I intend to mount 12 bicolor 60mm LED matrices, in a 4x3 configuration, but electronically it is going to be 1x12 with 8 PNP transistors driving the rows and 12 TLC5940s sinking the columns. Row selection will be done by a 3:8 decoder (74138) with a 1/8 duty cycle selecting the transistor to provide current for the rows, all wired together. The wiring diagram I am using right now for the transistors is here: SECARS Homepage - SECARS
Do you think a TIP42 is OK for providing current this application or might that transistor be too slow? What would be the best transistor to use? The choice of the TIP42 right now was rather arbitrary and I am not against using a MOSFET if that is the better choice. The switching of the rows will be, hopefully, at 800Hz for a complete refresh at 100Hz. Thank you.
That circuit is for 12V - what supply do you have (A 74138 can't drive 12V)?
Also how do you get from 12 bicolour LEDS (24 elements) to 4x3 (12 elements) to 1x12 to "8 PNP transistors driving the rows and 12 TLC5940s sinking the columns (8x12x16)... Something isn't adding up.
MarkT:
That circuit is for 12V - what supply do you have (A 74138 can't drive 12V)?
Also how do you get from 12 bicolour LEDS (24 elements) to 4x3 (12 elements) to 1x12 to "8 PNP transistors driving the rows and 12 TLC5940s sinking the columns (8x12x16)... Something isn't adding up.
I am just using the circuit link to demonstrate that I have a PNP transistor between the positive voltage and the load with a resistor on the base and a resistor pulling the base up. It's a 5V power supply, TIP42 transistors, and I know a 74138 works fine with TIP42s because I am already using it in a circuit.
About the matrices, assume * is a bicolor 8x8 matrix. You can wire it like this:
We have 64 columns (32 for each color) and 24 rows. The duty cycle is 1/24.
But you can also wire it like this:
And just place the matrices on the PCB in a 4x3 arrangement.
Now it is 192 columns (96 for each color) and 8 rows. The duty cycle is 1/8. The row current may be moderately high, though.
Is the TIP42 a good choice to drive such an arrangement? That is what I am concerned about.
So each column is driving 192 LEDs, at upto 60mA each, so max current of 11.5A? That's twice what a TIP42 can handle. Even at 5A a TIP42 needs about 500mA of base drive and will have Vce around 0.6V (dissipating 3W). A driver circuit for the TIP42's base would be needed and it would have to handle about 2.5W of dissipation itself.
A logic-level p-channel MOSFET with 0.01ohm Rds(on) would only drop 0.115V at the full 11.5A (1.3W) and would be a better choice. Still have issue of slow switching unless you boost the gate drive with a high-current MOSFET driver (like the MIC4422 and family) - the issue is the high input capacitance of power MOSFETs.
With a 5V supply you won't be able to use darlingtons as the voltage drop from them and the constant current drivers will be too much for anything but red of orange LEDs I fear.
Copious decoupling would be needed switching 11.5A, 0.1uF plus 10uF ceramic for each TLC5940, perhaps 2200uF electrolytic as well to bolster the powersupply's regulation.
MarkT:
So each column is driving 192 LEDs, at upto 60mA each, so max current of 11.5A? That's twice what a TIP42 can handle. Even at 5A a TIP42 needs about 500mA of base drive and will have Vce around 0.6V (dissipating 3W). A driver circuit for the TIP42's base would be needed and it would have to handle about 2.5W of dissipation itself.
A logic-level p-channel MOSFET with 0.01ohm Rds(on) would only drop 0.115V at the full 11.5A (1.3W) and would be a better choice. Still have issue of slow switching unless you boost the gate drive with a high-current MOSFET driver (like the MIC4422 and family) - the issue is the high input capacitance of power MOSFETs.
With a 5V supply you won't be able to use darlingtons as the voltage drop from them and the constant current drivers will be too much for anything but red of orange LEDs I fear.
Copious decoupling would be needed switching 11.5A, 0.1uF plus 10uF ceramic for each TLC5940, perhaps 2200uF electrolytic as well to bolster the powersupply's regulation.
In another thread I was given the advice to break up the string of 12 matrices into 4 of 3 each, drive each with it's own transistor. So the string of 12 TLC5940s stay the same, the 74138 drives the base of the four transistors so each row gets an independent current source. Does that help me on the current source front? Is there a better transistor to use if I used that setup than the TIP42? Also, I am planning on using this for text. I am not planning on lighting every LED on this at the same time ever. Do I really have to design as if I would do that? I am comfortable saying that there will be 20% LED utilization at the maximum.
The PNP transistors are current sources, they just clamp the anodes to 5V. Splitting into several transistors only increases the part count, the need for current gain is fundamental - if the 74138 outputs 5mA, this has to be multiplied by 100 to be able to drive the power transistor(s).
However a 74138 can drive a MOSFET (though only slowly - several uA perhaps - if no current amplification).
What multiplexing clock rate were you considering? If its slow enough the naive MOSFET circuit will be the simplest. The rather soft switching involved will make decoupling less critical too (but dissipate more heat in the MOSFETs compared to fast switching).
MarkT:
The PNP transistors are current sources, they just clamp the anodes to 5V. Splitting into several transistors only increases the part count, the need for current gain is fundamental - if the 74138 outputs 5mA, this has to be multiplied by 100 to be able to drive the power transistor(s).
However a 74138 can drive a MOSFET (though only slowly - several uA perhaps - if no current amplification).
What multiplexing clock rate were you considering? If its slow enough the naive MOSFET circuit will be the simplest. The rather soft switching involved will make decoupling less critical too (but dissipate more heat in the MOSFETs compared to fast switching).
I am such an idiot on these things I was basically going to work with the clock in software until it "looks solid/good" like I have done with my other LED multiplexing projects. I have done several of these, but all with no more than four to six single color matrices and one with one RGB matrix. I was planning on something like 800Hz so each row gets lit 100 times per second. Some simple math that I have done suggests I can keep the TLCs updated at this rate successfully -they, of course, need an update prior to switching rows.
What do you think of the TIP125? Supposedly the gain is 1000:1 because it is a Darlington. This web page, fourth schematic, has an NPN transistor driving a TIP125. The gain here should be tremendous:
The only time a Mosfet dissipates unwanted power is in the small time that the device is operating in the 'ohmic' range that is the transition time or rise/fall time which shouldn't be worse than several 10's of uS, relative to the total on time. It is a calculatable time to be added. The "time under the curve" is a difficult thing to accurately calculate so assume the worst and assume 5 to 10% of the total switching time (on and off transitions) and you have a 'first cut' at calculating the worst case power dissipation value. there are of course many other considerations but this will get you into the "ballpark" IMO