![](\"http:\/\/solarbird.net\/Livejournal\/2015-10\/groundlevel-before.jpg\")
Minionland and Chudville had Really Bad Wifi<\/center><\/p>\n
![](\"http:\/\/solarbird.net\/Livejournal\/2015-10\/groundlevel-after.jpg\")
Minionland and Chudville have Much Better Wifi<\/center><\/p>\n
Colder colours are noisier, blue is pretty bad, unevenness is generally not good. I’m showing signal-to-noise maps rather than raw-signal-strength maps because as long as you have enough signal to use, S\/N ratios are much<\/em> more important than raw signal power. A full-strength signal that’s 30% garbage is useless; a weak but audible signal that’s clean is just fine. So.<\/p>\n There’s less yellow super-hotspot area now, but you’re not going to get much throughput improvement between those strong green levels and the yellow. The evenness of field should help with reliability, of course – that’s lots and lots of yellowish green, which is pretty solid.<\/p>\n And most of all: no more blue. Blue in this software means problematic levels of noise. I was seeing S\/N headroom numbers as low as 20db, and regularly in the low 20s; that’s not disastrous, but it’s not good. Now we’re reliably in the 40db range, which is a huge improvement. dB is logarithmic, so that’s not twice as much headroom, it’s 100 times<\/em> the headroom. It’s a lot.<\/p>\n To get here, we’ve done a few things. One, we got seriously started on the RF noise suppression in the wiring through liberal use of filter units and ferrites. There’s still lots to do, but it’s a start.<\/p>\n Two, we moved the primary lower hub so that it’s as close to above the downstairs repeater as I could get without tearing into walls. (The location indicators on the display are wrong; it’s the software’s best guess, and it’s inaccurate sometimes.)<\/p>\n Three, I built a reflector for that primary hub to spread the signal around better on the ground level, the non-CHUD half of which is shown above. (The CHUD half is a secret.)<\/p>\n Also, we had to repurpose a functionally-useless AirPlay receiver as a second<\/em> repeater, one level up. That repeater is not shown. It’s so that the main level would have somewhat reasonable coverage from the lower network, which is useful for people going up and down stairs.<\/p>\n So, yeah, there’s that update. It’ll be a couple of days before the next set of line filters arrive – I burned through most of through my useful stock – but that will hopefully help with the noise a bit more.<\/p>\n Then I’m going to poke some at the upper network – officially the “west” network, though both “up” and “west” are true – to see if I can get some better south-end coverage out of it. It’s unlikely, but I’ll try. Mostly, I’m happy not to have done it any measurable harm by adding yet another transmitter to the lower network.<\/p>\n Will this solve all our problems? I doubt it. I’ve already found that we shouldn’t be using our upstream provider’s DNS servers. Our servers don’t use it, why should our workstations? So I’ve gone back to using our own, and that’s already helped. But the big job, I suspect, is getting IPv4\/IPv6 concurrency sorted. We’re running some<\/em> IPv6 now, on workstations, and I’m pretty sure some of the stacks are… not entirely ready. And I’m not sure what to do about that.<\/p>\n So much overhead in running a lair, I tell you.<\/p>\n
\n