Hack Chat Transcript, Part 2

A event log for LED Matrix Hack Chat

Let's see how deep this rabbit hole goes.

Dan MaloneyDan Maloney 06/09/2021 at 19:550 Comments

Steve Pomeroy12:23 PM
@Jason Moungey that's a disco stick if I've ever seen one!

Jason Moungey12:23 PM
Either way I am in love with it

Jason Moungey12:23 PM
@Steve Pomeroy :)))))

Garrett Mace12:24 PM
Once you do your transforms, you can step through the physical pixels you have, pull out their coordinates of interest, and feed that into your generation algorithm

Garrett Mace12:24 PM

Steve Pomeroy12:24 PM
I'm loving these explainer gifs!

Garrett Mace12:25 PM
I definitely recommend perusing some old math texts, maybe Khan Academy or really helps! And I do love my Python and matplotlib!

Orlando Hoilett12:25 PM
@Dan Maloney are we still saving HackChat transcripts?

Ben Hencke12:25 PM
@Garrett Mace what did you end up using to make these explainer gifs? good stuff

reynolds.garrett12:26 PM
What software are you using to simulate your end result?

Garrett Mace12:26 PM
Most of the patterns on that LED Nonagon are very simple. I have it run through the pixels and feed coordinates to generator classes. The core algorithm for many patterns is a oneliner

Garrett Mace12:26 PM


CRGB render(pixel_info pix) {


    int8_t color = (int32_t)(pix.r/12*(int32_t)(sin16(angle))/256) + pix.angle*256/PI*1;


    CRGB ret_color;


    ret_color = ColorFromPalette(palette, color+hue, 255);


    return ret_color;



Garrett Mace12:27 PM
Ben: I'm rendering them in Matplotlib, frame by frame, from within jupyter notebook

Garrett Mace12:27 PM
reynolds: I usually dev actual patterns on the device itself. However, there are some great FastLED simulator projects on github

Garrett Mace12:28 PM
I know hackaday posted one pretty recently?

Garrett Mace12:28 PM
I would like to figure out how to feed LED algorithms directly into Blender and render the results in EEvee

Garrett Mace12:29 PM
In any case....if you precalculate the math, use a "shader" style approach to generating physical-space patterns, you might forget the actual complexity looks like this:

reynolds.garrett12:29 PM
That would be slick, and thanks I will have a look around on github for some simulators

Garrett Mace12:29 PM

Steve Pomeroy12:29 PM
Do you do most of your project mockups/renders in Blender?

reynolds.garrett12:29 PM

Garrett Mace12:30 PM
Steve: I've done a few, but definitely not most. Lots of pencil sketches, and some guesstimations

Steve Pomeroy12:30 PM
Gotcha. I've been trying to figure out a good workflow for LED-based sim there. Sounds like Eevee? :-)

Garrett Mace12:30 PM
You can build the thing, and then measure where the pixels are, or where a strip starts and ends and what the pixel spacing is

LqqkOut12:31 PM
🤯 Incredible images! Maybe a chicken and egg question, but did you have the mathematics background when you started doing animations or did it evolve from these projects?

Garrett Mace12:31 PM
Eevee is a graphics-card based render option in newest Blender releases, it's a lot faster

Garrett Mace12:32 PM
LqqkOut I graduated as an EE with pretty heavy math along the way, but that was almost 20 years ago. Knowing where to look helped (hence this hack chat) but I did have to relearn a lot

Garrett Mace12:34 PM
Any other questions about the above projects, approaches to LED mapping, how to even use this, or random thoughts about the state of LED tech today?

Ben Hencke12:34 PM
I found this bit to be really good at explaining transformation matrix math:

Garrett Mace12:36 PM
That looks pretty good! I'm thinking about switching to Javascript (as in your link) because I think it'll be a bit more performant, easier to share, interactive

Garrett Mace12:36 PM
I mean look at all this cool stuff:

LqqkOut12:36 PM
(hold for later if it's coming up!) Any lessons-learned or tips re: power and power distribution? I imagine these can be pretty hungry

Garrett Mace12:37 PM

brainsmoke12:37 PM
are you using normal vectors for LED orientations in your 3d animations as well? (I've been using them to simulate artificial light sources, where the LEDs simulate reflective surfaces)

Garrett Mace12:37 PM
I don't have more loaded into the chat cannon, so we're well into freeform chat now

Garrett Mace12:38 PM
LqqkOut: I stick with 48V or lower distribution, and point-of-load DC-DC converters for large projects

Garrett Mace12:38 PM
Each node in that ceiling project needed to handle up to 400 watts

Dan Maloney12:38 PM
More of a hardware question. The link someone posted for the ceiling LED project mentioned "custom connectors". With the huge range of connectors available, why would it have to be something custom? And how does one get custom connectors made?

Dan Maloney12:38 PM



Garibaldi Mountain is one of the greatest natural beauties in British Columbia, towering majestically above the small town of Squamish. Every time you drive by or see it in the distance, the extraordinary view of the impressively peaked ridge inherently steals your breath away.

