Hello everyone,
I enjoyed reading these discussions of pulse generators:
https://www.eevblog.com/forum/blog/eevblog-306-jim-williams-pulse-generator/https://www.eevblog.com/forum/projects/transmission-line-avalanche-pulse-generator/In these threads, some readers suggested mechanical solutions for pulse generation. Unfortunately I couldn't find reports about how that approach performs. Some time ago, I made a generator with adjustable pulse width (adjustable by attaching different lengths of 50 Ohm coax as the charge line) from one of several mercury wetted reed relays, which I had lying around. With this post, I want to show results and tell why I finally switched to building an avalanche pulse generator.
Mercury Wetted Reed Relay Pulse Generator
Mercury wetted reed relay pulser, schematic.
The following picture shows the relay pulse generator at work. A charge line of 0.5m results in a 5ns pulse.
Mercury wetted reed relay pulser,
1ns/div, 500mV/div, HP54100A (1GHz), 50 Ohm input.
The incredible HP54100A with a bandwidth of 1GHz measured a rise time of 1.2ns. But I was not quite happy with this solution. The uncontrolled impedance of the reed switch and the wiring diminished signal quality. It was obvious that I should have put more effort into the setup. Disassembling the relay and separating the reed switch, optimizing the wiring and then using a suitable coil would have helped. But there were further issues with this reed switch. The pulse looked like it suffered from randomly occurring slight impedance mismatch. This varied with even minor tilting of the switch around its prescribed orientation. Other relays of this type behaved the same. While disassembling one relay I found out that the contacts weren't like in an ordinary reed switch where two springy parts move towards each other, but they were three parts with the moving one held in place only by the surface tension of the mercury. Perhaps this moving and in fact loose part is dangling around? The switch did not bounce, but it seemed, that the moving contacts cause variations in on-resistance, despite of being embedded in mercury. Reducing the coil current during the movement of the contacts in order to limit their speed before they finally touch, improved that behaviour, ... but I would have liked something more reliable.
So I discarded the idea of using a mechanical switch and decided to build a triggerable avalanche pulser with adjustable pulse width, based on Linear Technology AN47 by Jim Williams. See page 93 of
http://cds.linear.com/docs/en/application-note/an47fa.pdfAvalanche Pulse GeneratorInitially I used a 2N2369 as avalanche transistor, which delivered a rise time of 150...200ps, depending on the actually used transistor. Then I tried other RF transistors which were in stock at a local store: BFW425W, 2SC3355, BFP196E, BFP420, BFG541. The best results regarding rise time were achieved using BFG541. Rise time is about 75ps. However, pulse shape was best with 2N2369. With BFG541 the rising edge looks good, but the falling edge is distorted.
Avalanche pulser with BFG541, schematic.
Detail of pulse generator: Transistor, attenuator (R14, R15, R18)
and high voltage supply (R9) are directly mounted to SMA connectors.
The following two pictures show the output of the BFG541 equipped generator with 1ns/div and 50ps/div. Charge line length is 50cm. Rise time is about 75ps.
Avalanche pulser with BFG541,
1ns/div, 100mV/div, DS800 (8GHz), 20dB attenuator, 50 Ohm input.
There is some ringing visible, which is the result of an attempted correction of "dribble-up" before knowing what it is. For "dribble-up", see the below section "Trying to adjust the pulse shape".
Avalanche pulser with BFG541,
50ps/div, 100mV/div, DS800 (8GHz), 20dB attenuator, 50 Ohm input.
Same configuration as above. Rise time is about 75ps.
Measurement EquipmentMeasuring these short times was possible by using a DS800 8GHz sampling oscilloscope from IBZ Electronics. The following picture shows the DS800 scope and the pulse generator. The three cables attached to the generator are trigger output, pulse output and the charge line. In this photo the length of the charge line is 4cm, which results in a pulse width of 400ps.
DS800 oscilloscope and pulse generator.
Trying to adjust the pulse shapeWhat was common to all variations of my charge line pulsers, no matter whether relay- or transistor-driven, was the ramping up of the pulse soon after the rising edge. Jay_Diddy_B also encountered this while building his pulse generator and he identified it as "dribble-up". See his post
https://www.eevblog.com/forum/projects/transmission-line-avalanche-pulse-generator/msg606142/?topicseen#msg606142.
In the following picture I've marked a flat portion (region A) and the "dribble-up" (region B).
Pulse top,
500ps/div.
As long as the charge line is short enough, the pulse consists only of the flat portion (region A). The longer the charge line gets, the more of the "dribble-up" (region B) becomes part of the pulse. This distorts automated measurement of rise times, which are measured shorter, when the pulse width is low and are measured longer, when the pulse is wide and the pulse therefore includes "dribble-up". The initial rising slope itself doesn't change significantly with different lengths of the charge line.
Before I read more about "dribble-up", I thought that there is something wrong with the pulse generator. The first idea was to put a variable capacitor in parallel to the charge line. This produces an initial peak, extending the rising edge to an adjustable amplitude. This peak makes automated measurements more consistent. However, this doesn't cure "dribble-up", so I removed the variable capacitor. Lacking a clever plan, I tried putting stray capacitances in parallel to the charge line (short wire(s) sticking in the air). Somehow, this lifted the entire flat portion (region A) a little up, partially covering the "dribble-up". However, it added ringing. At least it seems that the averaged pulse top represents the true amplitude.
The following two pictures show pulse tops before and after attempted correction.
Pulse top, original,
500ps/div
Pulse top, after attempted correction,
500ps/divRegarding "dribble-up" I found this:
Tektronix Type 284 Pulse Generator. For "dribble-up", see page 2-12.
http://bama.edebris.com/download/tek/284/Tek%20284_v6.pdfAEA Technology AN204
http://help.aeatech.com/helpdesk/attachments/6005556252Maxim Integrated Application Note 5141
http://pdfserv.maximintegrated.com/en/an/AN5141.pdfPicosecond Pulse Labs Application Note AN-3a
https://kh6htv.files.wordpress.com/2015/11/an-03a-pulse-meas.pdf
Home
Help
Search
Login
Register
