sharpie7
MOSFET fixed threshold / no body effect
What model describes MOSFET behaviour using a "beta" and a fixed threshold? The few CMOS design textbooks' introductory chapters I looked at always model threshold as a function of the channel dimensions and substrate bias (among other things but those seem to be the most significant parameters). I haven't found a model that approximates a realistic variable threshold like this given a beta parameter and base threshold, though it might be possible since beta is determined by channel W / L.
Evidently some microprocessors use a dynamic body bias in places as a power-saving feature because subthreshold current is a function of the threshold voltage, which is a function of body-source voltage for a given Vds, so they can cut down on leakage in unused parts of a circuit.
Not that we really care about subthreshold action... There are probably better examples of where this matters, like maybe some N-ary logic could exploit that.
Hi, I got it from Pillage, "Electronic Circuit & System Simulation Methods", which got it from Muller and Kamins, "Device Electronics for Integrated Circuits".
It's described as the "quadratic model" here: https://ecee.colorado.edu/~bart/book/models.htm . That link doesn't mention beta but it seems to be the same model if you set beta = mu Cox (W/L). It also discusses the variable depletion layer model, which has a varying threshold voltage.
Someone has asked for the EKV model (#21) to be provided as an option, since it's much more accurate and can model things like subthreshold conductance.. And it would also fix your issue as well.
Wow that's a lot of parameters! I would have to do what ngspice does and run a converter to generate C from this verilog code, and then convert the C to java. Because the verilog code does not seem to calculate the matrix elements that I need.. It only gives the currents.
Yeah. I see some recognizable parameters from ngspice level 1 / 3 / 9 (probably level 1 since 3 doesn't have lambda for channel length modulation). So it could be meant to be easy to translate models (similar philosophy to sgp -> vbic). Anything that requires sophisticated "parameter extraction" and knowledge of process parameters of course can be rather annoying.
If I were choosing a practical model to port to Java I'd probably start from the ngspice level 3 / 9 (I've yet to spot a behavioural difference between those. I think level 9 is meant for experimenting with modifications since people depend on stable behaviour for level 3). Sorry I know nothing about verilog. The ADMS functionality looks crazy!
The only confusing bit is the way that some parameters override others. E.g. I've never actually seen a model that explicitly supplies kp for instance so they must always depend on that to be derived.
I could easily steal the mosfet level 3 code from spice. I have already done that with the spice bjt code. But is it sufficient to model subthreshold conductance? I thought you needed EKV for that, based on what little I know about it.
Hm I've not tested sub-threshold effects. I'd guess no. Or it might have some heuristic to guess generic subthreshold effects. But at least it gets body effect (what I was originally looking at back when I posted this).
EKV seems rather obscure. I only heard about it when you mentioned it here, and skimmed some of the old papers from the 90s, but no implementation was available until I noticed this one appeared in the latest spice manual.
Oh yeah... The interesting "analog reciprocal" example doesn't really work. It's super broken in spice because these "floating mirror" / "beta multiplier" patterns are really touchy with the more detailed models. I did a slightly more realistic one and forced the curve to roughly fit 1/x (with some DC offset) by using a weird valued depletion mosfet... Still not great. https://tinyurl.com/yjmf3tza
I noticed the BJT model... Very nice! :D
This should probably be a separate issue, but.. The analog reciprocal was contributed a while ago (by Felthry). He either designed it himself, or he got it from someplace he didn't remember. I should probably just delete it, if I can't replace it with a real circuit that's actually been used.
Right, I confused your MOSFET issue with another one, especially since you mentioned subthreshold current.
Hi @pfalstad . Kinda want the 4 pin mosfet with the subthreshold current. like in #21. found this that seems to explain in a simpler way https://studylib.net/doc/18221859/mosfet-operation-in-weak-and-moderate-inversion. maybe if i have time i will try to set up the java development environment...
