04/27/2015
AF5X – Why you owe it to yourself to use these radios for backhaul everywhere
Description

I've previously posted stories comparing the AF5X and other radios (both UBNT and from other manufacturers) showing how much better the AF5X is in throughput and in adjacent channel interference rejection. I will concentrate on another facet of the AF5X here - it's ability to drastically improve the noise floor and therefore the overall performance of the sites where you use them.

 

The AF5X is a purpose built radio rather than the majority of radios available today which utilize commodity WiFi based chipsets originally designed for indoor use in firewalls and the like. Until the advent of the AirFiber line of radios Ubiquiti also based their radios on Wi-Fi chipsets. With careful design, these can be capable radios. However they are not designed to have the same capacity, range, interference rejection and low noise generation characteristics as the AF5X. The test results shown here bear this out.

 

Test Settup

 

Block Diagram.png

 

This is the test setup that was used to gather the following data. The radios being tested were connected together using directional couplers and attenuators, emulating the Free Space Loss + antenna gain of a working radio channel and allowing the extraction of the signal for analysis in a controlled manner. Traffic was generated using iperf on outboard computers capable of loading the entire throughput capabilities of the radios. Each radio was further housed inside a shielded box to prevent any emissions radiating from the radios themselves from affecting the test. Two different radios were tested here: an AF5X and a dual-channel capable WiFi chipset based unit. The radios were observed with an Agilent N9010A Spectrum Analyzer to see what their respective output signals looked like under a range of circumstances.

 

Measurements

 

SA4.JPG

 

This is a picture of the actual display of the Spec Analyzer shows the output of the AF5X on a 20 Mhz wide channel. Note the flatness of the noise floor and the lack of sidebands or spurs (signals appearing outside of the desired channel itself). Also note the narrow sideband skirts and how far down in signal strength they are compared to the main carrier signal. Another thing to understand is that the limiting factor in this measurement at the noise floor is the analyzer itself – the AF5X is actually quieter than the analyzer. The analyzer is limiting the measurement here; the output of the AF5X is so good that it's legal to use antennas up to 35 or 36 dBi with it.

 

SAb5nrw.JPG

 

This is the same test of the WiFi chipset radio on the same channel and bandwidth. Note the spurs and raised noise floor – a product of the less expensive parts used in this radio. The big problem here is that this additional noise is being generated by every WiFi chipset radio you use, and that raises the noise floor for the entire site over the entire band at 5GHz.

Comparison4.png

This is a composite plot (computer captured from the analyzer) showing both the AF5X and the WiFi radio output together, with the noise floor in green. Note the addition of sidebands in the WiFi radio and the raised noise floor as well. This actually feels like a completely unfair comparison since the AF5X so totally outperforms the WiFi chipset radio.

 

All wideband radios (everything we use today) generate their RF signals at a low power level, and then combine the chains and amplify the result in a block using a wideband linear amplifier. The Power Amplifier (PA) needs to be able to boost the signal from less than 0dB (often a lot less) up to +20 dBm or more to adequately feed the radio antenna. And it needs to do this without distorting all the individual OFDM carriers plus not introducing phase changes or non-linearities in the QAM modulation underlying the entire system. And this is not easy to do, especially in low-cost radios.

 

Why is this important? Several reasons, but the most pertinent are to allow packing more data into the given channel, to introduce less noise both in the signal being sent out and in the surrounding area (you do use more than a single radio at most sites, right?) and to meet FCC and other regulatory agency criteria for Out-Of-Band Emissions (OOBE) and the individual Spectral Mask for the channel itself. If this isn't done right, the radio may perform badly, nearby radios can be negatively impacted, and you may be running illegally as well.

 

So where is all this noise coming from, and why isn't it a problem when you use these chipsets in other situations? Most of the noise comes from what is called Phase Noise in the oscillator which drives the entire signal chain in these radios, and that appears to be not too good in the WiFi radios. In a short-distance indoor environment the carrier is still 30 dB or more higher than the noise, so things work fine. But in an outdoor site, the noise will simply pollute the local environment for all the radios there. This is the one serious advantage of GPS Sync for Wi-Fi chipset radios – when all the APs are transmitting in sync, the noise simply isn't there when they are all receiving. Using AF5Xs as radios at sites bypasses all this since the AF5X just doesn't produce the noise in the first place. Also the AF5X's ability to reject adjacent channel noise from other radios coupled with it's built in GPS sync allows channel reuse at a site and channel selection to pack them in much closer together than other radios can do. Using AF5Xs at congested sites allows for more efficient channel planning and lowers the total noise at the site, while dramatically increasing the backhaul capacity on the same sized channel.

 

Dual channel WiFi chipset mode.

