Fast risetime/falltime pulsers

Discussion in 'The Completed Projects Collection' started by RichardO, Feb 8, 2015.

  1. RichardO

    Thread Starter Well-Known Member

    May 4, 2013
    1,238
    385
    Let me know what you think...


    Here is a project I have been working on to generate very fast rise and fall time pulses for testing high speed amplifier circuits.
    The circuits use a transistor array with 2 PNP and 3 NPN UHF transistors -- The HFA3096.

    These pulsers get transition times similar to an avalanche pulser but do not need the high voltage power supply. The power consumption is less than 50 mW.
    Since the output voltage is lower than an avalanche circuit, attenuators may not be needed to connect to an amplifier input.

    My work is partly based on the following:
    2N5271 avalanche transistor data sheet,
    Linear Technology application note AN94f by Jim Williams,
    Zetex application note AN8 for the ZTX415 avalanche transistor.
    and the stripline avalanche pulser here:
    http://www.eevblog.com/forum/projects/transmission-line-avalanche-pulse-generator/?nowap

    I simulated my pulser circuits using LTspice.

    The PCB's were made using the toner transfer method and copper tape. The small PCB's used the copper tape to make the ground plane. On all of the PCB's, the copper tape connects the grounds from one side of the PCB to the other.
    I printed the Gerber files using the free version of ViewMate:
    http://www.pentalogix.com/download_software.php

    My friend Scot did tests on the pulsers and here are his results (see the waveforms).
    They were done with the pulser into 50 ohms, no coax, straight into scope. His scope is a DSO with a 115 ps risetime.

    For the tail pulser: Scope claims fall time is 830ps I get 880 with cursors.
    for the coax pulser: Auto measure says 380ps I get with cursors 500ps fall time and 1ns rise
    For the stripline pulser: Rise time 550 to 660ps with cursors, 480ps auto. Pulse width is 5ns.

    As you can see, the rise and/or fall times are all less than one nanosecond into a 50 ohm load.
    The waveshapes of the coax cable and stripline pulsers could probably be improved by fine tuning the damping resistor and capacitor values.
    A faster oscilloscope than mine is needed. My scope "only" has a specified bandwidth of 250 MHz which is equivalent to a 1.4 ns risetime. This is not fast enough to see these fast edges in detail.


    How the pulsers work (short version):
    Transistors Q3, Q5 and the parts around them generate a high current pulse. This circuit is similar to a PUT (Programmable Unijunction transistor) oscillator.
    The pulse is a few nanoseconds wide at a repetition rate of about 100 KHZ.
    This pulse drives transistors Q1, Q2 and Q4. The transistors act similar to an SCR (Silicon Controlled Rectifier). Once the "SCR" starts to turn on, it does so very hard and very quickly because of the very high, positive gain, feedback loop.
    A capacitor, coax cable or stripline (a 50 ohm impedance PCB trace) is charged slowly from the battery through a resistor. The "SCR" quickly dumps the charge into the 50 ohm output resistor when it turns on.

    The tail pulser outputs a negative going pulse with a slow resistor/capacitor recharge tail.
    The coax cable pulser generates a narrow, more or less rectangular negative pulse.
    The stripline pulser puts a positive rectangular pulse.
     
  2. RichardO

    Thread Starter Well-Known Member

    May 4, 2013
    1,238
    385
    Here is more documentation:
     
  3. RichardO

    Thread Starter Well-Known Member

    May 4, 2013
    1,238
    385
    And finally, the references:
     
  4. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,292
    1,255
  5. RichardO

    Thread Starter Well-Known Member

    May 4, 2013
    1,238
    385
    The intent is to test output amplifier circuits for high frequency function generators. These amplifiers have slew rates of thousands of volts per microsecond. The ultimate goal is a slew rate of about 12,000 volts/us (12 v/ns). The output of my pulsers are fast enough to do this. Their main limitation is that the amplitude is too low to test an amplifier at low gains since the amplifier output will not be swinging the full voltage range. In that case the answer is still to use an avalanche pulser as shown by Jim Williams.

    Thanks for the link. Several ECL comparators are listed on page 3. It is easy to forget how fast ECL is! It is hard to believe that ECL was one of the first logic famiiles. (RTL may have been first).

    The risetime of the SP9689 is spec'ed at 180 ps and the falltime is just 80 ps!!! ECL has a small voltage swing of only about 0.8 volts p-p so it has even a harder time driving an amplifier to full output swing. :(
     
  6. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,292
    1,255
    Yes, that is a challenge.
    :eek:
     
  7. RichardO

    Thread Starter Well-Known Member

    May 4, 2013
    1,238
    385
    If it was easy then it wouldn't be any fun. :D

    p.s. A friend built a prototype of an analog multiplier driving an output amp capable of around 2000 v/us at 20 volts p-p into 50 ohms. It had an odd offset on the output that we did not understand but I got it basically working. After doing some testing and modifications I gave it back to him. When he looked at it some more he found that the offset would go away when he put a scope probe on the output the amplifier driving the output amp. The amplifier was oscillating at a bit over 1 GHz and the capacitance of the probe was stopping the oscillation. :eek:

    This is the highest frequency either of us has ever created and he was a bit reluctant to fix it. :)
     
Loading...