1-5ns Half bridge into 5-10pF load with 2,5kV Voltage on H bridge / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

1-5ns Half bridge into 5-10pF load with 2,5kV Voltage on H bridge

1 2 next

Move Thread

LAN_403

Linas

Tue Aug 18 2015, 11:31AM

Registered Member #1143 Joined: Sun Nov 25 2007, 04:55PM
Location: Vilnius, Lithuania
Posts: 721

Hello, I need to make half bridge device, that can charge and discharge 5-10pF capacitor with sub 5ns front on each side. Also, it must work at least to 200kHz to 2MHz repetition rate.

Control is extra simple just square wave

Small reward will be given, if some one will give me good start that will lead to device prototype.

This part is a good start, but in order to get 2500V switching voltage, i will need to use 5 in series. But how to drive this monster ? I only care of getting 5ns front, i can deal with delay with FPGA

Wolfram

Tue Aug 18 2015, 01:58PM

Registered Member #33 Joined: Sat Feb 04 2006, 01:31PM
Location: Norway
Posts: 971

The first challenge is the gate driver. To charge the gate in a couple of nanoseconds, you need around 5 amps of gate charge current. This is not really hard to find in a gate driver, but most gate drivers don't have a fast enough output rise time. The more powerful ones are often slower, and the best I am able to find quickly is the IXDD604. The graphs in the datasheet suggest a typical rise time in the order of 10ns. Since the transistor will start switching when the gate reaches a few volts, 10ns might be fast enough for the gate rise time.

Edit: The LM5114 also looks like a usable candidate, being slightly faster. It can't source as much current though so it might be no good.

I looked into using fast CMOS buffers to drive the gates, but you need very many in parallel to get the 5A required to charge the gate fast enough.

For the isolation, I would either use fast digital isolators (capacitive or inductive isolation barrier, available from Analog, Silabs and many others) or high bandwidth transformers like ethernet transformers. How long does the pulse need to be and what is the repetition rate?

A possible elegant solution could be the Si8235, which contains two channels of isolation plus gate drive. Its output rise time is somewhat worse than the IXDD604, but it could be worth a try. It all comes down to how fast gate rise time you need to get 5 ns output rise time. Do you have some of the GaN FETs to test with?

I notice that the devices themselves have significantly higher output capacitance than your load capacitance. If you can get away with some smaller output devices, like the TPH3002, it might be easier to get your required rise time.

...

Tue Aug 18 2015, 03:05PM

Registered Member #56 Joined: Thu Feb 09 2006, 05:02AM
Location: Southern Califorina, USA
Posts: 2445

Getting it to work at 2MHz will be tricky, you will have ~100w of reactive power into the load and probably much more running around in the switches (considering any transistor capable of switching your 10kw peak power is going to have much more than 10pf of internal capacitance). For inspiration you might look to commercial supplies that do it, ex

I believe will meet your specs.

The transform device might do the trick, I have never personally worked with their high voltage GaN, although I knew a few people at transphorm back when I was at UCSB who generally seemed to think highly of their products. Good luck finding them for sale, you may be able to sample a few of them if you are still associated with the academic community.

I would recommend doing a literature search for pockels cell drivers, there are a good number of published designs that do what you want using avalanche transistors in sort of nonlinear transmission line, ex

These devices have a hard time since they are pretty lossy but can get really quite quick switching times (down into the few to sub ns duration depending on load) and with active cooling can get up to megahertz rep rates.

Alternatively, you could use a bigass vacuum tube

It wouldn't surprise me if it works out being more efficient than the solid state solution due to the smaller capacitances associated with a tube.

Dr. Slack

Tue Aug 18 2015, 06:47PM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

This technique, and circuit, is in the right ballpark for speed, voltage etc

, as long as you can cope with transformer coupled

DerAlbi

Tue Aug 18 2015, 07:24PM

Registered Member #2906 Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727

Hi Slack, i was looking into this due to this thread.
Basically the problem here is that Solid State does not offer such rise and fall times even without load. Thats quite some problem here. I was thinking about "yeah dude, use smaller voltage and a full bridge and use a transformer at the output..." but the stray inductance will limit the rise time. (the risetime at the gate does not matter much here. even if this rises slowly the output is faster than that. A mosfet in source configuration has a voltage amplification after all)

Its the same with the distributed concept that you linked in: the transformers stray inductance will become a problem here.

