They don't give SPICE parameters in datasheets. (Except when they do.)
You have to look up the definitions for everything, and set up test cases in the simulator to produce a model fairly close to the real thing.
Example: IS = saturation current. Notionally, the cutoff current (e.g., Iceo). But this is rarely given in datasheets, with only some rated maximum provided (which is hopelessly high, e.g., 1uA). Typical values are still higher than what simulation models use (e..g, a 2N2222 might measure ~1nA at 25C, whereas the model gives 0.01pA).
In practice, IS is usually controlled to match Vbe at certain Ib or Ic conditions, and not to specify reverse or cutoff leakage current. If that parameter is desired, it can be approximated with a voltage or current dependent resistor. Related, N (emission coefficient), if present (e.g., diode model). See Ebers-Moll model.
BF, BR = Beta Forward/Reverse. BF is hFE or beta, more or less. I forget if this is approximated as constant, or if other parameters modify it. In practice, beta is NOT independent of various operating conditions, and especially is smaller at low currents, and higher at elevated temperatures. Reversed means, swap collector and emitter and measure hFE. This is used to manipulate Vce(sat), I think. See Gummel-Poon model.
FYI: many models specify erroneous BR, presumably to tweak better forward characteristics, at complete expense to reverse properties. BR=3 seems reasonable for a 2N2222; the PBSS303NX model gives a similar magnitude, however the actual device has BR more like 200 (and BF ~ 400)! This is one example where the model is only good for what it was made to do -- don't trust a model any further than you can throw it...
VAF = early effect voltage. Real devices may or may not behave as having a constant Early voltage, but YMMV, it's what you get. (Look up the definition of Early effect if you don't know.)
CJC and CJE are B-C and B-E capacitances (at junction voltage 0). There should also be an M factor for these, which describes how quickly that capacitance drops with voltage (they are nonlinear capacitances), and a VJ parameter which describes at what voltage that capacitance changes the fastest.
There should be an EJ parameter, which is bandgap. Related to IS, Vbe and whatnot. Default is 1.1 (silicon), so it's often left out. Other junction types, and semiconductor materials, often specify different values, but don't be surprised if the values aren't characteristic of the known physical properties of the materials -- ultimately, everything is a parameter, and whatever gives the best fit is the most correct model. (Example: silicon schottky is usually modeled with more like EJ=0.4, and SiC junctions more like 3ish.)
RB, RC, RE: parasitic or "spreading" resistance of the respective terminal. (Also, on the diode model, RS for series resistance.) I think these vary some as well (e.g., RB rising for base thinning effects at high Vce or current density).
TF, TR: forward/reverse transit time. Essentially junction reverse recovery, or storage time (forward = B-E, reverse = B-C; again, a forward/backward C-E swap case). Transit time is suggestive of charge diffusion through the base and junctions, but it's really just a name; it's modeled by a nonlinear capacitance, which is a "good enough" kind of model, but gets uglier the longer you stare at it. See also "TT" for diodes. Of particular note, forward recovery isn't modeled at all, so you will never see the terminal voltage on a diode junction (whether a .DIODE or a BJT or FET junction) overshoot when turned on. SPICE models tend to be about 2x overly optimistic about real switching speed (they might be designed around fT instead?), and even worse on storage (I've never seen a .MODEL that just sits there, stuck on, given excessive base current -- like a real device does).
I forget what IKF, XTB, ITF, VTF, XTF do. Things like saturation current crowding and noise factor, I think?
Your best bet is to create a "test bench" which recreates the standard transistor tests in the simulator, and adjust the parameters one-by-one until you get one which apparently matches the graphs and parameters in the datasheet.
In turn, you'll need to look up what the hell those measurements are in the first place, and how to do them. And what they mean, and whether a result is nonsense or real!
Overall, it's a deeply rewarding research project: BJTs will never go away, and the applied modeling experience will always be valuable! But do be prepared to study a huge pile of transistor terminology to get there. If you just wanted some damn model, well... oh well. You're probably better off modeling it like a dumb switch, or poking at 2N2222 parameters until you find the one thing that was limiting your specific particular case (e.g., faking a 2N2222 so it can draw 100 amperes?). If you're willing to go all the way and produce a *really good* model, you'll ultimately benefit not just yourself, but many others along the way!
Tim
Home
Help
Search
Login
Register
