[Radio Wrangler]

If you come from a valve background, you are used to reasonably well controlled parameters. The grid voltage for a given anode current is close to the spot value in the data book, and the slope (gm) of the curve is also close to the spot value in the book. Life is easy!

The bipolar transistor comes as a shock. Yeah, the base-emitter voltage is quite dependable, but the current gain (hfe) can vary widely from device-to-device. A 2:1 range is good in this area, 4:1 or more is common. Some devices were sold in graded versions and you still got around 2:1 range. Artful design is needed to make circuits which can tolerate these uncertainties, be sure-fire without selected devices and adjustments to be twiddled.

The FET then comes as a double whammy. Tons of uncertainty in the transconductance (gm) AND the bias voltage can be all over the shop, AND the current in saturation has huge uncertainty, all within one type number of device. Lots of FETs were sold in graded versions. Very early types like the 2N3819 had particularly wide specs. So the artful design needed by bipolar transistors is stepped up a notch.

Because the Junction FET biases like a valve with negative bias needed on the input, and gives transconductance sort of gain, valvey people felt more comfortable with them. This led to a lot of designs which were scaled copies of things which worked nicely with valves, but the designers hadn't understood that the nice solid foundation of tight specs had been jerked out from under them. There's a lot of amateur radio circuits which only work with selected FETs as a result of this. Semiconductors tend to be batchy. Buy a boxful and they may be close-ish. The designer can build several off of his circuit and things look well. Someone else buys FETs, gets a different batch and has trouble.

Professional designers learned to avoid JFETs, and only used them where they really helped.

JFETs are popular in oscillators because they're easy to get going. It's still difficult to make sure fire designs which don't need fiddling in some instances.

Oscillators involve you setting a device up to create negative resistance. Once you have negative resistance, ANY resonance with low-uish loss can oscillate. Oscillators can hop around several resonances and look like they're possessed by demons!

If you don't have a scope or spectrum (spectre?) analyser with enough frequency coverage you may not see what's really happening and just see rather foxy behaviour at lower frequency and DC.

Even mundane FETs work to rather high frequencies, better than period bipolar transistors. This opened the door to spurious oscillations up to much higher frequencies. So the wanted frequency and DC conditions would show sudden jumps and even hysteresis. Just a few centuries earlier, the blame would have been laid on beelzebub and the designer burned at the stake.

Getting a wide-range gate dip oscillator circuit to oscillate on the wanted resonance and not the strays is difficult. The coward's way out is to build a complete oscillator for each band. Transistors are cheap. Tuning capacitors aren't, but you can use values better suited to each band. A butterfly capacitor and a hairpin loop for UHF.

David