White LED Spice Model

If you have data points for other LED's... measure VF at many currents (*)... PULSED to avoid heating, send the data to me and I'll make a model for them.

(*) Measure VF at IF = 0.1mA 0.2mA 0.5mA 1mA 2mA 5mA 10mA 20mA 50mA ...Jim Thompson

I found my old HP-41C curve fit program card. Then I found my old HP-41C.

The model I get from your curve is:

.MODEL MyWhiteLED D

• IS = 4.4249E-9
• N = 8.3923
• RS = 7.7980

No subcircuits required.

That N looks frightening.

Jon

Yes, but it seems to work for the white LED (probably a blue LED) curve. I haven't gotten into the physics of the problem, but the emmission coefficient (N) seems to track the band gap of the junction involved. So this isn't as bizarre as it appears.

Here's a white LED Spice model that I created by fitting to data points of a real device:

Retrieve the original posting at...

Newsgroups: alt.binaries.schematics.electronic Subject: White LED Spice Model Date: Fri, 15 Jan 2010 09:18:31 -0700 Message-ID: X-Newsreader: Forte Agent 6.00/32.1186

to see simulation versus data points. ...Jim Thompson

I'm not sure if that will work over temperature.

It would be interesting if someone could take data at several different temperatures. It'll be tricky, you have to be quick or self-heating will screw up the result. Maybe with a plotting curve tracer? ...Jim Thompson

BTW: "your curve"? Whose curve? You clip too much, Paul. ...Jim Thompson

-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

That's a good point. N is often used as a 'fudge factor' for Vt, so for wild values of N (greater than 2.0 or less than 1.0) it would be reasonable to expect that nVt may vary with temp (or other conditions) in some different manner than kT/q.

Where I've seen weird values of N pop up, SPICE started to fall apart for anything other then approximate values. It was time to toss the standard model and go with something cooked up by the physicists.

Sorry. I took your spice model from your post, ran it for 70 current points and poked every tenth one into my program. I didn't have the table function values (the PDF you posted in a.b.s.e which I assume were your original measured data points). By the way, there appears to be an outlying data point in your graph.

So I was actually curve fitting to your curve fit.

EG is also a variable of interest. ...Jim Thompson

I finally figured that out ;-) I first thought someone I had filtered out, like from googlegroups, had posted some data ;-)

The "outlying data point" is an unknown, so PSpice stops and _holds_ at the last PWL data point. ...Jim Thompson

Right. That kicks in when you model diode's temperature dependence. But here's a question: What happens to the original subcircuit based model, which is based on a 1N4148? The activation energy isn't specified, so you're going to get Spice's default (probably for a silicon diode). As we are dealing with a (probably) blue LED, the chemistry will be diffeent, and we probably should find a more appropriate value for EG.

I agree. My subcircuit fit was accidentally fallen into while just noodling around ;-)

If someone will take data at 3 or more temperatures, I'll try to fit it. ...Jim Thompson

Evidently the PSpice people haven't learned anything about dry-labbing and tossing undesirable data points out from the climatologists. ;-)

;-)

PWL HOLDS at the last entered data point. Perhaps I can create one that continues the last SLOPE ?:-) ...Jim Thompson