Correcting effects of Humidity on sensors

Hi All

I have gas sensor, that gets effected by hummdity that needs to be corrected. So was hoping to see if we can correct this ?

How can i run my code on the this support forum with my data file, so it can be run?

Answer 1

[Star Strider]

What sort of correction do you want to do to your data?

Do you also have the humidity data?

Are there any published ways to correct the readings for humidity? If so, please share them.

To run your code with your data here, first upload the data file, using the 'paperclip' icon in the top toolbar (just to the right of the Σ). Click on the 'insert a line of doce' icon in the top toolbar (farthest left icon in the CODE section, or ALT+ENTER) to create a code line, then type or copy-paste your code in it. To run it, press the green arrrow in the top toolbar.

x = linspace(0, 2*pi);
y = sin(x) .* cos(x);

figure
plot(x, y)
grid

Your code should run here essentially the same way it runs on your computer, including reading the file.

,

[Dharmesh]

Thank you for your reply.

As humidity increases and decreases — or due to the rate of change — it introduces transients into my analogue signal. This behaviour is known to occur in electrochemical sensors. I would like to explore whether this effect can be corrected, possibly using a quadratic or adaptive algorithm that could later be implemented or ported into other software code.

If you run my code with my dataset, you’ll notice a clear correlation between humidity and certain variations in the signal. The signal is also affected by temperature, but that influence is much easier to handle, so my current focus is on correcting the humidity-related effects.

%% AAN-803-05 corrections: kT for NO, nT for NO2 & OX (O3)
% Dharmesh – edit the folder if needed:
folder  = 'G:\AVR_Project\MATLAB Projects\Air Sensor\Temp_test';
matFile = fullfile(folder, 'SensorLog_ALL.mat');
% --- Load the combined table ---
S = load(matFile);
fn = fieldnames(S);
T  = S.(fn{find(structfun(@istable, S), 1)});
% --- Time & environment ---
if ~isdatetime(T.Timestamp)
    T.Timestamp = datetime(T.Timestamp,'InputFormat','yyyy-MM-dd HH:mm:ss');
end
time  = T.Timestamp;
TempC = T.("Temperature (°C)");
Hum   = T.("Humidity (%)");
% ================== CALIBRATION CONSTANTS ==================
% ---------- ELECTRONICS OFFSETS (hardware bias; set if measured) ----------
ELEC.NO_WE   = 0.292;  ELEC.NO_AUX   = 0.259;
ELEC.OX_WE   = 0.229;  ELEC.OX_AUX   = 0.227;   % OX uses O3 columns
ELEC.NO2_WE  = 0.238;  ELEC.NO2_AUX  = 0.225;
% ---------- SENSOR ZERO BASELINES (clean air, include electronics if measured that way) ----------
ZERO.NO_WE   = 0.3158;  ZERO.NO_AUX   = 0.3017;   % NO-B4
ZERO.OX_WE   = 0.2366;  ZERO.OX_AUX   = 0.2278;   % OX-A431 (O3)
ZERO.NO2_WE  = 0.2396;  ZERO.NO2_AUX  = 0.2253;   % NO2-B43F
% ===========================================================
% --- Ensure expected columns exist (adjust here if names differ) ---
