SpenceKonde
commented
1 year ago Interesting. I think there may be something wrong with the 24 MHz implementation (I have seen oddities there, but thought it was fixed)- it seems to be unique to 24 MHz. The short term workaround, thus, is to use either 20 MHz or 28 MHz (or 32 MHz), all of which are achievable with the internal oscillator.
Which kind of LEDs exactly are you using? (if they're the "fairy light" style ones, a picture of which one would be helpful - as there are four different kinds and at least 2 of them and likely 3 of them use different types of LED. The part number used in listings for assembled (and less often, unassembled component) is frequently not what you're getting. There are a very large number of "compatible" WS2812 clones (and they generally are compatible), even though the only names you'll ever see mentioned in a listing are WS2812 and SK6812, there are at least half a dozen, not counting the low current versions that some of them exist in. I have samples (and in some cases far more than "samples").
I have a huge lighting project under way (though unfortunately it hasn't gotten to the point where I can light up any LEDs with the new controller boards. All the original controller boards wereburned out due to what turned out to be a combination of a flawed wiring topology (I do power injection by running 19v through thicker wire wrapped around the wires in the christmas light string type ones, and at the head of each string, a mini560 buck converter (like $1 on aliexpress) gives me the 5v to power that string. All the 5v rails are tied together (in case of a failure of one buck converter) and to reduce the difficulty of setup, I'd clipped off the 3 pin JST SM connectors and installed 5 pin ones. Unfortunately I hadn't considered 2 factors: 1. JST-SM connectors are utter shit connectors. They are poorly designed, and the fact that they're all chinese knockoffs doesn't help either (JST tried to save a bit of money by reusing one of the terminals from another connector). 2. When a buck converter is not connected to ground, the voltage on the positive input and output terminals is approximately equal. And the 5v from the head end was used to power the controller. So certain patterns of briefly interrupted connections due tothe shit connectors would putr 19v on the 5v rail, The LEDs would survive this, at least for a very short time. The ATmega328PB's I was using were not so fortunate. And I didn't put 2 and 2 together until I'd burned out my entire stock of the older designs, and was then working with an AVR DA-based on Aaah the lights, worked, the party cheered, and then as I stepped off the chair to put the controller down, suddenly the animation stopped, and I felt teh chip getting hot underneath my finger. At that point the penny dropped - It could only be caused by an intermittant connection, and it quickly became obvious from my schematics of the wiring which one it was. JST-SM was declared unfit for all purposes (still haven't decided what to do with the HUGE numberof JST-SM connectors I have - I got every size housing, buncha terminals. plus a bunch of prewired connectors. No way in hell am I using that garbage again. And so of the remaining working light strings (somewhere between 2 and 4 black wire ones failed, as did every single white wired one (the heat from the previous buck converter that I used generated enough heat that, in combination with the pressure from the ziptie and the squidgy low quality insulation on the cheap chinese wire, led to strands of copper working their way through the wire and shorting to another. (those converters were terribad - for the same pattern, they used a third more current than the mini560s (and cost 4x as much too), So running 255,255,255, I couldn;t run 4 strings from one 3.5A laptop power brick, I had to limit total brightness in software, which is no longer a problem). That failure mode was in evidence on all 100% of white wire strings (with shitty white zipcord) but occurred in 0% of the black ones, which used (apparently) better black zipcord. And so all the connectors were once more chopped off and replaced with MX3.0 conectors. The two ground pins were connected more rigorously as well. I still had failures from some sort of power excursion though. I moved to an AVR128DB48E instead (E's overclock better, and the faster I can generate the pattern, the more LEDs I can have for a given target frame rate. I'm hoping I can hit my targets for string length and frame rate in all animation modes at 32 MHz, but if not, there's the option of 40 or even 48 MHz (though I suspect not all parts, even E-spec ones, are stable at 48; I think at room temperature they'll all do 40 stably, (most I-spec ones will too, and virtually all I-spec parts can do 32). But, the firmware is in the middle of a giant overhaul.
cybermatrix84
after doing some research on local web shop there is different in series which is sk6812mini and ws2812mini, i will try to buy it and see if there is any difference on 24Mhz, also i will try on higher clock.
Hello thankyou so much for such a great core
i had flickering issue with sk6812mini which is no problem with common ws2812, maybe its timing issue?
im using 24Mhz AVR128DB32 hope there is any solution, thankyou so much