More fun with avalanche pulsers


 

I rebuilt the pretrigger and avalanche pulser onto one tightly packed board, and added SMA connectors for the output and the charge line to allow fast change. BNC for the ~65 volts from the power supply.
I'd discovered a GR 874 shorted termination in my bag of adapters, so I spent some time playing with shorted lines and reflections. Then I wanted to make a smaller toroid for the power supply - the junkbox one was (relatively) huge and wouldn't fit in a compact box. The Tek 130 LC meter got a workout figuring out AL for various cores. I was able to make a half-inch core work (think it was a T50-43) but the clock frequency required was high enough that it was getting into the unshielded pulser sitting next to it. An unknown green one saturated too soon and required very few turns - must be an iron powder one with high permeability. Finally a 3/4" yellow core (72 turns to get to 250 uH, that was tedious) worked at the original frequency and didn't cause trouble unshielded :)

A better RG-58 cable (SMA to BNC) worked well as a charge line, whereas a similar RG-174 cable was a droopy disaster. Putting the cheap BNC RG-58 cables on the end (with a barrel adapter) was also unsuitable. The "better" cable is too short by itself to be useful for risetime measurements on slower scopes, since the flat top is only about 3.5-4 ns long. As mentioned in the "3T77A tunnel diodes (again)" thread, I will try RG-402 (.141 semirigid) before calling it good enough. I doubt I have the skill or patience to build a really fast pulser to push the S-2(72 ps) or especially the S-4 (25 ps) to its limits.

Added a couple pics to the Sampling with 3S2 album of the breadboard so far and the pulse output.


Chuck Harris <cfharris@...>
 

A rule of thumb:

If a toroid is painted all over, it is powdered iron.
If a toroid is black, but not painted, it is ferrite.
If a toroid is covered with a form fitting plastic case,
it is a ribbon steel.

-Chuck Harris

Charles wrote:
...An unknown green one saturated too soon and required very few turns - must be an
iron powder one with high permeability. Finally a 3/4" yellow core (72 turns to get
to 250 uH, that was tedious) worked at the original frequency and didn't cause
trouble unshielded :)


 

Thanks for the tip!
It turns out that paint color is not useful as a guide to the material's specs though. I had expected a color-coding standard but there does not seem to be one. I have found some tables validating my assumption that yellow is low permeability and green is high, although not to be relied on.

Anyhow, I found that using a 33 ohm emitter load instead of 50 for the avalanche transistor (adding another 100 ohm in parallel) minimized the reflection and other distortion of the falling edge, although at the cost of some amplitude. 8 volts is still more than enough into 50 ohms.

Now to figure out the smallest Minibox I can cram it all into. I can't believe how expensive they have become! Price of aluminum, or price of labor? I think I may use a 10-turn pot to adjust the avalanche bias supply rather than an internal trimmer.


 

I bought ten BFR505 transistors which are supposed to be faster than 2N2369's. Unfortunately they come in SOT-23 packages (i.e. smaller than a mouse turd). I had a fun time installing one into my pulser. Dropped the first two (but found them later by lying on the floor for a really close look), the third sprang out of the forceps and I heard it "tick" somewhere across the room. Finally got the fourth mounted (emitter soldered directly to the SMA output connector).

I moved some components for the shortest possible lead lengths, but it still has a divot in the top of the pulse that I can't tune out. Could it be the open BNC hanging from the end of the charge line? Still likes 33 ohm better than 50 ohms of emitter load, too. Amplitude is decreased since the BFR505 avalanches at 30-32 volts instead of 50+ like the '2369.

The good news is that the risetime is indeed quite a bit faster. As best as I could measure, 10-90% is 400 ps with a 6.5 volt top :) This quite noticeable when using the 350 ps S-1 head, so I went back to the 75 ps S-2. Pics added to "Sampling with 3S2" album. (If I could figure out how to link individual pics in a post, I would!)

But the bad news (which others have noted) is that the jitter and noise is also increased. It's not really bad especially while smoothing, but I guess that's a consequence of a small, very fast (Ft=9 GHz) transistor?


jerry finn
 

Nice work.... I'm thinking of re-activating some equipment.


 

Thanks :) I stuffed it all into the 5 x 2-1/4 square Minibox, including a 6 ns semirigid charge line.
Displayed 10-90 risetime is just under 400 ps on a 75 ps sampling head.
Flat-top could be flatter but that really hurts the risetime...
I'd need to go to microstrip and leadless (surface-mount) parts to get any more performance. That's another project for another day!
Pics added to avalanche pulser album.


 

Any thoughts on what's causing the ring with the 2 ns period at the start of the flat top?
Maybe the short but unavoidable length of exposed center conductor where the charge line connects?
https://tinyurl.com/y9pemjh6


Jonathan Pyle
 

I don't have an answer, but I would note that the Leo Bodnar Fast Risetime Pulse Generator also has some ringing at the start of the flat top, and its circuit construction is extremely compact, so it might be due to something other than lead lengths.
http://www.leobodnar.com/shop/index.php?main_page=product_info&cPath=124&products_id=303


 

That is true, but Bodnar's pulser is a full order of magnitude faster than mine (30-40 ps vs. 380) and I'm sure nothing is what it seems at that speed.