need = ["NO WE","NO AUX","O3 WE","O3 AUX","NO2 WE","NO2 AUX"];
have = string(T.Properties.VariableNames);
if ~all(ismember(need, have))
    error('Missing columns. Expected: %s\nHave: %s', strjoin(need,', '), strjoin(have,', '));
end
% --- Initialise structs to avoid dot-indexing errors ---
WE = struct(); AE = struct(); WEo = struct(); AEo = struct();
% --- Convert zeros to sensor-only baselines (remove electronics bias) ---
WEo.NO  = ZERO.NO_WE  - ELEC.NO_WE;   AEo.NO  = ZERO.NO_AUX  - ELEC.NO_AUX;
WEo.OX  = ZERO.OX_WE  - ELEC.OX_WE;   AEo.OX  = ZERO.OX_AUX  - ELEC.OX_AUX;
WEo.NO2 = ZERO.NO2_WE - ELEC.NO2_WE;  AEo.NO2 = ZERO.NO2_AUX - ELEC.NO2_AUX;
% --- Raw -> electronics-corrected live readings ---
WE.NO   = T.("NO WE")   - ELEC.NO_WE;    AE.NO   = T.("NO AUX")  - ELEC.NO_AUX;
WE.OX   = T.("O3 WE")   - ELEC.OX_WE;    AE.OX   = T.("O3 AUX")  - ELEC.OX_AUX;
WE.NO2  = T.("NO2 WE")  - ELEC.NO2_WE;   AE.NO2  = T.("NO2 AUX") - ELEC.NO2_AUX;
% ---------------- Temperature compensation tables ----------------
T_pts   = [-30 -20 -10  0  10  20  30  40  50];
% NO (kT, ratio model)
kT_NO   = interp1(T_pts, [1.8 1.8 1.4 1.1 1.1 1.0 0.9 0.9 0.8], TempC, 'linear','extrap');
% NO2 & OX (nT, simple model) – replace OX row with your own if you have it
nT_NO2  = interp1(T_pts, [1.3 1.3 1.3 1.3 1.0 0.6 0.4 0.2 -1.5], TempC, 'linear','extrap');
nT_OX   = interp1(T_pts, [1.3 1.3 1.3 1.3 1.0 0.6 0.4 0.2 -1.5], TempC, 'linear','extrap');
% ---------------- Apply algorithms ----------------
% NO (Algorithm 2 / ratio): WEc = (WEu - WEe) - kT*(WEo/AEo)*(AEu - AEe)
WEc_NO   = (WE.NO) - kT_NO .* (ZERO.NO_WE ./ZERO.NO_AUX ) .* (AE.NO );
% NO2 (Algorithm 1 / nT):   WEc = (WEu - WEe) - nT*(AEu - AEe)
%WEc_NO2  = (WE.NO2 - WEo.NO2) - nT_NO2 .* (AE.NO2 - AEo.NO2);
WEc_NO2  = (WE.NO2) - nT_NO2 .* (AE.NO2);
% OX (Algorithm 1 / nT):    WEc = (WEu - WEe) - nT*(AEu - AEe)
WEc_OX   = (WE.OX ) - nT_OX  .* (AE.OX );
% ---------------- Plot sets ----------------
makeFig("NO (k_T ratio)",  time, TempC, Hum, T.("NO WE") ,  T.("NO AUX"),  WEc_NO);
makeFig("NO2 (n_T simple)",time, TempC, Hum, T.("NO2 WE") , T.("NO2 AUX"), WEc_NO2);
makeFig("OX/O3 (n_T simple)",time,TempC, Hum, T.("O3 WE") ,  T.("O3 AUX"),  WEc_OX);
%% --------------- Plot helper ---------------
function makeFig(name,t,tc,hum,WE,AUX,WEc)
f = figure('Name',name,'Color','w');
tl = tiledlayout(f,2,1,'TileSpacing','compact','Padding','compact');
% Top: Temperature & Humidity
ax1 = nexttile(tl);
yyaxis left,  plot(t,tc,'LineWidth',1.2), ylabel('Temperature (°C)')
yyaxis right, plot(t,hum,'LineWidth',1.2), ylabel('Humidity (%)')
grid on, title([name,' — Temp & Humidity'])
% Bottom: WE/AUX vs Corrected
ax2 = nexttile(tl);
yyaxis left
plot(t,WE,'LineWidth',1.2); hold on
plot(t,AUX,'LineWidth',1.2); ylabel('Raw Sensor (V)')
yyaxis right
plot(t,WEc,'LineWidth',1.5); ylabel('Corrected WE_c (V)')
xlabel('Time'), grid on
title([name,' — WE/AUX (left), WE_c (right)'])
legend({'WE','AUX','WE Corrected*'},'Location','best')
linkaxes([ax1 ax2],'x');
end

