----------------------[ Getting gigahertz from a 1N4148 ]-----------------------
Publication date: 2024-07-02


Let's make harmonics, lots of harmonics. This can generate evenly spaced
carriers for testing receivers, and create high frequency signals referenced
to a stable, low frequency reference. 

Harmonics are created by sharp edges on a waveform: Historically, this was
done with a diode or two, in a configuration like this: 

[IMG srd.png ]

As the diode moves from forward to reverse bias, it conducts in the reverse
direction for a short while before abruptly snapping off, creating a falling
edge of around a nanosecond even with a 1N4148. 

A small inductor converts the rapid fall in current into a very sharp voltage
pulse, resulting in harmonics up to hundreds of times the input frequency.
The largest gotcha is that many inductors will act like capacitors at higher
frequencies, sometimes even just at tens of MHz. This can be mitigated by
adding a small, hand-wound coil in series. 

It also takes quite a bit to drive the diodes, but a few volts peak to peak
from a 50 ohm source should do it. 

Specialized "step recovery" diodes can conduct in reverse for quite a while,
and snap off just a handful of pico seconds, creating strong harmonics up
to hundreds of gigahertz. Similar circuits can also be built with just one
diode, but those need a DC path on the input to bias it. 

Many diodes produce harmonics up to a few gigahertz, but the reverse transition
time is usually not in the datasheet, so it will take some experimentation
to find a good one. Normal switching diodes will have a weak output becuase
the reverse recovery time is too short to allow very much current to flow
in the reverse direction. 

That said, even some random 1N4148's generated harmonics up to 2 GHz, tapering
(but sill usable) off above 1 GHz. 

Nowadays, RF transistors capable of generating sub nanosecond rise times
are really cheap, making such diode tricks rather obsolete, but I still
think they're neat.