Anyway, I may just stop here rather than drive myself nuts trying to get a perfect square pulse! It's faster than any of my scopes can measure except the sampler. On my 7A26/7403 (the amp is 200 MHz but the mainframe only 65 MHz) I just see a little 1 volt rounded bump a few ns wide ;)


Chuck Harris <cfharris@...>
 

It might just be the math. You cannot get
a perfectly flat top even with infinite bandwidth...

Look up Gibbs Phenomena.

Basically, the Fourier combination of sine waves to
create a square wave will never quite get to a perfect
square, because there will be spikes and ringing at the
discontinuities.... even if you go out to infinite bandwidth.

-Chuck Harris


On Sun, 28 Mar 2021 09:23:01 -0700 "Charles"
<charlesmorris800@centurytel.net> wrote:
That is true, but Bodnar's pulser is a full order of magnitude faster
than mine (30-40 ps vs. 380) and I'm sure nothing is what it seems at
that speed.

Anyway, I may just stop here rather than drive myself nuts trying to
get a perfect square pulse! It's faster than any of my scopes can
measure except the sampler. On my 7A26/7403 (the amp is 200 MHz but
the mainframe only 65 MHz) I just see a little 1 volt rounded bump a
few ns wide ;)





 

True... but if this were the source of the ringing, then wouldn't I see substantially less ringing, and a corner closer to square, on the S-4 25 ps sampling head compared to the S-2 (75) or S-1 (350 ps) heads? The S-4 and S-2 traces look basically identical, and the S-1 looks bandwidth-limited ("softer" corner and ringing) as I expected...


Ed Breya
 

You can smooth out the flat-top by adding a little series R from the emitter to the output cable connection. This will trade some edge speed for more civilized appearance afterward. Remember, nothing is really impedance matched here, so trade-offs need to be made, depending on what features are needed. If the goal is to get more flatness and pulse width for viewing lower speed scope performance, the rise time and amplitude are probably plenty good enough already, so you can give some up. I didn't notice in the discussion anything about provision for pulse extension, just that there's about 6 nSec of 141 line inside. Do you have the open end of the charge line going to an SMA connector for external line? That would take care of pulse width, so cleaning up the waveform is next. I'd recommend trying a small SMD series R, like from a few ohms up to maybe 22 ohms, experimentally.

Also note that there is such a thing as too fast a rise time when checking lower speed equipment. When checking scopes that are native 50 ohm input, it should be good to go, but for setups with 1 meg inputs with an added termination, there will be a number of discontinuities and reflections - the scope is mostly a capacitive load, and a very short test signal rise time will aggravate the situation. This can make it difficult to assess the response to "normal" signals the scope would encounter, although it would show what it actually does at very fast edge speeds.

Ed


 

Thanks. For now I'm trying to achieve the fastest possible risetime and a not-too-ugly flat top. I have somewhat achieved that already.... I can reduce the overshoot and ringing by adjusting the trimmer cap. I have a calculated 380 ps 10-90% risetime now. What might a few ohms in series with the charge line do to that? (The 2N2369 did give a cleaner output as in the earlier photos, but only about 650 ps risetime).

I left the neatly-cut end of the .141 line open. If I wanted to continue over-stuffing this Minibox I could solder an SMA connector to it and add another piece of line. But I also don't want to overtax the BFR505 which is pulsed every 2 us. At slower rep rates (relaxation osc mode), other experimenters were seeing failures in a few hours with longer pulses than 16 ns...

I remember your previous advice on matching and dynamic impedance changes. By trial and error I found that paralleling another 100 ohm emitter resistor (i.e. 33 ohm emitter load) improved everything except output amplitude, of which there is still plenty (6-7 volts). All these traces are: SMA output, SMA-GR adapter, GR 10x attenuator, 3' of GR 50 ohm cable into the 50 ohm sampling head. So any mismatches in the cable or sampler should also be attentuated on reflection.

When removing the cable and plugging the attenuator directly into the head, the persistence of the ringing through the whole flat top is now visible without the cable attenuation... As expected, the risetime improved also (but naturally the overshoot is worse). 320 ps displayed (310 corrected) without the cable. 380/370 with it.
https://tinyurl.com/m6es2et9


Craig
 

Thank you Chuck! I was unaware that J. Willard Gibbs also contributed to electronics, but I am not at all surprised. My educational background is in Materials Science and he is will known as on of the greatest of the greats in thermodynamics, and chemistry. J Willard Gibbs is one of the greatest American science minds who ever lived. There is apparently very little in the physical sciences he did not touch.

Craig