 

Recently a radio has hit the market that uses a single chipset to produce 2 different MIMO channels at the same time. It is using a commodity WiFi AC based chipset, and while the dual-channel capability in a single radio isn't in-and-of-itself a terrible idea, unfortunately their implementation has some major flaws in it.

 

What they did was to take a 4x4 MIMO chipset and operate it as 2 2x2 MIMO radios on different RF channels. While not as spectrally clean as the AF5X when operating in single channel mode, it really has problems when operating in it's dual channel mode.

 

Since the WiFi radio being tested was designed to allow this dual channel mode, here's a view of that mode of operation

Comparison15a.png

.

This plot is comparing the AF5X to the WiFi radio in dual channel mode. The large spurs you see to the left and right of the main carriers are the IM products being generated in the PA by the two channels, and the noise generated by those products and the PA. Note that these products and noise blanket not just the 5GHz band but go all the way down to 4580 MHz and up to 6430 MHz, impacting licensed bands completely outside of the UNII spectrum. In fact, these are so bad they aren't even legal in the US with any antenna gain above about 6-10dBi...

 

Now someone will say that this is not a fair comparison; that the WiFi radio channels are at the farthest ends of the UNII band, and at a very high conducted power which gives a worst-case result. But that's exactly the point – when you take this radio out of the box and plug it in, and don't change any other settings, leaving it in “auto-everything” mode, that's exactly what the radio does – picks the farthest ends of the UNII band to run at and cranks the power all the way up. How many of these radios will be run this way by people who don't know what they're doing, let alone those who don't care that they may be running illegally?

Comparison20a.png

This plot shows the results of the two channels being on 5560 and 5670 MHz producing a large spurious emission right in the middle of the UNII-3 band, compromising the use of that entire chunk of spectrum. This is clearly not a good thing.

 

 

 

Conclusions:

 

In general, using commodity WiFi components to build an outdoor high power MIMO radio, while it works, is perhaps not the best way to go. The AF5X, on the other hand, since it uses proprietary circuitry built for it from the ground up and high end components designed specifically for the task, has much better performance . Which you use comes down to how well you want your overall network to function and how many problems you want to create for yourself and everyone around you.

 

Below are some additional composite plots and screen pics.

 

Comparison15b.png

Comparison14a.png

SAbad.JPG

SA5Xgood2.JPG

 

 Edit -  adding an AirView scan of the dual channel 5560/5670 output - 

ScrenShotDualChanAirView.png

 It shows roughly the same thing, but it's also the difference between the hardware in the $400 radio, which is good, but not a precision analyzer, and the Agilent which is a $20,000 lab grade instrument.

 

AF5X – Why you owe it to yourself to use these radios for backhaul everywhere

by ‎04-27-2015 08:39 AM - edited ‎04-27-2015 02:11 PM

I've previously posted stories comparing the AF5X and other radios (both UBNT and from other manufacturers) showing how much better the AF5X is in throughput and in adjacent channel interference rejection. I will concentrate on another facet of the AF5X here - it's ability to drastically improve the noise floor and therefore the overall performance of the sites where you use them.

 

The AF5X is a purpose built radio rather than the majority of radios available today which utilize commodity WiFi based chipsets originally designed for indoor use in firewalls and the like. Until the advent of the AirFiber line of radios Ubiquiti also based their radios on Wi-Fi chipsets. With careful design, these can be capable radios. However they are not designed to have the same capacity, range, interference rejection and low noise generation characteristics as the AF5X. The test results shown here bear this out.

 

Test Settup

 

Block Diagram.png

 

This is the test setup that was used to gather the following data. The radios being tested were connected together using directional couplers and attenuators, emulating the Free Space Loss + antenna gain of a working radio channel and allowing the extraction of the signal for analysis in a controlled manner. Traffic was generated using iperf on outboard computers capable of loading the entire throughput capabilities of the radios. Each radio was further housed inside a shielded box to prevent any emissions radiating from the radios themselves from affecting the test. Two different radios were tested here: an AF5X and a dual-channel capable WiFi chipset based unit. The radios were observed with an Agilent N9010A Spectrum Analyzer to see what their respective output signals looked like under a range of circumstances.

 

Measurements

 

SA4.JPG

 

This is a picture of the actual display of the Spec Analyzer shows the output of the AF5X on a 20 Mhz wide channel. Note the flatness of the noise floor and the lack of sidebands or spurs (signals appearing outside of the desired channel itself). Also note the narrow sideband skirts and how far down in signal strength they are compared to the main carrier signal. Another thing to understand is that the limiting factor in this measurement at the noise floor is the analyzer itself – the AF5X is actually quieter than the analyzer. The analyzer is limiting the measurement here; the output of the AF5X is so good that it's legal to use antennas up to 35 or 36 dBi with it.

