joshuaboniface/supersensor2
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Switch to SGP41 air quality sensor

Browse Source 
My testing with the SGP30 sensor proved fruitless, as various sensors
would mis-report (under- and over-) levels and generally not be
consistent with each other or reality. These sensors are simply too
flaky with zero consistency.

Instead, switch to the SGP41, which so far seems more robust and is used
in other tools like the AirGradient One. We leverage their calculations
for VOC Index -> VOC levels, as well as a generalized isobutylene-based
eCO2 calculation.
This commit is contained in:
Joshua Boniface
2025-06-21 23:31:09 -04:00
parent 0ba3b855d0
commit 3300ca2d8e
2 changed files with 88 additions and 202 deletions
Download Patch File Download Diff File Expand all files Collapse all files

111
README.md
View File

@ -42,15 +42,15 @@ and [my update post on version 2.0](https://www.boniface.me/the-supersensor-2.0)
## Major Changes from 1.x
1. Replaced the Bosch BME680 with the Sensirion SHT45 and Sensirion SGP30.
1. Replaced the Bosch BME680 with the Sensirion SHT45 and Sensirion SGP41.
The BME680 proved to be woefully unreliable in my testing. Temperature was fairly accurate (internal heating and offset notwithstanding),
but humidity was wildly off of what other thermometers/hydrometers would report. In addition, the AQ functionality of the sensor was a
source of much frustration and I was never able to get it to work reliably, either with the official BSEC library or with my own attempts
at self-configuration.
Thus, this sensor has been replaced with two Sensirion sensors which in my experience so far have been much more reliable and consistent,
and the cost difference is negligible.
Thus, this sensor has been replaced with two Sensirion sensors which in my experience so far have been much more reliable and consistent.
There is a slight cost increase due to these sensors, but not signfigant enough to outweigh the benefit of reliable monitoring they confer.
2. Replaced the SR602 PIR sensor with the AM312 PIR sensor.
@ -64,15 +64,12 @@ and [my update post on version 2.0](https://www.boniface.me/the-supersensor-2.0)
3. Completely redesigned the custom PCB around the above sensor changes, which is now more compact in a 50x55mm almost-square configuration.
4. Significantly cleaned up the ESPHome configuration, to support the above sensors and remove a lot of cruft that was caused by the BME680.
This includes a new set of custom AQ calculations based on the SGP30 and SHT45 sensors that, while not necessarily following the full EPA
IAQI spec, should still give a reasonable view of the air quality conditions of an interior room and not deviate wildly and nonsensically
like the BME680 did. Details of the calculation are provided below.
## Parts List
| Qty | Component | Cost (2025/05 CAD, ex. shipping) | Links |
|-------|--------------------|----------------------------------|-------|
| 1 | GY-SGP30 | $5.73 | [AliExpress](https://www.aliexpress.com/item/1005008473372972.html) |
| 1 | GY-SGP41 | $11.08 | [AliExpress](https://www.aliexpress.com/item/1005006746827606.html) |
| 1 | GY-SHT45 | $5.67 | [AliExpress](https://www.aliexpress.com/item/1005008175340220.html)* |
| 1 | SR602 | $0.81 | [AliExpress](https://www.aliexpress.com/item/1005001572550300.html) |
| 1 | TSL2591 | $4.59 | [AliExpress](https://www.aliexpress.com/item/1005008619462097.html) |
@ -84,7 +81,7 @@ and [my update post on version 2.0](https://www.boniface.me/the-supersensor-2.0)
| 1 | Female pin header† | $1.59 ($15.99/10) | [Amazon](https://www.amazon.ca/dp/B08CMNRXJ1) |
| 1 | Custom PCB (JLC) | $0.69 ($6.89/10) | [GitHub](https://github.com/joshuaboniface/supersensor) |
| 1 | 3D Printed case | $?.??‡ | [GitHub](https://github.com/joshuaboniface/supersensor) |
| **TOTAL** | | **$33.64** | |
| **TOTAL** | | **$38.99** | |
`*` Ensure you select the correct device on the page as it shows multiple options.
@ -112,17 +109,19 @@ SuperSensors in a single room and only want one to respond to voice commands.
