Please note: This project is no longer active. The website is kept online for historic purposes only.
If you´re looking for a Linux driver for your Atheros WLAN device, you should continue here .

Using Ubiquiti 400mW cards with Madwifi

I had some trouble getting an SR-2 card to work with Madwifi and had to have it exchanged. Here's my experience, in case it helps others. Ubiquiti Networks provided really great customer service in the process!

Original request for help, sent Oct 7th 2005

I purchased an SR-2 (SR2_REV002) in June from wisp-router.com and I have never been happy with it. Last week-end I was finally able to do a 1-1 comparison to a 80mW Wistron CM-9 and a 80mW Senao NL-5354 Aries2 and the SR-2 provides noticeably worse connections.

My set-up is a WRAP.2C x86 single board computer running Linux and using the MadWifi driver (I tried both a 6-month old version and the latest version). I am using a U.fl pigtail and a 15dBi Pac Wireless Vagi, and I'm swapping the radios in&out. The other end is 9 miles away with also a WRAP board, a CM-9 radio, and a 18dBi grid.

The SNR (as measured by the driver) at the far end is around 10-14dB, which I realize is marginal, but does allow connections at between 1Mbps and 11Mbps data rates. The SNR for all 3 radios is essentially the same, i.e., I don't see any systematic 7dB signal improvement at the far end. I do see about a 2x better throughput (measured using ttcp) when using the CM-9 or the Senao over the SR-2! I also see higher packet loss rates when using the SR-2.

I have used the SR-2 in other settings where I also swapped it against a CM-9 and it always performed worse (but in those situations it was harder to exclude experimental errors because my measurement tools weren't as well controlled). I never saw an improvement in any of my experiments when using the SR-2 over a CM-9.

I would really like to switch a number of my long distance links (see http://www.diablo.sbarc.org) to the SR-2 and SR-5, but given the performance of the card I have that doesn't sound like a good idea. Do I perhaps have a bad card? (I have tried a MMCX pigtail and saw no difference.) What can you recommend?

Ubiquiti's response

We received your card today and did some testing against CM-9 in our labs. We see a very clear 8dB link gain improvement. There is also a throughput advantage.

So, I don't know what is going on.

However, we have made some improvement to make our EEPROM programming more robust a while back and this seems to have helped some of the MADWIFI users.

Let me get you a couple new cards out and let's see what happens.

It is also possible your system setup might be a contributor. What SBC are you using? How about your power injector -- is it sufficient?

The end result: success! (Oct 29th)

I was finally able to test the returned cards. Many thanks for the extra card: that is immensely appreciated! In short, everything now works beautifully! It seems the EEPROM changes made all the difference. Over short distances it's hard to see the tx power improvement, but over a 9 mile link the approx 8dB are clearly there!

I have one station powered by a 12v supply through over 100ft of cat-5, and that doesn't work: the whole WRAP board resets if I do heavy transmit tests, but that's not too surprising. With a good 12v supply everything looks fine.

That was really excellent customer service! I hope their business is going well and wish them the best of luck!

A different response from wisp-router.com

Recommendations is to use only 1 SR card in a WRAP board and to use at least 18V if you are doing POE and even if you're not. 15W capable PSU is for most part enough with one radio card even when doing POE.

Also in the past we have noticed strange behavior with some cards on the WRAP.2C boards especially if you're using the mini-pci slot on the back of the board but mainly in conjuction with a card in the front slot. It seems that the front slot gets all priority and the back/bottom slot only gets "scraps".

Power setting for Ubiquiti SR2

Apparently what Ubiquiti does in the firmware is to program slightly fudged power settings. The actual output power corresponds to the output setting issued by iwconfig + 10db. So the 16dbm setting is actually the 26dbm (400mW) Tx output power, and setting 6dbm would actually yield 16dbm output power, etc.

- Thorsten

Power setting for Ubiquiti SRC

The programmed power for SRC is 19dBM with an offset of 6dB. This is straight from the hardware engineers. The card at 19dBm shown is actually trasmitting 25 dBm (300mW).

See Compatibility/Ubiquiti for more information about the TX power offsets.

New wrap2c since december 2005

The new boards got better onboard hardware to use 400mw cards in both slots.

- Steffen D. meshnode.org