But a flurry of rocket-fast bipolar RF transistors that cost just pennies. Also, if you don't need more than 10V or an Rdson under 10ohms consider UHF dual gate FETs. Some of these are impressive when used as switches. Best of all is that most of these also retail under a dime.
Your voltmeter is too slow. If it were fast enough, it would see 50 ohms for the round-trip delay of whatever length of cable you have. And for an ideal coax of any finite length, Rin jumps from Zo to infinity after the round-trip delay.
In real life, with a long piece of coax, the resistance looks like 50 ohms for some nanoseconds, and then starts to increase as ohmic losses pile up. I'm not sure what the r-vs-t equation is for an infinite 50 ohm lossy cable. Anybody know?
I've never got a bipolar transistor to switch super-fast. I tried driving one of the Infineon 45 GHz SiGe npn transistors with a 20 ps risetime pulse, and got a really mediocre edge at the collector. The only fast edges I've seen from bipolars was in avalanche mode.
GaAs fets and phemts are, by contrast, phenomenal switches.
I think patents like this deserve to be infringed. Hell, I came up with this "novel" circuit a couple of years ago when looking at generating an evil pulse, simply by reading up on the ancient art of magnetic pulse compressors, and (ICRW) reading about Grehkovs diode. And Im pretty thick really :)
And that aside, it still appears to be a pretty crap way of achieving the objectives, especially now that super-fast digital stuff is way easy.
Well, yeah. Transmission line. But how long does it take to charge the entire cable? Surprise: 100ns again.
Looks sort of like the first pi/2 of a highly damped LC oscillator which is exactly what it is. But it doesn't seem as if you could effectively reduce the gate drive risetime by more than about 10% (which might make more of a difference if the actual turn-on characteristics of the FET are taken into account).
That is not strictly necessary, at least for some value of resistivity of the dielectric. The constants for the resistance per unit length of the conductors and the conductivity per unit length of the dielectric just go into the equation for the characteristic impedance.
Z0 = sqrt((R+jwL)/(G+jwC))
from the Royal Signals Handbook of Line Communication, first published 1947
I guess that is now out of copyright (50 yrs for crown copyright???) so I suppose there would be nothing to stop someone from scanning it and putting it online. Maybe someone has.
I think it's more complicated than that. If you TDR a long lossy transmission line, it initially looks like 50 ohms, but the effective resistance increases with time, without limit, which makes sense physically... if a shorted 50-ohm coax is 5 miles long and has 1K dcr, it will measure 1K after things settle out.
I think what's happening is that the added series loss resistance creates dispersion, namely causes the impedance to become a function of frequency. The Royal Signals equation is approximately true only for short lines and high frequencies.
A telephone "600 ohm" twisted pair measures closer to 100 ohms at high frequencies, but acts sort of like 600 ohms for long audio runs.
Well, Terry, what I had in mind was specifically the RTLinux patent, which all low-level software people know is just a copy of half a dozen pre-existing real-time adapters for Windows, DOS, etc. Also, I had in mind the torrent of software patents of the past couple of years (you don't really believe 10K+ new software ideas have come in the last two years?).
But while I was collecting my thoughts, a run of silly and just plain wrong patents have been discussed on this group (indeed, you participated :-). Is that enough for you?
I thought you got several good answers, including 1) if you drive it properly, the 2n7002 is 10x faster than you think, and 2) if you really want to scream, use a GaAsFET.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.