Nelson's log

WiFi transmit power readings

I’ve got a router outside about 200 feet from my iMac with a clear line of sight through a window. I’m running 802.11 at 2.4GHz over the distance. It works surprisingly well. I’m in a quiet rural area. The signal strength isn’t great and the speed isn’t what I’d like. The Tomato firmware I use lets me override the transmit power in the router, numbers from 0-250. Default is 17 or 42, folk wisdom is that up to 70 is “safe” although other folks say any setting is safe. Those settings may be throttled too, depending on what country you’ve configured the router to be in. Also apparently too high and you start getting distortion. Or not. Wireless is spooky.

Here’s some readings on signal courtesy of the WiFi Signal Mac program. Routers are running Tomato and/or Shibby and are using stock antennae. Readings taken under light to no load. All readings are approximate / variable and sloppy enough to not be very good. It’s not clear what these settings really mean; people call them “milliwatts” but it’s not clear that’s true.

All this stuff is using channel 11. 20MHz wide, AFAICT, even though I configured for 40. Measured noise is -79 to -82 dBM. I wish I understood how noise worked better. I’ve been able to use this link even when reported noise was higher than signal. Is that a thing WiFi can do?!

iMac 2011 tests:

Linksys WRT54GL 802.11g, 200 feet away

ASUS RT-N16 802.11n, 200 feet away

a different ASUS RT-N16 802.11n, 10 feet away

Macbook Air 2011 tests

ASUS RT-N16 802.11n, 2 feet away

ASUS RT-N16 802.11n, 200 feet away.

No real conclusions from all this. Boosting power might or might not make a significant difference; measurably it does with the WRT54GL, not so much with the RT-N16. The RT-N16’s signal strength also seems a little better, which is sort of expected since it’s 5+ years newer hardware with fancier antennas, etc.

One thing I didn’t record systematically is the improvement 802.11n brings. With the 802.11g links my actual link speed seemed to be 2Mbit/s as measured by Ookla Speed Test. I can get the full 12Mbit/s with the 802.11n. The reported theoretical Mbps of the wireless link is also significantly faster, as expected. The way I understand this is that 802.11n makes better use of available signal than 802.11g.

See also this discussion which shows improvements from -73 RSSI to -60 RSSI as power went from 1 to 55. Another graph shows improvements from -80 to -72 RSSI as power goes from 0 to 80. This article is interesting as well.

Update: antenna arrangement

Woah! I rearranged my antennas and picked up 5-6 dBm, which when I only had 6 dBm over noise basically doubles the SNR.

All the tests above are with the 2 or 3 antennas on the router all pointing straight up. Turns out that’s dumb, you want them perpendicular. I arranged the 3 antennas on the RT-N16 into a W shape (45 degree angles) instead of all 3 straight up. Went approximately form -68dBm to -63dBm. Who knew?

Update 2: hot rod antennas

I tried some third party antennas, advertised as 9dBm gain (compared to 3dBm stock). Not impressed. Measured signal strength at my iMac was the same, about -63dBM with both sets of antennae. The new antennae are over 2x as long, which I guess is probably good? But the connector isn’t stiff enough and it doesn’t hold its own weight at the 45° angle I figured out was better. I think I won’t be using these.

Update 3: symmetric readings

I forgot that my RT-N16 wifi router 200 feet away has a report of RSSI too, so I can see the signal from the other side. These readings are all +/- 3dBm fluctuations.

These devices all report noise measurements too but its’ radically different, like -75 dBm on the iMac and -92dbM on the WRT54GL. No idea what that means.

Update 4: San Francisco tinkering

I tried to apply what I learned about improving wifi strength in San Francisco. Different environment: I’ve got 10+ networks in view of me (mostly 2.4GHz?) and the walls of the house generate a lot more interference, I think because of metal lath in the plaster. Anyway, what I learned:

I wish there were some easy way to measure reliability. I don’t care about max throughtput nearly as much as I do reliable packet delivery at whatever speed.