Very slow wifi-to-wifi speed on radio1 with 40 MHz

Turris Omnia
1

Client A: Atheros AR9485WB-EG, distance to router: 0.5m
Client B: Atheros AR9485, distance to router: 4m

Both running Win10 1607 with the same driver from Microsoft version 3.0.2.201, date 27.02.2016

TEST 1

radio0: 2.4 GHz, 40 MHz, 15 dBm, ch13
radio1: 2.4 GHz, 40 MHz, 15 dBm, ch13

radio0
A to B: 5 MB/s
B to A: 6 MB/s

radio1
A to B: 0.5 MB/s (!) … this starts happening after transfering some data back and forth
B to A: 4 MB/s

TEST 2

radio0: 2.4 GHz, 20 MHz, 15 dBm, ch13
radio1: 2.4 GHz, 20 MHz, 15 dBm, ch13

radio0
A to B: 2.5 MB/s
B to A: 2.5 MB/s

radio1
A to B: 1.5 MB/s
B to A: 1.5 MB/s

TEST 3

radio0: 2.4 GHz, 40 MHz, 15 dBm, ch1
radio1: 2.4 GHz, 40 MHz, 15 dBm, ch1

radio0
A to B: 3 MB/s
B to A: 5.5 MB/s

radio1
A to B: 0.5 MB/s (!)
B to A: 3 MB/s

2

amq,
To start, I would recommend using drivers from Atheros rather than Microsoft drivers. Atheros cards can be very picky with drivers, one small version change can impact performance dramatically. This goes for any wireless card, make sure your using the latest/recommended driver.

What application are you using to transfer data? Do you have any test results based on another wireless router or AP utilizing the same test setup? Are you utilizing channels that aren’t interfered/over-subscribed?

3

The driver is from Atheros (as seen in device properties), just supplied by Windows Update.

Well, yes. In the first post I have a comparison with radio0 under the same conditions.

Considering the above, I don’t understand how is this relevant.

4

What NYtimebomb is saying is that there might be neighboring stations which might cause interference. If you don’t live in a remote area this is probably the cause of slow transfer speeds. Other causes might be, placement of the antenna on the client and/or router, crappy drivers or hardware don’t liking each other.
Last one sounds kinda strange but I know from personal experience that some draft-n cards don’t behave that well when trying to talk to a router. Please note that I am not implying that you have a draft-n card, I am just stating my experience with this.

5

Please install the latest driver from atheros-drivers.com for your specific chipsets. I would suggest not relying on Windows Update for the latest drivers. I believe the latest driver relevant to you would be version 10.0.0.341

I specifically meant an entirely different router or AP, not the Turris Omnia.

As iddqd mentioned, if you are in an urban area with many wireless networks nearby there may be interference caused by other devices on the same frequencies.

I’d encourage you to check out the below post, as I specifically did the tests to try to emulate your issue. Alas I do not have access to the exact configuration of your systems, but I think you will find it informative.
https://discourse.labs.nic.cz/t/wifi-radio-test/1056/1

6

I was simply copying a large file in Explorer over a network share.

7

I would encourage you to explore other ways of testing throughput. Good luck!

8

Aside from (valid) considerations regarding drivers and test software, don’t forget the physics!

WiFi works with OFDM modulation, where throughput is basically determined by signal-to-noise ratio (SNR). ETSI and FCC regulation limits the transmit power to a certain value, 100mW EIRP in 2.4GHz. Now, that power is spread out over the channel width you use. If you take twice the channel width, you spread your power over twice as much, so you only have half as much signal for your power. On a logarithmic scale, that means you lose 3dB.

Now, at good signal levels, where you have more than enough SNR, those 3dB don’t really matter, so the benefit you get from doubling channel width more than outweighs the lower signal level. That’s why you get twice the throughput with radio 0 at 40MHz vs 20MHz.

But at bad signal levels, things are different. You need every bit of SNR you can get, so the 3dB hit from doubling the channel width really hurts. Worse, because you’re covering twice as much spectrum, you have twice the chance of picking up interference, so unless you are in a squeaky clean RF environment the penalty from lower signal level is far more significant than the benefit from doubling the channel width.

As for the asymmetry between the two devices, that is probably due to the difference in distance and so RSSI between the two, and the fact that you are testing unidirectionally. Assuming there is more attenuation to client B, located further away, it looks like receive sensitivity is (as usual) the limiting factor, so where B is receiving, throughput is lower - drastically so with 40MHz.

Best practice is to only use 20MHz channel width for 2.4GHz; for applications requiring higher bandwidth, go to 5GHz.

2 Likes