Looks amazing so far. Have you ever considered trying to put it in a loose / baggy sweater so it can conceal the electronics more?
I can’t really be wearing a sweater to a rave! I’ll burn up!
Everything went well last night - no crashes or glitches, and everyone loved it.
The QuickMedia buttons were an absolute blast - being able to say things to people from across the dancefloor resulted in some pretty funny moments.
I had to tone down the brightness of the animations down to like 25% as it was too bright otherwise and was dazzling people. As a result, the battery pack ended up at 3.9v / cell (For ~4.5 hrs of run-time) - I am starting to think that I don’t even need 4 batteries, 2 will do. In which case I would not need a custom battery pack, I could just use my Xtar PB2S, which can do 12v/1.5A and is way smaller and more compact. I would need a PD spoofer, but those are tiny. I could maybe take 2 spare 21700 cells with me, as it’s quick and easy to change the batteries in that thing.
1 Like
Why are you not using USB powerbanks with a 20V USB-PD port and then a step down converter like I’m doing?
I’m going to tell you right now that unless you remove all the batteries from your pack for every flight, you will be detained by TSA when they see batteries inside a device they can’t see inside and isn’t a normal looking powerbank.
Just wanted to mention that I’m very impressed by your design work and how much more polished yours looks compared to mine. Well done!
I never plan on flying with it. All the events I go to are within an hour or so of me.
Removing the batteries from the pack would be no problem anyway - they are all in cell holders, so it’s just a case of removing them like you would pop a AA out of a remote.
Thanks 
Oh, and by the way, if you want a copy of my frame, I could probably tweak it some to support 3 panels. You would probably have to print it in 3 parts though - I can only just fit a double height one on my 300mm high print volume printer.
I have been going at it pretty heavily on the UI front and have some nice new features.
I added a “Relative Brightness” value for each stream. You can set this so that when you reconvert a file, it will adjust the brightness by the specified percentage. So if you have one animation that is much brighter than the others, you can dial just that one down on any re-convert.
It also stores what value was last used for each file, so you know what it’s currently at
It also stores the matrix options that were used for the last convert
I then added “ReConvert all” functionality. You can reconvert all with a base Matrix Options and their stored relative brightness, or reconvert all with their stored matrix options, but supply a new relative brightness
And finally, I added an option to duplicate a playlist.
I have an event coming up soon that starts in the day but then goes on into the night, so now I should be able to quickly tweak the brightness for a playlist, or even just duplicate a playlist and reconvert the whole duplicated playlist with just a couple of clicks.
Thanks for the update. I put some screenshots of my UI here
I have 300-ish patterns with a crude UI to edit the playlist on which order to play them in, or turn them off for a pre-selected pattern list, and I have 2 levels of bestof (all, bestof1, bestof2). The updated UI editor now allows me to make an “on the fly” playlist for a special occasion (like one party, i could only display patterns with red)
Now, relative brightness for each pattern, is not actually a terrible idea it would help for a few of my animated gifs that are bright, but I can also use an editor and fix the gif to darken it with better software with gamma curve than just dimming all the pixels.
For your other points, I’ve never had the need to edit the matrix config on the fly, ever. Brightness is a runtime parameter and is the only one I need.
I do have dimming too, I typically work at 50% brightness (never below except for taking pictures with a bright screen that is overwhelming the camera CCD).
I do use 100% for 1 or 2H before sunset if I’m outdoor and it’s not quite fully dark yet. If there is more daylight than that, my outfit is not really usable in full daylight.
I very much appreciate your offer of a case. I’m indeed not well equipped to make my own and my current solution is minimalistic and lightweight. It would work perfectly if it weren’t for the fact that the panels flexing, causes them to fail and I have to replace them :-/
My current plan/hope is for @board707 and @hzeller to be able to merge in PWM panel support, which would allow me to switch to a newer generation of panels with better soldering and coating that should protect them.
I would of course love a light, custom made curved case, but am also a bit self conscious to have someone else go to the trial and error work to make one for me, even if I’d pay for materia/time/shipping
As for on the fly things, I’ve always wanted to make displays on the fly, but the work involved in making a face is involved on my laptop, never mind from my phone:
However, I do have a basic UI where I can write text, and it will calculate the max font size that will fit for that text
What “on the fly” pattern design do you have?
What, you have real-time dimming? How is that done?
All my stuff is played via the led-image-viewer, so I guess you have a custom implementation that can tweak frames in real-time?
Yeah I noticed while browsing aliexpress the other day that there are some panels which had the LEDs encased in something, but they were all PWM.
If / when we get PWM support, then maybe we could both buy the same make / model of the improved panels, and then I could supply you with CAD files for mounts that have been tweaked specifically for them. I also think it would probably be a bunch more cost-effective for you to have someone print them for you locally.
I would be very interested in a routine like that. Yeah, there is some stuff in the API for dynamic text, but it does not take that into account - I think there’s a scrolling text one too, but that does not really appeal.
