On the QuinLED site I noted that they suggest that the practical pixel count should be kept to around 2000 pixels for WLED and 3000 pixels for ESPixelStick firmware. Since eight pixel ports of 680 pixels each comes to about 5440 pixels, I'm wondering what led to that suggestion.
I wonder if when he wrote the copy on the web site, if he alternated between thinking of LED strings vs. LED Pixels, and that his use of grammar didn't make a differentiation between them, i.e. he didn't proofread what he'd typed.
the pixel stick limitation is driven by the internal buffer size. the buffer is 9000 bytes long. that maps to 3K pixels.
Are you sure the buffer is that small (and if so, why)? If the SPI module is used for outputting I would expect the DMA transmit buffer to be 5x that size, due to the likely 5x (or so) SPI output rate to form the WS2811 waveform.
The reason that I ask is that I'm playing around with using the ESP32-C3 to drive multiple output channels. I was wondering if I was going to run into whatever difficulties that resulted in such a low pixel count for eight output ports (although 3000 pixels is obviously satisfactory for four output ports).
Since I wrote the code, I am sure 🙂 For this implementation, the buffer must be in local ram. It cannot be in SPI ram due to how the data is removed from the buffer. I use ISR functions to remove the data from the buffer and schedule them into the HW transmitter. The ISR gets into trouble if it changes the page that the SPI RAM is using. This may have been fixed in more recent versions of the CPU, but the code was written a few year ago so that was the issue at the time of writing. The other issue is that the network nuffers do not seem to use PSRAM either. This results in an issue between using the network and having a large buffer.
