search

raspberrypi / firmware

This repository contains pre-compiled binaries of the current Raspberry Pi kernel and modules, userspace libraries, and bootloader/GPU firmware.
5.15k stars 1.68k forks source link

tvservice -n does not detect USB powered display name but dumps EDID correctly #1105

Open marcus905 opened 5 years ago

marcus905 commented 5 years ago

Hello,

I am working with a Raspberry Pi 3B+ and a 1024x600 touchscreen interfaced through HDMI and USB through a TI TFP401 decoder that needs a special config.txt line for it to work properly, as the TFP401 doesn't support scaling. I wanted to add an [EDID=...] filter in config.txt but to no avail as tvservice -n returns

[E] No device present

all the while EDID is read and parsed correctly through tvservice -d then decoded through edid-decode to validate correctness.

The display is powered by the Raspberry Pi itself through USB and this apparently is the same issue as reported in #401 but with the workaround of the bootcode_delay=10 entry in config.txt is not working as the USB hub is not powered (enough?) in bootcode apparently and it is therefore not reading the EDID.

Powering the display separately through a dedicated PSU makes everything work ok on the EDID side but since the display has a touchscreen and it is powered via the same USB cable as the touchscreen transmits data on, powering the display separately is not a very desirable option as it would entail not using the touchscreen at all.

Is there any chance that there could be a config.txt entry to early-power the USB in bootcode too, or at least power the USB before the delay is taken into consideration? Alternately, would it be possible to move the EDID detect phase later in the bootcode, after the USB powering event? Do you have any other suggestion to help with this?

Thank you!

pelwell commented 5 years ago

A 3B+ requires a 2.5A supply to function well on its own. Expecting it to power an external display sounds optimistic. I think you should power the display directly.

marcus905 commented 5 years ago

Ok, but that is not the issue though. If I set the specific config.txt entries unconditionally it works pretty well by itself, even under load, with a 5V 3A power supply. The issue is that the display is powered too late in the boot process to try to make it detectable at boot and therefore filter the options in config.txt accordingly.

I will however try to redesign it in production, if we ever get there, but for this specific reason I'd like to make it conditional so as to build a single image instead of multiple ones.

pelwell commented 5 years ago

I don't think we'll be redesigning the startup sequence for your application - getting the filters to work as expected is already tricky - but @popcornmix may be feeling more charitable.

marcus905 commented 5 years ago

Thank you, I'd really appreciate that a lot!

I think that the code that manages this is not open source, else I can try helping you out if you'd like that.

popcornmix commented 5 years ago

Powering the display separately through a dedicated PSU makes everything work ok on the EDID side but since the display has a touchscreen and it is powered via the same USB cable as the touchscreen transmits data on, powering the display separately is not a very desirable option as it would entail not using the touchscreen at all.

You may have to show a picture, but normally if you want to provide separate power from data for a USB device you use a Y cable. e.g. https://www.amazon.co.uk/StarTech-feet-Cable-External-Drive-Black/dp/B003HHK576

marcus905 commented 5 years ago

I will try with a Y cable as a stopgap measure and report back.