Hello Jack,
good pigtails and don't mount/position the antennas close together, then you don't see a big enough problem to counter the benefits of the four or eight port setup.
It's good to hear that the Atheros cards are well shielded however I'll bet that (unless ALL the issues mentioned above are successfully addressed) the collision/overloading/throughput reduction problem will become "big enough" to worry about when traffic on the box becomes heavy - the throughput can then degrade sharply.
A simple paragraph or two addressing this issue in your documentation would provide a very valuable service to the WISP community.
Yes, it is no problem to put this in our RouterOS documentation -- and perhaps an application note for the RB18, for those not using RouterOS.
The needs of people using this cards don't have to be -- full 8X more throughput. They can also need multple point to point connections with the full radio power of one card -- even if tx-holdoffs or receive collisions lower the throughput because of 'static' noise from other radios close.
If they need the full power of one card (e.g. they want a high EIRP), then, since the link must be recipricol, (or close to it) the remote (CPE) end must also be fairly high power (or have a larger antenna, whatever.)
Yes, I worked at Vivato. Yes, we did a design with 6 miniPCI cards on a single PCI bus, and make it work, but there are two significant differences between that, and what you're selling.
1) all the card ran on the same channel. While this is possible with your product, I doubt that more than 2% of your customers are doing it
2) (more important), the cards had some extra signaling that allowed us to keep a card from transmitting under certain circumstances. The circumstance most often in need of control was if any other card was receiving a packet (we had signaling for that too).
TX holdoffs are fine, the MAC knows how to deal with these and its not too big a penalty. However, smashing packets due to rx "collisions" is a bad thing.
One of the design points I've found important is that the wireless medium is "expensive". Packets that traverse it should be treated with care.
Thanks for the support. It would be better if this discussion could take place after you and Jim had both experimented in full with the multiport cards. If locating the cards close to each other is so bad, how is it that I can get 120Mb/s of throughput from a four port card with four 5GHz radios -- on the table? It is not nice being attacked by armchair flamers -- meaning Jim, not you.
Yes, you can make it work if you're very close, but here you have a very high SINR. You get 120Mbps of 'throughput' because you're close.
Try this:
Go to range with one card operating. (Range for any given modulation, say, 54Mbps.
Now turn up the second card, and see what the throughput with both cards is.
If necessary, repeat for cards 3-8, but my opinion is that you won't need to do so.
Please do let us know your results.