If enabled (the default), when overall presence is detected, the LEDs will
glow "white" at 15% power to signal presence.
### Temperature Offset (selector, -10 to +5 @ 0.1, -5 default)
### Temperature Offset (selector, -30 to +10 @ 0.1, -5 default)
Allows calibration of the SHT45 temperature sensor with an offset from -10 to +5
Allows calibration of the SHT45 temperature sensor with an offset from -30 to +10
degrees C. Useful if the sensor is misreporting actual ambient tempreatures. Due
to internal heating of the SHT45 by the ESP32, this defaults to -5; further
calibration may be needed for your sensors and environment.
calibration may be needed for your sensors and environment based on an external
reference.
### Humidity Offset (selector, -10 to +10 @ 0.1)
### Humidity Offset (selector, -20 to +20 @ 0.1)
Allows calibration of the SHT45 humidity sensor with an offset from -10 to +10
percent relative humidity. Useful if the sensor is misreporting actual humidity.
percent relative humidity. Useful if the sensor is misreporting actual humidity
based on an external reference.
### PIR Hold Time (selector, 0 to +60 @ 5, 0 default)
@ -230,70 +229,28 @@ is likely not useful.
## AQ Details
The SuperSensor 2.x provides 2 base air quality sensors (numeric), from which
4 human-readable text sensors are derived.
The SuperSensor 2.0 features an SGP41 air quality sensor by Sensirion. This is a powerful AQ
sensor which powers several commercial devices including the AirGradient One, which gave
us a lot of our configuration via their sharing of algorithms.
The goal of these sensors is to track general comfort and livability in a
room, not specific contaminants or conditions. Because the SGP30 can only
track TVOC and eCO2, we do not track particulates, CO, NOx, or CH2O, all
of which are required for a full EPA (I)AQI score. This means the best
we can do is approximate (I)AQI roughly, and since a scale of 0-500 based
on approximations seems pointless, I went with much simpler 1-4/5 scores
instead. I feel this does a good enough job to be useful for 99% of rooms.
The sensor provides two base readings: a VOC Index, and a NOx Index. These values are both
floating references centered at 100 (VOC) and 1 (NOx), where that value represents "normal"
air over the previous 24 hours. These sensors are very useful for any sort of quick-change
automations, e.g. turn on a fan if levels spike due to cooking.
We also cannot really debate whether the BME680 is actually any more accurate
in this regard, since their algorithms are proprietary and all that is exposed
normally is a single resistance value, so in my opinion this is actually
superior to that sensor anyways with two discrete datapoints (versus one),
even if it does still seem limited when compared to dedicated AQ sensors.
And that is to say nothing of the issues with that sensor (constantly climbing
IAQ values over time, poor calibration, etc.).
In addition, we leverage AirGradient's published forumulas to convert the VOC index into
actual VOC quantities, in both µg/m³ and ppb. While this may drift due to the sensor's regular
internal recalibration, I feel that following what AirGradient does is sufficient enough
for any real-world home usage. Further, we use a very rough conversion of the aforementioned
VOC quantity into an eCO2 reading, using Isobutylene as a reference gas. These sensors are
more useful for display purposes, to show the current levels in a room in a dashboard or
other such place, for human consumption. Note that no such conversions are done for NOx as
there are no (that I can find) published empirical calculations for this conversion, unlike
for VOCs via AirGradient.
### Base Numeric Values
#### IAQ Index (1-5)
The IAQ index is calculated based on the TVOC and eCO2 values from the SGP30
sensor, to provide 5 levels of air quality. This corresponds approximately
to the levels provided by the BME680 (0-50, 50-100, 100-200, 200-300, 300+).
5 is "great": the TVOC is <65 ppb and the eCO2 is <600 ppm.
4 is "good": the TVOC is 65-220 ppb or the eCO2 is 600-800 ppm.
3 is "fair": the TVOC is 220-660 ppb or the eCO2 is 800-1200 ppm.
2 is "poor": the TVOC is 660-2200 ppb or the eCO2 is 1200-2000 ppm.
1 is "bad": the TVOC is >2200 ppb or the eCO2 is >2000 ppm.
