iPad 3 Wi-Fi Performance Analysis – Throughput
Part 2 of 3: iPad 3 Wi-Fi Performance Analysis – Throughput Analysis (Simple Statistics)
In the previous post, iPad 3 Wi-Fi Reception Analysis – Transmit Power, we discussed the raw Tx power between iPads and various other devices. In those results it looked like the iPad 1 trumped the iPad 2, but the iPad 3 was stronger in both the 2.4GHz and 5GHz frequencies.
This post will cover throughput testing, both for Tx – uplink from iPad to a server, and Rx the downlink from the server to the iPad.
In order to make this as scientific as possible – given it wasn’t conducted in a Faraday Cage – all variables were held constant, and only the devices were changed. They all used the same testing software, all devices held in the same position, at the same angle, at the same location. Tests were rotated between devices as to mitigate potential external interference issues. Over 600 data points were collected to get these results.
These results are being shown with “standard” statistical measurements, things you’ve been familiar with since early childhood. Things like Average, Maximum, Minimum, and Range. I’ve also added a Standard Deviation.
This metric shows where the bulk of the collected data points fall. It can be used as a simple way of describing how ‘consistent’ the data is. The smaller the Standard Deviation, the tighter the results are to the average. This can show variability in the results in a numerical way.
Again, we are doing testing in both 2.4GHz and 5GHz frequencies what would be used in an Enterprise Wireless Network.
The remainder of this post will be heavy with graphics – I opted to go this visual route rather than bore you with the tabular data. I think it’s quite a bit easier to digest in graphical format.
First the 2.4GHz statistical results. See how even each of these are. The averages are very close. Also note how consistent the MacBookAir results are – very little variation between the Max, Min and Average.
Now on to the Rx statistics in 2.4GHz. Again, just like in the Tx area, the averages are very close.
And finally for the 2.4GHz – the standard deviations. The MacBookAir with its multiple spatial streams has the lowest standard deviations, and thus the highest level of client device consistency. But all of these aren’t too bad. Don’t be worried about the long bars – look at the scale and see there isn’t actually a lot of variation.
For example, the longest bar is for the iPad 3 at 1.7. This means that 68% of all data collected fell between Plus or Minus 1.7 around the Average. In this case that would be 68% of all results were between 12.3 and 15.7.
So the results for 2.4GHz were fairly consistent in actual data throughput.
Now on to the 5GHz results. Much more variability here than with the 2.4GHz. The averages between different iPads are fairly substantial. You can also see a stair-step result – each generation of iPad is better than the previous one. (Good news for those of you trying to convince your boss or significant other you should get a new iPad!)
The same holds true in the Rx statistics. Same stair-step increases in Rx throughput as you move from iPad 1 to iPad 3. With the multiple spatial streams of the MacBookAir with the most stable results.
Finally for the 5GHz Standard Deviations. Not only are there larger differences in average Tx and average Rx for each of these devices in 5GHz, but their consistency is also much worse, see the higher standard deviations in this frequency.
In addition to the first post’s results showing raw Tx power – here we can see the iPad 2’s weakness by -2dB compared to the iPad 1 is trumped by actually having better throughput.
Remember, it’s all about the throughput. Just having a stronger RSSI isn’t as important as delivery of real data across the link.
The 2.4GHz band is fairly consistent across the various models, but in the 5GHz band you can see the newer devices out perform the older ones.
In our final post of this series, I’ll introduce you to a different, more graphical approach to consistency statistics that will make it much easier to understand and visualize stability to client device throughput. Stay tuned!