natevw / co2-py

Notes and rough experiments with a T6603 CO2 sensor module

T660305R CO2 sensor notes/experiments

Rough "lab notebook" while testing and using some older CO2 modules whose official documentation appears to have begun bit-rotting.

General notes/links

Looks like these sensors were branded as GE Sensing Safeaire T6603-5 although made by Telaire even at the time. Amphenol may have bought Telaire since then.

The best guide left on the web might be: https://ecorenovator.org/forum/showthread.php?t=3600

There's an Arduino library here: https://github.com/LuisM78/T6603

And some discussion: https://picaxeforum.co.uk/threads/co2-sensor-help.25942/

The best PDF I can find is unfortunately not available from an original source: https://file.yzimgs[.]com/321940/2012082417090077.pdf [may or may not be a trustworthy download site]

There's a different PDF version with a slightly more specific provenance here as well: https://web.archive.org/web/20160416212536/http://www.sensorexpert.com.cn/UploadFiles/Others/20120910142843_18796.pdf

This may also be one of the original manuals although it has a different product number: https://web.archive.org/web/20101117090509/http://www.ge-mcs.com/download/sensing-manuals/T63182-003.pdf

The most relevant bits I see in those basic overviews are that the pins in the white plastic header should be (L-to-R, with the gold part on top and the header at the bottom):

• 1 — UART RX
• 2 — UART TX
• 3 — V+ [power input]
• 4 — V- [power ground]
• 5 — PWM or control signal output

V+ should get 5 volts (±0.5V), peak consumption 150ma.

UART defaults to 19200 baud, no parity, 1 stop bit.

The UART protocol seems to match the currently available documentation https://www.amphenol-sensors.com/en/literature/561-telaire-co2-sensors-uart-communications-protocol. Note the product-specific PDFs above mention a more specific "Refer to T6603_UART_Protocol document PD0081" which I've only found at some content farm thing https://wenku.baidu[.]com/view/b5f899b7a26925c52dc5bf17.html although possibly available for download to "VIP" users in exchange for email address and/or installing an app and/or giving credit card and/or who knows.

The closest new parts would either be:

• T6613: https://www.amphenol-sensors.com/en/telaire/co2/525-co2-sensor-modules/321-t6613
• T6703: https://www.amphenol-sensors.com/en/telaire/co2/525-co2-sensor-modules/3400-t6703

It looks like the MH-Z19C might be the current popular "go-to" sensor for hobby CO2 stuff these days.

https://en.wikipedia.org/w/index.php?title=Carbon_dioxide_sensor&oldid=1000144008 for general background on how CO2 sensors work. The T6603 is presumedly NDIR as it blinks a light inside its chamber thing each time it gives a reading.

Connecting via a USB TTL dongle

Those USB TTL serial UART adapters are pretty handy for communicating with this, since they provide the +5V needed and of course the serial itself. I guess the color coding isn't maybe always consistent between models but with the dongle I have it's:

• plastic header pin 1 — green
• plastic header pin 2 — white
• plastic header pin 3 — red
• plastic header pin 3 — black

Sample commands

For initial poking/prodding I started with Serial.app on my macOS, which doesn't have a great way to read binary data. What I do is set Raw Bytes in the "Serial > Preferences > Logging" tab, ⌘L to "Terminal > Start Log" and then ⌘T to "Terminal > Send String…" which does support hex stuff.

Here's some handy/useful ones:

via T63182-004-091614-web

0xFFFE020203
0xFFFE020203 - CMD_READ GAS_PPM
0xFFFE01B6 - CMD_STATUS
0xFFFE02C000 - CMD_SELF_TEST START
0xFFFE02C001 - CMD_SELF_TEST RESULTS

0xFFFE02020F - CMD_READ ELEVATION
0xFFFE04030F0180 - CMD_UPDATE ELEVATION 384


0xFFFE020211 — CMD_VFY_SGPT_PPM [was 1000]
0xFFFE0403110190 — CMD_SET_SGPT_PPM [to 400]
0xFFFE019B — CMD_SGPT_CALIBRATE

0xFFFE05004D41524B - CMD_LOOPBACK "MARK"

0xFFFE01BD - CMD_STREAM_DATA

Then to read the responses I open the log in HexFiend (using "Revert to Saved" to update when necessary…) and find the stuff manually. Definitely not ideal, so once that was working I moved on to the…

Python files in this repo

For some reason these only work once (?!). That is, if I run them and then Ctrl-C out of them I have to unplug and re-plug the TTL adapter in before they work again. Haven't really taken much time to troubleshoot why.

Anyway, the log.py was the first one, has some extra debugging/experimental junk in it and "polls" the device every 5 seconds. Then I realized the sensor can "stream" the data itself so log_stream.py uses that instead. You'll need PySerial installed, and then pass the command your serial port, e.g.:

python3 -m venv env
. env/bin/activate
pip install -r requirements.txt

./log_stream.py /dev/cu.Repleo-PL2303-00001014
# sensor doesn't output while it warms up, but eventually you should get output like…
2021-02-20T12:10:04,1430
2021-02-20T12:10:10,1446
2021-02-20T12:10:16,1466
2021-02-20T12:10:22,1484
2021-02-20T12:10:27,1570
2021-02-20T12:10:33,1571
2021-02-20T12:10:39,1578
2021-02-20T12:10:45,1575
2021-02-20T12:10:51,1575
2021-02-20T12:10:57,1583
2021-02-20T12:11:03,1586
2021-02-20T12:11:08,1587
2021-02-20T12:11:14,1592
2021-02-20T12:11:20,1597
2021-02-20T12:11:26,1599
2021-02-20T12:11:32,1603
2021-02-20T12:11:38,1609
2021-02-20T12:11:44,1607

…which you can open as CSV in Excel or Numbers.app or R or whatever and graph.

…or it might just crash and burn at the slightest provocation, usually because it didn't get any data back from the sensor (instead of the expected 5 bytes). Try turning it off and back on again or ???

Script buffering issues when logging via tee

It's nice to see the logs AND capture them. You can pipe into tee for this e.g. ./log_stream.py /dev/cu.Repleo-PL2303-00001014 | tee -a ~Desktop/co2_log.csv but that didn't really work "right" by default because pipes introduce a buffer (4096 bytes or so, probably depending on platform) and so you won't see data for a long time. The scripts now workaround this by a Python flag.

# (doesn't work... some potential explanation in comments and related posts)
# https://unix.stackexchange.com/questions/25372/turn-off-buffering-in-pipe#comment225838_25378
brew install coreutils
# the following does NOT work
./log_stream.py /dev/cu.Repleo-PL2303-00001014 | gstdbuf -o0 tee -a ~/Desktop/co2-overnight.csv

# … passing `-u` to Python *does* work (now in shebang of scripts) via:
# https://unix.stackexchange.com/questions/25372/turn-off-buffering-in-pipe#comment1037095_53445