WiFi Design and Deployment Methodology

In each of my weekly classes, I do a full walk-through of the “Define, Design, Deploy, Diagnose (Validate/Optimize)” process. It’s never a surprise when most of the students are excited to see a documented, vendor-neutral walk-through of the process.  I’m always asked for a copy of the flow chart (shown below) document so that the students can give it to customers and to sales professionals. As the old saying goes, “A shortcut is the longest distance between two points.” Shortcuts only cheat you, and my hope is that everyone starts to realize the value of following an efficient and proven process in WiFi design.

The Problem: Lack of Understanding

To help you understand the problem, I’ve taken a page out of Jeff Foxworthy’s playbook.

You might be a WiFi Magician if you’ve ever been asked to “survey, tell me what’s wrong, and tell me how to fix it.”

You might be a WiFi Magician if you’ve been asked to give an AP count, without having done any design work.

You might be a WiFi Magician if you’ve been asked where APs should be mounted, and you’ve never seen the facility.

As part of the design process, it’s common for an engineer to face such questions or requests, and being able to explain industry best practices benefits everyone involved.

The Solution: Best Practice Process

Admittedly, I have no graphical skills, as evidenced by the less-than-artistic flow chart below. I’m certainly happy to entertain volunteer help to make it Apple-esque.

The chart is broken into five distinct sections: 1) Define (red), 2) Design (blue), 3) Deploy (white/gray), 4) Validate (green), and 5) Optimize (yellow). The Validate and Optimize steps are interwoven, so we usually refer to them as a single step in the model.

 

Define

The two most important concepts in the beginning of a WiFi design are: requirements and constraints. Without knowing the details related to these two concepts from a project’s onset, it’s a given that time will be wasted somewhere along the way.

• Requirements – goals and desirable outcomes (technical and organizational) for the project
• Constraints – limitations (funding, timetables, vendor selection, etc.) for the project

In the beginning, floor plans are needed if available. Certainly, there are times when floor plans are unavailable, unusable, or outdated, but they are very necessary for planning/modeling and survey/validation work. This may mean that the architect or network administrator has to create or obtain usable floor plans by some method. Admittedly, this isn’t always a simple task.

Given reasonably accurate floor plans, with scale information, an architect should be able reasonably estimate the time it will take to perform the design and/or validation work. At this point, business-related documents, such as a mutual NDA, professional services agreement, and scope of work (SoW) should be exchanged between the architect and customer, as required. The SoW document should usually contain information having to do with the services to be performed, information regarding project requirements and constraints, the services pricing, and the timetables on design and/or validation services (depending on what the customer requested).

During the Initial Site Visit, which is considered a required step unless the facility doesn’t yet exist, vital information is gathered for use during RF modeling. I use an iPad Mini 4 with the PDF Expert application (which has cloud integration for moving files back-n-forth) to annotate floor scale information, wall loss values, time-stamps for my spectrum analysis recordings, and any special information about the facility (like high density areas). When you’ve finished annotating your PDFs, do not forget to “flatten” the PDF so that your annotations will be embedded into the floor plan PDF.

Stakeholder meetings.

• It’s advantageous to meet with the IT, end user, and management team stakeholders when feasible. This gives the architect a rounded perspective on requirements and constraints and often produces additional benefits for the customer.

Scale floor plans.

• CAD files are pre-scaled, so if you can get your hands on them, do. If not, then you’ll need a good laser to measure the longest hallway (preferably) on each floor plan.

Validate floor plans.

• Just because your customer could produce floor plans does not mean that they are accurate or to-scale. More often than not, floor plans are drawn incorrectly or have changed since the version that you were given. You should validate the floor plans with a simple walk-through of the facility. I often convert my floor plans to PDF and then open the files on an iPad Mini in the PDF Expert application, which has cloud integration for moving files back-n-forth. From there, I mark up the PDF where walls shown on the floor plans are inaccurate.

