Pulsed current generator

[FoxSky]

Hi all,

I'm writing you about a project I started working on and I need your help. It is a pulse generator (PWM) with the following requirements:

Modulated light is used not only in the LEDs illumination systems but also in the optical sensors. To evaluate such an optical radiation source the developer needs to have a pulsed current generator. This kind of generator will allow him to modify the current amplitude through the load and also the timing characteristics, the pulse length and the period. The proposed application is to develop a pulsed current generator with variable intensity and duration with the next requirements:

- variable current amplitude from 1mA to 2A

- variable pulse length from 1µs to 100ms

- variable pulse period from the pulse length to 100 times its length

- inverting output selection

- maximum output load 10kOhm

- input supply voltage of the generator 15...30Vdc


The adjustment of the current amplitude, pulse length and period will be implemented with potentiometers and the selection of the inverted output will be implemented with a push-button. The solution will be implemented without using a microcontroller. The PCB layout will be realized in one or maximum two layers and it should have maximum 80 cm^2.

To generate the pulses I thought about 555, being able to change the frequency and duration of the pulses: https://www.electroschematics.com/5834/pulse-generator-with-555/

The problem is that there will be a very high value beach. I do not know if I can charge 2A on a 10kOhm load with a voltage of 30V. I also need a current source. The work with reversing the pulses at the exit I have no idea how to do it (using a push buttom).

If you could help with the idea / tips / solutions, where to start what components (AO, Transistors, integrated circuits, etc) I could use (as uC can not use), if can reach the required values.

[dmills]

10k ohms @ 2A is 20kV and 40kW, you way wish to tweak your spec just a bit!

Regards, Dan.

[nctnico]

I'd make it microcontroller controlled. Use a DAC to set the current on a current source (opamp + PNP transistor and some passives) and a timer output pin to enable/disable the current source for the required duration. The biggest challenge is to get the current source to not ring too much.

I've build something similar for a entirely different purpose a couple of years ago.

[T3sl4co1l]

Oooh, no rise/fall time spec, this gon b gud!

Besides the 20kV range, 10kohms and 1us also implies capacitance much less than 100pF.  Not many transistors can do that, while offering a 2A range (though I don't think this is necessarily a show stopper).  Bigger annoyance: you need so many in series to handle the voltage.

Ironically, vacuum tubes look pretty good here, meeting the capacitance and voltage limits reasonably well.  Actually the one problem is: I don't know of any, that are reasonably available, that can deliver a pulse that long.  Radar modulator tubes are only good for about a hundred microseconds, at 2A.  They also top out around 10kV, not 20 (and really, you'd need ~25kV to ensure linear operation).

Note that 20kV * 2A = 40kW, quite a big supply and amplifier no matter how you cut it.  It could still be powered off 30V, but then the duty cycle must be quite low (say for 30A input, you need a worst case PRF below 0.25Hz), and you need quite a large capacitor, say 200uF at 20kV, to hold enough charge (40kJ) to complete a pulse.

Perhaps this is an X-Y problem.  You have an underlying problem X, and have formulated a plan Y which you say you need help with; but you haven't realized that there may be easier solutions to X, perhaps trivial even, which we could perhaps discover if you told us about X instead of Y.

Tim

[daqq]

Try looking at Behlke, they have all sorts of funny high voltage switching "transistors".

I do not know if I can charge 2A on a 10kOhm load with a voltage of 30V.

Well, you'd have to step up the voltage (using one of the wide range of DC-DC converters). That said, 2Amps @ 10kOhm is 40kW, which @ 30V is a little over 1kAmp. Depending on the duty cycle you could get away with a smaller average current (charge up caps then discharge), but that's not in your specification.

I think you are trying to have a too wide specification - do you really need 2Amps into 10kOhms?

[Marco]

FoxSky : You will need an UI and dicking around with potentiometers only to verify the accurate timing/amplitude on an oscilloscope is so 50 years ago ... so forget about the 555.

If you want to avoid a microcontroller, there's probably USB devices you can use so you can do your programming in a more convenient environment while just adding hardware to transform the generated pulses into current pulses. Maybe the Bitscope mini?

[David Hess]

As the others posted, 2 amps into a 10 kilohm load requires 20 kilovolts of compliance and 40 kilowatts which is going to be a problem.  If your load is made up of LEDs, then I hope the compliance requirements are within your 15 to 30 volt specification which is what I will assume.

With a minimum pulse width of 1 microsecond, what are you expecting for rise and fall times?  I would hope for 100 nanoseconds or less which is about 20 times better than one I made; see below.

Covering 1 milliamp to 2 amps with high accuracy at low and high currents may be challenging and is.  Are there any requirements for accuracy or resolution at either end of the range?

What about allowable aberrations on the rising edge?  How much overshoot is acceptable?

Simple output designs are going to have trouble producing a clean 2 amp output with 1 microsecond pulse widths.  High output current transconductance amplifiers (voltage to current) are typically not that fast but it is certainly feasible without too much complication.  I recall designing something using a single power MOSFET which was about an order of magnitude too slow for 1 microsecond current pulses but I think a two transistor MOSFET cascode and a high current gate driver with feedback could do it and that is where I would start for simplicity.  I think a schottky diode or power MOSFET source switched design could get down to 10 nanosecond transition times or better (I know a lot more now since I built my first one) but then would be limited by parasitics between the source and load.