#### Room Health Score (1-4)
The Room Health Score is calculated based on the IAQ, temperature, and humidity,
and is designed to show how "nice" a room is to be in. Generally a 4 is a nice
place to be, especially for someone with respiratory issues like myself, and lower
scores indicate more deviations from the norms or poor IAQ.
4 is "optimal": IAQ is >= 4 ("great" or "good"), temperature is between 18C and 24C, and humidity is between 40% and 60%.
3 is "fair": One of the above is not true, and IAQ is >= 3 ("fair").
2 is "poor": Two of the above are not true, and IAQ is >= 2 ("poor").
1 is "bad": All of the above are not true or IAQ is 1 ("unhealthy") regardless of other values.
Note that IAQ levels hold a major sway over this level, and decreasing IAQ
scores will push the room score lower regardless of temperature or humidity.
It is best used together with the individual sensors to determine exactly
what is wrong with the room.
### Derived Text Sensors
#### VOC Level
This reports the VOC level alone, based on the scale under IAQ Index, in textual form ("Great, "Good", etc.).
#### CO2 Level
This reports the eCO2 level alone, based on the scale under IAQ Index, in textual form ("Great, Good", etc.).
#### IAQ Classification
This reports the IAQ Index in textual form ("Great", "Good", etc.).
#### Room Health
This reports the Room Health Score in textual form ("Optimal", "Fair", "Poor", "Bad").
Note however that like all MOx sensors, the SGP41 does not differentiate gasses, and as
such cannot tell the difference between normal, everyday natural VOCs like those in
breath or from e.g. ripening fruit, and dangerous VOCs from e.g. construction materials.
These should be used only as a general indication of air quality over short periods, rather
than an absolute reference over long periods (much to my own frustration but inevitable
begruding acceptance).

179
supersensor.yaml
View File

@ -546,38 +546,65 @@ sensor:
cpu_frequency:
name: "CPU Frequency"
- platform: sgp30
eco2:
name: "SGP30 eCO2"
id: sgp30_eco2
accuracy_decimals: 1
icon: mdi:molecule-co2
- platform: sgp4x
voc:
name: "SGP41 VOC Index"
id: sgp41_voc_index
accuracy_decimals: 0
icon: mdi:waves-arrow-up
filters:
- sliding_window_moving_average: # We take a reading every 15 seconds, but calculate the sliding
window_size: 12 # average over 12 readings i.e. 60 seconds/1 minute to normalize
send_every: 3 # brief spikes while still sending a value every 15 seconds.
nox:
name: "SGP41 NOx Index"
id: sgp41_nox_index
accuracy_decimals: 0
icon: mdi:waves-arrow-up
filters:
- sliding_window_moving_average:
window_size: 20
send_every: 1
tvoc:
name: "SGP30 TVOC"
id: sgp30_tvoc
accuracy_decimals: 1
icon: mdi:molecule
filters:
- sliding_window_moving_average:
window_size: 20
send_every: 1
eco2_baseline:
name: "SGP30 Baseline eCO2"
id: sgp30_baseline_ec02
icon: mdi:molecule-co2
tvoc_baseline:
name: "SGP30 Baseline TVOC"
id: sgp30_baseline_tvoc
icon: mdi:molecule
window_size: 12
send_every: 3
compensation:
temperature_source: sht45_temperature
humidity_source: sht45_humidity
store_baseline: yes
update_interval: 15s
store_baseline: true
update_interval: 5s
- platform: template
name: "SGP41 TVOC (µg/m³)"
id: sgp41_tvoc_ugm3
icon: mdi:molecule
lambda: |-
float i = id(sgp41_voc_index).state;
if (i < 1) return NAN;
float tvoc = (log(501.0 - i) - 6.24) * -878.53;
return tvoc;
unit_of_measurement: "µg/m³"
accuracy_decimals: 0
- platform: template
name: "SGP41 TVOC (ppb)"
id: sgp41_tvoc_ppb
icon: mdi:molecule
lambda: |-
float tvoc_ugm3 = id(sgp41_tvoc_ugm3).state;
float tvoc_ppm = tvoc_ugm3 * 0.436; // ppb estimated using isobutylene MW (56.1 g/mol)
return tvoc_ppm;
unit_of_measurement: "ppb"
accuracy_decimals: 0
- platform: template
name: "SGP41 eCO2 (appr.)"
