The Picosecond Pulser , Designed by the late Jim Williams is a handy tool to check rise-time of circuits, perform TDR type tests and an overall handy little gizmo to have.
The base principle is to get a fast bipolar transistor to avalanche. A high voltage supply trickle charges a small capacitor until the avalanche voltage is reached. At this point the transistor goes into controlled breakdown and dumps the accumulated charge into the output. Depending on the transistor used this yields very fast pulses in the order of picoseconds. This design has gone through 2 iterations.
Picosecond Pulser MK-I
Picosecond Pulser MK-I
The MK1, depicted above right, was the base pulse generator as designed by Jim. The original design was constructed, dead-bug style using thru-hole parts, and wrangled in a pomona box.
My design updated the original design to modern-era surface mount parts and uses a printed circuit board. Further alterations include a different output connector and the battery used.
Picosecond Pulser MK-II
Stripline cap in MK-II
The MK2 built on this design but added a stripline charge capacitor to lengthen the pulse so there would be a nice plateau at the top. The stripline basically forms an open-ended coax and is used for its capacitive properties. when the transistor avalanches not all charge is released instantaneous. See it this way: the stripline behaves like a long tube holding charge. When the transistor avalanches the electrons close to the transistor are consumed first. The ones further in the tube take a while to roll out. So this behaves like a capacitor releasing charge over a time period, hence the nice plateau you get to see at the top of the pulse.
The original Jim Williams design could accommodate for this if you used a piece of coaxial cable. Due to size and the fact i wanted to keep this thing mechanically sound and easy to assemble i opted to calculate the equivalent coax as a strip-line and embed it in the circuit board. Some creativeness was required as striplines, actually a microstrip in this case, require a ground plane to be operational.
The schematic is pretty straightforward. The 3volts CR2032 coin cell K1 feeds the entire circuit through switch SW1. The LED D4, with its series resistor R2, indicate power is on. Capacitors C5 and C6 provide bulk capacitance for the boost converter built around U1, an LT1013-ADJ. the on board switching transistor in U1 periodically pumps current though L1. upon releasing the inductor the back-emf creates a high voltage at the D1-L1 node. This voltage is tripled using the diode-capacitor cascade built with diodes D1,D2,D3 and capacitors C1,C3 and C4.
The voltage created is sensed using resistive divider R4 and R5. So far this is a classic boost-pump, apart from the high voltage cascade. the peak voltage of this circuit approaches 90 to 100 volts. Capacitor C4 holds the final voltage.
The high voltage is used to trickle-charge a small capacitor. This can either be in the form of a standard surface mounted 2pF to 8pF in the form of C2, C2A, C2B and C2C; or in the form of the embedded stripline capacitor of the PCB.
The choice is made by installing the 10 MegaOhm resistor either in position R8 or R8a (waveguide) and closing the J1 bypass jumper.
The jumper J1 is necessary to disconnect the waveguide from the transistor.
The built-up charge in either the waveguide or C2 will reach a point where the transistor will go in controlled breakdown, also called the avalanche effect. at this point the transistor goes in full conduction and acts as a switch closing very fast. the charge held in the capacitor is injected into the 50 ohm resistor creating a very fast transient pulse. This pulse is brought out using a precision SMA connector.
Since we are dealing with very small capacitances, high value resistances and very fast pulses some special attention is required around the transistor.
In order to avoid any inductive effects it is critical that the 50 ohm resistor is a thin-film type surface mount part. To reduce stray capacitance and remove conductive leakage paths the transistor is mounted upside down, dead-bug style, with the emitter wire feeding the center pin of the SMA connector.
The 50 ohm resistor is soldered directly from this pin to ground.
The collector of the 2N2369 transistor is internally connected to the case so we can snip off the collector wire. The PCB is designed in such a way that the case of the transistor actually drops in a hole. the metallization of the board connects the case directly to the 10MegaOhm charging resistor, or , via the bypass jumper on the back, to the stripline capacitor.
The base of the transistor is bent down 180 degrees and directly soldered onto a pad holding the connection to the 10 KiloOhm base resistor.
The emitter wire is bent 90 degrees,trimmed to length, and inserted in the hollow rear of the SMA connector center pin.
A small drop of solder makes the electrical connection. Don't overdo this. We don't want big blobs of solder or residual flux on this connection. the goal is to keep this connection as clean as possible.
The 50 Ohm output resistor is a 1206 package and basically put on end. to install this first pre-tine the pad on the PCB. Holding the resistor with tweezers, heat the pad with the soldering iron and place the resistor vertically in the molten solder. The resistor should ideally touch the free hanging emitter wire.
Once the bottom solder has cooled and solidified you can apply a small dot of solder on the emitter wire connecting it to the free standing 50 ohm resistor.
Clean off any flux remainders around this section. also clean the backside of the board where the case was soldered down. essentially the entire area where the blue solder mask has been peeled back should be spotless. good solvents include isopropyl alcohol (IPA or rubbing alcohol) if you can't have dedicated flux remover. Rubbing alcohol can be had at any pharmacy, just make sure you get the 100% and it does not contain other agents such as aloe vera.
The Avalanche Effect
Avalanching is a semiconductor physics fenomenon. When a non-conducting semiconductor structure is brought close toits breakdown voltage a small current will begin to flow. In an avalanche transistor this current runs from collector to base. This current is large enough to create a base voltage across the base resistor bringing the transistor in conduction. This process triggers and electron cascade. The collector is fully saturated with 'holes' (lack of electrons) , and the emitter is fully saturated with electrons. The moment the transistor enters conduction all emitter electrons shoot to the collector.
The velocity combined with the massive amount of electrons released causes some of the moving electrons to actually knock-off additional electrons in the semiconductor crystal. The net effect is that now more hole-electron pairs exist than before entering conduction.
This is essentially a chain reaction. Even when the voltage has collapsed belwo breakdown the conduction continues as the effect is self-replicating. The electrons started moving at breakdown point but are continuing to knock off other electrons as they travel. The process only stops when no more electrons are being knocked off. At this point conduction stops , the remaining free electrons recombine with the remaining holes and the depletion layer is re-formed.
So here is the big question: how does it perform in real life ? Well, the only way to tell is to actually measure it ! Problem is, we are dealing with a very fast, physics, effect so we need a very fast machine to record such an event. My home scopes top out at 1GHz analog bandwidth with 4G samples/second so, while fast, they could not really show the real signal. After all this is what Jim had at his disposition as well when he did the original design.
Since this was a kit going to be sold with the option of a 'calibration report' it was time to call in a favor. But what to get ? I needed to establish once and for all what these pulsers generate so i would need the MOAO: Mother Of All Oscilloscopes, the Agilent DSA-X 93204.
This machine is mind-boggling. At the time this test was done (April 2004) it was the fastest realtime scope in the world. It has since been surpassed by a new model doubling the analog bandwidth, sampling speed and price.
32GHz analog bandwidth. And that is NOT the -3dB point !
80 Gigasamples per second on 4 channels
2.1 Billion points of memory, PER channel.
The image on the left shows the machine connected to the test-jig holding a pulser and running flat-out at 80 Gigasamples / second. The scope was calibrated to have the measurement plane sit at the end of the 3.5mm connection cable.
The scope has an on board cable compensation mechanism whereby tit uses the internal calibration pulse(15 picosecond rise-time) to compensate away the cable effects.