ONT Rx Power vs. Fiber Length Over Time
This post is the last in our short series exploring light levels (signal strength) in PON fiber networks.
In the first post, we looked at Rx power vs. fiber length. The second post explored attenuation (Tx power minus Rx power) and its daily variability.
This post looks at Rx power over seven days. The standard deviation plots in the second post showed that the light levels for many ONTs vary. For most, the variation is small, but there are some outliers.
The variation plots were based on an average of the per-ONT samples over two days and therefore, time was removed as a dimension. What happens if we look at that distribution over time? Are there any patterns we can observe visually?
This video represents a random sample of 2,500 ONTs, with five-minute samples for seven days. Each sample is the mean of the one-minute metrics we collect. 2,500 is a small subset of the ONTs in Preseem. Plotting all of them made the plot unusable.
This video shows a couple of interesting things:
- There are some ONTs that are highly variable (they jump around a lot)
- A portion of the distribution shifts up and down on a regular daily cycle
This raises a few more questions.
Why the regular daily shift? Temperature change?
What’s up with the jittery ONTs? Are they repeating on some shorter cycle that’s hard to identify with this animation? Conversely, what’s special about the stable ones?
For those interested, this was built with Python, Polars (data processing and loading Parquet files), Matplotlib, and OpenCV (converting plot images to video).
