Andy Burns wrote:
When the machine is on, the on-board LED is green, pin 2 is at +5V and
pin4 is at 0V, so the green external LED shows "on" as expected.

When the machine is asleep (or is shutdown) the on-board LED is amber
and both pins 2 and 4 are are at +5V which explains why neither colour
of the external LED across them is lit.

I can't see how the external LED can show red for "off" when connected
as per the manual. There's no setting in BIOS to let machine know it
has a bi-colour LED available.

The Intel doc only describes green and yellow states, never red.
The manual says:

FP PWR_P/SLP_N: Power/Sleep messaging LED terminal 1 with 510Ω pull-up
resistor to +5V_A voltage. Connect it to an extremity of a dual-color power
LED for power ON/OF, sleep and message waiting signaling. Please refer to Intel®
Front Panel I/O connectivity Design Guide, chapter 2.2.4, for LED
functionalities and signal meaning.
FP PWR_N/SLP_P: Power/Sleep messaging LED terminal 2 with 510Ω pull-up
resistor to +5V_A voltage. Connect it to the other extremity of the
dual-color power LED above mentioned

So it seems you can attach an R+G LED and get:
R=off G=on = green: power on
R=on G=on = yellow: sleep
R=off G=off = dark: off

The description agrees if there is a ground available, so a pin being pulled
up high turns the LED on. Though I would have expected an additional wire
on the connector for that, and the Intel doc suggests back to back G+Y LEDs.
Either way, you'll never get red.

If I pull up the schematic:
http://download.udoo.org/files/UDOO_...schematics.pdf
the signals are wired S3_LED (p4) and S5_LED (p2), with a 510ohm to +5V and
a MOSFET pulling low. They're driven from an STM32F100 microcontroller.
In ACPI terminology S3=sleep and S5=soft-off. I would therefore assume:

S3_LED=low S5_LED=low: machine on, LED=green
S3_LED=high S5_LED=low: sleep, LED=yellow
S3_LED=low S5_LED=high: soft-off, LED=red
S3_LED=high S5_LED=high: hard-off, LED=dark

which implies an R-G LED with separate ground, rather than an Intel
back-to-back G-Y LED.

Since it's firmware controlled, using either would be possible (with
different results).

Strangely, the on-board LED has a completely different setup that appears to
be based on reset and power rail state, rather than software controlled. So
its behaviour could well differ.

Theo