> Sims indicate the high-gain low-pinchoff-V parts are in fact terrible for FET compressor/limiter apps.
High Gm, low Rds, is not in itself bad, because it reduces noise voltage. Because noise current is about zero, this is lower noise power, and larger dynamic range, which we can scale to suit.
High pinchoff voltage would seem to be key. Signal voltage must be much less than Vp, so a high Vp gives larger dynamic range da da da.
For a specific device and process, Gm is generally higher for high Vp parts. Gm increases 1:2 to 1:8 over the whole range of Vp from one process, though less than the range of Vp or Idss.
The nearly-free variable is "size". Four devices in prarallel have the same max input but 3dB lower noise; one device on a similar process but 4 times larger has the same effect. Since FETs can be much cheaper than 6386 tubes, more/bigger FETs is practical, to some diminishing return limit.
> The power law was more like 4th or 5th instead of the square law
It appears that SPICE 2 and 3 JFET computations use the Shichman-Hodges model, "Simplifications such as gradual channel approximation and the square law for the saturated drain current are employed. Channel length modulation, was the only geometry effect and no sub-threshold conduction model is included." MOSFETs get 6 models, simple complex and empirical, but it seems all JFETs are the same to vanilla SPICE. A SPICE vendor could enhance that, but it seems nobody cares about JFETs, it's all about the MOSFETs.
Something new to play with: carbon nanotube FETs. Not yet at your local Radio Shack, maybe never as a discrete device, but possibly useful for jamming more wasted cycles into future CPUs.
High Gm, low Rds, is not in itself bad, because it reduces noise voltage. Because noise current is about zero, this is lower noise power, and larger dynamic range, which we can scale to suit.
High pinchoff voltage would seem to be key. Signal voltage must be much less than Vp, so a high Vp gives larger dynamic range da da da.
For a specific device and process, Gm is generally higher for high Vp parts. Gm increases 1:2 to 1:8 over the whole range of Vp from one process, though less than the range of Vp or Idss.
The nearly-free variable is "size". Four devices in prarallel have the same max input but 3dB lower noise; one device on a similar process but 4 times larger has the same effect. Since FETs can be much cheaper than 6386 tubes, more/bigger FETs is practical, to some diminishing return limit.
> The power law was more like 4th or 5th instead of the square law
It appears that SPICE 2 and 3 JFET computations use the Shichman-Hodges model, "Simplifications such as gradual channel approximation and the square law for the saturated drain current are employed. Channel length modulation, was the only geometry effect and no sub-threshold conduction model is included." MOSFETs get 6 models, simple complex and empirical, but it seems all JFETs are the same to vanilla SPICE. A SPICE vendor could enhance that, but it seems nobody cares about JFETs, it's all about the MOSFETs.
Something new to play with: carbon nanotube FETs. Not yet at your local Radio Shack, maybe never as a discrete device, but possibly useful for jamming more wasted cycles into future CPUs.
