Forum

  • While discussing with a professor (well known cellular networks engineer) about site surveys considerations, he asked me if Wi-Fi engineers use any good predictive survey method. I replied that there were, but they would not reflect the reality of the environment since APs behave differently, specially with technologies such as spacial multiplexing and beamforming. So the best way would always be to do the physical survey.

    He said that in the cellular world, they use predictive surveying almost 100% of the time, since it's impractical to build a tower just to "see if it will have enough coverage". So the math has to be as accurate as possible, both coverage and traffic-wise. Then he said "what if a building is in construction, and they want to know where to put AP's power and network support?". He has a point. I haven't been doing Wi-Fi for too long, but I am used to grab my laptop and walk around to see how my environment is doing. I didn't have a reasonable answer to give him, although I knew there is one.

    I would like to know everyone's opinion about predictive surveys and about RF math applied to Wi-Fi networks. Sorry about the long story, but his questions really got me thinking.. Thanks in advance!

  • My standard is, predictive survey for new constructions sites is for cable drop purposes only. During the predictive survey, do not be afraid to go a little heavy on the cabledrop placements. Then, once construction is complete, preform and active AP on a stick survey while choosing the closest cable drops for AP placements. You will have left over cable drops but you will have a GREAT and reliable survey!

  • Hi Luis,

    Have a look at this great free book (Math and Physics for the WLAN Engineer):

    http://www.wildpackets.com/elements/whitepapers/math_physics_jbardwell.pdf

    As I work for a WLAN vendor, we have guidelines and approximates that can be used with our predictive site survey tool in order to come up with an estimate when dealing with new constructions. Of course, if you can influence the cable drops, this is a great way to make sure your final pre-install survey will be reliable :)

  • So you guys are saying that you do use predictive surveying, but you always round up so you have enough room for changes afterwards? Is that correct?

    PsychoFin, I've already read this book and I understand a bit of RF math. What I'm really wondering is if and how it's used by WLAN pros to predict the size o BSSs, AP placement, and so on.

    Thanks for the info, guys!

  • I've used a couple predictive survey tools, one of which is made by the company I work for. The more complete and accurate the data fed to the tool, the better the output. Predictive surveys are usually rushed through as a pre-sales tool for budgeting purposes, so sometimes corners are cut and the final design and AP placement end up looking different.

    Predictive surveys, especially for buildings that aren't built, can't account for many factors, like asthetics, materials not in the blueprints (lockers, bookshelves, mirrors, furniture, human beings). I call "no fair" in comparing it to cell tower planning -- I've never done that, but I imagine that planning for outdoor coverage with licensed frequencies and different output power regulations is a whole different ball game.

    That said, if wi-fi designs were given the resource, time, and forethought of cellular designs, then the predictive designs would be more accurate as a whole. In a competitive bid under sales pressure, that may not always happen.

  • Thanks for the info, @@ron! I didn't think about how the licensed frequency band would make that easier. Also I haven't considered how much more Telcos invest in Cellular networks compared to Wi-Fi, since they get much more money back from their Cellular networks.

    Even though the power regulations are different, I don't think that weights too much because the clients wouldn't have enough Tx power to be listened by the cell sites, even if the cell sites have a lot of Tx power.

  • I assure you, cell tower engineers would do pre deployment active surveys if they could. I guess they don?t have a big enough AP on a stick. Lol Even though the Cell engineers do predictive surveys, they highly test the environment. Its not like they never visit the site before creating the prediction. Like @@ron said, they put a lot of effort into there predictive. Likely 5 times as many hours as it would to do an active wifi survey depending on the site. And it also helps that they are working with a licensed frequencies.

    Wifi deals with unlicensed frequencies anything from microwaves to radar signals can interfere with your proposed network.

    I know some customers can not afford a pre and post deployment active surveys. I also know that it could be hard to win bids when you bid to the customer that 2 active surveys need to take place.

    If money is the excuse, at least visit the site to pick out the AP placements to place in your predictive tool. Also this gives you a chance to test the environment.

    If money permits you from visiting the site at all before the predictive, all you are doing is basically guessing. Do you even know what the walls are made of? Whatever the case, always preform a post installation ACTIVE survey.

    If the customer is requesting that you produce a predictive survey for the bid or just after the bid, you should add some lingo in your bid proposal. Maybe say something like an active pre deployment survey is highly suggested and is ?BEST? practice. This proposal is for a pre deployment predictive survey only. If you would prefer an active pre deployment survey, please request a new proposal. Or just propose both solutions. It depends on the relationship you have so far with the customer if they will trust you are not. You could also even have references from the CWDP book or other sources to back up your suggestion.

