SarahD
Mersey Christmas Lights, maker and breaker
I've been running my show on a network of Pi zero W's, with capes. It used to run fine, but now that we can run 4 ports off a cape rather than two and my show has grown, I think I'm hitting some limits. I'm seeing some freezing for part of the prop as complex sequences run, which looks a bit like either a buffer overflow and a memory issue, or the processor can't keep up.
The offending units are running 1600+ pixels total across 3-4 ports, at 20fps. The worst one is also at the outer edge of my show at the front of our garden, which is about 15m from the nearest Wifi hub (outdoor Orbi).
Each sequence is uploaded individually to each pi in V2 sparse uncompressed format as local models only, which I believe means it sends everything needed for the local channels for that pi, and in a form that can be processed quickly with less computational effort - see screenshot below. When the show runs, the Master pi just provides synching packets over the wifi network. FPP file manager shows the uploaded *.fseq files to each be around 20-25MB.
I'm upgrading one of these fully loaded Pi zero W's to a Pi 3B+ and testing while I've still got the show set up. Hopefully that will fix it
I'm really curious to know if anyone has figured out what Pi memory / compute is needed to run how many pixels, and what the limiting factors are?
As one data point, I've run 1000 pixels of a Pi Zero no problems (which still blows me away every time I think about it).
The offending units are running 1600+ pixels total across 3-4 ports, at 20fps. The worst one is also at the outer edge of my show at the front of our garden, which is about 15m from the nearest Wifi hub (outdoor Orbi).
Each sequence is uploaded individually to each pi in V2 sparse uncompressed format as local models only, which I believe means it sends everything needed for the local channels for that pi, and in a form that can be processed quickly with less computational effort - see screenshot below. When the show runs, the Master pi just provides synching packets over the wifi network. FPP file manager shows the uploaded *.fseq files to each be around 20-25MB.
I'm upgrading one of these fully loaded Pi zero W's to a Pi 3B+ and testing while I've still got the show set up. Hopefully that will fix it
I'm really curious to know if anyone has figured out what Pi memory / compute is needed to run how many pixels, and what the limiting factors are?
As one data point, I've run 1000 pixels of a Pi Zero no problems (which still blows me away every time I think about it).