You are here

equalize power levels

12 posts / 0 new
Last post
equalize power levels

Our 5 GHz network has three stations connected to a mountain station.

The mountain node "Charts" sees different levels roughly as follows:

  • Node1: -53-54 dBm TX MCS: 13-15 RX MCS: 13-15
  • Node2: -63-64 dBm TX MCS: 13-15 RX MCS: 12-13
  • Node3: -79-70 dBm TX MCS: 12-15 RX MCS: 11-13

It has been suggested to reduce the power levels of the stronger nodes (especially the strongest) to obtain more equal signal levels as seen from the mountain node and that would give "equal chance" for all stations as well as preventing the one station (the closest and most powerful one) from overpowering the mountain node.

Would lowering the power as suggested be recommended and any other comments regarding such differences in power?

By the way, we are all running MikroTik devices.

Thanks, Mike - AB4YY

nc8q's picture
channel: contention and availability

Hi, Mike:
It was not mentioned, but I am assuming that Node1,2,3 are hidden transmitters to each other.
If you reduce power on Node1 and Node2 to have a MCS equal to Node3, then
Node1 and Node2 may transmit data consuming a longer time due to a reduction in MCS.

This may reduce channel availability.
Equal signal strength helps to ensure that, in the event of a collision, no data is transferred.

A solution:
It is not really a 'mesh' if there are 3 hidden nodes.
Use 'Charts' as an Access Point and Nodes1,2,3 as clients thus letting the AP poll the clients one at a time.

I hope this helps,

None of the 3 nodes can hear

None of the 3 nodes can hear each other.

When you say "Charts" I take it you mean the mountain node?  And if so, and we would have to revert to non-AREDN firmware to do the AP/Client setup, right?

Thanks, Mike

nc8q's picture
When you say "Charts" I take

When you say "Charts" I take it you mean the mountain node?
I was following your:  'The mountain node "Charts"'.

And if so, and we would have to revert to non-AREDN firmware to do the AP/Client setup, right?
AP/Client excels in Point-to-Multi-point networking wherein the clients are likely to be hidden from each other and the AP is likely to be exposed.

Another solution is to have separate point-to-point downlinks for each client.
i.e. a LHG-[HP]5[-XL] ($60-$110) for each client fed with one 30 watt POE ($16) and an Ubiquiti N-SW ($45).
The black POE will work with the N-SW.
If you need LAN services at 'Charts' a VLAN capable node-switch must be added (hAP? $50).

I sent you a PM.
I hope this helps,

Thanks Chuck.  Good info.

Thanks Chuck.  Good info.

We've toyed with the idea of going AP/Client but so far haven't done anything.  Maybe it's time for us to discuss again, especially for the 5 GHz part.

- Mike

K5DLQ's picture
i would start dropping tx
i would start dropping tx power until you see a decline in mcs rates.
I understand that but it it

I understand that but it it the normal way of fine-tuning a network and best practice?

Thanks, Mike

AE6XE's picture
Strong SNR neighbor

Strong SNR neighbor

It should not be a problem if one neighbor has a stronger SNR than others, in most situations.  The way to determine if the stronger SNR neighbor is an issue, is to review the /tmp/rssi.log file on the tower node.    Where there has been a problem historically, it is due to the fact that some neighbors have in contrast a ~weak SNR.  What happened, is that neighbors with low SNR, sufficient to do some communications, can be filtered out, or the tower node becomes deaf to that neighbor.   This is in the logic of the receiver trying to filter out noise or interference for the strong SNR neighbor.    A program was written to detect if a neighbor node was being filtered out, and the radio was reset to start communicating with the node again.   By inspection of this /tmp/rssi.log file, you can see if this is happening.  If the node is not becoming deaf to some weak signal neighbors, then no reason to reduce the performance of the strong signal node.

[after comment]  There is cause to reduce power, and part 97 relevant rules mandating lower power, if the node will continue to maintain same performance or if lower performance is still adequate for the usage. 


Thanks Joe.

Thanks Joe.

For our small 5 GHz network there doesn't seem to be an issue as far as I can tell looking at rssi.log.

To be clear, the radios (all Mikrotiks) should have no problem with with varied levels and that (exception for the part 97 power rules you mention) in this example the tower node sees nodes at SNRs of 39, 31 and 26?  That is, the tower radio's AGC would not be thrown off due to the relative great differences in received signal strengths?  (And there would be no need to tweak power to make the three look similar?)

Regarding your note on mandating lower power... I fully understand.  I have found that lowering from the rated power immediately starts to lower the iperf throughput.  I may do more tests along this line.

- Mike

nc8q's picture
Is AGC present?

