Yeah, SPICE is notoriously bad at charge storage. It's modeled as an exponential capacitor -- a battery, in essence, and the analogy is unsurprising.
A semiconductor junction tends to separate charges across the junction, so you get an excess of N on one side and an excess of P on the other. If there are charges just floating around (say from spontaneous pair production -- shining light on a solar cell for instance), they migrate across the junction and create a voltage. The V-Q characteristic is comparable to a battery, where it has a near-constant voltage, until charge is all but depleted, at which point it drops exponentially. The charge is just microscopic compared to an electrochemical battery. And, like a battery, it will self-discharge, usually on the order of 10s of us for silicon of average voltage ratings without special treatment.
A continuous supply of charge, say from continuous exposure to light, gives a V-I curve with a similar profile, for a similar reason; current is limited to the amount of pair production going on.
When it comes to BJTs, you need to slam the base around, which is all the more reason why you should (properly) think of them as voltage controlled devices: collector current is exponential with Vbe, and it takes current to change Vbe (pushing charge around). The only gross difference to a MOSFET is, the base conducts forward current, which is handy in some cases and annoying in others.
Another way to enhance switching speed is to avoid saturation completely. The latter is exemplified with ECL and CML circuits: the transistors go between 'off' and 'linear', but do not saturate. Schottky TTL works by clamping the last ~0.4V of saturation with a schottky junction (Baker clamp), shunting base current and preventing stored charge.
It's even more advantageous to avoid turn-off, because the circuit itself becomes very slow when insufficient current is handy: dynamic resistances all increase (e.g., r_e = Vth / Ic), bandwidth drops, and slew rate drops (I = C * dV/dt). The best analog circuits
Here's an example of the dependence of doping on recovery:
2N3904, B-C junction (E open), square wave input, series diode (DUT), 50 ohm load, 1:1 scale I think.
B-E junction (C open): higher doping, somewhat smaller junction.
In contrast, B-E with C tied to B (so collector current flows as the base is forward biased) gives almost nonexistent recovery.
Recombination: by adding impurities to trap free charges and effectively short-circuit them from within the semiconductor itself, you get a semiconductor (diode, transistor, whatever) that works more poorly (higher leakage, lower breakdown voltage, lower hFE), but runs much faster, especially in saturation speed. TTL chips and 2N2369 are examples of gold doped transistors -- keep in mind, TTL ran at 5V when almost all bipolar chips of the day handled 30V with ease (e.g., uA741). Today, you will occasionally see reference to platinum doping or electron irradiation applied to very high speed rectifiers -- same idea, applied very carefully so as not to impair the breakdown voltage (say for 600 and 1200V diodes!), while increasing forward voltage drop and reverse leakage slightly.
Among top shelf RF parts, supposedly, PHEMTs aren't so great in pulsed applications, but GaN FETs are okay. None of them are ever qualified or documented for conventional parameters, pulsed or time-domain operation, or sometimes even basic DC parameters.
On a related note, I would love if LDMOS devices became just a little bit more powerful (higher Idss, lower Rds(on)) and way cheaper (IRF510 is cents, while your average LDMOS of similar size is $15 and up) so I can use them for MHz+ converters. Super-Junction MOSFETs have made great strides in the 600V+ market, but the Cdss is still astronomical at low voltages -- it manifests as reverse recovery, even in the forward direction.
Tim
Home
Help
Search
Login
Register
