Inexpensive DIY 50ps Ultra Fast Pulse Generator

[starlino]

Inspired by  "EEVblog #306 - Jim Williams Pulse Generator" I decided to build my own ultra fast pulse generator.
I stumbled upon an interesting chip from Analog Devices ADCMP580 , it's basically a SiGe based comparator that delivers a very fast rise/fall time. It's advertised to be 37ps (20%-80%) so I prorated it to about 50ps (10%-90%). I did a test setup and confirmed my 500Mhz scope's bandwidth . However I have no way to confirm the actual 50ps rise time (I have no 8Ghz + scope !). Any ideas how I can do that?  Do I need necessarily a faster scope or is there any clever indirect method to confirm that ... ?
Is anyone with a fast scope able to do the same test and confirm the 37ps/50ps rise times ? If fully confirmed it would be an interesting tool for evaluating scopes' bandwidth and such ... The chip and the demo board I used is relatively inexpensive.
You can find the build details on my blog:
http://www.starlino.com/build-a-really-fast-pulse-generator-50ps-rise-time-using-an-ultra-fast-sige-comparator.html
I took screenshots from my scope and photos of the build.
Thanks !

[JuanPC]

However I have no way to confirm the actual 50ps rise time (I have no 8Ghz + scope !). Any ideas how I can do that?  Do I need necessarily a faster scope or is there any clever indirect method to confirm that ... ?

the only indirect way i can think of, to measure clock jitter & rise time, is using a sound card.
http://www.transaudiogroup.com/clock-comparison-files
http://blacklionaudio.com/product/micro-clock-mk2/
http://www.grimmaudio.com/pro_masterclocks_cc1.htm
http://antelopeaudio.com/en/products/category/master-clocks

but you need a much faster clock source.
really nice 3-way loudspeakers.

[BravoV]

However I have no way to confirm the actual 50ps rise time (I have no 8Ghz + scope !). Any ideas how I can do that?  Do I need necessarily a faster scope or is there any clever indirect method to confirm that ... ?

the only indirect way i can think of, to measure clock jitter & rise time, is using a sound card.
http://www.transaudiogroup.com/clock-comparison-files

but you need a much faster clock source.
really nice 3-way loudspeakers.

Huh ?  Are you sure you're not posting at the wrong thread ?

[JuanPC]

However I have no way to confirm the actual 50ps rise time (I have no 8Ghz + scope !). Any ideas how I can do that?  Do I need necessarily a faster scope or is there any clever indirect method to confirm that ... ?

the only indirect way i can think of, to measure clock jitter & rise time, is using a sound card.
http://www.transaudiogroup.com/clock-comparison-files

but you need a much faster clock source.
really nice 3-way loudspeakers.

Huh ?  Are you sure you're not posting at the wrong thread ?

http://www.antelopeaudio.com/blog/word-clock-sync-jitter/
http://en.wikipedia.org/wiki/Analog-to-digital_converter#Jitter

[JuanPC]

seriously this test board can become a Rise Time test reference, just like those form VoltageStandard.
http://shop.voltagestandard.com/main.sc;jsessionid=82D4359A155BB6590CCBD3F5C5ED0E54.m1plqscsfapp06

[Marco]

Could you create a rectangular pulse with two different length traces to feed the comparator inputs?

[Rerouter]

if you just want fast rise times you can get clock buffer / divider / driver chips with very very fast rise times, i'm currently eying a SY89312V which has a max of 160pS, you can easily feed this with any old clock signal as long as you make it differential,

[krenzo]

I played around with the ADCMP580 a few years ago.  I analyzed it with my Tektronix 11801B with the SD-22 12.5 GHz sampling head.  Attached is the only picture I still have of its outputs.  The timing offset between its two outputs was intentionally my doing.

[free_electron]

Sorry to burst the bubble but thing is slower than the jim williams pulse generator.

You are doing 400 millivolts in 37 picoseconds.  or  92.5 picoseconds per volt.

The JW avalanche generator shoots over 12 volts in 400 picoseconds. that is 33 picoseconds per volts ... the JW pulser has an almost 3 faster risetime ...
You need to compare apples with apples. So normalisation to picosecondes per volt is in order. to compare to your 400 millivolts : the JW generator climbs 400 millivolt in 13 picoseconds. much faster than that chip can do.

While this chip is not bad it is not nowhere near the JW pulser. it costs more , needs a signal generator to drive it and is not self containd. the JW pulser is small box running from a coin cell. ready to run.

[starlino]

I was thinking it can be made self-contained with a little feedback. See for example here:  
 w2aew is using an inverter shmitt trigger 74AC14 , I think a similar trick can be used with this comparator to get it oscillating.

Thanks everyone for the great feedback.

[Jay_Diddy_B]

Hi,

I was able to obtain a risetime of 70ps, measured with a 20GHz scope, with my version of an avalanche pulse generator.



The thread starts here:

https://www.eevblog.com/forum/projects/transmission-line-avalanche-pulse-generator/

I used a BFR505 transistor.

Jay_Diddy_B

[starlino]

free_electron: JW avalanche generator is fantastic, and yes it will win the competition if you're talking about slew rate (V/s)  .

However I would argue that  50ps rise time  @ 0.4Vpp is still better than  400ps @ 12V if we're talking about measuring bandwidth on the scope.

Let me start by noting that there's nowhere in the bandwidth formula  Bw = 0.4/Tr  , that we find the slew rate or the absolute Vpp value. That's because the bandwidth (frequency) of a pulse signal is determined by how fast it goes from  (low) flat  to (high) flat, and not really from how  low to how high it travels. In other words by increasing the amplitude of your signal to 12V you do not get a faster equivalent bandwidth or rise time ... If you amplify a  signal that has a 1Mhz harmonic (equvalent harmonic of 400ps rise time  is 0.4/400ps = 1Mhz)  to 12v  that harmonic will still be 1Mhz ...

The equivalent bandwidth of  50ps is  0.4/50ps =  8Mhz, thus you could evalute scopes up to ~8Mhz/5 = 1.6Mhz with it  , or use the more complex formula  Bw = 0.4/SQRT(Tr^2 – 50ps^2)   and measure even higher bandwidths, that takes into account the 50ps rise time of the test signal.

Let me know what you think guys.

[alm]

Let me start by noting that there's nowhere in the bandwidth formula  Bw = 0.4/Tr  , that we find the slew rate or the absolute Vpp value.

The bandwidth and slew rate of a system are largely independent. Scopes are generally thought to be bandwidth limited rather than slew rate limited.

The equivalent bandwidth of  50us is  0.4/50us =  8Mhz, thus you could evalute scopes up to ~8Mhz/5 = 1.6Mhz with it  , or use the more complex formula  Bw = 0.4/SQRT(Tr^2 – 50us^2)   and measure even higher bandwidths, that takes into account the 50us rise time of the test signal.

In theory you can use that formula, but you'll find that for Tr(scope) << 50 us Tr(scope) will disappear in the noise. For example a 1 MHz scope (Tr(scope)=350 ns) will have a Tr of sqrt(50us^2 + 350ns^2) = 50.001 us. Do you have the resolution to distinguish a 50 us rise time from a 50.001 us rise time? This is assuming the rise time of your pulse gen was known down to the nanosecond to begin with.

[Galaxyrise]

However I would argue that  50us rise time  @ 0.4Vpp is still better than  400us @ 12V if we're talking about measuring bandwidth on the scope.

Did you mean picoseconds there? Or was the initial post in error?

[starlino]

Galaxyrise: yes I ment ps not us. I corrected my post. Sorry , late night posting