Sparklan 139 does not draw a lot of juice but in any case probably first thing to check is amperage of power supply, integrity of crimps on POE.
Also I that the Sparklan wants 5v for the 802.11a PA stage so I’m guessing you’re on 2.4ghz here. I don’t think the RB153 is capable of supplying 5v.
If all that’s ok then maybe a driver issue? There’s another thread here with some users complaining of RB153 instability and several of 'em are using Sparklan boards. We’ve got a number of 153 boards deployed w/CM9, SRx and no problems (all running 48v).
I have two installations with rb153.
First have 18V power supply from PC Engines, second have buffered power supply 24V (2 x 12V 7Ah batteries) - 2.5A.
Different power supplies, 24v at 2A is plenty, meanwhile unlikely to have two bad crimps in separate installations. Sounds more like a driver or wireless card stability problem, eh?
On another thread, Uldis makes this suggestion regarding use of RB153:
“…which Mini-PCI slots were used? If you are using one or two card you can try to check if it helps when you are not using J5 Mini-PCI slot.” which hints at maybe a bus termination problem when not using all slots.
Sheer speculation on my part but Uldis certainly knows the product
Now, here’s a really spongy idea so you should probably just ignore it: Assuming that RB153 does not supply 5v and then really sticking my neck out and assuming the two 5v supply pins on the Sparklan end up unterminated as a result, I wonder if that could cause a problem? Does not seem likely but if the 802.11a PA involves logic circuitry of some kind I suppose it’s not impossible.