Hey LightGuy,
Maybe I can shed a bit of light on your situation.
The mechanisms in any broadband radio to measure signal to noise ratio are slower that some types of noise and interference that affects the data flow in any wireless system. For example, noise caused by other data systems is often bursty, meaning that most of the time, the noise is not there (when you are looking at a signal millisecond by millisecond). RSSI, and signal to noise ratio are measured during a scan…that is not continuously, but samples of the received signal are taken, and statistics calculated on that sample. To avoid disturbing the received signal, the sample time is short so the receiver can spend most of its time being a receiver and looking for signal to decode.
Secondly, to decode a stream of symbols that have been transmitted by the remote transmitter, that symbol needs to have a high enough signal to noise ratio to properly decode the symbol. QPSK needs 8 dB SNR, QAM 16 needs 13 dB, QAM 64 needs 23 dB, QAM 256 needs 28 dB, QAM 1024 needs 34 dB for most receivers. Errors occur when the momentary SNR is not higher than these values. For example, if the QAM 256 modulation has only 20 dB SNR, some of the bits in the symbol (the symbol represents 8 bits) will be errored. If the forward error correction cannot determine what the errored bits should be, then the symbol needs to be retransmitted. Symbols are always transmitted in a group (a resource block). If the FEC of the resource block cannot be determined, the resource block is re-transmitted. That is when you see errors in the statistics.
By forcing the modulation level lower, you are reducing the signal to noise ratio required to demodulate the frame. In your case, you have about 38 dB effective SNR…you do not have the 40+ dB to decode the QAM 1024 modulated symbols.
So how does this all come together? The AP has received a low modulation set of frames from the SM when the SM joined the sector. From the registration exchange, the AP knows the signal quality from the SM. Then, as traffic flows to and from the SM, when the AP transmits a frame to the SM, the SM responds with a signal quality or cell quality value. The scheduler in the AP uses that response to set the DL modulation level to the SM. When the SM requests UL bandwidth, the AP determines the modulation level that the SM should transmit the UL data. Normally, this multi-rate algorithm can handle signal fading, and most kinds of interference, and adjust the MCS (modulation coding scheme) on the fly, adjusting to allow the highest error free data rate in both DL and UL.
Where the MR algorithm breaks down is when the noise from an interferer is very short in duration, and very high in level. For example, we have a tower with 4 sectors. We have SM’s scattered all around the tower, connected to each of the sectors. Let’s say we are using Reuse = 2, meaning that east and west uses F1, and north and south uses F2. All SM’s facing south and north use the same transmit frequency. If we are receiving UL from a north facing SM (that is connected to the south facing sector), and a south facing SM connected to the north facing sector transmits at the very same moment, the south facing sector will receive the desired SM’s signal plus the signal from the SM on the opposite sector at the same time, momentarily raising the noise floor of the south AP. If that signal raises the noise floor by 6 dB, that may be enough that the desired SM’s data may be errored for that moment. A retransmission is called for, and a random timer changes when the frame is transmitted, and perhaps the multi-rate algorithm reduces the MCS for the retransmission, and the second attempt is successful.
I apologize for the long-winded explanation.
Setting the MCS level manually might stop the error counters, since there is no error. But, setting the MCS artificially low with also limit the capacity of the sector. With the MCS set to highest rate, and the multi-rate mechanism working, the only way to know what the effect of the errors (specifically retransmissions) are is a throughput test. I love the fact that Cambium builds the link test into the system. It is your best troubleshooting tool to see what the effect of any momentary interference might be. This momentary interference is called noise rise. Generally, the AP’s receiver (the uplink) suffers the most from it as the AP’s antennas are high in the air and can receive more interference from nearby devices than the SM’s receiver, which is generally lower in height, and has a more directional antenna.
If anyone wants to discuss this further, let’s keep this going. I hope this helps explain what you are observing.