Wireless Raspberry Pi - how things really are

Part II - In search of solution

    The task
The questions posed in the situation explained in part I are related mostly to how can be achieved an optimal performance. In order to measure the strength of the WiFi signal and its throughput I already had a tool good enough for the job - my Raspberry Pi equipped with its Wi-Pi module. But I also needed something else - a benchmark - a device with inherently good performance in terms of WiFi reception. Luckily my laptop has proven to be pretty good in this and I have this axiom in my mind, that it has a very capable antenna. This is kind of biased opinion but for my small experiment I guessed such bias could've been neglected.
On the other hand I already had a second Raspberry Pi unit, which could make a proper second WiFi-meter. 

The two measuring device prior to powering up

At that point it was clear to me that with three WiFi-meters I was going to test four devices. This means that the task has two dimensions:

1. Test and compare the reception properties of the two WiFi modules attached to their Pi units.
2. Test and compare the emitting properties of the two routers (the old G class, and the new N class).

    The setup and measurements

I picked nine points at home that differ in distance and visibility to the router. Their properties are given in the table bellow:

Point	Distance (m)	Visibility
1	1.0	direct
2	5.0	1 wall
3	8.0	direct
4	12.0	2 walls
5	10.5	1 wall
6	1.5	1 wall
7	5.0	2 walls
8	7.5	3 walls
9	13.0	2 walls

Point 4 here is the place that initiated this whole research. As can be seen, it is fairly distant and without direct visibility. And here is what wavemon detected in the first series of measurements - with the G class router:

wavemon readings for the G class router. The benchmark tools (laptop) is at the bottom.

The same spot covered by the signal from the newer router looked like this:

N class router signal readings.

It is quite obvious that in terms of signal strength the results are almost identical. The only features in which the N class router has clear advantage are the transfer rates.

The other obvious result is that the TP-Link wireless adapter has about 1.5 times better reception compared to the Wi-Pi.

    The results

In order to have a bit more objective results I took at least three measurements in every spot and made an average value out of them. So how the overall picture looks like? In the table below the overall results are summarized and put in order.

Distance (m)	1	1.5	5	7.5	8	10.5	12	13	15
Benchmark	69	70	58	47	52	37	43	33	50	Link Quality (X/70)	Linksys router
Wi-Pi	43	61	43	34	36	35	19	29	36
Tp-Link	59	59	52	44	41	45	35	32	33
Benchmark	-39	-31	-52	-63	-58	-73	-67	-77	-60	Signal Level (dBm)
Wi-Pi	-67	-49	-67	-76	-74	-75	-91	-81	-74
Tp-Link	-51	-51	-58	-66	-69	-65	-75	-78	-77
Benchmark	70	70	58	39	48	46	41	28	38	Link Quality (X/70)	TP-Link router
Wi-Pi	62	51	35	30	32	34	21	27	25
Tp-Link	64	50	47	26	52	40	34	37	33
Benchmark	-34	-40	-52	-71	-62	-64	-69	-82	-72	Signal Level (dBm)
Wi-Pi	-48	-59	-75	-80	-78	-76	-89	-84	-85
Tp-Link	-55	-60	-63	-84	-58	-42	-76	-73	-77

Some correlations can be found in there, but I'll stick to the obvious ones. The following charts could bring some visual clarity on the data interpretation and on how the different devices scale to each other.

In these two charts above, although there are some fluctuations, a general trends can be seen well:

• The laptop generally showed to be the correct choice for a benchmark - it gave best results in most of the cases.
• The Wi-Pi module proved my (undesired) hypothesis  - it gave worse results compared to the benchmark in all of the cases, and in almost all of the cases compared to the TP-Link wireless adapter.
• The graphs are not smooth curves. They all have drops at about the same places. This is due to the walls. For two points at the same distance, in the one behind a wall the reception will be weaker. That is how a very common and well known fact becomes visible in the data. So for the graphs to be smooth curves without the drops, on the X axis should be put some other dimension that is compound of the distance and the presence of obstacles. I didn't go in that direction, because although I can count my walls, I still don't have a tool that can measure their thickness. And at the moment I'm not quite sure what else would be involved in such a property. And I'm pretty sure someone has already invented it :)

The other dimension of the task at first surprised me with the nearly equal results for the both routers, but that also had to be expected in a way.

In the given distances (way below the routers' factory ranges) such similarity in the link quality and signal level should be expected. Despite that I was a bit surprised by the more consistent values the Linksys gave. Since the transfer rates are of no interest in this experiment I won't comment them. On the other hand charts that represent the Signal level measurement are also not given, not only for brevity,  but also because they look the same and the only difference with the above charts is the direction and the values of the Y axis.

    The verdict
Well, although the whole experiment may seem kind of scarce, to me it is enough to determine a direction for my following questions and activities. If I have to give at least a partial conclusion, it would be again in two dimensions:

1. In the case of routers comparison: Unless I've missed some specific tweaking while configuring the N class router, it didn't show any clear advantage in terms of signal strength over the G class one (results showed that the last one performs even more smoothly in the given short range).
2. About the task to determine a more capable wireless adapter: the hard number convinced me that for now I have to stick to adapters with bigger antennas, even though they would make a Raspberry Pi not so slick and tight in the looks.

So as a final verdict I'd say that my initial conclusion proved correct and the current line of production of the Wi-Pi wireless adapter is not as capable as I wished for.