[time-nuts] Fast risetime pulse generator

[jimlux]

On 4/12/16 1:30 PM, BIll Ezell wrote:
> (cross-posted to volt-nuts)
> After paying only limited attention to this topic, I suddenly have a
> need for a pulse generator that has <150 ps risetime and a pulse width
> of at least 2 ns. 100mv amplitude or more is fine. I've looked at the
> classic Jim Williams avalanche generator, but I don't want to have to
> deal with the (relatively) high voltage source needed.
>
> I've done microwave design using Gunn diodes, so I'm drawn to using a
> step-recovery diode. The topology seems very straightforward, and I can
> build it right onto a BNC connector, no PCB.
>
> I'm thinking using an SMD835 diode, biased at ~1ma. The (sketchy)
> datasheet claims a T of 20 nsecs and a Tr of 85 ps, Cj of 0.4 to 0.8 pf.
>

how about a fast digital buffer?
I don't think Potatochip semi is quite in the 150ps range (although you 
could look at their website and see).

But we use microwave dividers and such all the time at 8 GHz.  Check 
places like Hittite.




> Questions:
>
> The obvious, is it reasonable?
>
> Is the bias current reasonable? I'm assuming the bias current is
> actually dependent on the repetition rate, you need enough current to
> replenish the charge within one pulse cycle. I suppose I could compute
> it from the stated junction capacitance, but I'm not sure that's the
> only factor.
>
> Will the stored charge actually give me the desired transition rate into
> 50 ohms? Hmm, again I should be able to compute this, but any other
> factors ignoring the non-diode ones like cap inductance?
>
> How should I compute the coupling cap from the diode to the load? Use
> the impedance at the pulse rep rate? Seems reasonable. BTW, I don't care
> about droop in the  pulse, just the risetime.  (measuring overshoot in
> an HF amp). Again, just want to verify that the obvious answer is the
> correct one. I clearly need to be very careful about the inductance.
>
> Thanks, Bill
>