[aprssig] APRS Air Quality Monitoring

[KF4LVZ]

On 2018-11-11 18:30, Steve Dimse wrote:
> 
> 
>> On Nov 11, 2018, at 8:35 PM, KF4LVZ <aprssigZbr6 at acarver.net> wrote:
>>
>> You can scale the output by changing the airflow.  Swap the fan out for
>> something with 10x the flow (as small diaphragm vacuum pump with
>> throttle valve so you can tune it) and the 0-999 will actually be
>> 0-9990.  It reduces the resolution by 10x but you'll have your range.
>> Just put a filter in front of the vacuum pump to keep the dust from
>> fouling the diaphragms and valves.
>>
>> I've done this at work with commercial detectors to tweak the range as
>> needed without swapping out for other sense heads (expensive).
> 
> My guess is this sensor works differently than the others you have worked with. I believe it is a laser that measures the light scattered at different angles. The bigger the particle the more light will be scattered at larger angles. It has quite a large chamber (3 inches on a side) and uses a small fan to exhaust the chamber.
> 
> The first test I did of this was to blow out a match right next to the inlet. By removing the match quickly I give a pulse on input, and it reacts just as one would expect if it were based upon the average number of particles inside the chamber, with a gradual (meaning ~15 seconds) fall after removing the smoke input. If this were based upon something that was flow dependent (number of counts past the laser) I would expect the cutoff to be sharp. 
> 
> I'm curious about the sensors you are talking about. While I could understand methods that depend on a calibrated flow, e.g you could get particle concentration by counting particles passing a point, I would expect that increasing the flow would increase the count, raising the sensitivity of the sensor rather than decreasing it. I don't see how increasing the flow could decrease the sensitivity. What am I misconceptualizing? 

I've seen this type of inexpensive SDS011 sensor and similar and own a
few as well, they don't use angles, it's just a laser diode and a
photodiode with their optical axes at a fixed off-angle.  The intensity
of the back scattered reflection is binned to the size of the particle.
Here's a picture of the innards of the SDS011

https://aqicn.org/sensor/sds011/

There have been papers (I'll see if I can find a reference, I dug it up
at one point) that studied these and similar.  The result is that the
smaller PM values have to be offset by the larger PM values.  For
example, suppose you have a PM2.5 reading of 500 and a PM10 reading of
10 (raw readings).  The actual value of the PM2.5 is approximately 490
because the 10 PM10 also got binned into the PM2.5 reading.  The actual
offset would have to be calibrated out but the algorithms in the
on-board controller aren't well characterized however it is entirely
dependent on flow for the binning.  If you cut the flow down, a
reflection is assumed as a particle but it's assumed that was one
particle within a unit time (and therefore a unit flow) and then that's
converted to the ug/L reading or #/L (depending on the sensor software).

If you have a standard flow in the sensor as delivered of 10 L/min and
there's twenty photodiode reflections within a minute(let's assume only
one PM value) then the software will tell you it counted 2/L (20/min *
min/10L = 2/L).  If you slow the flow down to 1 L/min the software
doesn't know this.  Then if you get twenty counts under a 1 L/min flow
technically you have a reading of 20/L but the software will still tell
you 2/L because it still thinks it's at 10.  Speeding it up to 100 L/min
would go the other way of course.

As for the sensors I have at work, they're physically much larger and
use two methods, a diffraction grating  with a multi-pixel detector to
automatically measure the particle size based on diffraction of the
light around the particle as well as a second detector that measures the
light shadow (similar to the SDS01) and correlates shadow with the
particle.  The flow path is also very narrow to attempt to prevent more
than one particle at a time from crossing the laser path.  The SDS011
and others just have a wide open path.