[sdiy] USB Question

[rsdio at audiobanshee.com]

On Sep 23, 2015, at 10:59 AM, Ingo Debus <igg.debus at t-online.de> wrote:
>> Am 23.09.2015 um 19:22 schrieb rsdio at audiobanshee.com:
>> 
>> The Device connector power sense would need either a 5-V-tolerant pin, or a resistor voltage divider to bring the voltage down to the CPU I/O level.
> 
> Hehe, on NXP processors this pin is 5-V-tolerant - as long as the 3.3 volts power supply is present. Otherwise it is not. This is mentioned in a tiny footnote in the data sheet. 
> 
> So you need the voltage divider anyway. NXP published an application note where this issue is discussed - long after the processor was introduced. Probably many have gone up in smoke before that AN.

There is a somewhat similar mistake on the EK-TM4C1294XL LaunchPad. The TM4C1294 has a 5-V-tolerant pin that can handle direct connection of USB power, but it's always good practice to ensure that a current-limiting resistor is placed inline. I got started in USB-MIDI on the PIC, and their example circuits always have a 100-Ohm resistor in series for any GPIO that connects to a power source for sensing. Since there should ideally be no current flowing, the 100-Ohm resistor causes no issue. But, if there is a problem, the 100-Ohm resistor will limit the maximum current to 50 mA (for 5 V). In the case of the TM4C1294 chip, the GPIO is generally 3.3 V, but a couple of USB pins allow 5 V. Unfortunately, the LaunchPad has none of the typical protection on these pins, and they can latch up and get really hot (I think it happens if the voltage spikes above 5 V briefly). They do have an Errata (GPIO#9) for this, but I don't think it's a chip flaw - it's more of a standard protection that should always be done.

Funny thing is that if I had jumped straight to a custom board - which is my usual m.o. - then I would not have seen the error, because I usually remember the protection resistor. It's only because I decided to take a short cut and use the LaunchPad that I even saw this latch-up and overheating problem. It's rare to see such a mistake on an evaluation board, especially from TI.

Brian