I honestly can currently not think about a good solution.
I mean you are looking towards 200MHz + at least one 600MHz harmonic... awww

OK: the used IXZR08N120 is close to meeting the specification if used in a full bridge configuration. But the 10pf load capacitance.. will not matter anymore given the internal capacitances. maybe a small 1:1.5 transformer at the output.... and it could be designed to have ultra low leakage inductance and you dont need a high inductance anyways if you have continous 2MHz. But the layout will be quite complicated. You cant share heatsinks for example.

teravolt

Wed Aug 19 2015, 10:02PM

Registered Member #195 Joined: Fri Feb 17 2006, 08:27PM
Location: Berkeley, ca.
Posts: 1111

I dont think there is any semiconductor that switch that much voltage in that short of time. only thing that i know of might be like a crytron or thyratron. maby a bunch of low voltage high speed transistors in series. if you can find a fast enough switch use it to discarge a peace of coax with a delay time that is 1-5ns say 1.2 to 7 ft

Electra

Thu Aug 20 2015, 09:23PM

Registered Member #816 Joined: Sun Jun 03 2007, 07:29PM
Location:
Posts: 156

I think the circuit in this paper has good potential. Granted it's a hybrid design using tubes and semiconductors. But I kind of agree a lot with their approach, rf transistors are the only ones that have the capable bandwidth, though not the high voltage ability.

In their design there only using a supply of 560v. However if you were careful not to exceed the devices ratings or replaced them with higher current parts, the concept should be capable of higher voltage operation. As it mentions in the text it is a constant current circuit, so you'll need to increase that for higher voltage. The anode dissipation and that of its resistors may be an issue in longer pulse duration operation too. If you don't need it would have thought the dc offset section could be omitted also.

Hope the link works for you all " target="_blank"> Link2

Well we'll see, I think that went right!

edit; Nope try again its

2Spoons

Thu Aug 20 2015, 11:23PM

Registered Member #2939 Joined: Fri Jun 25 2010, 04:25AM
Location:
Posts: 615

I have no idea if this would work in your application, but have you considered using magnetic pulse compression? Essentially it involves using saturable inductors as high speed, high voltage switches.

Linas

Mon Aug 24 2015, 08:05AM

Registered Member #1143 Joined: Sun Nov 25 2007, 04:55PM
Location: Vilnius, Lithuania
Posts: 721

It is obvious that we are going to use it in laser system as pulse picker. (must be faster than 13ns oscillator , so 6ns is good rise/fall time

original system is based on 2x3 transistors ( simple low end, nothing special, can't recall correct model, from Fairchild, to-252 package) mounted on beryllium ceramic substrate, in half bridge configuration. What is strange that they use something with avalanche type operation. As example, system have 60Vdcdc converter for some reason.

But problem is, this is monopoly, only one company can produce pockels cell in size of very small box 10x10x5cm, with maximum frequency of 1MHz and it is water cooled. And availability is not that good, we buy as much as they can produce, and they can make price as high as they want, we still must buy it

Control extra simple, two inputs, one for high side, one for low side. Best case scenario that it would generate high voltage pulse for one oscillator period, that would be ~13ns, after that, it must go to zero, with as less ringing as possible. At the moment, we are running 150ns impulse, and as a result, contrast after RA (regenerative amplifier) is not that great, lot of oscillator pulses at ~80MHz get through pulse picker with 150ns on time

johnf

Mon Aug 24 2015, 07:42PM

Registered Member #230 Joined: Tue Feb 21 2006, 08:01PM
Location: Gracefield lower Hutt
Posts: 284

Why not use Belke high speed switches
I did detail how I used them here
viewtopic.php?172154

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob