# Linear Rise exponential fall pulse generator

For a neutrino detector related experiment, I need to design a circuit which generates the pulse similar to the detector. This is required to test the front end electronics.

Here are my final goals:

Pulse shape: Linear rise exponential fall, Rise time: 1-3nS, Pulse width: 10-20nS, frequency: 100Hz-10kHz (user control/ random distribution), amplitude: 0.05-15mV (user control/ random distribution)

But to start on a simple note I am trying to generate a simple linear Rise exponential fall pulse. Currently I am able to generate 1nS to 10nS square pulses (using FPGA), which mimics the linear rise part due to slew, but I need some circuit suggestion for exponential fall part. I tried to go with capacitor discharge but it seems that even RF switches are not that fast in terms of switching frequency.

Any suggestion is welcome.

• If money and complexity are not barriers, would a real fast DAC work? There are DACs on Digikey for under \$100 that have update rates above 2Gsps (there's one that can hit 5.6Gsps when interpolation is turned on), that'd give you 2-6 samples on the rising edge of the pulse. You'd have poor granularity, but you'd be able to program any shape waveform you wish. More importantly, what's the bandwidth of the input electronics? if it's less than a couple hundred MHz, it's going to act as a low pass filter and slow down the rise time of the pulse when is reaches subsequent stages anyway.
– Sam
May 19, 2016 at 7:30
• How is the linear rise done with "skew". What does this mean. Also, you give no detail about the exponential decay. May 19, 2016 at 7:33
• Will anything other than a neutrino trigger the detector's output, like an alpha particle? If so, do you have a spare detector, or a partially broken one that can still produce an output? May 19, 2016 at 14:31
• Thanks @Andyaka, to drive current in a further circuit, an op-amp is needed, buy limiting the skew of op-amp we can control the rising speed of output. May 20, 2016 at 7:53
• "skew" - do you in fact means "slew rate"? May 20, 2016 at 8:27

EDIT: I tried to to ismulate this, but something is wrong in circuitlab simulator.  You can controll a circuit similar to Marx generator. Where you have T_rise and T_fall, both exponential, but if input voltage is much hiher than needed peak, and R_rise is small enough you can get almost linear rise, perhaps adding an inductor can improve. simulate this circuit – Schematic created using CircuitLab simulate this circuit simulate this circuit

• This seems to be a similar approach I was trying, but there are quite some problem of real Life, eg. bandwidth of transistors used, parasitic etc. Anyway thanks a lot for your valuable time, I will try this in cadence simulator. May 20, 2016 at 7:48
• @PurnenduKumar WHy would you need a switch? You said 0.15mV peak voltage, this is less you output from FPGA, if your output is PMOS then the 1st picture is valid, see PMOS, if you have a push-pull output then you need a fast diode in between. If the slew rate of your output is enough, then consider the entire circuit made of coax cable and R_fall resistor. May 20, 2016 at 8:02

As your maximum voltage is 1v, and you can already generate fast square waves from your FPGA, consider this ... (if this isn't a solution, then in telling us why not, it may improve your ability to specify exactly what you want)

Have a parallel RC to ground, with a time constant appropriate for your falling edge exponential decay.

Switch a current into the RC for the rising edge. Although a current flowing into an RC will give you an exponential rise, the first part of that rise is more or less linear, say the first 1v of a rise to 5v. Control the height it rises to by the length of the pulse and the choice of the current source. Note that 5v (for example) is not a resistor from a 'real' 5v, but the Iout.Rshunt product.

Make the current source go high impedance for the decay part of the pulse.

Current output logic is very fast, and available to interface with most FPGAs. I used a 5GHz CML buffer from OnSemi some while back, NB4Nsomething_or_other, <200pS rise and fall times, to do something fairly similar. I've just had a quick look at the data sheet for NB4n11 (as an example) and you can get several volts of output compliance, so the output structure is good for 1v output swing. Obviously they are current sinks, so leading edge is -ve going, with an exponential trailing rise back to the quiescent level.

If you've got a fast pulse, then maybe just send it through a diode to an RC that will fix the decay time. (and some potentiometer after to set the height.) simulate this circuit – Schematic created using CircuitLab