Check out figure 142 on page 65 of Linear Technology application note 47 which shows a 1 amp current output design which gets down into the 20 to 30 nanosecond range.  Notice that it does not settle until about 50 nanoseconds, only has an output compliance of less than +/-5 volts, and uses UHF output transistors to get an idea of the difficulty.

(1) It settled in a couple of microseconds which was all that I needed and as good as I could do with a single operational amplifier, something like an LF356 or LM318, directly driving a power MOSFET.  I did not learn about MOSFET capacitance and charge models until 10 years later.

[Marco]

Lets say you want to switch 2A inside 100 ns ... I don't think you're going to manage that with a single MOSFET capable of dissipating 60W, with a heatsink obviously, except maybe some RF monstrosity costing a fortune.

With a diode switched design you can use a far smaller switching MOSFET since it has low voltage drop I guess.

[dmills]

I would have thought an enhancement mode GaN part would manage a couple of amps within a hundred ns fairly easily (That is only a 10MHz fundamantal, say 100Mhz or so of bandwidth for something that looks reasonably square, layout will matter but you are not building this sort of thing on breadboard anyway)?

No need for butch LDMOS, which in any case is far cheaper then it once was.

Regards, Dan.

[Marco]

The MOSFET which is diode switching, yes. The MOSFET which takes the voltage for the actual current source, that's going to need some serious cooling. Any TO-247 MOSFET is going to be way too slow to directly modulate, it might even be too slow as a cascode in a constant current source (say 500 pf output capacitance, not sure how fast that would adjust if it has to swing 30V when the switching MOSFET dumps a low current into a high resistance load).

If you tried to directly modulate the current source in closed loop feedback, you'd definitely need some gullwing type RF MOSFET to switch fast enough AND be able to dissipate 60W. Which will be expensive and awkward to heatsink.

[MF-jockey]

I think the thread was 1 hour and 29 minutes to late ;-)

[mycroft]

There is a nice Linear Technology Application Note 133 by Jim Willams: "A Closed-Loop, Wideband, 100A Active Load". Lots of details. May it serve as inspiration.

[CopperCone]

The easy way to do the control part of this is to get your self an old big wavetek pulse generator.

You can configure the rise/ time, duty cycle, pulse lenght and repetition rate easily, and it can put out 20v into 50 ohm. The max frq with one of these i think is like o0MHz, should run you like 100$. Only problem is that their real big mothers.

If you really wanna do the generator yourself, i highly recommend linear technologies timerbox rather then 555 timers. Much more integrated and neater.

From this point i would recommend you build a few different 50 ohm input impedance power amplifiers.

I have no idea how to build a 20kv voltage to current converter though.

 You might be able to make a modified howland current pump that has very high voltage vacuum tube buffer to cover the hv ranges but making it work sounds like a complete dog.

You can make a fast v to i converter with the lt1010 but your supply rails are limited to 30v even if you parallel them for performance with lower impedance loads.

I dont think you will be able to get the same loop response with high voltage parts.

If you dont care about edge rate adjustment, or implement some kinda current limiting (hoe at high voltages?), then you can just control a high voltage switch with pwm to make it a digital circuit.

The size of an analog circuit capable of doing 40kW would be pretty large. I think its the best way to do it for lab equipment or a test bench though since you can have absolute control over the waveshape, even with complex loads if you play around with a awg for inputs, theoretically.

Telling us what the slew rate reuirement/edge speed is would massively clarify things.

[mbless]

While OP's post does seem like a joke with the 10 kOhm load and no replies since, there are fast PWM current sources such as this https://directedenergy.com/product/pco-6131/. They are intended for laser diodes, but work for LEDs with forward voltages less than 20V.

Does anyone know how this can have high (125A) current regulation with pulses down to 100ns?

[nctnico]

While OP's post does seem like a joke with the 10 kOhm load and no replies since,

10k Ohm is a sensible spec when it comes to current sources because this will set the control loop stability requirements. Ofcourse you shouldn't expect to be pumping 2A into 10k Ohm 

[Marco]

Does anyone know how this can have high (125A) current regulation with pulses down to 100ns?

By cheating a bit about what they mean with pulse width, look at the fall time.

[mbless]

Does anyone know how this can have high (125A) current regulation with pulses down to 100ns?

By cheating a bit about what they mean with pulse width, look at the fall time.

Ah, I didn't catch that as I was remembering other ones. They have several short pulse diode drivers here.

For instance the PCO-7121 has rise and fall time of 12 and 10ns, respectively, with 50A output. The datasheet shows a pretty nice top-hat 50A current profile for 100ns pulse width. I'm curious as to how they achieve such fast current control.

[Marco]

For instance the PCO-7121 has rise and fall time of 12 and 10ns, respectively, with 50A output. The datasheet shows a pretty nice top-hat 50A current profile for 100ns pulse width. I'm curious as to how they achieve such fast current control.

It's not really a current source, it's a voltage source with ~1.5 Ohm current limiting resistance.