RPi4 not booting with two FTDI chips attached

[pierros]

Hello all,

A newly bought RPi4B is not booting when two FTDI chips are attached via USB.

By no booting I mean:
- Power LED comes on (and stays ON)
- ACT LED stays OFF

The FTDI chips are USB to Serial ICs with idVendor=0403 and idProduct=6015
Also it is important to note that both of those chips are not draining any V from the Rpi usbs (they are connected only on the D+ / D- and GND.

My RPi4 goes beyond (and boots successfully after this zombie mode) only if I do either of the following:
- Detach the USB cables to FTDI chips (only one of them would do)
- Attach a tty/usb interface to the UART pins , and then to another desktop/laptop (but it HAS to be all three pins TX/RX/GND)
Once it boots (and I re-attach the FTDI chips) everything works fine.

What I have tried:
- Powering the Rpi4 with all different ways (different usb-c cables and power supplies, directly on the 5V pins)
- Upgrading to latest EEPROM (rc3.3)
- All different combinations on where to attach the two USB devices (both on the USB3 controller, both on USB2, mixed etc)
- Isolated shield and GND on the USB cables attached
- Measured GND Voltage diffs to check for ground loops (none found)
- Tried many different cables

I am kinda at a loss Initially it looked like a ground loop, but then after isolating GNDs with chopped cabled, I think this is safely eliminated. Any ideas?

ps. Needless to say no other devices are attached.

[plugwash]

I saw a report of a similar issue with a USB switch box.

A compliant USB device must not drive any of the USB lines before power is detected on VBUS. Apparently on the pi4 leakage currents from non-compliant devices will stop the power rails dropping to zero when turned off which in turn stops the PMIC coming out of reset.

You should rewire your FTDI chips to be powered from VBUS (FTDI chips have a seperate IO voltage pin, so you should be able to power the FTDI chip from VBUS while running it's IO off whatever voltage you need).

[pierros]

Thanks for the reply plugwash!

I will be giving this a try. Can you please point me to the similar reports you are referencing?

Cheers!

[plugwash]

Unfortunately I believe the other report was in a private forum, so even if I could find the link you wouldn't be able to access it.

[6by9]

Reposting the information from that thread

It's what I expected it was. On the Pi I can see that the hub/switch is leaking power into the unpowered Pi (3V3 line charges to ~800mV) which will stop the power management chip turning on (as it's designed to wait until all aupplies are less than 200mV, so it guarantees supply up-sequencing). it's not leaking through VBUS so it'll be (almost certainly) leaking via a pullup to the D+ USB line.

The hub/switch is therefore behaving incorrectly - as it should never apply a powered pull to an 'unpowered' USB host - usually a peripheral never can do this as the peripheral itself is powered from the host's VBUS.