danwahl / adanet-co2-monitor Goto Github PK

The FeatherWing has an SD card slot, nominally for storing graphics, but this could probably instead be used for logging. Since we only log once every three minutes, even a small card could could effectively be used indefinitely. Some considerations:

• Assuming a ~3 minute cycle (how exact is this?), and knowing the system time when the device was programmed, we could sort of get "accurate" timestamps without having a real time clock
  • This would work less well if the device was reset frequently, and would probably drift over time
• Not sure how the SdFat library work, but each reset could potentially create a new CSV file, either with a datetime stamp or a sequential numbering scheme?
• The existing USB-C slot would need to be slightly enlarged to accommodate the width of the card, and it seems to be slightly off-centered on the FeatherWing as well
• The card would probably stick out of the case slightly, which seems fine

Currently using the "classic" fixed-space bitmap font, but FreeMono would probably look better

Scenario: I want to test it in the elevator and I mentally counted the seconds since the last update, but the trip inside the elevator is quite short.

Scenario: I like to get two readings in a place before I move on.

Pros:

• User doesn't have to wait as long to get a first or second reading

Cons:

• Battery life is reduced

Today I saw an error message for the first time.

“Invalid sample detected”.
I checked the code: it looks like the error is thrown if the reading is zero.

check out the photo

The Low Power Operation document says:

Power Cycled Single Shot Operation (SCD41 only)

But the firmware is not currently using single-shot operation, so I don't think this applies.

Still, worth verifying before purchasing SCD40s in bulk!

As per here.

Aranet4 has Good, Average, and Unhealthy levels at 420, 1000, and 1440 ppm respectively, and the current Adanet CO2_LIMIT is 1000 ppm, at which point the text turns red.

The pseudocode at the end is missing LC709203F

Automatic Self Calibration is currently disabled, as per the Low Power Operations document:

ASC is not supported for power cycled single shot operation.

It's not clear how often/badly the SCD4x goes out of calibration, but it might be necessary to create software/instructions for manual calibration.

Saw a comment about sparlines in another issue and thought it was a super neat idea! This is quite a high resolution display so it fits well with Tufte's original sparkline ideas.

To make room for a co2 reading sparkline I relocated the temperature field from upper left to lower right and got rid of relative humidity and pressure (never looked at them once but it could just be me...)

The 250 pixels across is just about perfect for a full day (24 hours) worth of data if we downsample by a factor of two. However, I ended up adding a label to clarify the timespan of the sparkline which took up 12 pixels leaving 238 pixel wide field. This is 2 pixels shy of the 240 we need but I figured it was close enough -- the full display ends up being 6 minutes short of 24 hours.

The vertical scale is determined by the maximum co2 reading over the last 24 hours which we already have on the display. Hopefully this relationship is clear because of similar labeling "24hr" and "max/24hr" and proximity of the labels.

I first tried a lightweight sparkline where each data point is represented by a single pixel:

But it turned out to be not super legible especially under less than perfect lighting conditions.

The second idea is a bar graph sparkline which looks a lot more legible:

@danwahl thoughts?

• The E Ink screen has some low-hanging fruit, since both the Reset and Busy pins are not currently wired up (see "Featherwing Connections"). See the MagTag section on Arduion Sleep for more info
• Could remove the LED on the SCD4x (and DPS310). Also switch SCD4x to Vin for sensor power?

Right now, the system will continue running until the battery is totally drained, which is presumably not good for the longevity of the battery.

Instead, the device could read the battery level on startup, and if it is below a given threshold (5%?), activate a low battery mode. The mode would set a persistent flag, and then display an error message before going back into deep sleep. Each subsequent start would check both the battery level and the flag, and only begin normal operation if the charge was above the threshold, otherwise going immediately back to sleep without updating the display.

This should prolong the healthy battery life considerably, though power draw in deep sleep is not zero, so a switch would be a better solution in the long term.

• Make neoPixel static, move initialization to setup() for consistency
• Remove deprecated altitude correction once pressure sensor is verified
• Improve (or in most cases, implement non-existent) error handling

Seems like the processor speed could at least be turned down to 80 MHz to save power (if not all the way down to 10 MHz, though I remember reading somewhere that this might cause USB serial issues?)

Should at least update to include:

• Limitations related to auto-calibration
• Info on "dangerous" levels, red text