 

SAb5nrw.JPG

 

This is the same test of the WiFi chipset radio on the same channel and bandwidth. Note the spurs and raised noise floor – a product of the less expensive parts used in this radio. The big problem here is that this additional noise is being generated by every WiFi chipset radio you use, and that raises the noise floor for the entire site over the entire band at 5GHz.

Comparison4.png

This is a composite plot (computer captured from the analyzer) showing both the AF5X and the WiFi radio output together, with the noise floor in green. Note the addition of sidebands in the WiFi radio and the raised noise floor as well. This actually feels like a completely unfair comparison since the AF5X so totally outperforms the WiFi chipset radio.

 

All wideband radios (everything we use today) generate their RF signals at a low power level, and then combine the chains and amplify the result in a block using a wideband linear amplifier. The Power Amplifier (PA) needs to be able to boost the signal from less than 0dB (often a lot less) up to +20 dBm or more to adequately feed the radio antenna. And it needs to do this without distorting all the individual OFDM carriers plus not introducing phase changes or non-linearities in the QAM modulation underlying the entire system. And this is not easy to do, especially in low-cost radios.

 

Why is this important? Several reasons, but the most pertinent are to allow packing more data into the given channel, to introduce less noise both in the signal being sent out and in the surrounding area (you do use more than a single radio at most sites, right?) and to meet FCC and other regulatory agency criteria for Out-Of-Band Emissions (OOBE) and the individual Spectral Mask for the channel itself. If this isn't done right, the radio may perform badly, nearby radios can be negatively impacted, and you may be running illegally as well.

 

So where is all this noise coming from, and why isn't it a problem when you use these chipsets in other situations? Most of the noise comes from what is called Phase Noise in the oscillator which drives the entire signal chain in these radios, and that appears to be not too good in the WiFi radios. In a short-distance indoor environment the carrier is still 30 dB or more higher than the noise, so things work fine. But in an outdoor site, the noise will simply pollute the local environment for all the radios there. This is the one serious advantage of GPS Sync for Wi-Fi chipset radios – when all the APs are transmitting in sync, the noise simply isn't there when they are all receiving. Using AF5Xs as radios at sites bypasses all this since the AF5X just doesn't produce the noise in the first place. Also the AF5X's ability to reject adjacent channel noise from other radios coupled with it's built in GPS sync allows channel reuse at a site and channel selection to pack them in much closer together than other radios can do. Using AF5Xs at congested sites allows for more efficient channel planning and lowers the total noise at the site, while dramatically increasing the backhaul capacity on the same sized channel.

 

Dual channel WiFi chipset mode.

 

Recently a radio has hit the market that uses a single chipset to produce 2 different MIMO channels at the same time. It is using a commodity WiFi AC based chipset, and while the dual-channel capability in a single radio isn't in-and-of-itself a terrible idea, unfortunately their implementation has some major flaws in it.

 

What they did was to take a 4x4 MIMO chipset and operate it as 2 2x2 MIMO radios on different RF channels. While not as spectrally clean as the AF5X when operating in single channel mode, it really has problems when operating in it's dual channel mode.

 

Since the WiFi radio being tested was designed to allow this dual channel mode, here's a view of that mode of operation

Comparison15a.png

.

This plot is comparing the AF5X to the WiFi radio in dual channel mode. The large spurs you see to the left and right of the main carriers are the IM products being generated in the PA by the two channels, and the noise generated by those products and the PA. Note that these products and noise blanket not just the 5GHz band but go all the way down to 4580 MHz and up to 6430 MHz, impacting licensed bands completely outside of the UNII spectrum. In fact, these are so bad they aren't even legal in the US with any antenna gain above about 6-10dBi...

 

Now someone will say that this is not a fair comparison; that the WiFi radio channels are at the farthest ends of the UNII band, and at a very high conducted power which gives a worst-case result. But that's exactly the point – when you take this radio out of the box and plug it in, and don't change any other settings, leaving it in “auto-everything” mode, that's exactly what the radio does – picks the farthest ends of the UNII band to run at and cranks the power all the way up. How many of these radios will be run this way by people who don't know what they're doing, let alone those who don't care that they may be running illegally?

Comparison20a.png

This plot shows the results of the two channels being on 5560 and 5670 MHz producing a large spurious emission right in the middle of the UNII-3 band, compromising the use of that entire chunk of spectrum. This is clearly not a good thing.

 

 

 

Conclusions:

 

In general, using commodity WiFi components to build an outdoor high power MIMO radio, while it works, is perhaps not the best way to go. The AF5X, on the other hand, since it uses proprietary circuitry built for it from the ground up and high end components designed specifically for the task, has much better performance . Which you use comes down to how well you want your overall network to function and how many problems you want to create for yourself and everyone around you.

