Rabbit 6710 version B ethernet not coming up reliably

Rabbit
, , ,
1

The custom interface board design we’ve used for 4 years (with Rev A1 6710 Rabbits) boots and connects via DHCP (or fixed IP) within seconds, using either Dynamic C 10.64, or 10.72 (with patched serial libs), as it does on the Digi RCM7600 board as well of course.

The new Rev B 6710 Rabbit also starts and connects within seconds on the same RCM6700 Interface Board, but fails to bring up the network interface at all, 90% of the time when plugged into our interface board - no Ethernet LEDs, no DHCP assigned, no ping response etc.

Given that there have been no software or other hardware changes, the only remaining difference is the Rev B of the 6710, which is not specifically documented other than the serial lib patch note.

We did numerous tests swapping hardware and looking at scope traces:
http://rjs.org/test/Rev_A_vs_B_6710.pdf
The only common factor is Rev B, which does have circuit/trace difference(s):
http://rjs.org/test/bottom_zoomed1.jpg

So, does anyone have new, reliable code fix (special settings?) or documented changes?

2

Hello, Did you inform this issue to Digi?

3

Yes, the last response was a quote for paid support. If we could at least get updated schematic and BOM it would probably be resolved.

4

The issue seems to be resolved by adding a 47uF cap to the 3.3V supply. The regulator is borderline spec for the 6700 and power-on surge appears to send it into a rest loop.
The new AT45DB081E flash has only ~6mA extra draw over the previous part
http://www.atmel.com/images/doc1113.pdf
but that might have been enough to droop below 2.93V when needed, and trigger a reset cycle.

1 Like
5

Is that issue present when using the RCM6710 in Digi’s development board? I’m wondering if it’s a design problem with your custom interface, or something on the RCM6710 itself.

6

Yes, it is only on our interface board; the chopper used has a spec for 300mA (which worked with Rev A) but the new flash/code on Rev B bumped it just enough such that boot was unreliable much of the time, but not always.
A 47uF cap smoothed the startup surge, and our next rev will have a 1A regulator.