Significant ringing with the Jim Williams 2N2369A pulse generator from AN47

[TimFox]

The 1 megohm input resistance is obviously negligible, but the scope’s normal 10 pF shunt capacitance at 1 GHz (relevant to 300 ps in time domain) is a measly -j16 ohms, which will screw up a 50 ohm transmission line (even if short).

[RoGeorge]

Just to be really clear this device is plugged into the 'scope BNC. No probe, No cable. It is a BNC plug with the components mounted on it.

It has a 50Ω resistor right across the centre pin to shell, so a 50Ω termination will make, and did make, no difference. The 'scope is however a 1MΩ input. Not sure what capacitance, but vaguely recall 10-15p.

I will experiment with settings as I need to bottom out what impact the 'scope may be having before I consume time on the pulse generator.

Use the probe on 10x, NOT the generator plugged directly into the oscilloscope input.  The oscilloscope input has 20pF, while the probe tip has only 4pF.  Follow the schematic.  Remove the additional 50 ohms between the center pin and GND.

At first, measure just the pulse generator alone.  Leave the oscilloscope on high impedance, not on 50 ohms, and use the 10x probe.  Put the acquisition mode on "Peak", and disable sin(x)/x interpolation.

Please attach the following two screen captures of the waveform:  one with the display in "Vector" mode, the other with the display in "Dots" mode.

[Zero999]

This was an SMD version, admittedly with a 2N3904, but it would not better 1250ps.

It was very small - mounted inside a phono plug that plugged into a phone-BNC converter directly onto the 'scope BNC.

Use the MMBT2369A, which is the surface mount version of the 2N2369A.

[T3sl4co1l]

The 1 megohm input resistance is obviously negligible, but the scope’s normal 10 pF shunt capacitance at 1 GHz (relevant to 300 ps in time domain) is a measly -j16 ohms, which will screw up a 50 ohm transmission line (even if short).

Usually the "pF" of a scope is the asymptotic slope of the impedance as it drops from 1M at DC.  What it drops to, isn't documented.  There is likely some ESR there, damping or perhaps even peaking the input.  But you'd have to look at the input circuit (or attach a reflectance bridge) to know.

The transient response and input impedance are generally best described when set to 50 ohms internally.  Even if you have a passthru terminator on hand.

Tim

[6SN7WGTB]

OK, probed with spring coil tip and probe on 10x. Dots mode. Vector gives same trace, but 'filled in' obv.

Can't see how to disable sin(x)/x on Rigol MSO5000.

So, 'scope and/or my bad technique are indeed ruining my day to some extent.

[G0HZU]

I don't know if this helps,  but a few years ago I made a time domain 'ping' generator based on the Jim Williams circuit but I used a faster BJT and I wanted something that didn't need a 90V PSU to operate. I made it on an old scrap bit of Rogers 4003C PCB using microstrip techniques and small SMD parts and a tight layout.

The result was quite good straight away, and I've not tried to improve it. I wanted a narrow ping signal to send through large wideband structures for educational purposes.

I just dug it out and powered it up. It needs about 36V to avalanche. See the image below. I don't have any decent modern scopes here with a high sample rate and/or lots of bandwidth. I have got a plot of it somewhere using a (works) Agilent DSO 80000 series scope with 40GSa/s and 13GHz BW.
However, all I can use at the moment is an old Tek MSO with 5GSa/s.

I think it will be possible to find a better BJT than the one I used and also the PCB layout could be optimised. I used a segment of an existing RF PCB after cutting it out from a much larger PCB. However, this should be seen as a starting point if you want to make something better. I made it in a hurry and I can show you the circuit etc if you are interested.

[G0HZU]

Note that my PCB circuit has an SMA end launch connector followed by a decent SMA 10dB attenuator (Suhner, rated to 18GHz) and this is then fed into the scope via an SMA to BNC adaptor with the scope set to a 50R input This isn't ideal, but it's going to be a lot better than using a x10 scope probe...

[Conrad Hoffman]

Maybe helpful to read the Tek TU-5 manual- https://w140.com/tekwiki/images/5/57/070-384.pdf

They certainly didn't do anything fancy to couple it to the scopes. I do like the JW circuit with the BJT better than tunnel diodes! Somewhere I have a small bag of 2N2369A parts, if I could ever find it.

[RoGeorge]

OK, probed with spring coil tip and probe on 10x. Dots mode.
...
Firts pic (direct to BNC, no probe):


Dots mode (with probe on 10x):

At 1ns/div and 8GSa/s should be 8 dots/division, don't know why are there so many pixels/division, my scope puts one dot per sample.

That oscillation doesn't look like wrong probing.  Those oscillations are about the same as in the first pic, so it must be something else, something that remained unchanged between the different tests.

Maybe the decoupling capacitor is not good enough (high ESR), and such making the power supply wires to appear as a transmission line.  Or maybe it's something wrong with the oscilloscope and that is its impulse response to a very short but high voltage pulse, because everything else has been changed during the previous tests yet the unwanted oscillations remained about the same. 

[iMo]

If you want to study the parasitics - here is the LTspice model (there are several avalanche transistors there but only this one models the physics, afaik) of the William's pulser from Bordodynov's library.
Simply do add the parasitics and the attached probe's or a BNC's input model and you will get the picture..

[tggzzz]

Maybe the decoupling capacitor is not good enough (high ESR), and such making the power supply wires to appear as a transmission line.  Or maybe it's something wrong with the oscilloscope and that is its impulse response to a very short but high voltage pulse, because everything else has been changed during the previous tests yet the unwanted oscillations remained about the same. 

ESL (inductance) is more likely to be a problem at these frequencies.

I suggest you create a Spice model of your circuit. First add all the ideal components. Second, measure the length of each wire and component body. Third, for each 1mm length add a 0.8nH in series with the ideal component.

Simulate, see what the output looks like, twiddle inductor values to see which make the most difference. At this point you will have a feel for what is more and less important.

[T3sl4co1l]

Mind, trace, lead and body length aren't strictly problematic: they have characteristic impedance, or if you want to extend the lumped-equivalent model accurately, some shunt capacitance that balances things out too.  It's not the length per se, but the mismatch * length that affects signal quality.

This all seems much more compact (short dimensions) than the scope's BW equivalent electrical length, so I don't have a problem with the build.

For comparison, here's the unit I showed earlier, with a ~2m pulse line attached:



350MHz bandwidth so it's probably sharper than seen here (with averaging enabled, measures 0.91ns rise, 1.28ns fall), and the peak-dip response is probably due to the dimensions (bulk, and height-above-GND) of the 2N3904 itself, and the attenuator built of THT resistors.

Tim