# Fast and accurate pulse generator

#### tiddlybiz

Joined Feb 10, 2016
5
Hello all, I am looking to build a circuit which can provide 6.5V pulses with a width of 10-50nm (or 20nm if fixed) and a frequency of around 10-50 microseconds. I'd like the rise and fall times to be as quick as possible with no overshoot. What do you think is the best way to do this?

I have considered different approaches from using 555 timers to FPGAs but I'm not sure what will provide the best results for the least cost and smallest size.

#### AnalogKid

Joined Aug 1, 2013
8,806
Please convert nanometers to seconds.

ak

#### Sonoran Desert Tortoise

Joined Oct 30, 2014
79
Please convert nanometers to seconds.

ak
How many furlongs in a fortnight?

#### #12

Joined Nov 30, 2010
18,216
Please convert nanometers to seconds.
About 6 thousand 10^12 hertz if I read the chart correctly.
0.000166 picoseconds?
I don't think we have any transistors that fast.

#### Kermit2

Joined Feb 5, 2010
4,162
A high speed op amp that has 1000 volts per microsecond slew rate is considered very fast.
It could be stated that it has a 1 volt per nanosecond slew rate. Such an op amp could get you a rise time of 6.5 nano-seconds for a 6.5 volt pulse.

That is ideal figures, other factors found in the real world would further lengthen the rise times.

Minimum pulse duration would be 2X the slew rate X the peak voltage. For the example minimum pulse length will be greater than 12 nanoseconds.

#### Alec_t

Joined Sep 17, 2013
11,797
It's certainly not going to happen using a 555 as the TS hopes!

#### AnalogKid

Joined Aug 1, 2013
8,806
10 us to 50 us isn't hard for a 555 to serve as the clock. If the output times are ns a suspected, that would take a separate pulse former clocked by the 555.

ak

#### Bordodynov

Joined May 20, 2015
2,733
See:

#### tiddlybiz

Joined Feb 10, 2016
5
Thank you all for the replies. Sorry the nm was a typo and should have been ns as you suspected.The transister circuit by Bordodynov simulates pretty well - thanks for that example. Currently I have been using an AWG to create the pulse train and the resultant pulse looks pretty ideal. The issue with it is that an AWG is too big and expensive to put in a box with the rest of the circuit. I think AWGs have a more complex circuit than I require with compromises due to the extra flexibility required. I would be more interested in the sort of circuit used in commercial pulse generators but I'm still trying to figure that out. I have also considered Schmitt trigger circuits from an alternating wave but I'm not sure whether there is any advantage to this method.

#### #12

Joined Nov 30, 2010
18,216
the nm was a typo and should have been ns
I was even considering Nautical Miles yesterday.

I have been using an AWG to create the pulse train
And today You're using an American Wire Gauge to generate a pulse train?

Maybe this would work better if you used less acronyms.

#### RichardO

Joined May 4, 2013
2,271
What rise and fall times do you need and at what voltage?

A 74AHC14 may work for "slow" rise and fall times at 5 volts maximum but an avalanche pulser like bordodynov shows will be needed for more than a few volts or less than 1 ns rise/fall times.

#### KL7AJ

Joined Nov 4, 2008
2,229
About 6 thousand 10^12 hertz if I read the chart correctly.
0.000166 picoseconds?
I don't think we have any transistors that fast.
At the shop we have an Agilent function generator that will do 40nS pulses. But it's a pricey puppy!

#### tiddlybiz

Joined Feb 10, 2016
5
AWG is an arbitrary waveform generator. Can't remember the model offhand but it is from Keysight. The waveform should appear reasonably square so for this to be the case a sub nanosecond rise and fall would be ideal. The power supply voltage from a computer supply so 12V max currently although I could potentially increase this. An avalanche circuit is possibly a little too high on the supply voltage though. I have been looking at step recovery diodes (SRD), which cut off quickly but still trying to work out how to implement them to enhance the rise and fall time.