Nothing built into the software itself. If I want to do something in the field, I can do it on my phone - I find that ezgif.com is pretty good at letting you asemble a GIF on the fly.
Would you maybe be interested in sharing? I am always on the lookout for new stuff. The last few sessions I have spent trawling the internet though hasn’t really yielded anything I liked. I am kinda picky. I saw a few in things in one of yours videos that I liked though. I am generally into the geometric and psychedelic stuff - the hexagon pattern at 3:38 is a good example of the kind of stuff I like
Dimming is supported on the matrix object at runtime
As per the link I just gave you, I wrote my own multi platform multi hardware framebuffer, write in it and framebuffer gets pushed to the rpi matrix for display.
Yes, the new matrices are all PWM :-/ eagerly waiting for support. Ideally when those come out, I can simply use my existing frame made with 2 paint stirring wooden sticks
Good point about ezgif, I could indeed do that and push it. Haven’t bothered so far (but most of the ones I make are high resolution high fidelity pictures of people, so not sure how well ezgif would do that over gimp which I currently use on linux)
Sharing, my stuff is already all online
and
I did see that, but it seemed to have no effect when I passed that matrix object to the content streamer
Thanks for the shares. I’ll put mine somewhere for you too
probably something is getting lost in your code somewhere, I can definitely confirm it works in C++, I don’t do the dimming in my code or matrix, I send the full FB to the driver, and tell it to change the brightness to x, and it does.
It works just fine with all the C# examples that directly draw to the screen, but it does not work when you pass the matrix instance to the content streamer API.
Are you using the content streamer API? If memory serves, there was no content streamer API (Even in C++) until I implemented it - It was just a demo app - in which case I suspect you are not.
Even with the content streamer demo app, if you prerender a GIF to a stream file with the -O option with one brightness, and then do a playback specifying another brightness, it has zero effect also - so I was hardly surprised when the Brightness option had no effect via the AP.
I’ll maybe revisit it with the AI tho - I guess that it may technically feasible to do, it’s probably just that the content streamer takes some different route (Calling something further down the pipeline, because the whole point of pre-rendering to a stream is that it writes everything to the file in the perfect shape to just be piped straight out to the matrix, meaning that zero pre-processing needs to be done, saving CPU).
I guess you are passing it a frame buffer (pixels), whereas a stream contains packets as they would be sent to the matrix.
Yes, I am writing to the framebuffer myself, pushing frame by frame and I do all the rest, including animations on my side.
Remember that in my case, my code works on 10-ish different hardware outputs and platforms from esp8266, esp32, teensy, rpi, and linux on laptop with display on LCD
A bunch of GIFs for ya: Gifs
thanks. I’ll see which ones can be nicely resized to 128x192, the resolution I use
These are really cool, I love them, excellent work!
My project uses 6x 64x64 panels and I’m using XT30’s, and the power over 18AWG works fine for full brightness (it’s rare that anything is nearly full white but still holds up, even with the fussy power requirements of the CM4). Only reason I mention that as (and I know this is contrary to what we are told in terms of power) if this might make some of the cabling and connectors smaller profile (XT60’s were too clunky for me and XT30’s were just the right amount of small).
I’d definitely consider consolidating some of the electronics on-board, the switch you could use a cheap mosfet so no batteries need disconnecting and the electronic switch could be triggered with so thin touch sensitive component, you could even layer a series of them and use the fabric of the t-shirt to swipe up and down for brightness.
I did suffer from the brightness control issue, born out of performance as I typically run from pre-rendered files that load straight into image-viewer, but needed dynamic brightness control and cross-fading of animations so needed to store them in a lightweight RGB format in the file which a fork of the image-viewer then reads and (using a shm in Linux) can accept dynamic run-time parameters. Given performance was the main issue I was worried about this, but the Pi4+ handles it flawlessly with seamless transitions.
Have you had a play around with the flexible LED panels, I know typically these are lower pitch\resolution, but might give you some additional options in terms of weaving them into the fabric?
Also one this I would highly recommend if you’re thinking of going down the custom circuit board route (and this would consolidate a CM4, driver, switching, buck convertor, etc) you could throw a simple audio chip (the CM108B is cheap and super easy to incorporate) then make your animations react to the sound.
My project is almost exclusively based upon audio reactivity, and the frequency of which you can analyse sound with FFT and drive the panels on a Pi4 (and above) is flawless.
Clive, sorry, I know I’m replying to multiple threads, but just to let you know I’m expecting the CM4 carrier end of week and will then drop a link etc, happy to help if you need anything.
Ross
The panels have XT30, the main control box accepts power via XT60 connectors.
The reason I went with XT60 for vIn was three-fold
- Panel-mount XT60s are readily available
- My hobby chargers that I have for my drones use XT60 (And JST-XH for the balance leads)
- XT60 for vbat (4S1P), XT30 for 5v - less chance of accidentally frying something
Interesting about your dynamic brightness. I use the led-image-viewer API and so need to re-convert if I want to change brightness. I may be interested in what you have. I wonder what effect it has on power consumption though. I currently get 10 hours out of a 4S1P 6500mAh Li-ion pack, which is about on the limit as to what is acceptable. I could go up to 6S1P or 3S2P I guess, but I would rather avoid the extra bulk.