That is, the tower radio's AGC would not be thrown off due to the relative great differences in received signal strengths?


I think that these radios do not have an AGC circuit.
I have operated them in extremely close proximity and on adjacent channels without adverse affect.

I suspect that the largest issues that will affect your throughput is the hidden transmitters and the exposed receiver.

  • Nodes 1,2,3 will transmit whenever 'Charts' is not.
    Especially even when 1 or more of the other Nodes are transmitting.
    Easily 'stepping on top of each other' at 'Charts' reception.
  • 'Charts' will not transmit as long as it hears NODES 1 or 2 or 3 transmitting.
    This may cause latency in initiating 'Charts' transmissions.

 IOW, If you have 1 exposed node and 3 hidden transmitters, you may not have a 'mesh' network.
It may be futile to adjust settings on 'mesh firmware' to compensate for a non-mesh environment.

In opposition to using the minimum power to establish communications,
being a good neighbor and transmitting less often is effected by using the fastest Modulation Coding Scheme.
IOW, the faster the MCS, the more often the channel is available to others.

$0.02, Chuck

AE6XE's picture
misc comments

Some misc details, maybe more than wanted, about how these radios work.

Here's a typical description of how the radio receiver works (from the AR9331 data sheet example).  Most of the smarts is done digitally.  Note, you're probably always be using 802.11n with OFDM modulation (not a spread spectrum modulation technique).  CCK referenced below is an older modulation technique from 802.11b  (a spread spectrum modulation), generally obsolete and not out there today (unless you're still running an old linksys WTG door stop :) ). 

5.1 Receiver (Rx) Block

The receiver converts an RF signal (with 20 MHz or 40 MHz bandwidth) to baseband I and Q outputs. The receiver operates in the 2.4 GHz bands to support CCK and OFDM signals for 802.11b, 802.11g, and 802.11n.The 2.4 GHz receiver implements a direct conversion architecture.

The receiver consists of low noise amplifiers (LNA1 and LNA2), diversity combiner inphase (I) and quadrature (Q) radio frequency mixers, and a baseband programmable gain amplifier (PGA). The mixer converts the output of the on-chip LNA to baseband I and Q signals. The I and Q signals are low-pass filtered and amplified by a baseband programmable gain filter controlled by digital logic. The baseband signals are sent to the ADC.

The DC offset of the receive chain is reduced using multiple DACs controlled by the MAC/ Baseband processor. Additionally, the receive chain can be digitally powered down to conserve power.

[after edit -- to correct that digital sampling occurs in the time domain and then it is converted to a frequency domain]
Here's some of the smarts that can be done, from the wireless driver code comments.  This is an example of how receiving the same signal is combined from the 2 vertical and horizontal antennas after converting the signal to digital format and from a time representation to a frequency representation.  For 802.11n, this is MCS0-MCS7  (MCS8-MCS15 sends different data on the 2 polarizations, so does not apply):

 * MRC (Maximal Ratio Combining) has always been used with multi-antenna ofdm.
 * With OFDM for single stream you just add up all antenna inputs, you're
 * only interested in what you get after FFT. Signal alignment is also not
 * required for OFDM because any phase difference adds up in the frequency
 * domain.
 * MRC requires extra work for use with CCK. You need to align the antenna
 * signals from the different antenna before you can add the signals together.
 * You need alignment of signals as CCK is in time domain, so addition can cancel
 * your signal completely if phase is 180 degrees (think of adding sine waves).
 * You also need to remove noise before the addition and this is where ANI
 * MRC CCK comes into play. One of the antenna inputs may be stronger but
 * lower SNR, so just adding after alignment can be dangerous.
 * Regardless of alignment in time, the antenna signals add constructively after
 * FFT and improve your reception. For more information:

The linux wireless ath9k 802.11n implementation in use on AREDN firmware supports RTS/CTS for hidden nodes.   This means that a node will send a Request to transmit to the tower node, which then in turn sends out the Clear to Send to everyone with the coordinated timeslot.  For a handful of nodes on the channel, this is negligible overhead and adequately handles the coordination.    if you install the tcpdump package on the tower node, you can capture this traffic to a data file, then see this handshaking on your desktop using wireshark.  I found it very interesting to see how this handshaking works in reality.

Note, exposed node situation is best to avoid -- don't put 2 radios at the tower site on the same channel, rather use 2 different channels and dtdlink the 2 nodes together.    Also, avoid this in the shack -- don't put a radio on the roof pointing to the tower, and another radio in the office-shack, both on the same channel.


Thanks for the good input

Thanks for the good input Chuck and Joe.

Lots of good info and reference material.  :)

73 - Mike

Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer