APA102 -- how many is too many

[phillipps76]

I would like to run between 14 and 20 1 meter strands of 144 APA102s in spectrum display and am not sure how many strands I can chain together. I will inject power as often as needed, power isn't the issue. I already bought 14 of the strands and prefer to use them as opposed to tossing them out.


There are posts that say 260 LEDs is as long as one sting of APA102s should be due to prorogation delays. And there are posts that say thousands are OK.


Here Paul highlights the problem: https://www.pjrc.com/why-apa102-leds-have-trouble-at-24-mhz/

Maybe that was not known when this was written saying it's not an issue: https://forum.pjrc.com/threads/27975-OctoWS2811-and-APA102-LED-strips



I am using the 74HCT245 for level shifting and a teensy 3.6 with an audio adaptor board and a teensy breakout board.

My plans are based on what I've been reading but hope someone can offer feadback

* It may make my life easier but I've decided to not use the OctoWS2811. I am also not sure how the OctoWS2811 would work with a Audio Adaptor, it looks like Paul is not suggesting doing so:
https://forum.pjrc.com/threads/4797...eactive-LEDs?highlight=OctoWS2811+audio+board

* I plan on using 40 pins for the clock and data over 20 strands to keep the length down to under 260 LEDs per given strand. I plan on testing before going to the trouble of soldering up 5 74HCT245

I'll keep reading and will do my own tests but I think I've scrubbed through every post on how long an APA102 strand can be before there are problems so any advice would be highly welcome.

Thanks!

 

[PaulStoffregen]

There isn't any single easy answer. It's like asking how many times can you use a photocopier to make a copy of a copy of a document. Each successive copy of a copy gets a little worse. Maybe you happen to have a really good copier. Or maybe yours is a little worse than the average. But one thing is for sure: you got your copier from a company in China that makes them as cheaply as possible!

WS2812 are slower but much more predictable, because each copy of the data is actually regenerated by the timing circuit inside each WS2812 chip. It's like copying an email. The data is digital and each 0 and 1 gets regenerated.

But APA102 regenerates the clock in a manner akin to analog copying. Each copy gets slightly more distorted. Some APA102s might be able to have a strip of 1000 running at 8 MHz. Others might only manage 2 MHz. Results might vary with temperature. It's very hard to predict. The specs in the Chinese datasheets aren't comprehensive and what little info they offer isn't trustworthy. It's really a matter of testing with the LEDs you have to see what they can really do.

I don't recommend APA102 for large LED projects, because of this uncertainty. WS2812 uses a slower bitrate, but generally the only problem you need to worry about is poor physical construction / soldering of the LED strips.

 

[PaulStoffregen]

I believe the entire "maker community" has quite a bit of lore built up around APA102 being technically superior. Adafruit is mostly responsible. They have published numerous tutorials that tout the advantages of Dotstar over Neopixel, portraying Dotstar as a slightly more expensive upgrade.

Most of Adafruit's dev boards lack non-blocking WS2812 libraries like you get with Teensy. This notion that Dotstar is better comes mostly from the lack of good software support for Neopixel. The answer to technical problems isn't better library support, it isn't better dev boards with more capable hardware, it's upgrading to APA102.

Really the only big advantage for Dotstar / APA102 is persistence of vision (POV) applications, where the higher internal PWM rate helps.

 

[phillipps76]

Thanks, there is no magic answer I just have to test. I will start with one strand of 144 and work my way up to 2016 LEDs and see where it fails then cut back by half for good measure.

You are right about the lore. It seemed like a better choice and probably is from a design point of view, if only the implementation matched.

Maybe I will replace my LEDs, it's painful. I will post what I learn by testing.

 

[MichaelMeissner]

Really the only big advantage for Dotstar / APA102 is persistence of vision (POV) applications, where the higher internal PWM rate helps.

I suspect another advantage is it can work on Linux single board systems like the Raspberry Pi and Beagle Bone out of the box without having to craft a driver that runs in kernel space to get the precise timing correct for WS2812.

But that being said, WS2812 comes in various shapes like different sized rings and matrixes, while APA102 seems limited to linear strips (and the Sparkfun single size rings).