Make USB Host More Reliable on Long Cable Connections

[ruvsudrn]

This may be impossible to achieve since I am asking for something that will mix between logic and physical layers, but here goes...

Can anyone suggest ways that the TinyUSB host library can be modified to make the host more tolerant of operation over long USB cables?

What I have here is an RP2040 USB host board, and a generic USB mouse as the device. I want to use a very long USB cable, and have a series of active cables here that are 10m, 15m, and 20m in length. When I test using these long cables I find that some of them work and some of them do not. When I test them with the same mouse on a regular PC they all work.

Is there anything in the TinyUSB library that I could tweak to improve reliability? Data throughput is not an issue. Thank you.

[braxlan]

Just a few comments, I'm not a maintainer of this project.

Active cables are just a few bus powered "single port hubs" in series. So you might be asking for hub support in general. I see two main things:

• If it is the hub support, you'll have cables which work and ones which won't work, depending on the hubs they use.
• It may be a power issue, as you have a series of passive hubs. You might try to use a powered hub between the RP2040 and the long cables to see if the long cable + mouse draws too much power from the RP2040. Keep in mind you can have a maximum of five hubs between the RP2040 and the device, and the active cables will have one hub per additional 5 m (n_hubs = length / 5 m - 1) iirc.

I personally used four levels of hubs between an RP2040 host (Raspberrry Pi Pico) and input devices without problems, but I used at least one powered hub in the chain.

Maybe you want to use a USB-over-CAT or a USB-over-IP extender for long cable runs, they might work more reliably.

[ruvsudrn]

Thank you for the comments, you are absolutely correct on all counts.

But I have monitored power at the host and device end and there is no sign of voltage sag or an inability to deliver current. And again, these same cables and device work fine when connected to a regular PC where, again, voltage measures fine at each end.

So my (unfounded) theory is that there could be some aspect of the TinyUSB library that assumes timing in some way, or, assumes handshaking happens in a certain way, which as result makes it less robust over a longer and more fragile link. Because the TinyUSB library is bit banging the USB exchange it must have quite a few dials that can be tweaked to adjust behaviour. I have looked at the code, but rather than stumble around without a clue, I had hoped that someone with intimate detail of the project could point me in the right direction to get started.

Thank you in advance.

[braxlan]

I don't know the details under the hood, but does TinyUSB do bit banging when using the RP2040 as a native host? I only know this when using Pico-PIO-USB to add an additional, non-native host which indeed uses bit-banging.

[braxlan]

I was asking because using Pico-PIO-USB makes a big difference in hub support compared to using the RP2040 as a native host. Which board and which port of it are you using as a host?