Bandgap Design

arsenide

silicon?

the

temperature.

still

Thanks for the citations Tim.

Yes Jim, the variation of dVf/dT over the various colors is less than I first expected as well. Lumileds specs an identical -2.0mV/C for their 1 and 3W LEDs in all colors: red, amber, green, cyan, blue, and white (see DS25 & DS45 at

I haven't measured these, for what that's worth.

Using an LED to bias a bjt current source is an old cheesy trick -- the LED drift mostly offsets the transistor's dVbe/dT.

Only surprise in Tim's numbers is that the blue and white are different, especially given that white LEDs are in fact blue LEDs painted with a bit of yellow phosphor.

As Jim Thompson pointed out, I only gave one signifcant figure. That was because the application didn't seem to care, and because that's about as good a generic answer as you can give. Your LED and YMMV. If it really matters, use a bandgap. For hobby purposes, a known or well-characterized LED might serve.

Ian's right too: LEDs are light-sensitive, so unless you want your bias to change with lighting, keep the LED dark.

James

That would be a surprise in itself. Maybe some devices have been made that way, but the spectrum from those the "incandescent" types is too broad to work that way.

Perhaps a good supply of darkons would help.

You have some example part numbers for these?

I read in sci.electronics.design that Larry Brasfield wrote (in ) about 'Bandgap Design', on Tue, 29 Mar 2005:

Yes, the '2' should be recognized for what it is; a value to *one significant figure*.

Hey, deja vu all over again!

John

I read in sci.electronics.design that Jim Thompson wrote (in ) about 'Bandgap Design', on Tue, 29 Mar 2005:

Nicely within the range of 1.5 to 2.4 implied by '2'.

The wide spectrum "incandescent" LEDs are diodes in that there are two electrode connections brought out. Internally, the device is a directly heated tube without an anode (plate) at all. The heater is run much hotter than in most tubes. On the more advanced versions of these, the doping is done with Cesium.

The vast majority of white semiconductor LEDs are blue ones with a phosphor. The tripple die ones require a higher operating voltage since the three devices are in series. It adds up to about 2V + 1.7V + 1.6V =

Indeed, the incandescent variety are approximately black-body emitters with a very high color temperature. I've made a few of these myself. Chief drawbacks are exceedingly brief service, and high currents required.

I hesitate to point out the obvious, but the straightforward source for darkons is the DED (dark-emitting diode). Fabrication? See above.

James Arthur

Could well be -- Tim didn't give any part numbers. The few blues I scared up on Stanley's site, however, were InGaN. No d(Vf)/dT info -- I couldn't read their .PDFs. Their site,

Regards, James Arthur

Which?

Vishay TLHW490 is the phosphor type

I think the Sharp GM5WA06 is the 3 LED type.

An non-drawback is that sometimes you get a NED - a Noise Emitting Diode - for free.

Tim.

Cree made the first blue LEDs, and they were SiC. We used them in our VME modules as bus activity indicators. They were horrendously inefficient, needing 50 mA at close to 4 volts to be decently visible; we drive them from +5 through 27 ohm resistors with paralleled 74S38 sections. They were eventually discontinued, replaced by successively more efficient GaN (I think) devices. So our VME led's kept getting brighter, and customers started complaining about being blinded and having to keep the rack doors closed. So we had to keep issuing ECOs to manufacturing to change the resistor as the led's kept getting better. We're up to 1K now.

I think Cree still uses SiC as their substrate, but the actual emitter is epitaxial something else.

SiC fets should be fun, when they get the yields up.

John

I read in sci.electronics.design that snipped-for-privacy@yahoo.com wrote (in ) about 'Bandgap Design', on Tue, 29 Mar 2005:

They must use different semiconductors. The blue one may be an SiC device.

I believe this is vaguely related to the Write-Only Memory.

Cheers! Rich

A sizable chunk of blue LED's on the surplus/hobby market are the old SiC. When you buy a random blue or white LED from the hobby/surplus outfits and they make no claims as to manufacturer or model number, you stand a 10-20% chance of getting a SiC.

SiC's are still manufactured new for some specialty applications.

Tim.

something?

Oh yeah-yeah-yeah- you're the big shot semiconductor physicist today, eh, pretentious idiot? I don't buy it. There is plenty of high quality resource material on the web covering this subject, and no one needs some unaccomplished bs artist like you expounding upon the diode equation, which you can hardly do at all.

What a total crock of garbage!

How could I forget - the FED, Flame Emitting Diode, is a distinct possibility too!

wrote (in

'Bandgap

info --

all.

tool

d(Vf)/dT.

from

Engineer

Thanks. The UB5306X datasheet pg. 7 reveals all: I get the same result (~-3mV/C) over the same temperature range as you did (-40C to

-2mV/C from roughly +20C & higher. (This part is InGaN.)

Windows being what it is -- delicate -- I distain installing new applications where possible. Here though, I took the opportunity to load the latest FireFox browser and a later version of Acrobat Reader, whereupon I was better able to load and read Stanley's documents. I do wish, however, that webfolks as a group would ease up on the bandwidth, and hang on to all those pretty pictures and nifty scripts. Please webweenies, just give me the facts.

James Arthur