Flaws of Wall Attenuation Measurements

Over the last 5 years, we’ve seen a huge increase in popularity the Wall Attenuation Measurement process. The idea is to go on-site and measure the actual attenuation of a wall in order to create a more accurate predictive model. However, I find this process and results fail to dramatically increase accuracy of predictive models, especially in disaggregated teams. In this blog, I’m going to outline the issues with this process and why it doesn’t ensure your RF model is relatively accurate as well as briefly describe the process I recommend instead of Wall Attenuation Measurements.


WAM Process overview:

There are lots of variations on this process. The general idea is that you place some sort of device advertising a Wi-Fi beacon several meters from the wall you want to measure. Some recommend 2,3,4 or 5 meters, with the idea being that you you want to have as little loss due to FSPL as possible. Then take measurements on both sides of the wall you are interested in and do some basic math to calculate the db loss of the wall.

Some examples for additional reading:

Flaw 1: Lack of resolution in measurement

One of the major limitations of this process is that I’ve yet to see a device that can measure and report RSSI in fractions of a dB. This means that your wall attenuation is always a whole number. Combine this with the fact that most common wall attenuation metrics are low values (1-8) this means the overall accuracy of the number compared to the scale of the number is fairly low.

Many folks recommend averaging these numbers, but since the resolution is so poor and it doesn’t support fractional values, you typically end up with the same values, which don’t contribute toward averaging. So if you have a wall that should have around a 2.2dB, you’re either off by .2, or off by .8 per wall. Multiply that by every wall, you can move the cell edge a long way.

Flaw 2: Sum of Micro measurements are not the macro measurement

You’ll often hear me refer to an WAM as a micro measurement. The reason for this is you are measuring a point in space. Folks think that if they take all these measurements for every wall that their model will be accurate. However, you generally are only measuring a single point in the vertical dimension. This means that you are not accounting for the affect of attenuation material above or below the measurement height.

Effectively, you are trying to predict how far a cell will propagate by measuring individual walls, but you don’t really care what the value the wall is. What you really care about is picking a value for walls that results in a reasonably similar cell size and shape. The failed premise is that if you measure all the micro measurements, they some how will line up to a good representation of what RF propagation does.

This is similar to “inside-out” passive survey technique. You can’t calculate the RSSI along every wall simply by knowing the RSSI in the center of a room.

Flaw 3: Accountability, Reproducibility, and Verifiability

While these are very different flaws, I group them because they are detrimental to performing Wall Attenuation Measurements in distributed teams, which is where i spent 10+ years of my career. Accountability means that there is not a great way to verify that the measurement was actually done and done properly. If using a process like recommended by Ekahau, you can actually walk the wall on both sides recording this data. This gives you some accountability, but not the other two. And this takes time. If you follow my recommended model, the delta is not that much and you get much more reproducible data than simple wall attenuation measurements.

Reproducibility means that anyone else could go and reproduce the same values quickly and easily, both in an actual deployment and in a predictive model. This is where these types of attenuation measurements fall down. I’ve had multiple instances where my cell edge is off by 10db, but measuring each wall. For reference, 10dB is enough to take an iPhone from being happy (-65) on it’s current connected AP, and it disconnecting from the network (-75). But when measuring individual walls, all the measurements line up.

Finally, verifiability means being able to use what was collected on-site to verify that our predictive models are relatively accurate. In this case, we have to assume that if our wall attenuation is correct, that the predictive module using those measurements is relatively accurate. However, these don’t directly correlate. Ekahau and many other tools don’t calculate reflection, diffraction, multi-path and more aspect of RF, leaving you trying to measure attenuation when these other factors are in play.

My Preferred Model

In my years developing “hybrid” processes, the goal was to ensure that someone could go on-site, collect data and that data could be both used to build and validate a predictive model.

The first major change to the process is to use a known quantity. Just any device that advertises a WLAN is not good enough. Pick something with an antenna pattern in Ekahau (or whatever tool you are using) and something you are generally comfortable with the antenna pattern on. By using this, you can compare a predictive model vs what you actually measured and see how close they are. This gives you the ability to compare and contrast cell sizing. When doing this it helps to have a standardized channel, width, tx power, height, etc that you know.

The second, is not to measure just the wall. Do a full cell survey of the AP in a location. This way you are creating an RF fingerprint of the building and how RF is affected. Repeat this as much as needed to take measurements of all “interesting” walls in a building. This is not intended to be a full AP-on-a-stick (APoaS) survey. Just enough data points to collect empirical data on the things that are interesting. Figuring out where to do these to optimize for the number of measurements is a thing.

Now with this data in your survey application of choice, you can create a predictive model that has similar cell sizing and shapes compared to the real data you collected. You know the location of the AP, the height, the orientation, transmit power, channel, etc. And you can directly see the effects the building had on the RF. Using this you can create wall types that result in predictive model that reflects what you measured on-site. And without the measurement limitation of whole dB, you can specify walls with very specific attenuation values. Using the data in Ekahau, often you can compare the channel, tx power, etc out of the recorded data to ensure that the equipment was setup properly.

Now have someone review your work. They can look at the APoaS survey and determine how well your predictive model matches those results and decide if it is “good enough” or if you need to refine your model. This gives you some flexibility to spend as little or as much time on the model to meet the customer requirements.

Final thoughts:

Wall attenuation measurements are a valid option to “get close.” They may even be the right option for the very budget-constrained and risk-tolerant customers. IMHO, it’s slightly better than guessing, but only just. And all too often I get asked to review someone’s work who’s used these processes and tell the customer how they missed so badly. Without the ability to review that data, the result and how it impacted the overall cell, I’m left telling customers that their predictive models just didn’t match reality.

2 comments

  1. Very interesting article Jake, thanks for writing it up and keeping the community always questioning the status quo. I totally agree that the typical wall measurement method risks some big misreads, but I can’t believe that with your recommended method “the delta is not that much”.

    I’m sure repetition of your method might lead to some efficiencies but if you’re surveying the cell edge of your temporary AP that is a lot more ground to cover and measurements to take (and review) than just jumping into the room next door. I presume you’re saying the saving comes from 1 AP location to cover multiple rooms rather than moving the test AP into every room to measure every wall.

    I question how your method differs to APoaS though. You did mention less positions than a full APoaS but I’ve heard many stories of wall measurements finding a normal looking plasterboard/drywall with massive attenuation because of hidden objects that your system risks missing.

    This article leaves me wondering which method is the lesser of two evils.

    Like

    1. If you are recording wall attenuation measurements on the map in Ekahau, it takes substantially longer than just taking measurements and writing them on a map. The time delta between performing that process and doing a full cell is not a large delta. So if you have accountability for the process, why not get the rest of the benefits?

      If you boil this down to doing a full APoaS survey (another design methodology) it can take dramatically longer.

      APoaS typically involves you putting the actual AP you intend to deploy in the actual place you intend to install it and with the power you intend to deploy it at. My process has none of those limitations, as you are trying to match a known entity to empirically collected data to provide feedback on how the environment affects RF.

      As to how many positions you need, that’s not in the scope of methodology. It’s flexible enough that you can do as little or as much of the building as you determine as necessary. So while I agree there is the potential to miss stuff, it comes down to picking where you measure and how much.

      Like

Leave a Reply to Andrew McHale Cancel reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s