I'm doing some testing with 900 MHz. We aren't a big metro area, and so far our noise floor for this band has been pretty low....95db on average.
This first test was just to prove the penetration capabilities of the band. Clear line of sight, as required by 2.4 and 5GHz, are much harder for us to achieve.
Node A was a Rocket M9, inside a brick house, on the second floor with a yagi pointed west towards town and another yagi pointed east away from town.
Node B was a Rocket M9, mounted on a 10' mast outside, with two yagis. One was pointed toward Node A, and the second was pointed towards town.
Node C was a Nanobridge without the big reflector - so using the internal antenna only with no focusing dish. Node C was taken into town, 2 air miles away and operated out the window of a car.
1. Node C (the roving test node) was first placed 1.5 miles away with clear line of sight, at 5' off the ground. Link quality was excellent, and Node C could see both Node A and Node B.
2. Node C was then placed 2 miles away with only trees interfering. Node C could still see both Node A and Node B with 100% link quality.
3. Finally Node C was placed 2.5 miles away, inside of the town. Multiple trees, multiple houses, even a dirt berm were in the path. Node C could not see Node A, but it could still see Node B.
I'm very pleased with these initial results. Our next step is to put one of the Rocket M9 units up high - we've got a grain elevator that we're targeting, which will put it at 180' AGL. We're going to use the yagi antennas for now - we may switch it to more of a sector antenna depending on the results.
Mike
Good report! Thanks Mike. I suspect a sector antenna will perform better. Will be good to see the test details for it.
If you have a chance, could you run some piece of equipment such as camera, IP phone etc and publish results. Our group has seen good node performance numbers with marginal results when equipment is added.
My summer environmnt is forested and we are looking at M9 as a possible backbone.
Thanks - Vance
"2. Node C was then placed 2 miles away with only trees interfering. Node C could still see both Node A and Node B with 100% link quality."
Can you help us understand the amount of tree blockage with still 100% LQ? Not sure how we'd quantify this, but are we talking 1 tree, 5 trees, or a forest over the 2 miles?
Great data points!
Well, it wasn't a forest and it wasn't open. It was a typical small town with oak and maple trees...all kinds of homes. I'm going to run a more scientific test soon. What was most promising for me was being able to shoot through ANY obstructions. In the end, our main node for the area will be up high, so line of sight is at least more likely.
Perhaps try IPerf on the experimental paths - https://iperf.fr/
Bill
Hey Michael;
Clarification......What was the distance between nodes a & b with the yagi antennas?
Nodes A and B were only 500ft apart. What was interesting was that node B could be heard from Node C, even though Node C was located inside a brick house.
I think your results are very consistent with what we should expect. I would expect that
you could get ever greater range with the rocket/yagi configuration. Now the range with
the bridge is also about what we should expect, bearing in mind that the only antenna
that the ham firmware allows us to use is the internal antenna. (which we should not
expect too much of because of the size of it, a antenna of this size can not be very
efficient at 900Mhz)
All this goes to show the absolute need for a new firmware build that will enable us to
utilize external antennas on the bridge / loco ect.
I installed our first permanent location M9 rocket on a grain elevator at 150' yesterday. There are two 11 element yagi antennas, one pointed north, one pointed east. I have two other nodes that I used for testing. The results this time were not as impressive as my first outing. ** I would not recommend 900MHz at this time without further analysis!!! **
Using another rocket with two yagis pointed directly at the elevator, 2 miles away, direct line of sight, I was only able to create a stable connection when using all 20MHz of the 900 MHz allocation. I started by using a 5MHz channel at 922MHz. Even through the S/N was 50db, the link would not support any traffic. I tried all of the other channels and had the same results. I then moved to 10MHz and finally 20MHz as the channel width. With a S/N of 30db and a width of 20MHz I was able to have a stable link capable of no packet loss. I wasn't trying to saturate the link...I was doing simple pings, pulling up a web page, and using the AREDN website.
The other node was a nanobridge without any type of reflector, mounted on a tripod lifted up 10'. The results from it were abysmal. Within .5 miles of the elevator I had a stable link with anywhere from 30 to 20db S/N. Below 20db it just didn't work. It didn't seem to matter if I had a clear line of sight or not.
Why did 20MHz work and not 5MHz? I'm guessing interference. There are a number of SCADA systems, even in our small town. The noise floor was at 95db. Even changing the channels made no difference. But....
