Wireless is hard, 5dBi to 7dBi antennas

[azazel1024]

So I got the 7dBi antennas yesterday. Spent most of the night testing and playing with them. Unlike the 3 to 5dBi antenna swap on the AC1750 router which was basically a positive, and generally a large positive in every single location (or at worst, one neutral/ambiguous result), the 5 for 7dBi swap on the WDR3600 was VERY mixed.

In its current location, it sits in the playroom exactly in the middle of the fireplace/chimney. Which also dawned on me, that right there, it also sits right behind the METAL pellet stove insert AND the SS flue liner. Yeah, execellent planning. With some testing I actually figured out it was pulling poorer results there than where it was located previously when we had our entertainment center there and it was located nearer the edge of the fireplace. I digress, the testing.

Location A, in the playroom, Location B at the dinning room table, Location C in the living room, location D sitting in bed (most results are from my laptop with Intel 7260ac in it, the tablet results are from my Asus T100 with a 1:1 2.4/5GHz 11n adapter in it)

A showed slightly flatter results, but no faster in 5GHz and in 2.4GHz Tx, but a loss of 10-15% 2.4GHz Rx. Tablet, ambiguous (tablet is hard to test because the numbers are ALWAYS so spread around, unlike the laptop which is almost always +/-5% from test to test and across an entire 1+GB transfer, where as the tablet can be +/-15% from one transfer ot the next, so unless a result is pretty far out and repeatable...). Tablet may have been a little slower in 2.4GHz Rx and Tx, no 5GHz difference.

B showed slightly 2.4GHz increase of about 5-10% in Rx and Tx. Around a 20% boost to 5GHz Rx and 10% to Tx. Tablet may have actually be a little slower in 2.4GHz Rx and Tx, no 5GHz difference.

C showed slight loss in 2.4GHz of 5-10% Rx and Tx, neutral 5GHz Tx, loss of about 25% in Rx. Tablet showed a loss in 5GHz Rx and Tx slightly and 2.4GHz was neutral.

D showed no change in 5GHz (can't connect), 2.4GHz showed neutral Rx and 20% gain in Tx, didn't test with tablet.

Seriously man. Just gazing at signal strengths a little, all over the chart. In room they seem to show the same numbers between 5 and 7dBi antennas for both bands. At location B, it shows the same for 2.4GHz, but a 10dBm loss for 5GHz (even though it is faster). In location C, it shows a 3dBm loss in 2.4GHz and a 1dBm gain in 5GHz (even though 5GHz is generally slower too). At location D, 2.4Ghz and 5Ghz both show a 3dBm gain (5GHz was too weak to warrant trying to connect to either way as it went from -91dBm to -88dBm. Windows could connect in the later case, but pages would not load and file transfers would sit for 10-15s doing nothing at all. Would not connect at all at -91dBm/5dBi antennas).

So I then tried moving the WDR3600 around. First toward the outside of the house on to one of the new built in shelves emily added between the fireplace and the outside wall.

All locations showed a loss in all frequencies except location C, which saw 20-110% gains in performance depending on the band.

Then I tried moving it to the other side of the fireplace sitting up on a storage unit thingie that the kids toys sit in at about chest height. In room, Location A with the 7dBi performed better than it did with the WDR3600 in any other position, but still perhaps 3-5% slower in some of the tests than with the 5dBi antennas, but the same in others and all of the transfers were flatlined, as opposed to having some small variation in them. Location B was 5-10% faster in 2.4GHz and 40-55% faster in 5GHz. Location C was 40-50% faster in 2.4GHz and 180-250% faster in 5GHz (sitting in the original position all I could get was 4-5MiB/sec in 5Ghz, new location about 14-16MiB/sec). Location D was 80-120% faster in 2.4GHz and I can connect in 5GHz and get around 5-6MiB/sec (which is inifininitely faster than the 0MiB/sec of before).

Roaming still works well between the AP and the router in the basement, at least with the laptop and tablet. Laying in bed 2.4GHz signal strength went from -80dBm with the 5dBi antennas, to -78dBi with the 7dBi antennas and moving the WDR3600 to the final location then bumped signal strength up to -58dBm (and -69dBm 5GHz). Changing between the 5dBi and 7dBi antennas nets a difference of 6dBm laying in bed now in 2.4GHz and 3dBm in 5GHz.

So, at a guess, with some of it, not only is it having the AP unmasked by 4ft of masonry and the flue/fireplace, but I think that the metal of the fireplace and flue were/are causing some really nasty radiowave reflections and its made worse by having stronger antennas on it, but some brief speed testing between the 5 and 7dBi antennas with the WDR3600 in that final, fastest location saw a net gain in speed at ALL locations using the 7dBi antennas. Mostly very tiny gains in the 5% range or so (5GHz was closer to 10%), but every test saw either neutral or a 5/10% gain, except laying in bed where there was a ~10% gain in 2.4GHz and an 18% gain in 5GHz.

Why the heck does wireless have to be so damn hard? I mean, I know that "final location" would likely be the fastest, I just can't leave the AP there right now because I'd have to run the network cable 12ft and around stuff to plug it in, so I need to put a jack in that wall (there is an outlet already fortunately). No biggie, I'll get to it one of these weeks/months, and I can live with the AP in its current location...but still man, danged annoying.