The XT30s on the panels are primarily there because the stock power plugs stick out too much and raise the whole thing too far off my chest. I don’t disconnect them between uses.
WRT the buttons, I prefer them on the control box on my belt - I can very subtly press them and people don’t even realize I triggered that animation. Had a hilarious moment at the last rave I went to - a guy comes up to me and says “You look like some kind of Teletubby”, so I press a button on my belt and my teletubby animation plays:
Blew his mind Bro thought I was controlling it with my mind or something.
These are flexible. v1 was curved at a 250mm radius, and I am in the process of printing off a v2 with a radius of 200mm (I want it to blend even more into my body). The manufacturer says that the tightest recommended bend radius is ~300mm, so I am exceeding spec by quite a bit.
I think it’s worth the risk tho - one of the things I love about this version is how it’s so hard to tell that it’s not just fabric that happens to be a screen. People are consistently surprised when they touch it and realize that there’s actually a normal-ish panel there - they think it’s some futuristic tech that they’ve never heard of.
My first t-shirt was semi flexible and ultra-thin, but stuck on the outside of the t-shirt and so was obvious. My 2nd t-shirt was rigid and again on the outside - very obvious as it really stuck out at the corners - people thought it was an iPad or something stuck to my chest.
The other reason I want it to be really subtle is that I have to wear the bloody thing on the train on the way to raves, and especially with the hot weather lately, a coat to cover it up is often not an option. With my 250R one, I ended up just having to wear nothing over the top of it, and got some really funny looks on the train 
The 250R one started intermittently glitching out after 9 raves, suspected driver chip coming loose. That was only at the start of the night though, it was OK for most of the night.
With the 200R v2, I have put silicone conformal coating over all the chips - hopefully this will help stop them from coming loose.
If I can get 10 raves out of each one before they die, then I am happy - I pay ~£70 for a pair of panels, so that would be £7 per rave, which I am cool with - that’s not even the cost of a drink.
Yeah, I thought of that, would be pretty cool. I also thought a Shazam t-shirt would be an amusing idea (Shows name of track being played), but to be fair, Shazam utterly fails to identify the kind of stuff played at the raves I go to 
That’s amazing, very impressed - clearly and meticulously thought out detail, and I have questions! The first one that springs to mind is the kind of music they’re playing at these raves? I would imagine some kind of EDM? Like most of my generation we grew up driving to fields listening to early old skool hardcare and jungle, although unlike most I still listen to it to this day - and my project (nearly there!) reflects that, and the love of the Amiga and those demos!
Wearable tech is cool, and there’s nothing quite like building it and pioneering it yourself, that is amazing. For me the XT30’s suited a purpose, but definitely it’s horses for courses, and especially given the differing voltages that’s a wise move!
I did think those flexible LED matrices, which aren’t really matrices (well, they ARE, but such low res you could recreate them with strips), would look nowhere near as good as the panels, maybe one day! Being able to switch animations is cool - and that did make me laugh, that guy must be talking about that to this day!
You clearly have some cool ideas, would be good to explore them. For me the biggest leap in recent months has been porting my use of AI to my personal projects where I have more scope, less constraints, and it’s been transformational. Claude Code in Auto mode is bananas, if you trust it, but that’s more about trusting yourself to be explicit in prompts, build memory guardrails etc. Point being the audio reactive, shazam (if it worked for your tunes)etc stuff is done in five minutes, exactly how you want it, it’ll develop it, push it to the Pi, compile it, debug it, then test it (as much as it can). It really is that good. Anyway I don’t want to sound like an AI fanboy so will shut up. 
Excellent project Clive, excited to see what you do next…
Yep, that stuff. Not jungle though, I don’t like that. More the house and hardcore side, '88 to '93 were the golden years.
Hardcore was very UK specific AFAIK, so I am guessing you are in the UK? If you’re anywhere near London, you’re welcome to come with if you like - it’s all very civilized these days, most events are daytime and done by 11PM.
I used AI a bunch for this, I have GitHub copilot and the integration with VSCode is really nice - you give it a prompt, it makes changes, but you see each block of changes as a diff and can accept or revert it out - even run it, see if it works, and if it doesn’t, revert it out - so effectively you have two layers of commit - the AI one sits on top of your normal git commit.
I made a new version of the frame and have replaced the panels (Same type of panel though).
The panels had started glitching out a bit, I got 9 raves out of it, which IMHO is acceptable.
The manufacturer says that maximum recommended bend radius is 305mm, so I am exceeding the spec, so hardly surprising.
v1 was curved at a radius of 250mm, and it stuck out slightly at the corners, v2 is 200 radius
To combat the extra bend, I put silicone conformal coating on all the chips to try and hold them in place. As long as it holds up as long as v1 did, I am happy.
I also added extra straps at the side, to hold it more snug to my body
I am really pleased with the result, it blends in way more
1 Like