|Back to projects|
Fast pulses are sometimes required when measuring slew rate or propagation delay, and for sampling. Avalanche breakdown of a bipolar transistor in this circuit produces pulses with sub-nanosecond rise and fall times. Pulse width is controlled by the length of a transmission line. Output is 5V into 50-ohms, and a very fast oscilloscope (not mine, alas) is required to do it justice:
The step-up switching regulator produces a +130V collector bias for avalanche transistor Q1. Collector voltage rises as the capacitance of the open line slowly charges through R2. When it avalanches, Q1 discharges the line into the output network. Over 1 Amp flows through Q1 for the duration of the output pulse:
Free electrons are accelerated by the high collector-emitter voltage. Avalanche is the collision of high-energy electrons with atoms, releasing further electrons causing a chain reaction in the semiconductor material. The free-running repetition rate is irregular, as the collector voltage (yellow) reveals:
Linear Technology application notes [1,2] show how the rate can be controlled by biasing the collector just below the point where random avalanching ceases, and applying trigger pulses to the base. Also, note their use of damping at the collector in reference 2.
The action of the delay line is illustrated by the following trace, which shows voltage at the open end (purple) against collector voltage (yellow) and output (green). To lengthen the pulse, the solid coax used to produce the short (3ns) pulse pictured above was replaced with a much longer cable. The inferior frequency response of the cable is evident from the trailing edge:
Collector voltage rapidly falls from 120 to 60V at the onset of avalanche. It's like the closing of a switch, connecting the 50-ohm transmission line to the output attenuator. The source impedance of the line and the output load form a potential divider. A -60V step propagates and reflects from the open-end of the delay line back to the collector. Pulse width depends on the length of the line. Despite the heavy load, the collector is held steady at 60V until the reflection arrives. If the line is mismatched, the initial step is not 60V, and there are multiple reflections.
A load of about 30 ohms from emitter to ground proved about optimal, suggesting a contribution of 20 ohms from the transistor. The T-attenuator presents about 30 ohms to the emitter and 55 ohms at the output, whilst simultaneously dropping the pulse amplitude to 5V. This is the best I could manage with the resistors I had in stock.
The SMPS was quickly lashed-together, without much thought, using a pot-core taken from my junk box. It works, but the efficiency is rather poor: current consumption is just under 100mA at 10V; and the RC4190N is slightly warm to the touch.
Its rise-time puts this circuit's output well in the GHz realm, for which my construction technique is barely adequate. You could do a lot better with SMD components and controlled PCB track widths. My resistor selections for the output matching network are sub-optimal; and those enormous through-hole parts are just a lot of stubs at GHz. Look at the consequences:
Collector damping, as used by Linear in ref 2, may go some way to cleaning this up.
Having recently moved to a new QTH, I'm in the process of erecting an HF receiving antenna in my back garden, and I needed a coax run to my upstairs-front-bedroom shack. The last few meters were already covered by an old cable-TV coax, which I extended a further 20 meters to the antenna feed-point. 75-ohm cable/satelite coax is well-screened, and has low-loss at HF. Before connecting the antenna, I used the avalanche pulse generator as a time-domain reflectometer to check cable integrity:
The change in height after 40nS is due to my running 50-ohm coax from the pulse generator to a 75-ohm cable-TV junction box on the shack wall. The little blip 10nS left of screen centre is where I joined my extension to the old cable. From here, a delay of 160nS for a 40-meter round-trip equates to a velocity factor of 83% - which is about right.
|Copyright © Andrew Holme, 2006.|