Hi, there are some methods available, but depends on the kind of application you are looking for. Let say that your specification is 40v, 400mA at variable pulse, then rather amplifying the signal, you can directly use 40v DC and a MOSFET. Using any of the microcontrollers you can get a desired digital pulse and to trigger the MOSFET. If you are looking for amplifying the signal from 5v then you might start looking for voltage doubling or voltage multiplier circuits.

Yes, this is one way in which I can do it and I'm really fond of it, but I can not find fast enough MOSFET transistors for this - most of them are able to produce pulses with duration of several hundreds of nanoseconds (even if rise-time is in order of tens of ns, the fall time is over 100 ns or even much bigger). I need a pulse as short as possible, so a 100 ns pulse is too much for me. I was thinking on utilising an EL7158 from Intersil (this is an FET gate driver, although the current is fair enough for my case) for the powerstage, but the maximum voltage of this driver is around 18 V.

you can use OP AMP (operational amplifier ) with some capacitors , resistors and inductors to generate what you need.

Hi, can you make a bit more elaborated specification of the pulse desired - maximal duration, rise time, fall time, polarity, repetition time (frequency). How do you intend to transmit the current to the load - such a short pulse would require extremelly short rise and fall times which may lead to enormous losses during the transmition (antena).

Yes I agree with Ivan Kojouharov.

Ivan, the pulse duration should be as short as possible - I am aiming for ~10 ns, with as short rise and fall times as possible, so the pulse will be almost square. Repetition time will be adjusted in order to obtain the best results, but it should be in the order of 10 MHz.

The current will be transmitted to the sample using metal bars as short as 2 cm, or so.

Nikodem, to have a pulse current of 400mA, with a rise time of 10ns reaching 40V the problem will not be the choice of the semiconductor, which probably exists on the market with such figures, but the inductive (in particular for the 400mA which is a huge current) and capacitive effects (10 ns is a quick slope) of the layout and external interconnection. This kind of performance would be easily possible inside a VLSI circuit, difficult in the discrete world. Parasitic effects are a huge limit. Need to create a quite complicate model for simulation including all parasitic effects to design the circuit. A pull-up CMOS symmetric configuration can be the base design, with a square wave 10 MHz generator input.

I found that DEIC515 driver IC from IXYS should be good in this application. It has maximum voltage of 30 V and current high enough to charge all the parasitic capacities (15 A). What do you think about it?

One-shot circuit may be a trip. a transistor, a transformer and a capacitor.

40/400ma is not problem, But 10MHz may be a lttle trouble.

Holger,

I cannot download your attachments, seems that our netwotrk has some problems. Would you please, e-mail them on [email protected]. Thank you.

Nikodem, I guess that 10 Mhz pulses with this power realy may generate too much heat and cause problems, as Jek says. If I do not mistake 10x10^6 x 40V x 0.2 A (assuming that 400 mA is a peak current and some average of 200 mA to be taken for the power calculation) yields in 80 MW power (efficiency=1) ?! If only single shot pulse or a train of few pulses is the goal, then it would work. Or if the average current (single shot power) is much lower. The power constraints have to be carefully considered as well.

Dear Sir,

Regarding your question fot producing a short pulse of 500 mA, I am afraid that with inductive effect of connections and the inevitable loop that you will have with a current loop, nearly it is impossible to have a sharp pulse of 10ns duration. the only way is to either increase the pulse duration or its sharpness. for the voltage the problem could be solved !

Hi Nikodem, of course I had a mistake, the average power is about 2 W. However, the layout is a probles as above was mentioned.

Hello. Suggestions by Holger is what I was also thinking. Check out high speed switching BJTs with >1GHz. This would be suitable for 1ns and less applications. And since the power rating, when averaged over a time period comes to a lower value, there would be no damage in using small signal BJTs that has drive current of 500mA. You can operate the BJT as switch and drive with a short pulse. To increase the load current you can parallel up the BJTs.

Dear Nikodem,

I would like to ask you for some additional information. You mention pulse duration, but not: requirements of pulse shape, repetition frequency, type and characteristics of the load? Wire length, type of wires between generator and load? These are all very important for the final circuit that you would have to use.

