Any reason for this

[FFMan]

On my 128a1 based project i have multiple com ports straight wired to a robust connector. this includes 5v, gnd, tx, rx and an IO line for DCD etc

I've been using the OpenLog microsd module and set one up to act as a debug trace capture device, so when in the field, i can capture the debug without the need for a laptop. So i have an openlog wired to a connector which i can move from port to port.

One odd thing, when i put it on com9, it behaves slightly differently to com10 (which exhibits the behaviour i expect). Its as if com9 is slightly slower to trap characters. i had to increase the sleep time a fraciton to make it work on com9, whereas on com10 it was happy (same setting works on com7 too).

To move from one port to the other, all i do is amend 2 variables (port no and dcd pin) so i don't think it can be a coding issue. I have tried alternate pin for dcd, and checked the operation with a meter but i can find no difference to explain what I see.

Itrs fixed now by increasing the wait a fraction, but is there any logical explanation for this behaviour ?

[dkinzer]

[...] is there any logical explanation for this behaviour ?

None that I can think of. Do you think you could produce a minimal test case that demonstrates the issue?

[FFMan]

I have a test program that exhibits the behaviour, i'll email it over later.

I'm using one of these https://www.sparkfun.com/products/9530

[FFMan]

so when i came to isolate the problem in a test program i discovered the real problem.

the code fails when i let it run following a download. if i reset the processor it works what ever the port, so there is no port difference, just me not noticing a sequence of events.

sorry to have troubled you