I do need to do more roaming tests once I get the AP relocated to see if devices still roam well. With the 7dBi's, walking in to the basement works really well. Signal strength is 3-5dBm lower in the basement in the 3 places I spot checked (because it is WELL below the AP in most of the locations, so below the half power beam width), compared to the 5dBi antennas. Walking in to my bedroom/the kids bedroom/bathroom, my tablet seems to want to roam right over even though the signal strength is pretty decent still, but I don't know how well the laptop, iPad 2 and phones are going to behave. Laptop I assume will roam well also as windows and roaming is set to Aggressive just like my tablet. Phones might hold on to it, but at the same time, they also tend to prefer the 5GHz band, which is weaker enough from the AP still that they'll probably roam instead of holding on to the AP (I mean, the new signal strength is high enough it isn't a big deal, but I'd still rather devices connect to the WAP closest/strongest to them, even if there isn't a huge signal difference).

Since the performance and general behavior seems a lot better with the 7dBi antennas (in the final location), I'd prefer not to swap back to 5dBi (also because the AC1750 has now claimed them along with one 5dBi from my spare WDR3600). I guess one of the things I could do is play with setting the Tx power on the WDR3600 lower so that clients see lower RSSI and roam more readily if I need to. I'll just have to do both roaming and performance testing. Since 5GHz seems to be golden in terms of performance AND in signal strength I shouldn't need to tweak that, but I might need to set 2.4GHz to 75% power or something and that likely won't mess with actual performance much if at all, but might encourage better roaming (and it'll allow the increased Rx performance of the router since it'll still have the big antennas on it and all of the nice 5GHz benefits).

 

[azazel1024]

To add since it is a wall of words and my head hurts re-reading it all (sorry), I think the issue with the first location between the 5dBi and 7dBi antennas is that I am getting nasty reflections off of the metal fireplace and flue and going to stronger antennas accentuates the issue, which is why there are mild to modest losses in a number of the locations, but also some gains, where reflections aren't as much of a concern. Moving the WDR3600 to the side, both unmasks it from the chimney, but also moves it far enough from the fireplace and flue to dampen reflections significantly (instead of being 1-2ft from the flue and 2-3ft from the pellet stove insert in the fireplace, it is now located around 6ft from both, plus not having 4ft of masonry in the way from any of the test locations). So with that out of the way, the 7dBi can actually function better than the 5dBi.

Site surveys are a B.

 

[thiggins]

Play with your antennas too much and they'll fall off...

Seriously, you are pissing against the wind. Today's gear has implicit and explicit beamforming and RF chains and rate adaptation algorithms ultra-tweaked to squeeze as much performance as possible without violating FCC limits.

Throwing anything into that mix is a recipe for what you are experiencing. Not to mention highly variable RSSI that seems to be part of AC.

When bringing up the first octoBox setup, I tried correlating location to path loss using three different methods. Each method produced different results. RSSIs at lower levels didn't correlate at all. Readings flattened out surprisingly early.

The attached IxChariot plot shows RSSI at 0 and 50 dB of attenuation with an ASUS RT-AC68U router. and PCE-AC66 STA. See what I mean? The 50 dB RSSI shows an -52 dBm average, but look at the range!

 

[azazel1024]

I'd agree on the pissing against the wind. At least until midnight or so all I was doing was getting splattered.

The wierdest bit to me was the 11ac router handled an upgrade from 3-5dBi antennas gracefully and saw big performance improvements with basically no tweaking, threw them on and it ran with it. The 11n (n600) router on the other hand was a fist fight in an amputee ward swapping the 5dBi for 7dBis.

I still think a lot of it had to do with odd reflectance issues, but maybe not. I do know once I finally found a better location, it was all roses with the bigger antennas.

Since I've gotten "serious" about digging in to wireless, I think I've spent significantly more time on learning about wireless, testing, setup, etc than I ever had on wired networking, even though I have about 10x longer experience than with wireless. Mostly because with wired networking, 99/100 times it works just fine. Wireless, "the theory is good, but the implementation sucks" seems to be the way it goes. Maybe 10 in 100 times it works as I expected or better than expected, 80 out of 100 times I spend playing with it until it works the way I had hoped and 10 out of 100 times I just have to throw up my hands and walk away.

 

[RMerlin]

If you want to go hardcore, you could go as far as this guy:

http://jasmcole.com/2014/08/25/helmhurts/

 

[azazel1024]

That is awesome. I think the only issue I see with that is, without a lot of empirical testing, there is no way to get a matrix that'll accurately comform to the actual setup of the environment. That said, it could be somewhat instructive/illustrative to work with as a first step in determining placement, changes, etc.

The environment is just too complex as none of the walls are going to be simple (and it doesn't look like he even used accurate refractive/absorbtion numbers for the materials/composition in use). Even if perfectly modeled, it won't comport to reality, because you know that contractor actually spaced the one stud a quarter inch of 16" on center, or the wiring is run on the other side of the stud, or one stud is doubled up, etc. Though probably at that level it is nit picking and might not have any impact on the real world results based on the modeling. Then you've got the impact of furniture on everything, and external RF sources, reflection from things OUTSIDE the flat from leaking RF inside the flat and so on.

I still go back to "that is really F-in neat" and that I could see how that could be somewhat useful, especially if a bit more effort was put in to plugging real numbers in to the matrix for absorbtion and refraction.