Harrie, I am glad that you ask these questions. I have mentioned 10 ns of duration and 10 MHz repetition rate (although the second is not a real need only a 'want' coming from the convenience of the device).

The load is almost resistive. The sample is purely resistive, but it will be connected using some aluminum foil and bars so I guess that some parasitics will go into that. No wire will go between the sample and the power stage. The power stage will be connected to the generator using a standard 50 ohm coaxial cable.

Dear Nikodem,

Thanks for your elaboration. However, I still miss information about pulse shape, rise and fall times (V/ns; V/ps?). Even for resistive loads, this can be problematic due to parasitic capacitance or inductions, leading to overshoot and poor pulse shape, ringing. You should avoid that the resonance frequency of the parasitics combined lies in the range of the pulse frequency.

Harrie, I would love to elaborate on these parameters, but I do not know what parameters are realistic. I want first to test a device based on a commercial MOSFET driver - DEIC515 or IXDD604PI - as they offer an pulse length in order of 10 ns. Do you think that there might be a better solution?

Nikodem,

I think both circuits are not suited for your application. First, their Vpp max is 30 resp 35 V. Furthermore, tr and tf are 8 resp 20 ns, resulting in spikes rather than nice square block pulses.

I was thinking about using a simple HF video circuit using a bipolar transistor as driver. What matters, is your load connected to ground, or can it float? Floating is usually more complicated. MOSFETs are difficult to control due to the capacitance presented by the gate, which makes timing and charging more complicated than bipolar transistors, which are controlled by current at the basis.

Appreciate your reply.

Maybe I was not clear enough on that topic, but the parameters (voltage and current) are only approximate and 30 or 35 V is high enough for my experiment. What really matters for me is the simplicity of the design.

On the other hand I though about a common base, paralell transistors amplifier, but for that I need transistors working with 40 V, at least 100 mA and bandwidth of at least 1 GHz or better. And I'm not sure if thiss will allow me to get 10 ns pulses with full amplitude.

And in both cases I get a similar result, only the second (bipolar transistor) solution is significantly cheaper.

If simplicity is the issue, I think I can help you. Give me 24 h and I will come with a solution. Cheap, simple and proven.

Hello!

If the shape of the pulses is not important I would suggest to build a low voltage (5V to 12V) RF oscillator using a transistor and an L-C circuit and to use a transformer to increase the voltage.

If you do not need a continuous operation, then use a gated oscillator that is activated only during very short periods of time, just long enough to make the measurements.

Doing so will surely get around the thermal problem of power and high frequencies.

Use a circuit simulator for the design and double sided PCB with tight routing for the realization. Take also great care for the connexion to the load.

More, if you need the adjust the frequency precisely, you can, in most oscillator circuits, synchronize the oscillator to a clock signal from a generator,

More more, if you just need a single pulse, with a low repetition rate (less than a few kHz) then you do not need an oscillator, but instead, a power stage like in a switching power supply (using a 100V power mosfet for e.g..) could be triggered by a pulse generator.

It is really interest discussion.

Dear Nikodem,

I ask for a time out: other matters need my attention. Will be back ASAP.

Hi Nikodem,

It took some time, but I have managed to collect all the equipment for testing the circuit that I am going to try to provide to you. Attached the picture of my old Tektronix pulse generator (type 115) and Philips scope (PM3267): all analog as you can imagine ;).

Settings: 50 ns pulswidth, 10 ns rise/fall time (tek115); 50 ns/ div (Philips).

I know, it is not exactly what you are asking, but in this way I will be able to show some of the characteristics of the various circuits I will have for you to consider.

Greetings,

Harrie

Harrie, were you able to do any work in the topic? I would be extremally glad for any help at tthis point.

Currently I'm building a prototype based on the IXDD604 RF MOSFET driver.

I think this can be done with a TLP (Transmission-Line Pulse) system. It charges a coaxial cable and then discherges it. The result is a short voltage pulse with flat top.

Mr. Napoli, is it possible to control amplitude and time of such a pulse?