 

Below are some additional composite plots and screen pics.

 

Comparison15b.png

Comparison14a.png

SAbad.JPG

SA5Xgood2.JPG

 

 Edit -  adding an AirView scan of the dual channel 5560/5670 output - 

ScrenShotDualChanAirView.png

 It shows roughly the same thing, but it's also the difference between the hardware in the $400 radio, which is good, but not a precision analyzer, and the Agilent which is a $20,000 lab grade instrument.

 

" How can anyone trust Scientists? If new evidence comes along, they change their minds! " Politician's joke (sort of...)
"Humans are allergic to change..They love to say, ‘We’ve always done it this way.’ I try to fight that. "Admiral Grace Hopper, USN, Computer Scientist
":It's not Rocket Science! - Oh wait, Actually it is... "NASA bumper sticker
":The biggest problem in tech I see right now is that most users don't want to do things that are hard. That doesn't bode well for the industry or the society.": (me. actually ;-)
Comments
by
on ‎04-27-2015 09:07 AM

Very nice explaination Jim!!

You should be writing books about this stuff!

by
on ‎04-27-2015 09:55 AM

Mega KUDOS to you sir! I am pretty sure you have forgotten more about wireless that most of us have ever learned!

by
on ‎04-27-2015 11:32 AM

Great piece Jim

by
on ‎04-27-2015 11:34 AM

This shows some pretty strong evidence on OOBE emisions from cheap radios being a real problem, and purhaps why some gear is not getting some certifications as OOBE is part of the emissions and so counts against the radio's in-band.

 

so Jim, have you done this same test with a RocketAC-PTP/PTMP so we can see how the airPrism chip affects the OOBE?

by
on ‎04-27-2015 11:39 AM

Looks like AF team needs to get into ptmp Man Wink

 

Thanks for sharing this Jim.

by
on ‎04-27-2015 11:44 AM

Doesn't AirPrism only affect what is received? I'm guessing putting it on the transmitting side wouldn't do a lot of good if its the amplifiers creating most of the OOBE?

 

I'm guessing this test is going to be met with some skepticism, but it seems like an easy thing to reproduce with AirView on a running link?

by
on ‎04-27-2015 11:56 AM

No sure where airPrism sits, pre or post amplifier and if it does anything on tx.  

 

airview doesn't really provide a whole lot for this type of comparison IMO.  It cycles through channels and takes samples rather slowly unlike a scope that is sampling at a vastly faster rate.  Makes seeing some of this information much less obvious.  That said, you can see airFibers pretty clearly on the 802.11 based airMAX gear's airview.  100% duty cycle negates the slow sample rate.

by
‎04-27-2015 11:57 AM - edited ‎04-27-2015 11:59 AM

 

    Doesn't AirPrism only affect what is received? I'm guessing putting it on the transmitting side wouldn't do a lot of good if its the amplifiers creating most of the OOBE?

 

    I'm guessing this test is going to be met with some skepticism, but it seems like an easy thing to reproduce with AirView on a running link?

 

Yes, the Prism chip affects the recieve side only.   It helps to reject interference in adjacent channels, but it's never going to be as good as the AF5X - see http://community.ubnt.com/t5/airMAX-Stories/Radio-Shootout-Pt-2-let-s-try-a-whole-bunch-of-them/cns-...

And people can be as skeptical as they want - the facts are the facts.   Although that doesn't always matter to folks... 

https://www.psychologytoday.com/blog/happiness-in-world/201104/the-two-kinds-belief

Jim

by Ubiquiti Employee
on ‎04-27-2015 11:59 AM

Great analysis!

by
on ‎04-27-2015 12:24 PM

the double edged sword here is that adding tx filtering (and adding airPrism for that matter) starts to make the cheap/junk atheros chips price point unimportant.  No one wants to pay $250 for an 802.11ac (ac-lite$ + airprism$ + tx_airprism$) unit when the AF is $400.  That $150 is an incredible value.  Also, ~$850 for the mimosa is a bit steep.  Ubiquiti got the AF5X out just in the nick of time I think.  Now I don't see mimosa getting a foothold with the B5 series.  We still don't know how the B5-lite will perform, but it's the same chipset as the B5c with some features stripped out so I don't expect it to do any better in Jim's tests here.

 

so that race is starting to look like the race to beamforming/MUMIMO for PTMP.  So, are we going to see an RocketAC_X with a MUMIMO antenna near the release of the mimosa A5 series?  That will be an interesting battle I think.  'cheap' AC + airPrism vs 'quality' AC, probably comparable antenna design/quality.  or are we going to see an AF PTMP product? or a hybrid (AF PTMP speaking something airmax like to AC radios?)