Current balancing resistors

[throbscottle]

Here's another one that's had me scratching my head from time to time.
Suppose I want to put several of some kind of semiconductor in parallel. I know I need to put a resistor in series with each device to ensure the current through each is roughly equal, but question is, how do you work out what value of resistor you need? Is it based on how much inequality you are prepared to tolerate? Or is there something more empirical to go with?

Thanks in advance

 

[audioguru]

High current voltage regulators and high power audio amplifiers balance the output transistors with series emitter resistors.
If the product is made "over there" then they don't bother balancing anything. If (when) the product fails then they simply throw it away and replace it.

 

[jpanhalt]

Suppose I want to put several of some kind of semiconductor in parallel.

I suspect the solution may be different depending on the function of the semiconductor and the resistors. For example, the answer for LED's might be different than the answer for mosfets. Here is an answer from Fairchild for mosfets. My notes on the source document indicate that it is no longer on the Fairchild site, so I have uploaded the whole document. Fairchild recommends 4.7 Ω for each gate drive.

John

 

[EinarA]

I haven't actually worked through this engineering problem, but feel it is mostly a balance of how much current imbalance you can tolerate and how much power or voltage you can afford to lose in the resistors. The two goals conflict, a very common situation in circuit design.

 

[KeepItSimpleStupid]

It's not so much equal, but you don't want one of the transistors to be doing all of the work. The resistors compensate for the difference of Vbe which might be 100 mV or so and carry the full current. You end up with very low values of resistors. Otherwise, you sort of have the situation like batteries in parallel.

Initially, you would try to match the gains with bipolar transistors. Matching might be within 10 or 20% of each other.

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=9&ved=0CEUQFjAI&url=https://www.newark.com/pdfs/techarticles/irc/irc_powerSupplyStudy.pdf&ei=G-7kU_iOEMKUyASF0IBg&usg=AFQjCNE7KMqSQ6joszaGPyJGxIhx_WSP_g&bvm=bv.72676100,d.aWw&cad=rja

 

[throbscottle]

AG - where's "over there"? Not over here I hope!
John - that makes very interesting reading, and an angle I'd never thought of. I see they have chosen the value that lies at the kink in the efficiency graph. I wonder if a similar exercise with pn junctions would yield a similar result. There's an experiment to do here...
ElinarA - suck it and see, then!
KISS - Ahhh, I see the answer!

I once had a power supply from a mainframe computer when a lot of them were being scrapped (1983/84). It had for the 5V pass element maybe 6 or 8 2N3055's on a big heatsink. Each had a resistor in series with its emitter, They were aluminium cased, bolted to the h/s, 1 ohm I think (or it might have been 0.1 ohm). Never really thought about how they worked out that value, at the time, often thought about it since, but especially since I've been interested in boost converters with high input currents, and using paralleled devices becomes attractive.

 

[audioguru]

AG - where's "over there"?

In the country where some workers are little kids and get paid a bowl of rice per day.