Hey all,

Just need someone to reinterpret some details for me from this datasheet for a 24 channel constant current led driver, DM163.

Basically the device lets you shift in data which is used to output PWM to 24 LEDs. I want to use it (or many of them) with an FPGA to create a small LED display.

Here is my major issue:

If I wanted to use just 8 bits of data for each LED and didn't care about the dot correction, can I shift in 6 ones into the dot correction register at the start of the program and then forget about that register for the rest of the program? Or is it a case of reloading the dot correction every time I put any new data in?

I just don't want to waste half my clocks shifting in ones every time for something I don't really need/want, and I know that my display size is dependent on how long it takes to shift a frame into the drivers.


Also, can I do it the other way around and make the 8 bit register all ones and use the 6 bit register for image data?

Hopefully one of you bright sparks will sort this out for me.

Cheers, Josef

 

If it's anything like the tlc59401, you just set the dot correction once upon initialization. After that, you just send the PWM data. And yes the data sheet says that either bank (dot correction or grayscale) could be used for PWM data. So not having used this chip personally, it seems your understanding is correct.

 

I just figured out that I can't even find anywhere to actually source the part anyway.
I originally wanted to use one of the TI chips. They have built in EEPROM that automatically reloads the dot correction data.

The problem with the TI chips is they are all 0.65mm pitch (or less for the QFP) which makes it difficult/impossible to solder by hand.

The TLC5940 comes in a DIP package but I was hoping to avoid something so big (and through hole), since they will take up so much (expensive) board space.

Does anyone know of any other manufacturers that make a similar chip? Preferably in a standard SOIC or SOP package.

Maybe I'm being too picky

 

I have learned to solder the tssop (0.65mm pitch) chips out of necessity, to play with a couple of the led drivers. A dap of the flux pen, pre-load a small bit of solder, drag with a smallish screwdriver soldering tip, cleanup bridges with solder braid if necessary.

I see the Macroblock MBI5040 has a SOP24 package (0.1mm pitch) if you can source that from where your at, supposedly Tenrod in Australia.

 

Thanks alot for that, I think the Macroblock chip should work perfectly for me.

The other reason I was avoiding the TSSOP was the thermal pad, how would you go about soldering that? Or can you not solder it and use some heaksink compound?

 

I've been avoiding/ignoring the thermal pad issue with prototyping. It hasn't been a problem with a 20ma current limit. Even had to put a piece of mica under the chip, to isolate the pad from the dip adapter pin traces, to keep from shorting out.

I would think a toaster oven, or skillet, with solder paste methods would be required to solder that thermal pad. No experience to pass on there, it looks to be a hassle.

 

Thermal pads are best soldered by applying the main heat to the board from the bottom, not always the easiest to do. A hotplate can do as can a calibrated toaster over (just make sure the temp is correct and stable and shove the board in).

I've only done thermal pads on tiny leadless 8 pin ceramic devices where it
is easy to get the pad to flow.