I’ve worked through to the end of a project (lucentcube.com) and the non-PWM panels using chipset FM6124 and these work really well, albeit with a slower refresh rate. Given the success of the changes from @kingdo9 that @marcmerlin has managed to leverage for his LED bodysuit project I’m hoping to experiment with PWM based panels. Specifically my issue is partly to do with the current refresh rates effect on the ability to film the panels, but generally I just want to see how the PWM based panels look!
Marc has kindly shared his supplier in China that can provide a variety of panels (my target being P2.5 64x64), his offerings come in FM6124 (non-PWM) and FM6373 (PWM) flavours.
I just wanted to confirm this assumption is correct, and obtaining some FM6373 panels will let me start investigating with these PWM panels, using kingdo9’s updates to support these, and attempt to resolve any issues along the way?
PWM panels kind of work thanks @kingdo9 but his code is not merged, there is a lot of black magic going on (like dozens of potential init sequences per chip that can be sent and should be tried one by one), and I’m pretty sure it also takes more power on the Pi3 to generate this than PWM panels (only matters if you’re running on batteries).
In theory, PWM panels are supposed to refresh super fast on their own and give better pictures, but in my limited testing, it’s not a given with the current code quite yet (but I’m doing 128x192, you’re doing 64x64 which is 6 times smaller and faster) and I have no idea how quickly it refreshes the panels themselves
I opened this
I went to PWM because Azerone did not have any P2 128x64 panels anymore, unless they were PWM, so I had no choice.
Honestly, for simplicity and stability, I would pick the old style panels (FM6124) if available.
And this is with infinite appreciation for @kingdo9’s work who has literally saved my bacon and is allowing my outfit to work, but this code is not on the same production quality than the old code (which makes sense of course, it’s new and on top of that, every PWM chip seems to be different and incompatible, which is maddening).
Thanks Marc for the insight, I really appreciate it. I think I’ll give it a go, I’ve asked our mutual friend in China for details on the PWM panels, I’ll grab some just for fun, even if I only assemble a single version with these panels I reckon it will solve some refresh rate versus camera capture rate woes I’d been having. Will keep this thread updated.
Yeah, for 64x64, you will probably get 400-500Hz at most for a 64x64 in row/scan mode. I just have no idea what actual speed I get from my PWM panels right now
The FM6373 driver belongs to the most thoroughly studied group of sPWM panel chips to date, which also includes the ICN2055, ICN1065, SM16380, and the FM6373 itself.
All of these drivers share the same data loading protocol and differ only in their configuration register sets. All of them are supported by the @kingdo9 project.
As for refresh rate, it is currently limited by code efficiency. Given that kingdo9’s code is primarily an AI product, it is quite difficult for human programmers to optimize. It is likely that there are some suboptimal parts in this code. Furthermore, AI generated it without consideration for its subsequent integration with the main HZeller library, this also complicates further development.
Thanks @board707, it is a problem we’ll see more and more, the ability to be able to resolve such issues through AI is going to generate a ton of source which most repo managers aren’t going to want being included for obvious reasons. I’ve bitten that bullet and am resigned to accepting the code generated by AI is most likely never going to see a public repo, but everything is changing so rapidly, who knows what the future brings!
The supplier also said he didn’t have FM6373 panels in stock, but did have SM16380 which is supported under SM16380SH in @kingdo9’s library, so I presume there are issues with this variety as opposed to the more standard FM6373 variety, but I’m thinking I may be able to work through some timing issues unless people have hit a wall with these!
I’ve pulled the pin on getting some P2.5 64×64 SM16380SH/TC7262FJ based 1/32 panels which I’ll test with, hopefully the work @kingdo9 has done will give me a huge head-start! Thanks all.
I just wanted to come back with my experiences with PWM panels, the support given and work done by @kingdo9@board707 and @marcmerlin, this has been transformational. It truly is night and day as far as refresh rates are concerned, the panels are almost TFT grade now, with kindo9’s optimised kernel and the PWM panel support, before I couldn’t even look at the panels close range (and one of the driver so wrapping them in chemcast diffuser panels), now they’re smooth and rock-solid.
There’s a couple of small, unnoticeable (unless you’re looking) blips\glitches but these are related to CPU isolation (beyond all the documented tweaks in the repo), video footage of results below.
Just wanted to say a massive thank you to you all, it’s terrifying getting a load of panels then the second shoe drops and you start questioning whether you’ll even be able to use them!
(Oh and @marcmerlin Paul came through with those SM16380SH panels and delivery was under a week!)
looks sweet, glad Paul got you the right panels and that they work well enough for you.
As you said, PWM is supposed to refresh faster, but compared to a single 128x64 panel I’m not sure if mine are self refreshing faster (I was getting around 400Hz before on the old panels). See this bug
Until we get some feedback on the command line on the panel self refresh speed, I’m just guessing
Thanks kingdo9, seriously, I couldn’t get to this level without your hard work, everyone has been so helpful, this project wouldn’t have been anywhere near the quality without everyone’s help! It’s running a CM4 (Pi 4) which handles everything cleanly. The only thing I was unable to get 100% right was Amiga demo emulation, but this was solid on a CM5 which works on the same carrier board, I’ve added support for stock CM fans as the CM5 cannot run without a heatsink and fan (instantly gets hot, then throttles).
@marcmerlin asked for the hard numbers and I’ll try to paste them in a bit, but this video shows the night and day effect as far as I’m concerned, this aurora audio visualisation (one of two hundred available ) was impossible to film - now I can enjoy Donna Summer without the anxiety.
Check this (left set is the PWM panels, middle of that set has the chemcast diffuser over it, the right set is the non-PWM panels): https://youtu.be/2LX1c-k6VQc
Thanks kingdo9, yeah despite the changes I still have an occasional flicker but have been working on other things and hadn’t focused specifically on that, but you’ll love these panels I’m sure. I’ve bought another 12 as I need a couple more prototypes, that Paul took a week to respond (probably chinese holiday season) so got a little worried, but he got back to me and all is well!
What do you do for a living? Just being nosey - I’m in Infrastructure, working for a company called Fruugo, so this is a busy time of year for us, peak period!
) was impossible to film - now I can enjoy Donna Summer without the anxiety.