried some drivers rated for 4A+ sink and source current, and they were lower than the totem pole. According to LTspice the gate current was peaking 2A for the totem pole and 1.1A for the drivers, so not using the full 4A. Again, not sure if it's a spice problem.

At this point I'm just going to design a test PCB and pick various drivers and mosfets with the same package to see how they perform.

The Hill designs are all push pull, if the load can just be relied upon to pull itself up again after the pulse such as in the circuit you posted in the original post push pull is needlessly complex.

That's a good point. I'm looking at adopting the bottom of the design 3x.115 as it provides a good isolation and buffer stage.

[David Hess]

The other part I can't buy is the 501N04 mosfet that was made by Directed Energy, who was bought by IXYS, who was then bought by Littlefuse. Well, Littlefuse has discontinued the IXYS RF line, so I have no chance of buying that mosfet. The closest one I can find is IXYS IXZH10N50L2A. Its typical gate threshold is 2V higher (5.4V vs. 3.4V) than the 501N04, and the input capacitance increases from 570pF to 611pF. Again, I'm worried the slightly worse specs will slow down the output slew rate. Also, it's a THT and not SMD, so I imagine the increased lead inductance won't help.

That has been a continuing and worsening problem.  Many older power MOSFET families which had lower capacitance have been discontinued for more area efficient designs.

Thank you for any feedback! I also welcome suggestions of more modern designs that use different drive architecture or different mosfets like GaN FETs, etc.

I would use a completely different topology but you seem to be restricted to improving this one.

The first change I would make is to replace the original output MOSFET with a two MOSFET cascode configuration.  The high voltage cascode MOSFET then shields the high voltage swing from the drain of the driven low voltage MOSFET reducing the charge required to switch it yielding better high frequency performance.  And the driven MOSFET can be lower voltage anyway so lower capacitance for the driver to handle.

[mbless]

I would use a completely different topology but you seem to be restricted to improving this one.

After I have a working baseline, I can change as much as I want as long as it's reliable. So suggest away 

[David Hess]

I would use a completely different topology but you seem to be restricted to improving this one.

After I have a working baseline, I can change as much as I want as long as it's reliable. So suggest away 

I would replace the drain resistor with an active pull-up to make the output push-pull.  AC coupling from the drive to the lower transistor can drive the upper transistor.  However this change might not provide any advantage in your application.

An example of this sort of design can be found in the z-axis amplifier used for analog oscilloscopes which easily meets your performance requirements but with a much lighter load.  Z-axis amplifiers are also linear so stable in their linear range which is not required in a pulse amplifier which leaves open the possibility of faster performance.

[mbless]

I started laying out the baseline PCB and had some layout questions.

1. Is there an issue with a ground plane surrounding high-speed, high-current switching? This PCB is switching 500V into a 10Ohm resistor for 50-500ns with estimated 10ns edges. The attached images show a 3D view of the AC-coupled MOSFET gate driver, the top PCB layer (red) and the bottom layer (green). The blue line in the Top Layer image is the high-voltage path from the connector, through capacitors, through mosfet, and to resistor. Because the top layer is rather bare, I have a ground pour taking up the remaining space. (Note that I haven't done any via stitching yet.) My concern is if the fast, high-current switching will inject noise into the ground planes surrounding the high-current path. Most of what I have seen deals with removing ground planes around circuits sensitive to parasitic capacitance, so I am unsure of this scenario.

2. The Art of Electronics x Chapters discussed earlier have a high-voltage monitor circuit shown in the last attached image. What is intended with the "shield." Is it supposed to be a via-stitched ground trace surrounding the components with little openings for the input & output?