Why did the link work no matter if we had clear line of sight or not? Well, it's all about signal strength and the fresnel zone. With the elevator at 150' and the test nanobridge at 10', the fresnel zone was into the ground even though we were only 2 miles away. Get this....even at .5 miles, with the test node at 10' we're in the zone.
So what does this mean? If you're thinking about using 900MHz, the noise floor may not be the only thing you need to worry about. Intermittent SCADA systems and a REAL BIG fresnel zone may be good reasons to avoid this band.
For my next test I will be putting the second rocket up at another site for testing. They will be 12 miles away but both will be up over 100' which keeps us out of terrain. My initial expectations were not realistic. I should have run all of my scenarios through airlink.ubnt.com. You're not going to have a 900MHz antenna down low and make this work.
Very nice report. Thanks for the info/testing.
KG9DW,
From this comment, "the fresnel zone was into the ground even though we were only 2 miles away."... While the fesnel zone diamenter would be about 40' to 45' (radius of ~20'), this would be the max and in the middle of line of site. with a little trig, the ground clearance under line of sight in the middle would be 70' and plenty of room. The zone diameter close in to the antennas is significantly less (it's a football shape with the point at the antenna). There shouldn't be any objects in the freznel zone unless there's a hill or a other object above the ground somewhere. We'd have to run the math to see how big the zone is 20 ft. from the antenna--but shouldn't be into the ground at that point.
The data you collected is interesting. I'm suspecting that there is reflection surface(s) greatly affecting the results or antenna issue. If there were competing signals, and you tried all the channels at 5Mhz bandwidth, if none worked, then the competing signals are through out the freq range. going to 20Mhz shouldn't have worked with competing signals, because this spreads the same power out over a larger freq area and lowers the chances of the signal getting through the compeition. (same power in 5Mhz bandwidth is a bigger amplitude signal on a given frequency than across 20Mhz.)
If this is a reflection issue, you might try moving the node closer/farther away 1/4 wave, 1/2 wave, etc. and see if that makes any difference. The reflected waves' phases can add or subtract. No practical way to predict the paths and outcome.. Maybe the signal is distorted on the node/antenna. since you are using 2 physically different antennas on the rocket how are they oriented? 90 deg phase delta, exact same distance, etc.? They do connect up with a usable link under at least 1 senario, correct?
[Follow up thought after 1st post] What was the LQ on both ends? Might be 100% LQ comming in one node, but 5% getting into the other node the opposite direction. The SNR listing doesn't say what neighbor signal this is--were there 3+ nodes live to cause confusion? Here is a modified /usr/local/bin/wscan, which shows individual neighbor node signal strengths in wifi scan if more than 2 nodes are in the equation: https://drive.google.com/file/d/0B2bEy75HhwWhRTl3aHpLanEybFk/view?usp=sh...
Joe AE6XE
Joe brings up some good points.
A few other points I would consider as well is the fact that the rocket on the one elevator was using two antennas pointed opposite directions from each other. This is a BAD practice, the radios in these units (and really its been ALL the units including the Linksys units for those who have played with BBHN) are designed to have BOTH antennas point towards the SAME signal. Users had done this in the past before we created the DtDlink feature but since than the use of a node with two antennas in opposite directions has become really a bad thing to do.
It is possible the grain elevator node may have been reflecting off something as Joe suggests, or even the grain elevator itself was detuining the antenna and affecting the radiated pattern or bouncing the rear end of the signal causing a multipath itself perhaps (I'm not 100% sure on this one w/o some modeling)
My concern though is the fact that it was not both antenna pointed the same direction in opposite polarity (and yes as Joe mentions if there was a 3rd node in the mix) phased to match (feed line's exact same length and appropriate separation) that this alone may be part of the issues in performance while in your first test the other ports it sounds like were basically dead with no antenna?
I can't explain the 20MHz VS 5MHz right now that really seems counter intuitive to me (unless were talking reflection which are weaker at the wider bandwidth maybe)
SCADA and AVL for sure could be an issue (some areas have a lot of it) and AVL is PRIMARY on the band HAM is secondary (and part 15 is of course dead last)
LQ is the big question, the Signal Quality meter will stay up with only 1 packet per IIRC 10 seconds (would have to go to the code it might be 5 seconds but I think it was set to 10)
I had a link I ran it was a 20DBM link in the models, was around 25DBM in the real world, Except for one problem, a tree 50foot from the node (which was not in the model) would blow its branches into the path causing a 94% packet loss. (we would only git signal when the wind blew the right way)
Also since we have so little experience with the 900MHz Ubiquiti gear, Ill even throw out (and this is a wild throw out not based on any known fact) the possibility of the fact that maybe we are over driving the link for the downconverter and that backing off the RF power by 10dbm might help. Its counter intuitive but it may be happening.
Oh and I'll also question what the Distance parameter was set to as well.
Either way I agree with you 900MHz is a band that needs some analyzes, I don't have real world 900MHz experience with the mesh to know how it will perform and appreciate hearing the feedback and hope you will keep experimenting so we can find out where this band will fit into the mix on performance.
Good comments...thanks for the insights. I had 6 hours of driving this weekend, and the more I thought about it the more I'm thinking too much power may have been the problem. Otherwise, moving to 20MHz doesn't make sense. My previous conclusions about the intereference don't make sense...
The gain elevator is made up of many metal legs (to move the gain between silos), but the primary structure is 4 concrete silos that are 130' tall. Detuning isn't likely. Both antennas at the elevator are mounted to be horizontally polarized. I had the distance parameter set to 0 on all units.
Today we have light rain and fog. With the nanobridge in the 2nd story window of my house, pointed toward the elevator 2 miles away, with a metal pole barn in the path, I have a signal of 28db at the nanobridge and 22db at the elevator. (the other node is turned off)
Thinking more about the antennas at the elevator, I switched the north facing antenna to transmit and then both for receive. The east antenna points directly at my house. The LQ immediately sank into the 50% range with the ratio at the nanobridge down to 11db. The elevator ratio remained unchanged. Turning both antennas back to diversity recovered back to 100% LQ.
I then turned off the north antenna entirely - the LQ went down (makes sense during the change over) and then stabilized at 100%. The signals stayed at nearly the same ratio - the nanobridge receive went up only one 1db. So diversity seems to be working quite well in my case - the unit is using the result of both chains to get as good of a signal as it can.
I'm going to take the other rocket into town to the top of a parking garage 12 miles distant and see what happens. I'm also going to experiment with reducing the power to see if I can make the 5MHz channel stable.
"So diversity seems to be working quite well in my case - the unit is using the result of both chains to get as good of a signal as it can. "
What it sounds like you have is a non-diverity system, you have a MIMO (Multiple Input Multiple Output) device running in a 1x chain communications model (even though it thinks its a 2x chain deployment)
These nodes are designed to operate with both chains pointed to the same path, this deployment may be loosing signal capability because only one of the two antennas is picking you up, in other words it is a NON diversity system for that link, SNR nor Signal strength would change in this case.
Second to that, your likely also are not taking advantage of of 802.11n in this deployment (requires diversity) which dictate that when BOTH chains are in sync they will actually work together and increase the bandwidth of the link by treating each polarity as a separate signal (This could become important on 900MHz because of the limited bandwidth)
Unlike the old Linksys devices many users have worked with in the past which (from what I have been told but have not verified) uses an antenna switch between the two antennas, Ubiquiti gear actually has two decoding/encoding "radios" onboard, one for each chain
Ah, ok. That's interesting information. I thought the chains were normally used for different polarities...so same direction yes, but one H one V. They both won't receive the same signal - there will be reflections, or at least a difference in strength because of cross polarities.
I'll try pointing both antennas in the same direction for one test and see what results.
I posted pictures of my first mesh node install up on amateurradio.com.
http://www.amateurradio.com/aredn-first-site-on-the-air/
Great pictures and a wonderful view.
Nice writeup! I'm really curious how it will perform with both antennas aimed in the same direction.
It's time for my favorite Friday Fun Activity! That's right folks, it's time for Can We Make 900MHz work!?!?!
Today's adventure started with a nanobridge pointed at the elevator site. For those just joining us, the elevator site is up 150' in the air with two yagis, horizontal, pointed in different directions. The LQ was 100% and there was 20DB of snr.
I went up to the elevator and took the east facing yagi and put it in line with the north facing yagi, about 19" apart. I also switched it to vertical polarization. Back at the farm, the nanobridge went in and out of connectivity, with less than 10DB of snr. Yikes!
No fear, we've played this game before. Taking the nanobridge and putting it on at 15 foot mast outside with a single horizontal yagi also connected to it, the link has been re-established. LQ is coming back up to 100%, and the snr on the nanobridge is at 22DB. But....(yes, if you've read my other posts you know there's always a but....) the snr at the elevator is down to 14DB. Guess what...I increased the noise floor at the elevator by 8DB when I moved the second antenna to the vertical plane.
Keep in mind, the farm is 2 miles east and I'm trying to hit the elevator off of a side lobe. These are 11 element yagis so they are very directional. All in all, not a bad test.
Here's what's next:
1. Take the nanobridge on a drive and see how things work with both elevator antennas pointed in the same direction. I'm betting I've got a rockin' signal to the north with a narrow beamwidth!
2. Head to the parking garage 12 miles away and see if I can pickup a signal. Fingers crossed....
3. Take the nanobridge up to the top of the building that is my second permanent location and see if we've got a signal. It's also 12 miles away and 90' higher than the parking garage.
The adventure continues....73 for now from Mike KG9DW...putting the AMATEUR back in amateur radio!
(this was posted from the Mesh!)
"The adventure continues....73 for now from Mike KG9DW...putting the AMATEUR back in amateur radio!" Love it!
just add...all this using AREDN.... At the center of prepAREDNess...
Here's where we are at for those just joining our ongoing 900 MHz experiement:
node 1 is a Rocket M900 mounted at 150' with two yagis pointed north, one horizontally polarized and one vertically polarized.
node 2 (installed today) is a Nanobridge M9 without the reflector mounted on top of a building at 190' pointed south. It is 12.2 miles away from node 1.
The mesh is alive between the two nodes, but it isn't very happy. SNR from node 1 to node 2 is at 18DB. SNR from node 2 to node 1 is 14DB. LQ is running between 63 and 83%. While it technically is up, it isn't very usable.
Here's what's next:
1. I've got a 900Mhz ubiquiti sector antenna coming from ebay. It was a pretty good deal, so its worth a shot. I'm not confident that my cheap yagi antennas are doing that well. While this may hurt my long shot up to node 2, it will undoubtably provide excellent coverage throughout the small village of 2900 that it will be looking down upon.
2. After trying the sector, I'm going to add an 11db panel antenna to the nanobridge. Doing some research, it looks like the nanobridge was never meant for use without the reflector. I'm betting your reading this and thinking "well duh" but hey I thought I'd give it a shot. I've probably got very little gain without the reflector. This was another ebay purchase and the reflector wasn't available. Even if it was, the site I'm at is not at all interested in something like a dish being installed up top. With the winds we get, putting that up there wouldn't be my first choice either.
3. If #1 or #2 aren't successful, I may try a smaller yagi at node 2, or even go with a rocket that I have for testing with a smaller sector(21" tall). With two sectors pointed at each other, 12 miles apart, will it work? Airlink says it will work.
4. Finally, if none of this works, I'm putting all of my 900Mhz gear up on ebay and wait for the 3GHz support to come!
I'd really like to get 900MHz working as I've invested in this gear already. But I'm not opposed to throwing in the towel and starting over with 3GHz. What do you all think?
I can think of one reason why a 20 MHz wide signal would work and a 5 MHz wide signal wouldn't. Assuming this is WiFi or a WiFi-like signal, most of the newer WiFi modulation modes are OFDM, Orthogonal Frequency Division Multiplex. It's a large collection of closely spaced carriers, each carrying a piece of the data. Error correction is applied in both time and frequency so that if one piece of the signal gets nulled out by multipath or perhaps a CW interferer, the missing data can be regenerated without a retransmission.
It might be that the multipath is such that you need to spread over considerably more than 5 MHz to get enough of the signal to pass to overcome the parts that don't.
Should be able to rule in/out this 'frequency selective fading' by trying out the different 5Mhz channels over the 20Mhz BW. Haven't run the math, but probably couldn't have fade destructive nulls of the OFDM sub-carriers uniformly across the 20Mhz BW and still have this be the explanation.
"I started by using a 5MHz channel at 922MHz. Even through the S/N was 50db, the link would not support any traffic. I tried all of the other channels and had the same results."
this suggests that frequency selective fading was not the issue between 5Mhz and 20Mhz channels--if all the 5Mhz channels were affected, then all of the 20Mhz channel would have been affected too. But, weirder things have happened before, one might not expect.
Two basic parameters define a fading channel: coherence time and coherence bandwidth. Coherence time refers to how quickly the channel changes, e.g., how fast ionized clouds in the ionosphere move or how many wavelengths per second a mobile user moves. That's not directly relevant to a fixed link, except for second-order effects like airplanes, wind-blown trees, etc.
Coherence bandwidth is determined by the multipath time spread. If two equal-strength components arrive a microsecond apart, then the coherence bandwidth is on the order of 1 MHz, i.e., the largest chunk of bandwidth you can reasonably expect to not have a null in it is 1 MHz. Multipath spreads can be quite long in a long-haul point-to-point link, so the coherence bandwidth could be quite small.
This calls for symbol rates considerably less than the coherence bandwidth, and that's what OFDM is all about.
The standard mitigation technique for both impairments is diversity: time diversity, frequency diversity and spatial diversity. WiFi uses all three: error correction coding over time, error correction coding over frequency (with OFDM), and multiple antennas. But these techniques have practical limits, such as maximum allowable delay, maximum available bandwidth, and an obvious cost to multiple antennas, so it's good to be aware of these impairments and try to reduce them. One good way to reduce multipath on fixed links is through the use of circular polarization, as all odd-order reflections will have the opposite polarity.
I should have added that using more bandwidth can actually improve things when you have a small coherence bandwidth. This may seem counterintuitive, but it's true with a WiFi-like OFDM signal. The more frequency bins you use, the greater the chances that not all of them will fall into a frequency null, and the error correction coding can reconstruct the bins that are notched out. That's why a 20 MHz signal might do better than a 5 MHz signal.
It's fairly rare for a terrestrial point-to-point signal to be limited mainly by signal-to-noise ratio (i.e., transmitter power) especially for a system like WiFi that can vary its data rate over a very wide range (1 Mb/s to 54 Mb/s for 802.11g). Multipath is the real killer.
Similarly, it's somewhat counterintuitive that using more time can combat a channel with a short coherence time, but that's also true. By adding error coding and interleaving (spreading) the encoded data out over time, you improve the chances of filling in any deep fades that may occur during your transmission.
All good information points for me to consider. I'm continuing with a swap out of the beam antennas to sector antennas. The 922MHz frequency slot is pretty much useless for my area due to other carriers (I'm betting they are SCADA systems). The experimentation continues....
You might be better off with 3GHz band devices coming in the near future.
Is circular polarization an option? That's the most practical suggestion I can make out of my little theoretical aside. A right-hand signal reflects into a left-hand signal, and vice versa. So if you have a strong multipath component generated by a single reflection (which seems likely) using circular polarization will very likely cancel it at the receiver.
Tighter antenna beamwidths can also help. Even when you already have a strong signal, higher gain might help suppress a multipath component coming from the side.
This example used a Rocket M9 MIMO, consequently both LH & RH polarization antennas are available to connect in--should be an option. This helps against scada applications using linear. Would be interesting how circular compares to linear since reflections would cross feed into the other channel. I found at least one paper showing that circular achieves better throughput over linear for MIMO. But, now why don't we see Ubiquiti offer circular polarization options? Price-point? Diversity gain not compelling?
Good question. Circular antennas are more complicated than linear, and maybe they didn't think most of their customers needed them.
900 MHz just didn't work for me. I switched to sector antennas on my Rockets and could not maintain a consistent link 12 miles apart. The larger fresnel zone and the other users of the frequencies were my downfall. I did really like being able to shoot through trees and even some buildings! If I wanted to provide coverage in a remote location or if I was wanting to provide coverage within say a 2 mile radius with relatively good paths, 900 is a band to consider.
I'm selling all my 900 gear and moving on to other bands for my next experiments. I'll post another thread about the gear if you're interested in buying me out!
It would be nice to test circular polarization before giving up. If a single ground reflection was your problem, CP really should have solved it.
CP wouldn't do much to reject SCADA using linear, though. The isolation is only 3 dB. If linearly polarized interference (not multipath) is your problem, then orthogonal linear would be best.
Philip, that may have been a possibility and may have been a solution - but compared to using the more off-the-shelf components for the other bands AND the lower cost radios for the other bands, moving away from 900 MHz is the best decision for me. I had only started down this path because of the potential availability of two large Nextel antennas that were abandoned long ago already mounted on the side of a corporate building I had access to. But the use of those antennas wasn't approved (they sit unused still), and I had already purchased two sets of 900 MHz gear.
Anyway, your ideas are helpful and especially your feedback as to why using more bandwdith produced the results I was seeing. If I was more involved in RF, I may have come to these conclusions quicker. It's been over 20 years since my BSEE (in RF design, actually) and my memory continues to fade! Getting involved in this project has certainly been a learning experience and it has rekindled my passion for experimentation.
73, Mike KG9DW