Read this on summit | Tangible Interaction

Garrett Mace12:39 PM
brainsmoke: I haven't needed to worry about normal vectors yet, but on that summit project they were very important. We had the three sides of the sticks modeled, and wanted self-shadowing effects so we needed to know what direction each LED pointed

Garrett Mace12:40 PM
The person who set that up was a data scientist, I definitely recommend getting one on the team if you have a similar scale of problem!

grcornwe12:40 PM
Any thoughts on MicroPython on the microcontrollers you use?

Garrett Mace12:41 PM
Dan: The custom connectors were mostly to handle the long strips we were using, the default connectors were not strong enough or high enough current for a semi-industrial application. So custom was mostly "hey Ray Wu, can you put these on instead" and he said yeah

Garrett Mace12:42 PM
grcornwe: Loving MicroPython. I haven't used it for a high-update-rate LED application yet, but for small info display type projects it's so nice to not dig deep into C to handle network comms, text processing etc

Steve Pomeroy12:44 PM
One of the things I've heard people talk about in terms of pixel mapping is the idea of using a network of smarter pixel drivers + more parametric control signals instead of raw pixel data. Have you explored that at all?

grcornwe12:44 PM
Thanks, I've been struggling with a higher-update-rate application, drove me back to C :)

Garrett Mace12:44 PM
The micropython and circuitpython implementations have high speed pixel drivers under the hood. Might be cool to do a similar approach with a lookup table and shader kernels written in C

brainsmoke12:45 PM
I've written a jitter-hiding driver for micropython, to hide garbage collection cycles :-)

Steve Pomeroy12:45 PM
(I imagine the idea of JPEG-like frames being sent to a controller instead of raw pixels.)

Garrett Mace12:45 PM
Steve: Sort of a game of life approach? Rules-based pattern generation? I haven't done that but see great things on twitter with interchangeable modules

Garrett Mace12:47 PM
That is a fourth approach to the problem though. I know that Paul does demos with a rectangular array from a Teensy, and a computer pipes pixel data to it over USB. However, you could do the heavy processing on the computer and send pre-mapped data to pixel chains or strings

LqqkOut12:47 PM
Are most of your projects for specific clients or are they personal projects? If clients, do you get to design collaboratively or is it more prescriptive?

Garrett Mace12:47 PM
There are a lot of commercial applications that do this, mostly with DMX but some Artnet

Garrett Mace12:48 PM
LqqkOut: It's definitely more personal projects these days, or no-money projects (collab with Burning Man sculptor etc)

LqqkOut12:48 PM
[thank you for all of the answers, I wish the chat had response icons to show the love]

Garrett Mace12:49 PM
Maybe someday

Garrett Mace12:49 PM
It'll be next to the edit button

Steve Pomeroy12:49 PM

Ben Hencke12:49 PM
good stuff

Ben Hencke12:49 PM
thanks, @Garrett Mace

Garrett Mace12:50 PM
If it's all right, I'm going to grab some lunch. If you'd like to continue discussion over the next few days, or have questions for many other experienced LED artists and coders, hop over to the LED Art Ongoing/Weekly Chat

Steve Pomeroy12:50 PM
Indeed, thanks so much!

Garrett Mace12:50 PM


Public chat for LED Art Ongoing/Weekly Chat |

Participate in the public chat for LED Art Ongoing/Weekly Chat!

Read this on Hackaday

Darrin B left  the room.12:50 PM

Garrett Mace12:50 PM
I'll try to spend some more time in there too!

Jason Moungey12:51 PM

Garrett Mace12:51 PM
Thank you all for a great discussion and great questions!

Dan Maloney12:51 PM
Sounds goos Garrett, we're almost at the end of the hour anyway.

Dan Maloney12:51 PM

Louis Beaudoin12:51 PM
Great chat Garrett, and thanks for linking to the LED Art Chat group!

Dan Maloney12:52 PM
Thanks so much for dropping by today, interesting stuff. I never thought there's be so much math involved, but it makes sense once you think about it.

Garrett Mace12:52 PM
Thanks for inviting me, Dan. Was a lot of fun!

Ben Hencke12:52 PM
favorite c matrix math library? I've picked up - seems a good mix of simple, capable, and fast

Dusan Petrovic12:52 PM
Thanks @Garrett Mace and everyone!

grcornwe12:52 PM
Thank you Garrett!

Dan Maloney12:52 PM
You bet. And thanks all for turning out today with great questions.

Garrett Mace12:52 PM
Ben: I'm doing the math in Python so that I don't have to do it in C :) Uh, eventually will probably need that though. Thanks for the link!

Dan Maloney12:52 PM
Next week we'll be talking with Cory Collins about Art with Technology:

Dan Maloney12:53 PM


Art with Technology Hack Chat

Cory Collins will host the Hack Chat on Wednesday, June 16 at noon Pacific. Time zones got you down? Try our handy time zone converter. As hackers, we naturally see the beauty of technology.

Read this on Hackaday

Dan Maloney12:53 PM
Transcript for today's chat coming right up...

Paul Stoffregen12:54 PM
thanks - I only caught little parts of this - looking forward to reading the transcript later tonight

Garrett Mace12:55 PM

Garrett Mace12:55 PM
Er, same