You are here

Link Quality

9 posts / 0 new
Last post
Link Quality
The AIROS has settings for max speed and for alternate algorithm for setting the speed.

Is there any way to get to these config settings under AREDN?

I want to improve Link Quality at the expense of speed to improve VOIP quality on some links.

k1ky's picture
And more bandwidth options?
And for the sake of discussion, would it be possible to have more bandwidth selection options both higher than 20 Mhz and lower then 5Mhz?
The current system (minstrel)
The current system (minstrel) already takes quality into account. It calculates the fastest throughout calculating in lost and retransmitted packets are considered a "decrease" so it maximizes the link to maximum throughout (or goodput)

On WIFI networks (like MESH) slower speeds can in many cases actually DECREASE quality.

A slower packet while it has a little more power (depends on the speed and where it places on the power chart) also has a significantly higher probability of being corrupted in transit due to longer air time.

In addition longer air time decreases the quality of all other devices that can hear the traffic being sent.

You actually really want to get the traffic out as quick as you can, the faster it clears the air the faster the packet is delivered. This is especially true on VOIP (REF: Jitter)
AE6XE's picture
Here's hwo the AREDN firmware
Here's hwo the AREDN firmware determines what rate to use.  A table is created to determine what rate to transmit to each unique neighbor (not one table or necessiarly one max limit setting to have).  

type         rate     throughput  ewma prob   this prob  retry   this succ/attempt   success    attempts
​HT20/LGI     MCS11          30.3      100.0       100.0      0              0(  0)         1           1
HT20/LGI  t  MCS12          38.7       99.8       100.0      6              0(  0)      2045        2078
HT20/LGI T   MCS13          44.0       99.5       100.0      6              1(  1)      1345        1477
HT20/LGI     MCS14          30.5       58.7       100.0      6              0(  0)       487         625

the "ewma prob" is a measured and weighted average (recent traffic) of packet success transmitted to the neighbor.  The OLSR LQ/NLQ is less accurate to forecast how VOIP traffic is going to perform.  In this live example from my mesh node, MCS13 yields the highest throughput and is 99.5% probability of success in transmitting a packet.   If we arbitrarily dropped the rate to MCS11,  then while we get to 100% success, we've unnecessarily limited the data that can be transmitted by ~30%.  

Note, the difference between MCS13 and MCS12 is going from QAM64 (with 33% correction bits or non useful data) to QAM16 (with 25% correction bits).  MCS11 is QAM16 (with 50% correction bits).

The way the rate is selected in AREDN should be the optimum to use for voip and other applications both UDP and TCP.   I've not seen any instances where the error rate is sufficiently low, that there'd be a latency issue with voip given the retries.   When I see voip not working, it's generally because there's not sufficient SNR to begin with for the link--the rate in use is already at the bottom.


Thanks for all the info, I figured out how to monitor the Minstrel rate/LQ status on a node.
Does AIROS use a different rate setting algorithm that allows for some tweaking? (and Minstrel does not)
I believe AIROS just does a
I believe AIROS just does a swap out to some of the other quality algorithms to do the extra tweaking along with removing supported nodes from the broadcast data of the AP. What your actually doing with the supported speed is crippling the hardware (tell it not to advertise that it's capable of faster speeds) instead of tweaking the algorithm.

Really these features won't be used too often even in the AirOS world but have been made available because AirOS a is intended to be used by wifi network engineers. This is in contrast AREDN mission to provide a simple solution that is compatible with each other. Many of these AirOs tweaking setting (but not all) can actually create incompatibilities or differences in behavior. When you have a trained network team that controls every single piece of hardware this is less of an issue, when you have field devices coming in however it all can stack up and create weird issues in the real world.
k1ky's picture
40 Mhz Bandwidth
Thanks Joe, I read the ticket - very good stuff!  Can you help me out with the definition of 40 Mhz MIMO as used by regular Wi-Fi Routers?
My Cradlepoint and many of the commercially supplied Internet modem/routers supplied by Comcast and AT&T have a 40 Mhz mode that "appears" to show up on 2 channels when performing a Wi-Fi Scan.  Obviously this takes up 6 channel "spaces" of spectrum.  Is this what the AREDN impementation would look like also?  I've seen the MIMO term used interchangably when I see router specs - like the Cradlepoint MBR series whcih still uses "Vertical" antennas.  I assume that they really aren't doing dual polarity as we are doing with the Nanostations, Rockets and Nanobridges.  More like diversity antennas in order to reduce multipath issues? So, in effect if we choose 40 Mhz, do we  we actually get 4 independent data paths or just 2 with a wider channel occupation to support more bandwidth?

Inquiring mind wants to know!
It is generally recommended
It is generally recommended not to change the subject of a post created by someone else as it makes it harder for others to find the same information in the future (who would expect a discussion on 40MHz channels inside a thread on reducing speed ?) 
AE6XE's picture
K1KY,  there's a wealth of
K1KY,  there's a wealth of knowledge in the openwrt documentation.  Take a look at .   Specifically the 'htmode' parameter.

In 802.11n, the 40MHz channels 'pairs' 2 x 20MHz channels.   This was an easy engineering design to stay backward compatible with devices, but means the channel # is not at the center of the bandwidth in use.  Rather, continue to select the channel of a 20MHz signal, then there's a setting to pair the "+" or "-", above or below channel.   Caution and stay away from band edges as this would not have been tested, we'd expect, e.g. in 3GHz, there to be signal distortion with power drop-off going another 20MHz outside current edges even if still in part 97. 

While it's trivial to make a couple character change for this setting, this is new territory,  issues will come up, e.g. what channels should not be selected, does it work in the part97 extended channels, etc.?   If you are testing this, please post your experience in the ticket so the issues are captured and can be addressed.  KE2N and I were testing this a while back and had some success.

Note, from an information theory and RF perspective, for the same amount of power in a perfect channel, the same data rates are achieved regardless if using 20MHz or 40MHz channel width.   There's 2x energy per Hz to achieve 2x higher rate when using the half size channel.   I suspect that 40MHz channel will only improve rates when there is low noise at shorter distances--for the SNR needed to beat out the 20MHz channel.  The larger size channel yields more opportunity to have noise and not perform as well as a half-bandwidth channel.   While transmit power stays the same and stretches across the bandwidth in use, the noise power is increasing as the channel bandwidth increases.


Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer