dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
3023

AnonDOG
@12.160.x.x

AnonDOG

Anon

Mesh Network Simulation -- RMDLX

I've done the study I promissed a couple of weeks ago.

Here is the pdf:
»www.freeantennas.com/Sim ··· work.pdf

superdog
I Need A Drink
MVM
join:2001-07-13
Lebanon, PA

superdog

MVM

Great work!, and a very interesting read. I am thinking about lighting the downtown area where I live, and this gives some interesting points to ponder. I am wondering how using one central location with 2 AP's that have 180 degree sectors running channels 1 and 11 with 4 AP's attached running WDS, and 1 more AP attached to the second group will work. I do realize that with each hop, throughput drops in half, so if I keep the signal strength up, and get them to hook at 802.11g rates, I should be able to deliver about 1 meg at the outer radios??. Any thoughts on this?. My town is fairly small, and I only have a short section of the city I want to cover. Using Tropos for a deployment of this nature would break my piggy bank, and I am sure that using DLB2300's or HighGain 2400Maxx AP's would do the trick in a small setup like this?

AnonDOG
@12.160.x.x

AnonDOG

Anon

I have to agree with a number of pundits who have said they can't fathom why Earthlink decided upon Tropos.

Any one radio mesh solution starts off with a 50% handicap.

I have built and deployed these:
»www.locustworld.com/

They work and are as cost effective as it gets.



Contact me via E-Mail, maybe I can help you with some maps.

dcshobby
Premium Member
join:2004-11-24
Rochester, MN

dcshobby

Premium Member

I would like more information on how you deployed your Locustworld mesh system. Thanks
lutful
... of ideas
Premium Member
join:2005-06-16
Ottawa, ON

lutful to AnonDOG

Premium Member

to AnonDOG
Excellent analysis and report, Michael.

Back in 2003/4, we did some urban mesh research and came up with a 3-level architecture shown in this post:
»Re: Trango's New Mesh Products

We envisioned 3-level architecture:
Level 1: resilient backhaul ring using 5.8Ghz or 24Ghz
Level 2: WiMax PtMP using 5.4-5.7Ghz
Level 3: Multiple localized StellarMesh networks

StellarMesh was an attempt to avoid the dreaded 50% handicap of random single radio mesh. Sadly our FPGA radio board was never built to test the promising simulations and there was no usable WiMax gear in 2004.

However, radio prices have come down so much that I have started exploring GPS synchronized multi-radio mesh designs.

polk5
join:2001-12-29
New Orleans, LA

polk5 to AnonDOG

Member

to AnonDOG
Michael, Thanks for posting this! It makes things so much clearer. What happens if you throw a indoor CPE into the mix? Wouldnt that greatly reduce the number of nodes needed? I still see a problem with the ROI in any city wide network when you factor in franchise fees, deployment cost, maintenance,and bandwidth cost. I also see a small take rate in an area where you are competing with bell and cable. Bell can bundle their service and practically give away the data side of things. I think it would be a nightmare to try to deliver voip and data over a mesh network today. Im sure earthlink has done their homework on this and they seem very confidant they can make the deployment work well here. It will be interesting to see how it goes.
Nowireneeded
join:2004-02-11
Montoursville, PA

Nowireneeded to AnonDOG

Member

to AnonDOG
Did your study take noise floors into account?
bryandj239
join:2002-08-15
Bay City, MI

bryandj239

Member

AnonDog..

I'd be interested in how you deployed LocustWorld. I currently have one Locustworld node up and running in a small downtown area. I'm curious as to how yours may be set up.

AnonDOG
@12.160.x.x

AnonDOG to Nowireneeded

Anon

to Nowireneeded
Did your study take noise floors into account?
Excellent question, I know of no way to predict noise floor but if you can help me find a way I would greatly appreciate it.

What I did in my study to compensate for that which I did not make clear enough, and which I will be correcting in the pdf, is I ensured I looked at the laptop to the AP path when projecting coverage on the last hop, rather than the AP to the laptop path.

The reason that I did that is that the laptop is modeled at about 18 dB EIRP, whereas the AP is modeled at about 36 dBi EIRP. Since the laptop is operating at less than 1/32 of the power levels of the AP we can reasonably assume that will be the weakest part of the link.

Now I do realize that all paths are not really symetric; however, I also know that it is impossible to model them with any greater fidelity. Therefore I model the weakest transmitter on the link as the probability of path being that asymetric is very unlikely.

If you consider that the receive sensitivity of the 5210 is modeled at -87 dBm, and the transmit power of the device is at 36 dB EIRP, you will realize that you can subtract the difference of the transmit powers (36 - 18 dB = 18) from the receive sensitivity of the AP (-87 dBm) to estimate the received signal at the laptop on the link. In this case we get -87 dBm + 18 dB ~= -69 dBm.

Most 802.11b cards operate at 11 M/Bit/sec at -69 dBm and most 802.11g cards will operate at a bit better than that at -69 dBm.

I did not consider noise floors because nobody can predict noise floors but then I added in an 18 dBm fade margin for the laptop...
AnonDOG

AnonDOG to bryandj239

Anon

to bryandj239
I have deployed a handful of LocustMesh nodes built on old HP desktops with Prism two cards in them.

They worked well to cover about one quater of a square mile in our area. They were a test bed. I liked their performance. I didn't care for Wiana.

There is no magic in our setup which you can't discover at the Locust world site.

We also deployed about fifty Nokia mesh units back around 2001 or 2002. We still have three or four of them running. That was some great hardware. Expensive but it was the only radio I ever bought that had a heater in it.
Nowireneeded
join:2004-02-11
Montoursville, PA

Nowireneeded to AnonDOG

Member

to AnonDOG
Noise floors in Phila were between 75 and 88 a year ago.
If my memory serves me correctly, Miami was documented 71 to 94 with the Civitium study that was done down there last year. We found that cell sizes could be cut by 30-50% depending on how bad it was for any given node. Noise floors change which means these networks will have to be constantly monitored and retooled.
Nowireneeded

Nowireneeded

Member

On another note for those of you following Wimax. 70 mesh nodes per square mile at $3,000 per unit plus the backhauling of every 3rd or 5th node makes the economics sound quite unreasonable. A 4 sector Wimax base station with 4 10 MHz channels using a diversity antenna system (smart antennas etc) in the $20k to $30k per sector range can cover the same square mile. Obviuosly there aren't any PC Cards available yet and indoor self install cpe are just starting to get tested.

AnonDOG
@12.160.x.x

AnonDOG to Nowireneeded

Anon

to Nowireneeded
Noise floors in Phila were between 75 and 88 a year ago.
If my memory serves me correctly, Miami was documented 71 to 94 with the Civitium study that was done down there last year. We found that cell sizes could be cut by 30-50% depending on how bad it was for any given node. Noise floors change which means these networks will have to be constantly monitored and retooled.
Then -69 dB should give the laptop some fade margin, on the other hand -87 dB is probably too generous for the AP and I may need to adjust the model.

I think it is important here to realize that these noise floors you mention were measured and that in the first case they vary by 13 dB whereas in the second case they vary by as much as 23 dB. If you can show me how to model that in any reasonable manner I'd appreciate it.

The beauty of a mesh is that it is self anealing. Any network needs to be monitored and occasionally tweaked. The ease of installation of a "patch" on a mesh network is appealing. You just stick it up where you need more coverage. Have a unit die? No worries, just go stick up a new unit in the same place when you get a chance, until then you will have some as opposed to no coverage in that spot.

The largest cost in operating a network is the recurring cost of highly skilled laborers. If you can reduce the number of engineers required to maintain the network by deploying smart devices. You save big money in the long run.
On another note for those of you following Wimax. 70 mesh nodes per square mile at $3,000 per unit plus the backhauling of every 3rd or 5th node makes the economics sound quite unreasonable. A 4 sector Wimax base station with 4 10 MHz channels using a diversity antenna system (smart antennas etc) in the $20k to $30k per sector range can cover the same square mile. Obviuosly there aren't any PC Cards available yet and indoor self install cpe are just starting to get tested.
Physics are physics. Vivato made similar claims and failed to deliver spectacularily. It is not fair to WiMAX to make equally unreasonable claims. There is no magic bullet. There is no city on this planet that a single Wimax base station operating at the legal maximum power levels is going to achieve 50% indoor coverage over one square mile.

Nowires what I have here is a model. It is just a model. It is useful to try to understand the problem. In order for it to be useful you have to read it with an open mind. The nice thing about a model is that you can tweak it and learn from it but you first have to be willing to learn.

The model has shown me that even at 70 nodes per mile the mesh is going to be thin. It is blatantly obvious that the earlier densities were way too thin. From working with the model I've come to the conclusion that fifty percent more density would be necessary to hit the 90 percent indoor coverage target.
v_lestat
The Blood Is The Life
join:2002-11-09

v_lestat to AnonDOG

Member

to AnonDOG
you will never accomplish what you want to with Tropos.

those people will find that out if they continue down the path using tropos single radio nodes. they dont work and there is plenty of evidence to support that.

There are much better dual radio Meshing products with the backhaul devices built in that make the tropos radio look like a dlink router.
lutful
... of ideas
Premium Member
join:2005-06-16
Ottawa, ON

lutful

Premium Member

said by v_lestat:

There are much better dual radio Meshing products with the backhaul devices built in that make the tropos radio look like a dlink router.
I am thinking that single enclosure with 4x 2.4/5Ghz combo miniPCI radios will provide enormous flexibility in large-scale mesh deployments.

3x 5Ghz backhaul + 1x 2.4Ghz AP
2x 5Ghz backhaul + 2x 2.4Ghz AP
1x 5Ghz backhaul + 3x 2.4Ghz AP

Somehow Locustworld MeshAP has fallen behind the times but WILI mesh firmware looks very promising. WILI already supports IEEE 802.11i (WPA2/AES) over WDS, 802.11e QoS and will support IEEE 802.11s mesh protocol in the future.

Hopefully others will follow WILI example to give us some interoperability that is lacking in the current proprietary mesh marketplace.
Nowireneeded
join:2004-02-11
Montoursville, PA

Nowireneeded to AnonDOG

Member

to AnonDOG
Noise floors are all over the place on any given day. This is one of the challenges of mesh. Cell sizes shrink over time as noise floors increase. Self healing doesn't mean anything if cell sizes aren't what they were 6 months ago. Dead spots exist and therefore need to be redeployed.

"Physics are physics. Vivato made similar claims and failed to deliver spectacularily. It is not fair to WiMAX to make equally unreasonable claims. There is no magic bullet. There is no city on this planet that a single Wimax base station operating at the legal maximum power levels is going to achieve 50% indoor coverage over one square mile."

Don't mix the word Vivato with Wimax smart arrays, beam forming technologies, or adaptive arrays. Not the same. Wimax in the licensed bands WILL do 1 mile cell radiuses. It's already being done with Nextnet and Navini systems. Wimax products will be even be better. Link budgets in the 160's are quite the norm.

I did read your model with an open mind....I'm giving you real world deployment feedback from projects I'm on and responding to. Your on track with the plus 70 ap per square mile.
Nowireneeded

Nowireneeded to AnonDOG

Member

to AnonDOG
Reality Bites Muni WiFi
unstrung.com/blog.asp?blog_sectionid=269&WT.svl=unblogger1_2

AnonDOG
@12.160.x.x

AnonDOG to lutful

Anon

to lutful
Lutful;

Here are some of my thoughts.

If there is anything the model shows with certainty it is that the mesh node needs to be dirt cheap and brain dead simple. The ideal mesh node is going to be really inexpensive so that we can deploy as many as we need.

Ideally it will be so cheap that we can imbed it in indoor access points as well as outdoor APs. In that form factor they become as common as the light bulb on the front step and as easily replaced. No more paying an RF engineer $150K annual recurring.

It will be a dual radio device only if that does not drive up the cost because backhauls (injection points) are not the real driver of system cost with this architecture. Ten injection points on a square mile might cost fifteen thousand dollars but 100 mesh nodes at $350 each are $35K.

The price point of the mesh node is more important than the receive sensitivity or the transmit power or the number of radios because they absolutely have to be deployed at very high densities to get reliable coverage and to reduce mesh link distances enough to esure that the installers can ignore the RF link calculations.

The mesh should be deployable by the same guy that replaces the light bulbs on the lamp poles. That is the main attraction of mesh, after the engineers are done designing the injection layer, the mesh can be deployed without a lot of additional thought. That can only happen with very dense mesh deployements and very dense mesh deplopments.


Gotta run ... more later
bryandj239
join:2002-08-15
Bay City, MI

bryandj239

Member

AnonDog...

Yep. Wiana definately blows. I wouldn't mind diving into running my own RADIUS, but there's so little documentation on how to implement it with LW. I guess I was hoping you had gone that route.
lutful
... of ideas
Premium Member
join:2005-06-16
Ottawa, ON

lutful to AnonDOG

Premium Member

to AnonDOG
Click for full size
Click for full size
said by AnonDOG :

If there is anything the model shows with certainty it is that the mesh node needs to be dirt cheap and brain dead simple.
Michael, I am in total agreement on this. Also, it is curious that we always wanted to deliver the StarAPs packaged as replacement street lamps.

However, we encountered a serious problem trying to design a $250 "uniform" mesh node in 2003. The 166Mhz MIPS CPU inside Atheros AR5002-AP SoC could not do a good job running both secure WiFi and mesh routing.

So, I decided to go with a non-uniform mesh design concept where a $500 StellarPortal (Via Eden 533Mhz) will do the dynamic route calculation using association requests, observed RSSI and traffic congestion feedback from a constellation of $100 StarAPs.

The StarAP was developed on same FPGA/Ubicom/Agere radio platform as our WISP CPE. There are photos of the PCB layout in an old post but we never managed to build it.

Currently, Atheros AR5007 SoC platforms ($25-$35) will be good for the StarAP while multi-radio StellarPortals can be made with Gateworks boards for ~$250.

I could not post the whole file, but I am copying a small section on the Stellar Mesh concept together with the overall 3-level architecture for large-scale urban deployments.

AnonDOG
@12.160.x.x

AnonDOG to Nowireneeded

Anon

to Nowireneeded
Don't mix the word Vivato with Wimax smart arrays, beam forming technologies, or adaptive arrays. Not the same.
Vivato's claim to fame was indeed smart arrays and beam stearing technologies... EXACTLY the same. Exactly the same power levels, simpler modulation schemes (read better receive sensitivitys).

From here:
»www.pacificafund.com/blo ··· /19.html

This:
quote:
Vivato is creating a high end WiFi based 'switch'. Unlike so-called WiFi switches that are simply a means of clustering and controlling cheap access points, Vivato's system works by beamsteering the WiFi RF energy to separately reach machines located along radial paths from the antenna location. Each such beam has the full complement of WiFi channels and bandwidth, and has a range about five times of a conventional AP in a similar setting - that can reach the kilometer class for outdoor line of sight. The system works with off the shelf client side cards. The beams are tight enough (5 - 7 degrees) that Vivato is able to qualify for the FCC point-to-point rather than the omnidirectional power and gain rules. Jim was a bit evasive on the total throughput of each system, but marketing slides suggested it might be about the equal of 12 conventional APs.

Wimax in the licensed bands WILL do 1 mile cell radiuses. It's already being done with Nextnet and Navini systems.Wimax products will be even be better. Link budgets in the 160's are quite the norm.
Kinda eliminates the economies of scale that an unlicensed product brings to the table. The point of using WiFi is that it is unlicensed. You are comparing apples to aardvarks here. Link budgets in licensed spectrum are allowed to be larger but you still have to power that card in a laptop.

EVDO can not cover a one square mile area and get 50% indoor penetration in downtown urban areas. What do THEIR link budgets look like?

Agreed, give me WiMAX hardware to build a mesh. It will be a better mesh. I will probably only need about 2/3 as much mesh density to make it work at the street level in high density urban construction. That might mean I can get away with only 60 mesh nodes per square mile, but then I would not have access to forty million installed WiFi clients...

AnonDOG

AnonDOG to AnonDOG

Anon

to AnonDOG
Still ...

We've been thrashing around ideas about mesh density and so I've been thinking about what the model tells us and how we can use the model to extrapolate required mesh density based upon noise floor. It isn't pretty.

If my model shows that the APs can talk to each other at -87 dBm and we decide that the noise floor is going to be more like -77 dBm (which is entirely possible as Nowireneeded said). Lets estimate the required node density to meet the new noise floor.

We know that 6 dB doubles our range and that 3 dB increases our range by about 41 percent. This means we can extrapolate what mesh density would be necessary to meet the noise floor.

We know that I modeled 70 nodes per square mile. That means that the average distance between nodes is 2*(sqrt(((5280x5280)/70)/PI)) = 712 feet.

If we want to increase the received signal by 10 dB we need to divide the range by two and then reduce that result by 40 percent. So we divide 712/2 = to get about 356 feet and we multiply that by 60 percent to get about 214 feet. This means the same hardware which sees the other node at -87 dBm at 712 feet will see the other node at -77 dBm at about 214 feet. And this means that each node is covering an area that is about 107 feet in radius.

Now we need to figure out how many nodes will be required to cover that square mile. The area of a circle 107 feet in radius is PI*107*107 ~= 35,968 square feet. We divide this into the number of square feet in a square mile (5280*5280)/35,968 ~= 775 nodes.

Now that is definately ugly. If our mesh needs to communicate at -77 dBm then our nodes need to be about 214 feet apart. There is no doubt about that number. To actually cover a square mile with that kind of coverage you have to have about 775 nodes per square mile and we have only just met a -77 dBm noise floor. To build reliable mesh backhaul links a fade margin is required. This means that we get the same kind of mesh density requirement if we require the mesh to have a 10 dB fade margin and relax the noise floor back to -87 dB.

So, the more I look at the model, they more ugly things look for those who are agreeing to make these deployments. We now are looking at node counts which require about 70 insertion points per square mile to get good thruput on the mesh.

Clearly these numbers destroy any business plan which expects ROI to be measured in months rather than years. Admittedly this is the absolute worst case; however, it is easily within the realm of possibility.
lutful
... of ideas
Premium Member
join:2005-06-16
Ottawa, ON

lutful

Premium Member

said by AnonDOG :

the more I look at the model, they more ugly things look
Since MIMO mesh could operate even below noise floor, maybe the scenario will change in a few years?

AnonDOG
@12.160.x.x

AnonDOG

Anon

LOL

Ayup. Operation below the noise floor makes things interesting.

This was the idea discussed in another post a while back. There are technologies which can recover data under the noise floor.

First time I ever saw such a technology was at a little company in Ohio, near Wright Patterson AFB. IIRC that was in 1992...

I don't remember how far below the noise floor they were recovering the data and if I did I probably couldn't tell you, but then I wasn't supposed to be given access to that data either... so....

Yes, sir, you are absolutely correct! Technologies which operate below the noise floor change the rules and technologies which change the rules are DISRUPTIVE.

DISRUPTIVE technologies change the world, especially when attached to smart routing.

Expand your moderator at work
Nowireneeded
join:2004-02-11
Montoursville, PA

Nowireneeded to AnonDOG

Member

to AnonDOG

Re: Mesh Network Simulation -- RMDLX

I am not going to argue Vivato's "claims". What they had and what vendors are doing with Wimax are "apples and aardvarks". This is well documented that their "claims' were self serving.
Nowireneeded

Nowireneeded

Member

BTW, With my comparison if Wimax vs mesh coverage modeling I should have stated "licensed Wimax". And yes I have seen EVDO work at more than one mile in-building. Several of my clients have Verizon or Sprint service and they constantly monitor results. The point is a licensed Wimax play in a metro area will have a better ROI than mesh.... when it's available.....less installed gear, less to maintain and therefore less costly. The question is when?, battery life, and if the spectrum holders deploy.

AnonDOG
@12.160.x.x

AnonDOG to Nowireneeded

Anon

to Nowireneeded
I am not going to argue Vivato's "claims". What they had and what vendors are doing with Wimax are "apples and aardvarks". This is well documented that their "claims' were self serving.
Please, don't be offended. That was never my intention. Vivato was IIRC the first company to apply beam steering antennas to WiFi. I believe they were the company which convinced the FCC to relax the PtMP rules for their APs.

