This is part of a series of post on the design and technical steps of creatingHimblick, a digital signage box based on the Raspberry Pi 4.
One full day of crazy debugging, and the result is that if the Raspberry Pi 4 outputs HDMI at a resolution of (x) , the WiFi stops working.
Any lower resolution we tried, from (x) down, does not show this problem.
We did not try any resolution above 2560 x 1440, for lack of a monitor that would do it. *****
We reproduced this:
- on both microHDMI outputs
- with two different cables: one with a microHDMI to HDMI dongle adapter; one direct microHDMI to HDMI. The latter is the one I bought in the Raspberry Pi Store in Cambridge together with one of the Pi 4 units.
- with three different RaspberryPi units
- with 4 different power supplies: one rated at 2A, one rated at 3A, one rated at 3A bought in the Raspberry Pi shop in Cambridge, and a laptop USB-C charger
- with stock Raspbian Buster Lite
- with stock Raspbian Buster
- killing every process in the system, starting the network manually with
dhclient, and starting X manually with
- with two different SD cards
- connected to an AP some meters away, and connected to a phone hotspot next to the Raspberry PI
At the bottom of this forum thread(
guestxyzdated Aug 07) someone mentioned screen resolution, which is what finally prompted us to try that. Thanks,
Further things left to try after chatting about it on IRC:
- whether disconnecting the HDMI cable from the Pi end (with X still started at high resolution and everything) make the WiFi work again
- switching the monitor to another input while the Pi is at 2560 x 1440
- letting the monitor go into power saving mode while the Pi is at (x)
- cable chokes
- Hacker News thread
- mentioned the test with a different distance from the APmadengr’s comment reminded me we also did that
- Mike Walters took and posted some measurements
- Android Authority had a story, where Raspberry Pi’s Eben Upton chimed in
- Mentioned that one of the cables we tried was the one that came with the Pi 4; added photo.
- Mentioned that we did not try to go to higher resolutions than (x) .
This post seem to have become quite popular, so to be clear, the intention is to document the issue that we spent a whole day chasing so that others won’t have to.
It’s been fantastic to see how this has made the rounds and resulted in actual RF measurements being made, and better pinpointing of the issue than we could ever have achieved.
I greatly admire Raspberry Pi’s work. Thanks to them we get to have an affordable standard platform with a wide use base, which over time gets extremely well known, and as such can be a very well mapped ground to build on.