I.ve made the ledarray and decided to be diffrent and went 20 yellow 20 orange 20 green 20 blue and finished it with red. So now the green and blue don't work properly due to the their color more than the ma's. Just to be extra creative (stupid) i went with the cheapest superbrite thinking i was getting regular leds in various ma's spread degrees. Doubles as a light show against the mirror on the wall. I'm just starting to get a handle on fets or mosfets. This is what i think will fix my lights, they don't got a mosfet calculator that i know of. There's four columns where i would add them i presume after the chip. I tried taking out the voltmeter and gave it 7.5vdc and she got a little too excited. There you go just the value of the "fets" or your idea. thx

fjc

You'll be running into a similar problem as me, except my problem is that I'm dealing with over 200 LEDs with standard 20ma, versus your blue and green LEDs requiring more with a smaller number of LEDs... There are several issues that come up...

Adding FETs (or BJTs or whatever) is on the right track, except there are a few "Extra" considerations with it... First is the nifty type of multiplexing the nk folks do where they drive two columns using one pin... That's great when you're driving the LEDs with just the chip since it has the relatively unique ability to be able to sink or source current... Once you throw a FET in there things change since it won't have that ability... That means each row will have to be drivin seperately...

Doing that works, but now instead of driving 100 LEDs with 15 pins, you'll now need 25 pins... Not exactly a "good" situation...

So now we have to be creative in how we multiplex it... In my case, driving over 200 I'm having a problem generating enough current for a single LED when my duty cycle is 1/40th... (Each column can only be on for 1/40th of the time...)

I haven't had time to do much research or experimenting with this lately as I've been tied up with other things, but right now I'm leaning towards doing a fairly complex form of charlieplexing or rather a crazy arse variation of it, in order to light more LEDs at a time an drop my duty cycle down... Unfortunately that still won't work for you since that is still going to require the tri-state ability of the MCU pins...

The only thing that quickly comes to mind would be using some sort of latching shift registers to drive them seperately... That way you could have each column driving via a fet and be able to drive with the right current... I don't remember the part number, but if I recall there is a special FET driver for LEDs that packs 8? fetlingtons in a single chip... That would be a good choice for that portion... A 595 is probably the best shift register for the job as well...

I haven't completely given up on my other ideas and convinced myself that the charlieplexing is the best way to go for my situation, so you might want to keep an eye on that thread and see what I come up with... It might help you with this one as well...

This Fet idea is quite the thing i need a MPF-102 N-channel to measure static electricity. Like brush your hair and the light dims.How cool is that? But for mine i didn't think it would be as simple as sticking one of these inline and expect it to work. It's the charlieplex wiring is the issue. I could just change them to "std" leds, or make another one but just thought i'd ask anyway. Thanks for the imput.