#### Sonoran Desert Tortoise

Joined Oct 30, 2014
79
Hello all, I am looking to build a circuit which can provide 6.5V pulses with a width of 10-50nSec (or 20nSec if fixed) and a frequency of around 10-50 microseconds. I'd like the rise and fall times to be as quick as possible with no overshoot. What do you think is the best way to do this?

I have considered different approaches from using 555 timers to FPGAs but I'm not sure what will provide the best results for the least cost and smallest size.
...The waveform should appear reasonably square so for this to be the case a sub nanosecond rise and fall would be ideal. The power supply voltage from a computer supply so 12V max currently although I could potentially increase this. An avalanche circuit is possibly a little too high on the supply voltage though. I have been looking at step recovery diodes (SRD), which cut off quickly but still trying to work out how to implement them to enhance the rise and fall time.
I am entertained by the requirements. You did mention a few other things that are important to you but you didn't give specifics. Could you add details on
- physical size limitations
- how you will assemble this (DIY toner transfer pcb/other; smd/through-hole; ...)

#### tiddlybiz

Joined Feb 10, 2016
5
The FPGA provides an excellent 20ns square pulse but has an overshoot on the falling edge which affects the rest of the measurements. It is also very expensive. Physical size is not massively important within reason - a few inches square would be fine. Shouldn't need to be any larger. I will initially probably build the circuit on a prototyping board of some variety and then if it works reasonably well I will more carefully design a PCB. Below £100 would be ok for the parts cost - hopefully less.

Emitter coupled logic looks quite promising and is already built into some laser diode driver chips which are pretty cheap. I'm yet to discover a circuit which is exactly what I'm after but the general technology seems to be along the right lines.

#### atferrari

Joined Jan 6, 2004
4,169
I cannot recall precisely the details. In youtube, look for a video by (Alen) W2AEW about an oscillator which later he uses to do some TDR practical measurements.

From what I recall he is using a 74AC or 74C gate, or I could be dead wrong. Not sure if that could fit your needs.

#### Sonoran Desert Tortoise

Joined Oct 30, 2014
79
This chip will delay a square wave by 10, 20, 30, 40 or 50 nSec (depending which of the 5 taps you draw from).

Then connect to an XOR gate to get a pulse at the leading and trailing edge of your square wave.

Since you get two pulses per full wave, then your carrier wave needs to be half of the planned pulse rate.

This chip only gets to 5 volts, so connect to a fast XOR gate that can meet your voltage need.

http://www.digikey.com/product-detail/en/DS1100Z-50+/DS1100Z-50+-ND/680524

#### hp1729

Joined Nov 23, 2015
2,304
Hello all, I am looking to build a circuit which can provide 6.5V pulses with a width of 10-50nm (or 20nm if fixed) and a frequency of around 10-50 microseconds. I'd like the rise and fall times to be as quick as possible with no overshoot. What do you think is the best way to do this?

I have considered different approaches from using 555 timers to FPGAs but I'm not sure what will provide the best results for the least cost and smallest size.
74120 Pulse Synchronizer set in single pulse mode? I don't know about overshoot and undershoot.

#### Attachments

• 276.5 KB Views: 15
• 36.5 KB Views: 11

#### slackguy

Joined Feb 11, 2016
67
they sell inexpensive $20,$40,\$80 function generators online that are 2-20 MHz, which covers your freqency range. they are meant to provide "nice signals" in the range.

some are adjustable pulse width or "duty cycle" adjustable i beleive - though unsure if a 20 MHz at (their supported minimum duty) will be within your nm range or show up on a scope as a "slower rise time" than you wish

i assume you know quality generators are lab equipment, expensive, and require certified repair

on the other hand you might resort to, instead of electronically creating the signal - to find a naturally occuring event of the same charateristic to (probe and amplify) as the signal; kind like an atomic clock does but using a different phenomenon.

or maybe the amplified output of a timing crystal pulse (ie of computer), duty cycle adjusted (ie, if it were a few GHz signal, could be on 1/100 of a MHz, and on the rest off), could suffice.

#### slackguy

Joined Feb 11, 2016
67
oh dang - i didn't see that you had a "trying to fit it in a matchbook" requirement and have AWG already