I am trying to count the output of a photodiode. It outputs 2V 20ns pulses. My counter is able to read anything above 3.3V and about 60ns.

I've built a comparator circuit able of amplifying the signal to 4V but the duration of the pulse is still too small.

Therefore I've tried adding an RC filter both to the input and the the output, separately.

For the first two images, the yellow is the input to the comparator and the purple is the output of the comparator. Here I tried to use the RC filter on the input to give the comparator more time to produce the wave, clearly didn't work. VRef is 0.5V.

For the last images this is what you recommended (although on FI4) with the resistor and cap going from FI4 input to ground. It's doing what it should but unfortunately the voltage is significantly lower.

Where do I go from here, is there any nice fix I can do or do I have to add a secondary amp or comparator?

If there is no filter the signal is just a delayed and amplified copy of the input. (with no negative part)

Edit:
Would something like this help (monostable)?
http://cache.national.com/ds/LM/LMC555.pdf

 

You have images of an O-scope, but no schematics. Without seeing schematics, we're not going to be much help.

Unless you're using an ultra-fast comparator like an LT1016, you'll have difficulties recognizing signals that short in duration.

Inductance in your wiring will be a problem at those speeds. You will need to have your wiring as short as possible.
1/20nS = 50MHz. There are very few 555 timers that will even reach 2MHz.

 

Ok then.

I use a high speed comparator which is up for the task (see scope output).
http://www.farnell.com/datasheets/23167.pdf

These are my three circuits.

They correspond to the oscilloscope readings. In each case, channel 1 measures Vin and channel 2 measures Vout

I guess these setups are parallel and wont do much. I've tried a series RC circuit but then the voltage just goes to 5V and stays there.

 

Seems your addressing another thread I am unaware of, sorry if I am asking redundant questions.

I'm looking at your first two images trying to determine what the raw input it. Is it a series of about 6 pulses for 80 ns, repeating 20nS later, or is each pulse a valid signal? Looks like you are satisfied with just that first (largest) pulse in each series from the yellow output trace. You have one clean output pulse from each input series, though it may be small.

To stretch a signal you can drive a parallel RC (as you have been doing) thru a diode; the diode charges the cap fast and the resistor discharges it slowly, but your input signal isn't big enough to drive a diode on.

You may be best off by amplifying your input signal first, then pulling the pulse out of it. At least that gives you a larger signal to work with, and more options.

I once built an amp with 60dB of gain at 3MHz using a MAX4413 (4412 is single amp version) that ultimately gave me a 5V digital signal from a minute quartz transducer signal. The trick was to keep the input resistor small, less then 50 ohms least you cause a low RC break with the input resistor and the pin input capacitance.

These specific parts are tough to work with as they only come in very tiny surface mount packages.

 

Yeah I only care about the first large 2V pulse. But an actual signal to be counted is like you say about 150ns in length. That is due to the filter though as I am trying to prolong it. If you look at the third and fourth picture (the yellow) that is what the input signal actually looks like.

Can I not just use a monostable timer here? Or is the output from the comparator to short to successfully trigger a timer?

Edit:
Also, 2MHz would be more than enough for my application. The signal I show you is with the lights turned on. I will actually be doing very low fluorescence spectrometry where I will be running at maybe 10-100kHz.

So I only need to make sure that I can catch the 20ns signal and turn it into something longer.

 

From your oscillographs I can't tell much, I'm not sure what the horizontal or vertical is set to.

20nS is small. A MC74HC4538A monostabe has a minimum trigger of 15 to 18 nS, so you may be hitting that if you amplify that 2V pulse, but it's still very close.

There may be faster monostables. I doubt a 555 would see that trigger pulse (even if you inverted it, as a 555 wants a hi to low trigger)

 

Well, I missed the exchanges after reply #3 while I was fiddling around with trying to stretch the pulse.

Putting a 100pF cap on the diode output or comparator output is a large peak current requirement. Also, you need to charge the cap quickly, and discharge it slowly.

I had to use an RF transistor as an emitter follower from the output of one comparator to charge an RC circuit, and then use another comparator with a different voltage reference.

The diodes I had in my model library had too much junction capacitance (even BAT54's and 1N4148/1N914's), and the same with the NPN's I had (cje too high; 4pf was way too much) so after looking around a bit, found the MPSH10/MMBTH10, which seems to do the trick. You can't just slap in any old NPN transistor and expect it to work.

Note that to keep the schematic less cluttered, I've omitted many necessary bypass caps. L1/R1/C1 are just there to "ballpark" emulate your input signal. The AD8561 has a slower rise time than the LT1711, which gave a nicer output signal; but since you have the AD8561, that's what I modeled.

Have a look at the schematic/simulation. The output u2out is ~100nS wide. As always, this is just a simulation; your mileage will vary - layout/bypassing will be critical.

 

Many thanks for your input SgtWookie, your circuit looks good and I managed to implement in LTSpice after learning some new tricks. I've ordered those parts and I was going to install it tomorrow morning however the 2MHz 555 arrived today. I agree with you that it shouldn't really work but apparently it does.

I could probably play with the RC to significantly reduce the one-shot length but for now it is more than sufficient as I am counting at about 10kHz. Could probably also filter out my 555 output to make it more clean.

So for anyone who needs to widen a pulse, either try your luck with a monostable or use something similar to SgtWookies circuit! I'll probably try it tomorrow as well just to see if it works as predicted.

Cheers guys