Generating short current pulses

[elki]

Would anyone have a suggestion for the possible approaches of building a simple voltage controlled current source that is capable of generating short pulses of the order of 100-200ns and the current values of 1-100uA (example attached)? One thing I was thinking about was to use an NPN transistor triggered by short pulses created by a schmitt-inverter circuit. In case someone has a reference or other ideas, I would be very grateful to learn more.

[Manul]

Honestly, I think it's very hard to do. Would seem simple on paper, but in reality, with such fast edges and currents as low as 1uA, any parasitic capacitances (causing unwanted coupling, charge injection) will mess up the shape of pulse. Of course, that's subjective, depends on how accurate you want it to be.

But that's just one part of story. What will be your load (current sink)? If the load is a low parasitic capacitance resistor connected with very short trace, then it's ok. But most likely you use it for something else. So if PCB trace is a bit longer and the load has some capacitance, then you current pulse will just integrate and not be a pulse you want. 5pF capacitance will charge or discharge only 20mV during 100ns at 1uA!

Maybe it could be done with a laser or LED coupled to a fast photodiode or something.

[David Hess]

Fast bipolar transistors in a differential configuration can easily switch faster than 1 nanosecond, and their outputs are currents, but performance degrades at low currents.  What could work better is a differential pair switching a current from 101 to 200 microamps, and an opposing current in parallel with the output of 100 microamps, so the transistors operate at reasonable current levels and the output current is 1 to 100 microamps.

[antenna]

Not sure if it would work, thinking out loud, but what if you had a 2N3904 in avalanche mode using a coax as a timing element?  The HV supply resistor and coax length adjusts the pulse width and frequency, and since the transistor avalanches at the same high voltage each time, the same lower voltage pulse is created across a low value emitter resistor.  Since you only need 100mA max, even with a load impedance of 10kΩ, you only need 1v max source voltage, so the emitter resistor on the avalanche transistor can be pretty small ensuring a really low impedance to drive whatever the load is you are connecting.  Perhaps with an adjustable constant current source in series with the load it would work?

I'm just guessing here.  Every time I hear ns pulses, i think of my 3904 TDR pulse generator

[elki]

Thanks a lot for all the ideas! Very appreciated.

[Berni]

You need to also specify the compliance voltage of the load and how sharp of a square wave you need (rise time, ring out time..etc)

If your compliance voltage is 0V then the solution is take a regular digital signal, resistive divide it down to a 1mV pulse and then send that pulse trough a 1000 Ohm resistor towards the load and you get a 1uA pulse. Parasitics are taken care of by allowing the use of low-ish resistance values for both the divider and the series resistor.

If you need to do this with a load compliance voltage of +/- 100V and still keep the pulse sharp. That is going to be a very Very VERY difficult task, likely involving some very custom analog ASICs.

[elki]

Indeed, this could be an interesting solution. There are no strict requirements on the pulse sharpness, etc. Would it be possible to produce a short pulse coming out of e.g. attiny?

You need to also specify the compliance voltage of the load and how sharp of a square wave you need (rise time, ring out time..etc)

If your compliance voltage is 0V then the solution is take a regular digital signal, resistive divide it down to a 1mV pulse and then send that pulse trough a 1000 Ohm resistor towards the load and you get a 1uA pulse. Parasitics are taken care of by allowing the use of low-ish resistance values for both the divider and the series resistor.

If you need to do this with a load compliance voltage of +/- 100V and still keep the pulse sharp. That is going to be a very Very VERY difficult task, likely involving some very custom analog ASICs.

[elki]

Another thing I was thinking about was to use a transconductance op amp, e.g. LM13700.

[Berni]

Yeah the ATTinys can run at 20MHz and can run their timer peripherals at the same speed so a 100ns pulse should be possible. Tho you won't be able to fine adjust the pulse length to any sort of sensible resolution. If you need to adjust the pulse length down to 1 nanosecond resolution it is likely easier to do it in the form of a analog pulse extender circuit that is then digitally tuned trough a DAC or something.

[elki]

Thanks! I will give it a try with ATTiny.

[David Hess]

If you need to do this with a load compliance voltage of +/- 100V and still keep the pulse sharp. That is going to be a very Very VERY difficult task, likely involving some very custom analog ASICs.

I am not sure it is that difficult.  Take the existing low voltage design and add high voltage cascodes.  Rise and fall times can certainly be below 10 nanoseconds.  They actually make specialized transistors for this which are high voltage, fast, and have low reverse transfer capacitance.  They were intended for CRT drivers but also work very well for the low distortion VAS stage of an amplifier.

[Berni]

The problem is that at such high voltage swings and low currents combined you get big problems with parasitic capacitance where having just a few pF of parasitics on the output would turn the 1 uA square pulse into a triangle wave that doesn't even reach full amplitude.

[elki]

Thanks again for all the feedback. I have a bit of dilemma now, which most likely arises from the lack of my education. Now that I have short current pulses, the value of which can also be adjusted by a resistor put in series, is it possible to measure it with a transimpedance amplifier? Wouldn’t the voltage divider and any other resistance in the input current pulse circuit create an effective big impedance at the entrance of the amplifier?

[Berni]

The classical opamp transimpedance amplifier doesn't really care about what impedance you feed the input with.

It works by nulling out any current you push into the input using its own feedback resistor, no matter what that current comes from (as long as the amplitude is within range). The voltage at its input will be stubbornly kept at whatever the reference input is connected to (this is usually 0V). So if it is fed by this simple resistive 'current source' it will work fine since that also requires the output to be held at 0V.

This is actually the ideal thing to drive as a load. Any other load that actually requires some compliance voltage will not work well at all with the resistive source.

[elki]

Thank you for the clear explanation!

The classical opamp transimpedance amplifier doesn't really care about what impedance you feed the input with.

It works by nulling out any current you push into the input using its own feedback resistor, no matter what that current comes from (as long as the amplitude is within range). The voltage at its input will be stubbornly kept at whatever the reference input is connected to (this is usually 0V). So if it is fed by this simple resistive 'current source' it will work fine since that also requires the output to be held at 0V.

This is actually the ideal thing to drive as a load. Any other load that actually requires some compliance voltage will not work well at all with the resistive source.

[Picuino]

Try using the CTMU module of the PIC microcontrollers.
It is a module specialized in delivering currents between 0.55uA and 55uA with controlled times that can be very short.

CTMU general information: https://ww1.microchip.com/downloads/en/DeviceDoc/61167B.pdf

PIC18F24K22 family processors with CTMU included: https://ww1.microchip.com/downloads/aemDocuments/documents/MCU08/ProductDocuments/DataSheets/PIC18%28L%29F2X-4XK22-Data-Sheet-40001412H.pdf


Edit: Microchip parts selector for finding CPUs with CTMU
https://www.microchip.com/maps/microcontroller.aspx

[Jason Henry]

Consider using an NPN transistor triggered by short pulses from a Schmitt inverter circuit for a simple voltage-controlled current source capable of generating 100-200ns pulses with current values of 1-100uA.