@Stuart_Taylor exactly what I mean. He should inject for best results, each meter 5V and GND.
@Mike_Thornbury I’ve now had 3 of ~650 LEDs on these strips fail (two of them the red LED was very dim, 1 of them the red LED is always on). I’m wondering if it’s a low-quality strip. I’m just waiting for more of them to malfunction. Looking online now for a better source (also considering buying the APA102 non-c model – 2x more expensive).
@Jurgen_Skrotzky I’ll pick up a level shifter and see what that does. And yes I’ll try injecting power/ground at both ends.
Can anyone recommend a known good source of these strips? On Aliexpress I’m not sure what I’m looking for (ie, “manufacturer” is all over the board).
You can buy direct from GreeLed, but I have no idea what their manufacturing is like. I have only bought WS series LEDs and I got those from WorldSemi direct.
AP sell strips direct. Might be more expensive than China, though.
I have found this: http://www.aliexpress.com/item/5v-24-36-60-96-pixel-5m-apa102-addressable-led-smart-strip-taiwan-chip-non-waterproof/32285255288.html
which seems right. I’m not sure if any manufacturers are cloning the square-window LEDs or not. In fact I’m wondering now if the APA102c (round window) can possibly be anything but a clone. Still confused. I’m also not sure if other manufacturers clone the APA control chip, or if they use an original and just do PCB/LEDs themselves.
Do you have another channel on the scope, I think it would be interesting to look at +5v on the LED that you are looking at CLK OUT. Connect the scope ground to the same LED.
@Tom_Davidson I do – I have 4 channels. That’s a good idea. I’ll add this to the list for the next time I get back to the lab. I couldn’t make it today. I also have to source a good logic-level converter before I go back.
Having very similar problems with some apa102 leds my self, have some custom pcb’s, 100 apa102’s per panel and the third panel is getting glitchy.
I can get all three to operate correctly in a straight chain if I lower the clock speed to 2 mhz in arduino ide. If I go up past 4-5, the second panel even starts glitching out. Seems as I up the speed, the number of leds that reliably work get shorter and shorter.
Is the consensus that the clock does not get regenerated but the data does?
Is it perhaps bad clones only or all apa102’s by design?
In the block diagram on page two here:
http://lednexus.de/ebay/apa102/APA102.pdf
it shows the data line going through an extra block that the clock does not.
Perhaps this is indication of the clock not getting re generated?
I wonder if a high quality logic level shifter would regenerate the clock signal accurately? If so one could theoretically include it every N leds so you always have a clean signal?
This seems to be the only drawback to the APA102’s. Fix this, and you could run as many LEDs at a high speed as you want :).
I’m pretty sure the clock signal is regenerated on each APA102. I’d look at your GND and power at the far end of the panel, how clean is the GND and +5V ? You might need an isolated o-scope to see if the GND is moving. If you do put a buffer on the CLK line after 100+ LEDs, remember to put the same buffer on the DAT line, this will keep the two signals in sync since any buffer introduces some delay to the signal being amplified. ie http://www.onsemi.com/PowerSolutions/product.do?id=NLU2G17
Thanks for the ideas!
Another thing I’m unsure about:
I’m connecting the data/clock between panels using a shielded 2 conductor cable, where wire 1 is data, and wire 2 is clock.
Is it possible that this configuration is causing some sort of cross talk?
Would it be more appropriate to have a cable for data, and a cable for clock, and have ground coming back down the second wire in each?
I recently bought some ‘loose’ apa102c LEDs. My supplier asked did I want Taiwan or Mainland LEDs. The price was about 30% higher from Taiwan.
I suspect, then, that there are many clones of the original and if quality is important, go Taiwanese.
I think you’ll be OK for a 20MHz signal, maybe put some series terminator resistors in, (70-100 ohms) at the source of the signals. Ideally, use twisted pair for each signal, use one wire for the signal, and the other as a ground return for that signal. I’ve used Cat5 ethernet cable and connectors on projects. It wont carry the power though, you’ll need much heavier wire, my infinity mirror pulls 6Amps@5V on all white with about 5 foot of 144LED/Meter strip.
+Tom Davidson
I did what you said and added the 100 ohm resistor as well as rebuilt my patch cables (going between apa102 panels) to use a twisted pair for clock, as well as data, with the ground tied into the main ground…
Night and day difference!
With out that config, I was getting flickery behavior at 3 panels at around 3 mhz
I’m able to go past 10 mhz, as high as 15-16 with all three tested so far working.
Try running separate power and ground to both input and output ends of the panels if they are organized with premade strips. Then run all the 5V and GND leads to a power bus made with copper ribbon or welding cable. A big capacitor, 100uF at each panel’s power and ground will help clean up the power rails. The goal is to have a low resistance and inductance path to +5V and GND at each panel, so your signal GNDs wont be carrying any LED current.
Looks good, Tom. At some point I’ll get back to the lab and put this together.
I’m guessing that even if the clock signal isn’t regenerated, there’s a lot I can do to keep it cleaner for longer.
I’ll post back here as soon as I’ve done new experiments. I just started a new job so it might be a while.