For the past few months my Purple air II has been reporting less than 100 percent concordance between A and B. I’ve cleaned it out multiple times with a vacuum and even sprayed compressed air into the 8 holes. Both A and B data curves of pm2.5 basically look the same but the A line is always a bit higher than the B line. Does this mean that my A sensor is dirty/bad? When is it time to replace the sensors?
I am sharing the general steps we do in case this could help your decision:
We like to do a linear regression of channel A (x-axis) and channel B (y-axis).
We have seen one channel report slightly higher then the other but have an R2 >0.9.
This is considered ‘good’ for a PurpleAir sensor.
If the R2 is <0.7 for a prolonged period of time (typically more than 2 weeks), 1st cleaning and then replacement if cleaning doesn’t improve the R2 value.
Another consideration for replacement is if one of the channels gets stuck on reporting very high values (>1000) for PM2.5.
Note, one can replace the laser counters instead of the whole sensor. The procedure on how to do that is in other community forum posts.
1 Like
Thanks. It’s been stuck at 88 to 94 percent no matter how much cleaning I do.
I’m wondering if I should pull out the sensors as if I was replacing them and clean them out when they are out of the white shell? And then if that doesn’t help then replace them…m. I bet there’s a bug or something stuck in there.
I took it apart and gave it a thorough cleaning and a good suction with the vacuum cleaner. How long should I wait before declaring success or failure?
just as a follow up I took it apart and gave it a good cleaning with my vacuum cleaner, and now it’s reporting even lower – 74%. I ordered new laser counters, coming tomorrow. We’ll see how that fairs. I think it’s been several years since I’ve installed them, so maybe it’s time…
Is there a way for me to send back the counters to purple air (I can pay for shipping) and have them conduct a post-mortem, or is that not helpful for anyone?
Hi, I may be a bit late in replying. We have some generalized criteria for the R2 of the channel A and B scatter plot with linear regression. If an R2 >= 0.9; we’ve been considering that ‘good’; an R2<0.9 but >0.7 something to watch, and if prolonged R2<.7 likely need to clean and/ or replace. We typically do a channel A / B linear regression weekly.
Here is a another good resource page on the community forum:
it’s up to 96%. By R2>=0.9, does that translate to the 90%+ reported by the Purple Air? (ie if mine is reporting 96%, then R2 = 0.96?
Hi Eric,
I do not know how PurpleAir calculates their confidence score - the % you see reported.
They do need two counters to determine a confidence score, a channel A and B.
The confidence score may be related to the R2 of a linear regression analysis. I did a quick search in the community forum and didn’t see an answer as to how the confidence score is calculated, just that high is ‘good’ and low means likely a problem with one or both channels. Perhaps a question for PurpleAir as to how the confidence score is calculated.
Regards
1 Like