I did a recent blog post on WiFi performance based on power. I was going to post the full post here but it doesn't allow images correctly. Tom, maybe Im doing something wrong ? Anyway in the post I share how excessive power degrades the radio performance. Keep that in mind when designing and tuning your network especially more so with PtP.
Thank you for the post. Could you tell us a little more why the link isn’t able to reach 256-QAM with the higher power levels?
You’re saying noise is increased, but the underlying noise should be constant as its not affected by the increased power levels? Could there be any ISI problems due to a much more powerful reflected multipath signal? Would you experience the same without SGI? And why does it say 1024-QAM? Is this not 802.11ac at all?
Jan Vidar Rognsvåg
I'd like to see more detail and variation in the data.
The difference is obvious, but without more explanation it is hard to make any conclusions as to why this is happening.
Some details on the particular 1024 QAM specification would also help. And might the two modulations have different restrictions on their power levels by default ?
First the 1024 QAM is specific to Ubiquiti radios.
Secondly as we backed away the bridges at different power levels and at different points physically we seen higher connected rates.
What you see here is distortion of the signal. This isn't the first time, Ive seen this with other bridges as well. Im benching Cisco 1530s now actually. Same thing ..
BTW here is the spec sheet. http://dl.ubnt.com/datasheets/airfiber/airFiber_DS.pdf
You can see we should have been able to hit 1024 QAM in all 3 radio power setting test.
I think terming it "front end overload or receiver overload" might be a more accurate description, than "too much power". Too much signal will overwhelm the receiver and cause receiver distortion. But that's because you are too close. What I suggest is to do a test at further distance (like 1000 feet) and then you'll see if it was too much TX power or it is too much signal.
From my experience BOTH are true. With Ubiquiti and EnGenius you'll find in their datasheets, that MCS rates are not maintained at the highest power level settings. This is from a transmit (PA) limitation (linearity cannot be maintained at higher modulation complexities at higher power).
Here is a link to an example datasheet of an EnGenius ENH500. Notice full MCS rates are only up to 23dBm even though the device can do up to 28dBm. https://www.engeniustech.com/images/stories/ENH500_Datasheet_Medium_Resolution.pdf
The same is true with receivers. Receivers have a "dynamic range". Dynamic range is the minimum and maximum signal range the receiver can handle. There typically is an AGC circuit (automatic gain control) that will back of receiver gain when the signal is very strong, but it has a limit. I've seen for most lower end to medium end devices, a -30dBm or stronger will start to introduce some front end overload (i.e. too strong) and your throughputs will actually go DOWN!
I tell installers, for very short links (a few hundred feet or less) to turn down the power, or you can purposely mis-align the devices a little, if needed, to get signal down a bit, so signal is not stronger than -40dBm.
Thanks for the feedback Dan!
I test radios at the PHY level, and can almost guarantee that the Dynamic Range of any .11ac radio is better than what older .11b/g/n versions exhibited - some by as much as 9 dB at the .11b rates.
There was not much improvement with .11n over .11a/b/g, even at the low .11b rates. Not true with current 802.11ac radios. Compared to .11n radios, the dynamic range for the .11b/g rates of the /ac radios is very impressive.
Maximum received power levels greater than -10 dBm are now common place. This is the upper limit in the standard, but every radio I have tested in the last year has met this with a healthy margin - which markedly contributes to the Dynamic Range measurements.
Newer radios are touted for higher rates and throughput, but it is good to know that they can deliver even better reliabilty than they could previously - when the lower rates "kick" in.
I concur with Dan. The spectral overgrowth increases considerably with higher rates, indicative of poorer linearity. This often happens when non-sensical marketing types get their paws into the product, and enginners who know better are too timid to hold their ground.
The issue certainly is radio distortion becusee of excess signal at higher power levels. I've seen this issue happen on other close ptp links with very high gain antennas. The purpose of the article is to get folks thinking more power, more signal isn't always a good thing.
I agree that more power is not always the answer, but unless the range is extraordinarily short, I have rarely found that receiver overload is the problem. Too much power is definetly not a good neighbor policy, but the more likely conclusion in this case is that this is a transmitter issue.
This is consistent with the reduced power levels specified for these radios at the higher rates. There is nothing "wrong" in this behavior. It should be expected with higher rates and more complex/dense QAM schemes. The manufacturers are probably just pushing the power levels upwards to compete in the "specsmanship" arena, when in fact they realize that they should be used at lower levels.
Increasing the output power of a PA beyond linearity creates spectral overgrowth (which is briefly described in the CWDP study guide). This is in turn increases the transmitters EVM levels which need to be lower for more complex modulations.
Overgrowth is obvious in Spectral Mask measurement plots and the loss in linearity is often visible in EVM I/Q diagrams, where the data points are "pulled-in" towards the middle of the diagram. As a visual effect, this can be a very subtle, and hard to discern.
Rarely would a system manufacturer advertise these plots - and I can't measure 1024 QAM signals. But they would be interesting to see.
We know that with 1024 QAM there is almost no room between symbols, and adding overgrowth would muddy a received signal even more. Add just a small amount of interference and you'd have an unrecognizable signal at the receiver.
Every directional antenna, including a parabolic dish, has some number of sidelobes and I would expect they produce a measurable amount of multi-path interference at close ranges.
My take away is to remember George's original premise and not use too much power, and to realize that high output power levels can be more "marketing speak" than a useful number for those people configuring PTP networks.