id: sgp41_eco2_appr
icon: mdi:molecule-co2
lambda: |-
float tvoc_ppb = id(sgp41_tvoc_ppb).state;
float eco2_ppm = 400.0 + 1.5 * tvoc_ppb;
if (eco2_ppm > 2000) eco2_ppm = 2000;
return eco2_ppm;
unit_of_measurement: "ppm"
accuracy_decimals: 0
- platform: sht4x
temperature:
@ -622,51 +649,6 @@ sensor:
return (b * alpha) / (a - alpha);
update_interval: 15s
# IAQ Index (1-5, 5=Great))
- platform: template
name: "IAQ Index"
icon: mdi:air-purifier
id: iaq_index
lambda: |-
int tvoc = id(sgp30_tvoc).state;
int eco2 = id(sgp30_eco2).state;
if (tvoc > 2200 || eco2 > 2000) return 1; // Bad
if (tvoc > 660 || eco2 > 1200) return 2; // Poor
if (tvoc > 220 || eco2 > 800) return 3; // Fair
if (tvoc > 65 || eco2 > 500) return 4; // Good
return 5; // Great
update_interval: 15s
# Room Health Score (1-4, 4=Optimal)
- platform: template
name: "Room Health Score"
icon: mdi:home-thermometer
id: room_health
lambda: |-
float temp = id(sht45_temperature).state;
float rh = id(sht45_humidity).state;
int iaq = id(iaq_index).state;
bool temp_ok = (temp >= 18 && temp <= 24);
bool hum_ok = (rh >= 30 && rh <= 70);
bool iaq_ok = (iaq >= 4);
int conditions_met = 0;
if (temp_ok) conditions_met++;
if (hum_ok) conditions_met++;
if (iaq_ok) conditions_met++;
if (iaq_ok && temp_ok && hum_ok) {
return 4; // Optimal: All conditions met and IAQ is excellent/good
} else if (iaq >= 3 && conditions_met >= 2) {
return 3; // Fair: IAQ is moderate and at least 2 conditions met
} else if (iaq >= 2 && conditions_met >= 1) {
return 2; // Poor: IAQ is poor and at least 1 condition met
} else {
return 1; // Bad: All conditions failed or IAQ is unhealthy
}
update_interval: 15s
- platform: tsl2591
address: 0x29
update_interval: 1s
@ -795,59 +777,6 @@ text_sensor:
mac_address:
name: "LD2410C MAC Address"
# VOC Level
- platform: template
name: "VOC Level"
icon: mdi:molecule
lambda: |-
int tvoc = id(sgp30_tvoc).state;
if (tvoc < 65) return {"Great"};
if (tvoc < 220) return {"Good"};
if (tvoc < 660) return {"Fair"};
if (tvoc < 2200) return {"Poor"};
return {"Bad"};
update_interval: 15s
# CO2 Level
- platform: template
name: "CO2 Level"
icon: mdi:molecule-co2
lambda: |-
int eco2 = id(sgp30_eco2).state;
if (eco2 < 500) return {"Great"};
if (eco2 < 800) return {"Good"};
if (eco2 < 1200) return {"Fair"};
if (eco2 < 2000) return {"Poor"};
return {"Bad"};
update_interval: 15s
# IAQ Classification
- platform: template
name: "IAQ Classification"
icon: mdi:air-purifier
lambda: |-
int iaq = id(iaq_index).state;
if (iaq == 5) return {"Great"};
if (iaq == 4) return {"Good"};
if (iaq == 3) return {"Fair"};
if (iaq == 2) return {"Poor"};
return {"Bad"};
update_interval: 15s
# Room Health
- platform: template
name: "Room Health"
icon: mdi:home-thermometer
lambda: |-
int score = id(room_health).state;
if (score == 4) return {"Optimal"};
if (score == 3) return {"Fair"};
if (score == 2) return {"Poor"};
return {"Bad"};
update_interval: 15s
button:
- platform: ld2410
restart:
name: "LD2410C Restart"
icon: mdi:power-cycle