    Predictive survey is not a survey? it is a prediction. And depending on the tools and information you have to go into those tools, it could be a pretty bad prediction ;-)

    As a VAR I never, never , never did a pre deployment active survey? I did not know the benefits at the time to do so. I was IGNORANT!

    Remember you and the customer benefit on an active survey if the price is negotiated correctly. You get more hours or more money because your solution is ?Premium? compared to others. And the customer gets the 802.11 network they deserve. Most customers are ignorant in that they just do not know. You are the expert. Get the satisfaction of being an expert by creating the BEST WLANs possible.

  • Also, with that being said, as long as you have emphasized the benefits of an active survey to your customer and they still do not want to ?BUY? it, do not feel bad about not being able to design to the very best. If the customer does not want to buy it, sucks for them. Just like when I take my car to the mechanic. I might not believe or ?BUY? what he is saying. He could very well be telling the truth. But it depends on the cost and how much I believe them if I will listen? lol

    Sell it better next time lol

  • After a decade of doing Wireless Designs all over the world, in all sorts of situations... my recommendation is to ALWAYS do a post-installation verification survey. That is a minimum. You'd never pay your cable installer without testing to verify if your cable plant matched specs - so don't ever turn over a WiFi system without verifying it meets specs.

    I've been behind probably a hundred people who did a pre-deployment AP-on-a-Stick "Survey" as their excuse for a design, then never did a post-install verification. When the customer complains its a very quick test to see their precious WiFi system doesn't come close to what they were promised.

    Designing a Wireless LAN is so much more than merely delivery of RF for coverage! You MUST include all the client device's specs - not merely the minimum of some arbitrary RSSI figure.

    Like I've said teaching hundreds of people advanced wireless design classes... "Coverage is Easy" - it just needs more AP's and/or higher power. Just because a client devices can 'see' the AP at a given RSSI value does NOT mean the Wireless LAN is working properly.

    You've got to consider load balancing, primary and secondary coverage, latency, jitter, co-channel interference, capacity issues, density issues, among many others. Most folks do NOT think about all of those issues when performing an AP-on-a-Stick "Survey"... they are usually after only Coverage and then call it a day.

    Using Predictive tools is just that... Predictive. The better your input into the RF Math system, the closer your results will be to reality. I like to go on-site to gather site information, measuring the wall thicknesses in dB loss, prior to doing the predictive survey. Then doing some on-site testing to confirm if the predictive design actually works as predicted in the real world. Then you can finish the design off-site.

    All the suggestions mentioned in above posts, like having additional cable drops, adding maintenance loops at the end of each drop, etc. are valuable as well.

    But the MOST IMPORTANT step in any Wireless Design must be the post-installation Verification Survey. How else do you know if the Wi-Fi is working properly? Talk about 'guessing' - doing an AP-on-a-Stick "survey" and calling it a design is truly 'guessing' if you don't verify it post installation.

    (By the way - one of the worst flagrant errors I see coming behind AP-on-a-Stick induced designs is the total disregard for co-channel interference. It is never measured or allocated for. Yet it causes some of the worst Wi-Fi outages and client problems... yet is never even considered in the "design" process when you're only looking for Coverage)

    Sorry to sound a big harsh here... I've just had this message engrained in my psyche by constantly fixing other people's "coverage-only" designs. By the way - there is NOTHING WRONG with AP-on-a-Stick processes... as long as its only a part of the WLAN Design process - it gives you a coverage model, but as a professional, you have to include ALL the design criteria, not merely the baseline of delivery of RF energy to an area.

    Sometimes in our industry people use the word "Survey" - when they really mean "Design". A Survey is just one phase of the design process.

    Keith Parsons
    CWNE #3
    http://wlanpros.com
    @keithrparsons

  • Like usual, nicely said Mr. Parsons!

    I also mentioned that whatever the case, a post validation survey is a MUST!

    Now days, giving the use case, most sites should not be designed for coverage. They should be designed for capacity. Dual radio APs, cutting lower data rates, and decreasing cell sizes by reducing AP output power are some ways to design for capacity.

    Trying to do this while providing for use case (example -67dB for voice), mitigating co-channel interfearance, providing redundancy, engineering a balanced link budget, and designing for client specific devices can be difficult. It comes with practice and lots of planning.

    Anyone can turn up some APs and connect but you are the expert so do it correctly for that customer's specific needs.

    Check out the CWDP book for some good information and a great read!

Page 1 of 2