No.
You reduce R3 or C1.
Then why are you differentiating the pulse?
Do you want the transistor to be on only for a short time at the positive pulse edge?

Yes, that is exactly what I want, I would like to get 10nS 'on' period or less.

 

Turn a transistor on (and off) with the pulse.

Do you want to turn on the transistor when the input signal is high and off when the signal is low?

 

Yes, that is exactly what I want, I would like to get 10nS 'on' period or less.

Then try a RC time-constant of 10ns or less.
But note that the op amp may not be that fast.
Also the square-wave rise time must be much less than 10ns.
For a shorter pulse output you might try using a fast comparator or a VH or AHC series logic gate.

 

Do you want to turn on the transistor when the input signal is high and off when the signal is low?

I did not yet decide, can work with both, depends on what pulse is simplest to produce.

 

But note that the op amp may not be that fast.

By "fast" do you mean "slew rate"? I'm asking out of ignorance.

Then try a RC time-constant of 10ns or less.

Where in the circuit would you put the RC?

Also the square-wave rise time must be much less than 10ns.

It is.

For a shorter pulse output you might try using a fast comparator or a VH or AHC series logic gate.

Yes, but I'm thinking the same can/should be done with a few transistors and passives. I want a simple and robust design that can survive abuse by a 300 pound gorilla. I tried the 74121 which is advertised as min 35 ns, but I can't make it go below 80 ns, so I'm not so inclined to use IC's in general for this. And ref VH or AHC, could be - I was actually thinking of looking into ECL or PECL if my analog engineering skills fail me.

 

Then try a RC time-constant of 10ns or less.
But note that the op amp may not be that fast.
Also the square-wave rise time must be much less than 10ns.
For a shorter pulse output you might try using a fast comparator or a VH or AHC series logic gate.

Ref RC, are you thinking something like this? Perhaps with a schmitt-trigger input on the inverter?

 

I did not yet decide, can work with both, depends on what pulse is simplest to produce.

This is going nowhere fast.
Draw a timing diagram of what you are trying to produce, i.e. show input signal and output signal, with proper voltage and time scales.

 

This is going nowhere fast.
Draw a timing diagram of what you are trying to produce, i.e. show input signal and output signal, with proper voltage and time scales.

Would this be satisfactory?

Where
A: 0 - 5V
B: 0 - min 0.7V
C: 10ns
D: 1us

in the ideal world I would like both C and D to be variable, where C would be 0-1us, and D 50ns-1us. But that would complicate things at this point when my main focus is to make pules width of 10ns (or less), which seems to be a challenge without going into exotic components. And I don't mind if the desired signal is inverted or not, that is not a problem.

 

Now that is perfectly clear. (I am going to assume that you aren't concerned about rise-time and fall-time.)
I'll breadboard a circuit for you shortly.

 

The shortest pulse I can generate with the off-the-shelf components that I have in stock is 30-50ns from a 74123 monostable multivibrator.

 

Ref RC, are you thinking something like this? Perhaps with a schmitt-trigger input on the inverter?

View attachment 139824

That should work if you use fast logic parts.
You may not need the RC delay as the delay of the inverter is likely sufficient to give near a 10ns pulse, depending on the logic family you use.

 

The shortest pulse I can generate with the off-the-shelf components that I have in stock is 30-50ns from a 74123 monostable multivibrator.

Thank you for your efforts, that's about the same experience I had, so I made a mono vib with two transistors, but then I got in trouble with the high pass filter, so here we are. Today I will venture down the logic gate path.

 

Thank you for your efforts, that's about the same experience I had, so I made a mono vib with two transistors, but then I got in trouble with the high pass filter, so here we are. Today I will venture down the logic gate path.

The logic gate suggestion seems like a good solution. I too will investigate this.

 

With the two-gate solution (and no RC) I get 7ns pulse using two 74HC00 NAND gates.

Edit: You can vary the pulse-width by varying R (in between the two gates) and leave out C.

 

With the two-gate solution (and no RC) I get 7ns pulse using two 74HC00 NAND gates.

Edit: You can vary the pulse-width by varying R (in between the two gates) and leave out C.

That's clever, using the input capacitance for RC. And if you swap the resistor for a transistor you can control the pulse width from your MCU.

 

That's clever, using the input capacitance for RC. And if you swap the resistor for a transistor you can control the pulse width from your MCU.

But the control would be very non-linear, being very sensitive around the transistors threshold voltage.

 

But the control would be very non-linear, being very sensitive around the transistors threshold voltage.

Yes, but I meant it as a quick and simple solution, I'm confident you could use one of those digital potentiometers instead. Anyway I still find it wired that I could measure a voltage on the input of the opamp and nothing on the output, just because the input frequency was below cut-off frequency. And I find it wired that nobody else find it wired, or maybe it's a natural phenomena that I'm ignorant about, anybody else seen this before and can explain it?

 

Anyway I still find it wired that I could measure a voltage on the input of the opamp and nothing on the output, just because the input frequency was below cut-off frequency.

I believe the word you want is "weird" (although if you are on an upper, you could be "wired").

I thought we determined that the resistor value to ground was too low and that's why you didn't get an output?

 

I believe the word you want is "weird" (although if you are on an upper, you could be "wired").

I thought we determined that the resistor value to ground was too low and that's why you didn't get an output?

Actually no, after I changed C3 to 1nF something began to show up on the output, but the weired thing is/was that there was a voltage of 2V on the input (before changing C3), and since the input has infinite (1000GΩ) impedance, there should (theoretically) have been an amplified output, but there was zero, why? MrChips said it was "derivative of the square wave", but it still showed up on the scope on the input.

Ref. wired/weired: Sorry, not native English speaking person here, time for a downer ;-)