Spectrum analysis.

• I suggest performing a spectrum analysis of the entire facility, using spectrum analysis software, prior to performing any modeling work. The purpose of this walk-through is to identify, locate, and remove any RF interferers prior to deploying the new WiFi network. If you decide to use dedicated spectrum analysis software, you will have to tie the data to a map in some way. I use my iPad Mini with PDF Expert, as mentioned above. When you perform the spectrum analysis, start a recording. Whether you annotate interferers in real-time or review the recording later, when you see interference sources, note a time-stamp on the PDF.

Find/document alignment points.

• In order to align floors for a 3D building design, you will need to note your alignment points on each floor. In Ekahau’s Site Survey tool, you will need to have at least three alignment points per floor, spaced out around the floor. I use a “big, fat triangle” approach, normally locating stairwells, elevator shafts, and building corners/edges, depending on what’s available on each floor. Finding and documenting (on your PDF floor plan) these alignment points is FAR easier when on-site than it is when looking only at the maps. When on-site, you can “go there” and see where each stairwell and elevator exits on each floor.

Measure wall loss at 5GHz.

• In order to measure wall and door loss (in dB) of each wall (or set of walls), I use a DLink DIR-510L (also called an AC750) dual-band AP/router. It has an integrated battery that lasts for a really long time, and it’s highly configurable. I usually use a reasonable amount of output power (on 5GHz) for taking these measurements, like 13dBm (20mW), a 20MHz channel width, and an easily-recognizable/unique SSID. For taking measurements, you can use most any dual-band device that has a scanner. You can use a smartphone, tablet (such as an iPad Mini with the AirPort Utility’s scanner function), or purpose-built diagnostic device. I’m always asked, “do you measure every wall/door?” The answer is either “yes” or “mostly”, depending on the facility’s construction.  I like to tell students that “it’s wrong unless it’s measured.”

Understand the client device specs.

• When asked for a list of all client devices in use on the network, customers resist, but when asked for a “top-10” list, customers usually accommodate. Being transparent, you often design WiFi networks around the least-capable, most-important client device. This is, many times, a 1×1:1 device, such as a smartphone, since it’s the “worst case scenario” for receive sensitivity and processing power. Sometimes, the design is focused around a 2×2:2 tablet or low-end laptop, such as in the K-12 market, but of course, that means that 1×1:1 devices may have problems roaming properly. Some devices may be 2.4GHz-only, and some services may only be provided on 2.4GHz, such as RTLS (location services for tagged items), so keep that in mind during the design.

Understand cabling locations/limitations.

• There are times when cabling is not really part of the design equation because the customer has up-to-date cabling in all or most of the locations where APs will be mounted. Other times, cabling may be run inside conduit, and one of the customer’s constraints may be that no new cabling will be run, which could certainly impact the RF design. In some facilities, such as when there are thick, pour-concrete walls, running cables may be very difficult, expensive, time-consuming, or impossible.

Understand aesthetic requirements.

• Before deciding on where APs should be located, you should find out from the customer where APs are allowed to be located and mounted due to any aesthetic requirements/restrictions. This could drastically change the RF design.

Optional: find & document existing AP locations.

• Some customers do not currently know where some (or any) of their APs are physically located within/around their facility. For this reason, if they are going to remove/replace the existing APs, they need to locate them. When I am asked to locate and document existing AP locations, I typically do so using a PDF floor plan on my iPad Mini. It sometimes helps to use a WiFi scanner to help locate them, but oftentimes you need to perform a thorough physical exam of the facility to find them all.

Predictive model validation.

• By temporarily placing an AP (like the actual APs that will be installed) within a facility, usually as an “AP-on-a-Stick (APoaS)”, configured for a modest amount of output power (EIRP), e.g. 10-14dBm on 5GHz, and measuring the desirable signal level boundary using the “least-capable, most-important” client device, a signal level offset can be understood and documented as an adjustment for the RF modeling software. For example, if a temporary model (with correct walls drawn in around the AP location) shows that the -65dBm boundary extends to a specific set of locations around an AP placement, then measuring those boundaries with a real device should show if the RF model matches the FSPL calculations of the modeling software. If the two values do not match, the average differential (in dB) should be used to adjust the RF modeling software.

Design

Once the on-site RF and building-related information is collected, then the 2D (single floor) or 3D (multi-floor) RF model can be constructed in software. The RF design should focus around the requirements and constraints given by the customer, keeping in mind any client density/capacity requirements, minimization of CCI, client roaming, RF redundancy, and/or RTLS enablement. Once the RF design has been completed, then an RF Design Report and/or the project file(s) should be given to the customer and another copy archived for safe-keeping. It is also my practice to give the customer all of my source documents, as I believe this promotes trust and transparency with my customers, while simultaneously giving them adequate technical information for upgrades, modifications, and/or redesigns at a later date.

Deploy

Deploying the network can logically be broken into three phases: quotation (for equipment and/or services), installation, and configuration. Per the RF Design Report, the number of APs and controllers can be determined and added into the equipment quotation. If the customer requires professional services, then this also can be added into the same quotation. Next, the WiFi related network components should be installed, e.g. controllers, WNMS, NAC, BYOD on-boarding, and other system components installed into the data center(s) and APs hung in their correct locations. Thereafter, the system can be configured according to the security/performance needs of the customer and the RF Design Report.

Validate

The validation phase can be broken into two distinct sections: RF Coverage and Device/Application. When validating an RF design and deployment, then coverage and CCI are checked by performing a passive (listening-only) RF site survey and optionally an active RF site survey (moving data to assess performance characteristics). The second validation is for how specific devices (as listed during the stakeholder meetings) and applications are performing (coverage, throughput, roaming) on the network.

It’s important to note that an RF site survey shows the coverage and CCI, but does not give the RF engineer an indication of how to repair the problems found. If the validation shows that the RF requirements and constraints are met, then the network is considered acceptable and the validation report can be created and given to the customer. If adjustments are necessary, then the optimization process will begin.

Optimize

The optimization process is broken into two sections: Fine Tuning and Coarse Tuning. The fine-tuning steps are configuration-based (as listed in the flow chart above) and come first. They are the simplest steps that can be taken to adjust the network to meet the customer’s requirements and constraints. If fine-tuning steps do not adjust the network to within the customer’s requirements, then course-tuning steps, which are typically physical adjustments (as listed in the flow chart above), must be taken.

Both fine-tuning and course-tuning steps are very slow in comparison to network modeling, so great care should be taken during the initial on-site visit to collect accurate information. As the old saying goes, “measure twice, cut once.” Adjustments in the real world are very slow indeed, and so the RF model should be, at a minimum, 99% accurate. Before you argue that a 1% error margin isn’t feasible, consider that in a 5,000 AP network, like a large hospital, a 1% error margin means physical adjustment of up to 50 APs.  Physical adjustment of 50 APs could take a week or more. Today, existing RF Modeling (RF Planning) tools and RF Site Survey tools are not bi-directionally integrated. This means that when performing an RF Site Survey, the software does not indicate what changes could be made to repair RF problems – nor do they allow for an RF Site Survey to be virtualized into an RF Model for adjustment in software (which is really fast). For this reason, we have to make sure to get the up-front data collection right the first time. Running the tuning loops over-and-over will practically guarantee wasted time during the project.

Summary

As the old saying goes, “the longest distance between two points is a shortcut.” If you try to shortcut the RF design process, everyone involved gets cheated: manufacturer, VAR, and customer.  This process is as efficient as it can be given today’s modeling and surveying software capabilities, but of course, that’s always improving. WiFi is an ever-evolving technology, so as software capabilities are enhanced, so too can the RF design process be enhanced.