Hi there!

great forum.

Im loving the peggy 2.0 board. nice work guys.

Im wondering whats the limit of the current each led can use using the peggy 2.0 board?
Can it use 1w leds or do i need to make any changes to the board.

best regards

niklas

Peggy 2.0 is designed to drive ultrabright low-power LEDs, not high-power LEDs. 1 W LEDs will work, however they will not be driven anywhere near their maximum power. In fact, the only real change if you use them is that it gets much more expensive, not any brighter.

The on-board driver hardware is configured, as the kit ships, to supply each LED with up to 20 mA, peak. If you change the current-limit resistors, that goes up to 60 mA. If you swap out the transistors as well, you can probably supply up to 90 mA to each LED. (That's ~300 mW for white LEDs.) However, that's the limit of the on-board hardware. Supplying more current to that many LEDs takes different hardware.

If we were to redesign the Peggy boards to drive high-power LEDs, the cost would probably go up by a factor of 10 or more; it would be a very different design. Also, the heat sinking requirements could be substantial.

Windell, couldn't one drive slightly higher wattage LEDs (I'm thinking superbright, 30mA @ 2.2V, non-diffused so we can use a larger physical diffuser on more widely space pixels) by using higher wattage PNP transistors (Fairchild MJD45H11?) on the rows and using the 2STX2220s (flipped) on the columns? This way you wouldn't be splitting the max 90mA out of the LED drivers between up to 16 LEDs each for the columns, and you could support more than the 900mW that the 2STX2220s are capable of on the rows.

I tried using the 2STX2220s on a couple columns and it was a lot brighter when multiple columns were lit than when simply reducing the value of external resistor on the STP16DP05. But it was a fairly limited test, so maybe there's something I'm missing there.

Maybe that's crazy, and I assume there's a reason you didn't do it that way in the design, like expense, availability of fast rise/fall but high-wattage transistors, PCB layers, and tileability of the design, but I may give it a shot so I can milk a few more precious millicandela out of these LEDs.

I'm thinking superbright, 30mA @ 2.2V

The reason that we have it set, by default, for 20 mA per channel is that not all LEDs are rated for higher. You can of course turn it up to use 30 mA instead (or all the way up to 60 mA), without additional hardware changes besides the resistors-- *if* your LEDs will handle it.

This way you wouldn't be splitting the max 90mA out of the LED drivers between up to 16 LEDs each for the columns, and you could support more than the 900mW that the 2STX2220s are capable of on the rows.

Not sure where you're getting this idea. The LED driver chips handle about 90 mA per channel-- almost 1.6 A max, for all 16 channels-- we *are not* splitting 20 mA 16 ways.

The standard transistors handle 1.5 A total. If you need a bit more current, the right transistor to use is the STX790A-- it's pin compatible and rated for 3 A. In looking at alternative transistors, you need to consider not only the power rating, but also the collector-emitter saturation voltage. The ones that we use have a suitably low value at the currents that we use.

If you want to dump the LED driver chips all together, you will need to add load resistors in series with each column, and make sure that the maximum current limits that can be achieved are safe. That part of the problem is normally handled by the LED driver chips, which do regulate the current, keeping it safe for LEDs of different colors.

Now here's a real trick if you want better brightness: Notice that the display is multiplexed, so that each LED is actually on only 1/25 of the time. We *normally* drive the LEDs at 20 mA (or 30, in your case) because the Peggy becomes "uncontrolled" when the microcontroller is being reprogrammed, and tends to drive one or several rows steady. If, however, you eliminate that concern by reprogramming off-board, *and* you are very careful with your programming, you could wire it up to drive your LEDs at the full 90+ mA, at the times when they are driven. Nearly all LEDs come with a factory spec on "pulsed mode" operation, the ones that we have are all allowed to run at 100 mA, at 10% or lower duty cycle, 0.1 ms max pulse.... which means that they would be happy running at roughly four times the current, so long as it's in the multiplexed mode.

Not sure where you're getting this idea. The LED driver chips handle about 90 mA per channel-- almost 1.6 A max, for all 16 channels-- we *are not* splitting 20 mA 16 ways.

You're right. Somehow I skimmed right over the "Max Ground Sink Current: 1600mA" in the spec sheet of the driver. Derrrr.

I don't need any more than 1.5A. I'll only be using 750mA at the most in my upgrade. I was overly focused on the wattage rating of the transistors, which I realize now doesn't matter that much in switching applications.

I'll retain the driver chips.

If, however, you eliminate that concern by reprogramming off-board, *and* you are very careful with your programming, you could wire it up to drive your LEDs at the full 90+ mA, at the times when they are driven.

Huh... I've built your circuit from the schematic on a breadboard and wired it to an 8x8 R/G matrix and have seen the behavior you describe, but isn't it lighting up those lines because the first multiplexor shares pins with the TX/RX pins? Instead of programming off-board, couldn't I also just put two physical DPDT switches that turn on and off the current to the supply pin of the multiplexors and drivers?

Wouldn't upping the duty cycle to closer to 10% result in slower refreshes to the board?

> but isn't it lighting up those lines because the first multiplexor shares pins with the TX/RX pins?

The TX/RX pins are available for reprogramming Arduino-style through the bootloader, but they are used as regular I/O pins as soon as programming ends. If you need to stream data to the board while it's running, you should use the I2C (TWI) interface instead.

>couldn't I also just put two physical DPDT switches that turn on and off the current [...]

Yes and no. That will probably work, BUT, chips usually specify that none of the inputs should be exposed to voltage when the power supply is not applied. They may not like it.

>Wouldn't upping the duty cycle to closer to 10% result in slower refreshes to the board?

I have not suggested messing with the duty cycle.

Thanks for the tip re: off-board programming.

>I have not suggested messing with the duty cycle.

Ok, so what you meant by "being careful with your programming" was to multiplex properly and not ever tell a whole row to be on over 1/25 of the time, for example?

Ok, so what you meant by "being careful with your programming" was to multiplex properly and not ever tell a whole row to be on over 1/25 of the time, for example?

Exactly.

Hrmm.... Isn't this basically the key to doing 50x50 with one microcontroller? Obviously not with the 168 unless you want to give up all inputs.

I guess there would be dimming associated with a 1/50 cycle too. But if you interleaved two sets of 4 (or 3 if 45x45 is sufficiently large) LED drivers you could have two rows on and driven at a time.

> But if you interleaved two sets of 4 (or 3 if 45x45 is sufficiently large) LED drivers

This also comes at the expense of reduced refresh rate.

I guess if we were going to lower the reference resistor to increase brightness then I can't think of any reason we couldn't bump up the resolution to 30x30 without any additional components, except 5 transistors and resistors. That seems like a good compromise between 25x25 and 50x50. I'll give that a shot.