Current mirror to match parallel LED string currents.

[Flyback]

Hello,
The following shows how a current mirror can be used to equalise led string currents....

https://ledlight.osram-os.com/wp-co...rentDistributioninParallelLEDStrings.Web_.pdf

..it obviously works on the principle that matched BJTs with equal Vbe's, will have equal Ic's.

However, if greater string current flows in the led string which contains the "reference BJT", then the current mirror doesn't work.
....is this because in such a case, the "non-reference BJT's" would always have at least as much Collector-emitter impedance as the "reference BJT"?

 

[Nigel Goodwin]

Sorry - but what would be the point?.

Simply use the transistors as individual constant current sources (add an emitter resistor in each one, and feed the bases from a common reference).

 

[Flyback]

..I don't understand?...that is basically what is happening anyway.
All the bjts bases are connected to the collector of the reference bjt.
Yes I agree that small emitter resistors would improve the matching of the bjts.

 

[Nigel Goodwin]

..I don't understand?...that is basically what is happening anyway.
All the bjts bases are connected to the collector of the reference bjt.
Yes I agree that small emitter resistors would improve the matching of the bjts.

Emitter resistors (and a base reference) would make them all constant current sources - removing the need for the overall one.

 

[Flyback]

removing the need for the overall one

..theres no "need" for an over all one....that reference led string is going to be there anyway.....why would I want to add a reference?

Are you talking of putting the voltage across one of the emitter resistors into a negative feedback loop and regulating it....then feeding that base voltage to all the other bjts?

In any case, its bad news for all of us because it appears that matched *power* NPN bjts are simply not available with Ic >100mA.....this has surprised me...I suppose i can just buy some power bjts and *assume* that because they are off the same reel they are matched?...but that sounds a little worrying....

I wonder why osram put that app note up when there are no matched power bjt's on the market?

 

[Nigel Goodwin]

..theres no "need" for an over all one....that reference led string is going to be there anyway.....why would I want to add a reference?

Are you talking of putting the voltage across one of the emitter resistors into a negative feedback loop and regulating it....then feeding that base voltage to all the other bjts?

No, just a voltage reference on the bases (referenced to chassis - one resistor and two diodes would do), and a resistor in each emitter to set the constant current.

It's a standard constant current circuit, used almost everywhere - particularly in power amplifiers.

 

[Flyback]

I think if all the paralleled series led strings were running off a constant voltage rail then I see your point...but all these leds are being supplied by a single current source.

I believe you are talking of 2 series leds, from base to emitter of the bjts....anode of "top" diode to base.(?)

 

[Nigel Goodwin]

I think if all the paralleled series led strings were running off a constant voltage rail then I see your point...but all these leds are being supplied by a single current source.

No, you don't need a constant voltage, as long as you know roughly what the voltage is, that's all that matters - but it would be a complete waste of a constant current source to feed the individual ones.

I believe you are talking of 2 series leds, from base to emitter of the bjts....anode of "top" diode to base.(?)

No, two simple diodes are all that's required.

 

[Flyback]

sorry I meant "2 series diodes" not "2 series leds"

 

[Nigel Goodwin]

sorry I meant "2 series diodes" not "2 series leds"

That's OK, you could of course do it with two LED's, or even one LED for that matter - but two simple diodes is cheaper.

 

[crutschow]

The current mirror technique is a relatively efficient way to match LED strings driven from an efficient (switching) constant current source. Adding a reference voltage and resistors in the emitters of each transistor would reduce that efficiency. Also that would mean you have two constant-current devices in series and they would be fighting each other.

But adding a small emitter resistance would help matching, as noted in the article. Its value likely could be sufficiently small to have only a small effect on efficiency since you probably would only need a few 10's of millivolts drop for adequate matching.

 

[Nigel Goodwin]

The current mirror technique is a relatively efficient way to match LED strings driven from an efficient (switching) constant current source.

My point was that you could do away with the overall constant-current source, and have individual ones per string, with VERY few extra components.

 

[crutschow]

My point was that you could do away with the overall constant-current source, and have individual ones per string, with VERY few extra components.

That's OK if the constant current source is a linear device, but if it's a switcher constant-current source for efficiency, then you'd need a constant-current switcher for each leg to maintain that efficiency.

 

[Flyback]

in the current mirror circuit, the reference bjt has its collector and base shorted together. Surely this means that the reference transistor can therefore never become saturated?..because the base-collector junction can never be forward biased.

 

[crutschow]

in the current mirror circuit, the reference bjt has its collector and base shorted together. Surely this means that the reference transistor can therefore never become saturated?..because the base-collector junction can never be forward biased.

Technically, yes. The voltage across the collector-emitter obviously cannot be less than the base-emitter voltage or about 0.7V for a silicon BJT.