Unsure about accruacy of cell site locations

[BShorten]

What steps will reproduce the problem?

What is the expected output? What do you see instead? I expect to see cell locations inline with how a typical network would look on a map. Instead, there seems to to be a concentration around transport routes and far more than I would expect to see. Am I really looking at true cell location or something else, such as measurement points of coverage along transport routes, for example?

What version are you using? On what operating system?

Please provide any additional information below.

[wish7code]

Currently we estimate cell position by averaging all cell measurements for an individual cell. A individual cell is characterized by MCC/MNC/LAC code/cell id/cell tech (2g, umts, lte).

Given the fact, that most of people scan on while walking or driving along roads or rails, this explains the bias for transport routes alignment. Unfortunately more sophisticated approaches (e.g. angle of arrival measurements) are not supported by current cell phone sensors.

[BShorten]

Thank you for this explanation. Just so I am clear. Are the point markers that are displayed and which can be downloaded an estimate of the cell location or where the measurement was taken from? If it's the cell location, does the transport routes bias mean that some cell sites which are not located near transport routes are not captured?

[wish7code]

Are the point markers that are displayed and which can be downloaded an estimate of the cell location or where the measurement was taken from?

--> estimate of the cell location

If it's the cell location, does the transport routes bias mean that some cell sites which are not located near transport routes are not captured?

Not sure if I understand your question correctly - the scanner app saves all 'visible' cell tower information together with the current position.

If the cell tower is 5km away from the road, in the worst cast we would have a 5 km mismatch for the cell tower location estimate. But since we take the average on all measurements mismatch is usually lower.

In case you want to experiment with the data, you might be interested in raw measurement data, which is also provided as download. Drop me a private message, if you need extracts for your region..

[BShorten]

Hi Again.

Apologies if I was not fully clear. In short, I really want to know what points I am looking at. Is it cell site locations or points where the cell phone user was located or both? When you way "- the scanner app saves all 'visible' cell tower information together with the current position," what exactly do you mean by "current position?" I am really only interested in the cell site locations, as I assume most people would be. Is there a way of downloading only the cell site locations? And then, if the road network bias comes into play, how representative are the cell site locations shown of what exists in reality>

[wish7code]

Hey!

Is it cell site locations or points where the cell phone user was located or both

It's the geographic midpoint over all measurements for a given cell.

what exactly do you mean by "current position"?

user's GPS position

I am really only interested in the cell site locations, as I assume most people would be

The problem here is that cell phones have omnidirectional antennas (cell phones don't know from which direction the signal comes from --> no angle of arrival available), so using cell phones there's no easy way to estimate real cell tower position. Nevertheless given the fact, that at least in urban areas, cells become smaller and smaller (micro-cells) and the fact, that typical network infrastructure is often aligned to transportation network (i.e. cell towers are installed near common travel routes), that bias might be smaller than expected in practice.

There are also some ways to further reduce bias, in particular weighted averages on measurements, or simply using strongest signal as proxy for cell tower location (works surprisingly well), but currently that's not implemented.