I worked hard to quickly deliver a comprehensive Omron Sensor Evaluation Board 2JCIE-EV RoadTest Review as it seemed to be a rather straightforward. This was because I had other things to worry about including the current RoadTest I'm undertaking and dealing with the fall-out from a previous RoadTest that is still awaiting its ultimate conclusion. One of the key reasons I wanted to evaluate the Omron was for the B5W-LD0101 air quality/dust sensor, as I have never dealt with a pulse-out LED-based optical sensor. It does have some vague similarities to photoelectric smoke alarms, but my experience is usually with the laser-optical particulate monitor sensors which are more tuned to monitoring fine dust.
As I discussed in the RoadTest review, the two sensors actually do different things. The B5W-LD0101 has two analog outputs, one that counts the number of particles of 1um or larger, the other for 2.5um or larger although the upper limit is not specified. The SPS30 instead has plenty of digital outputs which come courtesy of its internal computations that supply mass and counts per cubic meter of air for 0.3um through to 0.5um (count only), 1.0um, 2.5um, 4um or 10um (both count and mass). Note that the measurement tops out at particles up to 10um, thus is probably not going to register larger dust particles.
One of the disappointments were that the MKR 1010 Wi-Fi had dropped off the network prematurely, spoiling the intended data analysis which was going to be over a longer period. As a result, after the publication of the review, I reset the apparatus and tried again. I'm happy to report that this time, I was able to achieve 24 days of uninterrupted data capture from both B5W-LD0101 and SPS30 sensors and perform some analysis to see how well they correlate (if they do). Another helpful thing was the commencement of bushfire prevention back-burning which introduced periods of higher particulate levels to provide some extra data.
As with the shorter period of monitoring during the RoadTest, the B5W is seen to have a "floor" value around 200, but there are occasional spikes below. The B5W also had a number of spikes which seem uncorrelated with the SPS30 - likewise the SPS30 had an occasional spike that doesn't correlate with the B5W even though they were both placed behind a fly-screen on the outside of a window, ingesting outside air. The SPS30 figures seem rather clear with regards to trends and some comparison with our Department of Planning, Industry and Environment's monitoring stations shows that the SPS30 is accurate.
A raw scatter shows that there is a slight correlation, but this is perhaps quite weak and spoiled by the "noise" which results in large spikes. There is a bias towards low particulate matter levels, as the majority of the time, the air quality where I am is rather good.
I've tried to remove some of the obviously outlying data (e.g. >400 counts on the B5W for <50 counts on SPS30, and >1200 counts on SPS30). The resulting automatic linear trend line has a poor fit and seems to be quite a bit steeper due to the bias in the low-concentration data. The visually estimated best fit shows a relatively low sensitivity of the B5W to increasing PM10 counts, meaning that the B5W is probably not a replacement for a laser particulate monitor even though it is much cheaper and seems to do a similar thing. Perhaps the B5W is more sensitive to larger dust which is throwing off the figures or the majority of the dust is very fine particles below the minimum detection capability of the B5W. Perhaps the data from the B5W could be improved by using some filtering and scaling to try and better match the SPS30's outputs. But at least, now I am satisfied with knowing that the two sensors are not going to give data that is entirely equivalent.