# Avalanche pulse generator, and some scope porn.

in builds by | 9 comments

By now we are all sure you know about Jim Williams’ avalanche pulse generator from his famous application note 47 we even posted about Kerry Wong’s pulse generator using jellybean transistors. And we can’t forget The EEVblog covering a build sent to him. Today we will be exploring our build, that is not quite like the others.

This is the classic Jim Williams avalanche pulse generator. C4 is charged through R11 when the voltage reaches the avalanche voltage(below blue) C4 is discharged through T4 and the impedance matching resistors giving a fast rise time pulse(below yellow).

Producing the needed voltage to create an avalanche condition is where we strayed from the pack. Instead of using a boost regulator IC, we opted to use discrete components forming an astable multivibrator(above) using the same 2n3904 transistors used on the avalanche generator. The resultant PWM from the multivibrator was used to drive a simple switching voltage booster that can produce about 50 volts from a pair of AAA batteries. this voltage was then feed to a chain of diode multipliers(below) to produce the high voltage needed to create an avalanche condition.

Also included in this build was a simple low voltage detector created from a pair of, you guessed it, 2n3904 transistors. T3 turns on when the input voltage exceeds the forward voltage of the diode plus the turn on voltage of T3(in this case several 1n4148 in series.) When T3 is not turned on the base of T5 is pulled high turning on the led indicating insufficient voltage to power the circuit.(with a little headroom)

We all know that rise time can be used to measure Gaussian bandwidth using the simple formula:     BW = 0.35 / Tr     where BW = bandwidth and Tr = rise time. Now let us measure a couple digital oscilloscopes.

on the MDO4014-3:  0.35 / 2.2ns = 159MHz about what we expected.

on the Atten ADS1062CA with Siglent firmware: 0.35 / 2.4nS = 145MHz not bad!

We had a guest stop by the lab. John Seney from Teledyne LeCroy came by with a drool worthy 20GHz oscilloscope.

This is a snapshot from A LeCroy SDA 820Zi-a, An impedance mismatch is surely slowing us down, rise time was only 747pS. looks like we will need to try this one again. with proper attenuation. We might even swap out some transistors to see what we can get with what is lying around the lab.

This entry was posted in builds and tagged , , .

1. Sleepwalker3 says:

Interesting that the tail end of the pulse varies so much with the different scopes. Did they bring a towel for you with the LeCroy? – Certainly Drool worthy

• SQKYbeaver says:

the MDO was the only scope that was properly terminated,

I will have to arrange another try with a better setup. or try to modify the circuit to not drive upwards of 40volts. i was well past the 8V peak to peak max on the SDA

2. Sleepwalker3 says:

Have you published the complete circuit (as one thing, rather than separate sections) anywhere?

• SQKYbeaver says:

i had a drool towel handy, i have not yet posted a complete schematic.

3. J. Peterson says:

By the way, in most schematics, transistors are labeled “Q1″, “Q2″, etc. The “T” designation is reserved for transformers, since they arrived on the scene a few decades before.

4. SQKYbeaver says:

seems there is an error in the library I am using.

5. Reg says:

Interesting project as these are required for scope calibration. Long ago I tried to get sub ns rise time square waves using a 10 MHz clock feeding an ECL divider chain. In the end I concluded that my board fabrication was defeating me. Ultimately found a surplus Tek generator and dropped work on the project.

I suggest you study the “Digital Black Magic” books by Johnson & Graham. Construction details (e.g. surface roughness of the foil!) matter a lot when the speed goes up.

A low cost generator design suitable for use to 200-300 MHz scopes would be a big boon for hobbyists trying to repair and maintain old gear.

6. SQKYbeaver says:

the pulse is only about 2.5ns (best estimate) the idea is to have a relatively square pulse lasting about 3.5 to 4ns with a max of 10VPP, much harder than it looks.

I have got a much lower peak by dropping C4 but that also shortens the pulse width.

I agree this can be very useful as long as we can guarantee its performance, and at this point we can’t(no budget for a \$150K scope)

7. Reg says:

I was referring to rise time as that’s what is important to verifying the bandwidth. I was trying to build a combination timebase & risetime tester. Hence the clock and divider chain.

Do you really need 10 VPP? That’s going to be *really* hard to do. For a modest 1 ns risetime (not very square for a 4 ns pulse) you need a slew rate of 10^10 V/s. Calculate the instantaneous current required for various capacitive loads. I assume from the spec that your objective isn’t scope testing and calibration.

My first pulse generator design (several years earlier) was a 7400 series quad nand gate that when triggered turned itself off. That gave me single pulses of a few nanoseconds from a cheap part. Which was important because I was a poor grad student trying to get a PhD in geophysics. But I wanted a fast single shot pulse for another hobby project.

I think an SMT dongle design scope tester with limited functionality should be reasonably reproducible, but I agree it’s not a trivial project. A USB controlled device with a suitable connector coming directly off the board would allow doing a lot of cool instrument calibration tests.

Have Fun!