Where can one get them, and is there a wide range of transistor type represented?
How good are they, eg: E-B breakdown correct, beta at given sets of Ic VS temp correct?, Vbe VS temp and VS Ic correct? Are there models for each manufacturer?
As for the BVbe, so far, at least in Ltspice, that value isn't set in the stock parts and I found a few circuits I had in my junk folder that did not work after setting this parameter. Correction was easy for most of them.
if you look at the list of parms you can set, over their default values, it may help in getting a model that gets closer to a real life operation if you're having that type of problem.
Someone here helped me with a short cut that became a time saver if you want alter a value a stock model. .model YourNewNameforTheTranny AKO:TheOriginalNameOfTranny BVbe=6
You may already be familiar with this but others may find it interesting.
Jamie
Most manufacturers publish Gummell-Poon models for their transistors.
The data can be plugged into LTSpice, and LTSpice includes the Gummell-Poon data for all it's stock transistors.
Gummel-Poon isn't good at modelling transistors operating in the inverted mode - which is to say using the collector as if it was the emitter and the emitter as if were the collector.
VBIC is better, but the model data is commercial-in-confidence. I've never been able to get hold of any. LTSpice can run VBIC models - if you can get the data.
-- Bill Sloman, Sydney
Thanks;been over what you referred to. My main interest is how to describe in a model, beta VS temperature; i think leaving the rest of the stuff alone is OK. Tried beta constant and nothing else specified in a model, a plot of Ib and Vbe looked reasonable from 25C to 204C. So...how does one toss in a curve or data points for beta over temp?
the best I've seen in terms of accuracy and being able to converge when you use them comes from Alex Bordodynov. He posted all his models for people a while ago, some 35MB of zipped content! I'm pretty sure he matches the models very closely to data sheet.
There's a Russian site you can download them from.
if you can't find it let me know.
Found Google Groups, and in middle Jim Thompson
9/11/13- show quoted text - Here you go...
Kill the pop-up on the lower left, then you can download. Problem is, did not find the pop-up or models (yes, tried the "find" link). Any pointers? Thanks
Well, noodled around a lot and in
Measured a few transistors beta VS Ic; anywhere from almost flat to like a hill. Yet another relationship that the models fail to fit.
Am guessing the models were meant to work for only one temperature and one collector current.
type
I would like a copy of that myself.
?-)
I just sent an email to Alex giving him your localnet.com email address asking him to help you.
Are they really "behavioral" models? ...Jim Thompson
-- | James E.Thompson | mens | | Analog Innovations | et |
Don't know. Alex has designed semi's but he constantly discusses how to make the models fit a datasheet, so don't know.
it's c .SUBCKT ModelName Node(s) [PARAMS: ...]
it's behavioral ...Jim Thompson
-- | James E.Thompson | mens | | Analog Innovations | et |
Oh, I always called them models [even if they're a subckt] and behavioural [because the subckt is based upon observation, not basic principles]
For example, I always thought behavioural was made up of a conglomeration of mathematical components that 'acted' like the component. not based upon basic, deep down principles of a particular mask set using a semiconductor process. Fr me, a true model of an IC is a series of little components connected like the mask and all the pieces act like each little piece, even parasitics, which also can make for some huge subckts - but those subckts are pretty accurate during simulation.
Again, an example of a 'behavioural' model for an IC is made of current sources, voltage sources, impossible feats like negative poles, and a lot of other crap just so the thing acts like the data sheet under most conditions [think Analog Devices Models, where they simply plcae stage gain after stage gain, oops except for this one thing, and oops except for this other thingy].
I don't see how to make a true model of an OpAmp without using subckt etc. or maybe I missed the nuance of that (PARAMS: ...)
Not necessarily... a .SUBCKT can be a conglomeration of individual pieces, each described by a .MODEL statement or even other subcircuits.
PARAMS is simply a way of customizing. An _excellent_ example is my Configurable OpAmp on the Device Models & Subcircuits Page of my website >:-}
There are LOTS of bad models out there... mostly constructed by the manufacturers themselves... utilizing green PhD's just out of school, or marketing "engineers" >:-} ...Jim Thompson
-- | James E.Thompson | mens | | Analog Innovations | et |
I also think that fitting to a datasheet is almost the best way; far easier than getting a bit of expensive equipment for a DIY on hundreds of parts. Gotta be rich and have lots of time for that...
I disagree on the "fitted to the datasheet"; beta does not follow in a linear manner WRT temperature, AND beta WRT Ic _also_ does not follow in a linear manner WRT temperature. The models look like fantasy.
What models are you referring?
My comment was that one can fit BJT model parameters to a datasheet.
You jumped to a false conclusion. ...Jim Thompson
-- | James E.Thompson | mens | | Analog Innovations | et |
No i did not jump to ANY conclusion. I was/am referring to any transistor in SPICE; 2N2219A, 2N3053,
2N2369A, 2N3904, and 2N3906 specifically. Maybe someone might be able to fit BJT model parameters to a datasheet, but one sure does not see those.
It gets into a grey area as, for example, there are parameters in standard spice models that are there empirically, not because of a 3D physics based analysis.
Probably a bit of individual preferences for the name "behavioural". For me I agree with what you say. I usually only refer to "behavioural" as mathematical expressions and ideal components. However, in Spice this will always be achieved by putting the description in a .subcuit. If one then makes say, a diode by an exp() function, one might end up just using a diode, so it morphs into mix of mathematics and Spice parts to make, what is still, essentially, a "behavioural" description.
In SuperSpice I actually a set of components, like edged clocked d-types using nothing more than tanh()
Kevin Aylward
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.