I was always one of their detractors, from day one. Still I never thought anyone at Vivato was lying or being self-serving. In my NSH opinion they simply deluded themselves. The best RF engineers sometimes forget that "physics are physics" and that the real world is a very different place from the math world.

See my further comments about the model, and the mesh density required to hit -77 dBm RSSI. Those comments were prompted by your question about noise floors.
BTW, With my comparison if Wimax vs mesh coverage modeling I should have stated "licensed Wimax".
Ok, no problem. Still not the same thing. The object with the mesh is to exploit the existing installed base of WiFi enabled mobile devices.
And yes I have seen EVDO work at more than one mile in-building. Several of my clients have Verizon or Sprint service and they constantly monitor results.
The comment I made was 50 percent indoor penetration over 1 square mile. It is one thing to hit someone ten miles from the tower thru a window and quite another to hit fifty percent of every indoor square foot within a square mile.
The point is a licensed Wimax play in a metro area will have a better ROI than mesh.... when it's available.....less installed gear, less to maintain and therefore less costly. The question is when?, battery life, and if the spectrum holders deploy.
Mesh is a tool, just another tool. WiMAX is a tool, just another tool. Of course we are going to try to pick the best tool for the job. The best tool for hitting WiFi enabled mobile devices may be mesh. Then again it may not. The only way we are going to positively know is to deploy some and see what happens.

The next part is just an opinion. It isn't directed at anyone in particular. It is intended to place the arguments which inevitably follow the words "mesh deployement" or "muni-wireless" in a context I personally believe to be true. Again, it isn't directed at anyone in particular.
quote:
This carping by the WISP community about the evils of the Earthlink deployements is two things. First it is protectionism. This is because there are hundreds of WISPs who have built pretty large networks using WiFi hardware. They stand to loose a lot if Earthlink deployes a hundred APs per square mile underneath their networks. The pundits who told them it was a good idea to use ISM and WiFi to build out these networks are the most outspoken critics. I wonder why? We all knew that 2.4 GHz was ISM spectrum and that anyone could deploy pretty much anything they wanted in it. We all knew that we had to endure interference. Now there are hundreds of WISPs and dozens of pundits in the community feeling possessive about spectrum they can never own... so they use their voices to argue deployements in their cities saying they are completely untenable. To those people I say what Earthlink will inevitably say, "There is no right of first use in the ISM spectrum, you knew that when you installed your first AP."

Second this carping is fear mongering. Who cares if the Philly deployement gives Earthlink or even WISPs a bad name? If you are a good ISP and your customers support you, Earthlinks success or failure will have little to do with you. Neither is that success or failure going to stop me from putting up our next tower, or the one after that. One thing I have learned, if I put up the infrastructure, the customers will come and once they get a taste of good service they never leave. John Q. Public could care less whether or not Earthlink fails in San Francisco and I'm riding with John.

lutful
... of ideas
Premium Member
join:2005-06-16
Ottawa, ON

lutful

Premium Member

said by AnonDOG :

There is no right of first use in the ISM spectrum
Excellent quote, AnonDOG.

I like to think that newcomers with sound business plans will have a chance to compete in unlicensed bands even if they are late to the game.

That is definitely not the case with licensed bands where legacy matters. For example, someone has to buy out »www.look.ca/ and rip out their 5-year old 3.5Ghz LMDS setup with thousands of TV and internet customers before they can install something better.
Nowireneeded
join:2004-02-11
Montoursville, PA

Nowireneeded to AnonDOG

Member

to AnonDOG
AnonDOG, I have said all along that mesh is just another tool. I have also added that I am definitely NOT convinced Mesh is the 4th leg of broadband behind fiber, cable, and dsl. IMHO it doesn't fit on projects surpassing it's limits such as total coverage blanketing major metro areas. It should be used in the restaurant districts, parks, business blocks, etc as a "convenience" factor or "hot spot" extension. Instead Mesh is now seen as the 4th leg of broadband and a bridge for the underprivelaged to cross the digital divide.......it's way out of hand and unrealistic. You can now throw in the fact that Muni's want mesh umbrellas built for free and somehow the provider is supposed to make an ROI with advertising? And some RFP's state there will be revenue sharing for using the cities assets? It's totally unrealistic. But then again I'm waiting for someone to prove me wrong.