[Star Strider]

My pleasure!

You provided the .m file for your code, however not the data ('SensorLog_ALL.mat'). (I checked -- the data are not included., at least that I could see.)

It would help to have it.

[Dharmesh]

Sorry its attached now

%% AAN-803-05 corrections: kT for NO, nT for NO2 & OX (O3)
% Dharmesh – edit the folder if needed:
%folder  = 'G:\AVR_Project\MATLAB Projects\Air Sensor\Temp_test';
matFile = 'SensorLog_ALL.mat';
% --- Load the combined table ---
S = load(matFile);
fn = fieldnames(S);
T  = S.(fn{find(structfun(@istable, S), 1)});
% --- Time & environment ---
if ~isdatetime(T.Timestamp)
    T.Timestamp = datetime(T.Timestamp,'InputFormat','yyyy-MM-dd HH:mm:ss');
end
time  = T.Timestamp;
TempC = T.("Temperature (°C)");
Hum   = T.("Humidity (%)");
% ================== CALIBRATION CONSTANTS ==================
% ---------- ELECTRONICS OFFSETS (hardware bias; set if measured) ----------
ELEC.NO_WE   = 0.292;  ELEC.NO_AUX   = 0.259;
ELEC.OX_WE   = 0.229;  ELEC.OX_AUX   = 0.227;   % OX uses O3 columns
ELEC.NO2_WE  = 0.238;  ELEC.NO2_AUX  = 0.225;
% ---------- SENSOR ZERO BASELINES (clean air, include electronics if measured that way) ----------
ZERO.NO_WE   = 0.3158;  ZERO.NO_AUX   = 0.3017;   % NO-B4
ZERO.OX_WE   = 0.2366;  ZERO.OX_AUX   = 0.2278;   % OX-A431 (O3)
ZERO.NO2_WE  = 0.2396;  ZERO.NO2_AUX  = 0.2253;   % NO2-B43F
% ===========================================================
% --- Ensure expected columns exist (adjust here if names differ) ---
need = ["NO WE","NO AUX","O3 WE","O3 AUX","NO2 WE","NO2 AUX"];
have = string(T.Properties.VariableNames);
if ~all(ismember(need, have))
    error('Missing columns. Expected: %s\nHave: %s', strjoin(need,', '), strjoin(have,', '));
end
% --- Initialise structs to avoid dot-indexing errors ---
WE = struct(); AE = struct(); WEo = struct(); AEo = struct();
% --- Convert zeros to sensor-only baselines (remove electronics bias) ---
WEo.NO  = ZERO.NO_WE  - ELEC.NO_WE;   AEo.NO  = ZERO.NO_AUX  - ELEC.NO_AUX;
WEo.OX  = ZERO.OX_WE  - ELEC.OX_WE;   AEo.OX  = ZERO.OX_AUX  - ELEC.OX_AUX;
WEo.NO2 = ZERO.NO2_WE - ELEC.NO2_WE;  AEo.NO2 = ZERO.NO2_AUX - ELEC.NO2_AUX;
% --- Raw -> electronics-corrected live readings ---
WE.NO   = T.("NO WE")   - ELEC.NO_WE;    AE.NO   = T.("NO AUX")  - ELEC.NO_AUX;
WE.OX   = T.("O3 WE")   - ELEC.OX_WE;    AE.OX   = T.("O3 AUX")  - ELEC.OX_AUX;
WE.NO2  = T.("NO2 WE")  - ELEC.NO2_WE;   AE.NO2  = T.("NO2 AUX") - ELEC.NO2_AUX;
% ---------------- Temperature compensation tables ----------------
T_pts   = [-30 -20 -10  0  10  20  30  40  50];
% NO (kT, ratio model)
kT_NO   = interp1(T_pts, [1.8 1.8 1.4 1.1 1.1 1.0 0.9 0.9 0.8], TempC, 'linear','extrap');
% NO2 & OX (nT, simple model) – replace OX row with your own if you have it
nT_NO2  = interp1(T_pts, [1.3 1.3 1.3 1.3 1.0 0.6 0.4 0.2 -1.5], TempC, 'linear','extrap');
nT_OX   = interp1(T_pts, [1.3 1.3 1.3 1.3 1.0 0.6 0.4 0.2 -1.5], TempC, 'linear','extrap');
% ---------------- Apply algorithms ----------------
% NO (Algorithm 2 / ratio): WEc = (WEu - WEe) - kT*(WEo/AEo)*(AEu - AEe)
WEc_NO   = (WE.NO) - kT_NO .* (ZERO.NO_WE ./ZERO.NO_AUX ) .* (AE.NO );
% NO2 (Algorithm 1 / nT):   WEc = (WEu - WEe) - nT*(AEu - AEe)
%WEc_NO2  = (WE.NO2 - WEo.NO2) - nT_NO2 .* (AE.NO2 - AEo.NO2);
WEc_NO2  = (WE.NO2) - nT_NO2 .* (AE.NO2);
% OX (Algorithm 1 / nT):    WEc = (WEu - WEe) - nT*(AEu - AEe)
WEc_OX   = (WE.OX ) - nT_OX  .* (AE.OX );
% ---------------- Plot sets ----------------
makeFig("NO (k_T ratio)",  time, TempC, Hum, T.("NO WE") ,  T.("NO AUX"),  WEc_NO);

makeFig("NO2 (n_T simple)",time, TempC, Hum, T.("NO2 WE") , T.("NO2 AUX"), WEc_NO2);

●

makeFig("OX/O3 (n_T simple)",time,TempC, Hum, T.("O3 WE") ,  T.("O3 AUX"),  WEc_OX);

%% --------------- Plot helper ---------------
function makeFig(name,t,tc,hum,WE,AUX,WEc)
f = figure('Name',name,'Color','w');
tl = tiledlayout(f,2,1,'TileSpacing','compact','Padding','compact');
% Top: Temperature & Humidity
ax1 = nexttile(tl);
yyaxis left,  plot(t,tc,'LineWidth',1.2), ylabel('Temperature (°C)')
yyaxis right, plot(t,hum,'LineWidth',1.2), ylabel('Humidity (%)')
grid on, title([name,' — Temp & Humidity'])
% Bottom: WE/AUX vs Corrected
ax2 = nexttile(tl);
yyaxis left
plot(t,WE,'LineWidth',1.2); hold on
plot(t,AUX,'LineWidth',1.2); ylabel('Raw Sensor (V)')
yyaxis right
plot(t,WEc,'LineWidth',1.5); ylabel('Corrected WE_c (V)')
xlabel('Time'), grid on
title([name,' — WE/AUX (left), WE_c (right)'])
legend({'WE','AUX','WE Corrected*'},'Location','best')
linkaxes([ax1 ax2],'x');
end

[Star Strider]

Torsten -- Thank you!

@Dharmesh -- I am having a bit of trouble understnading this.

What are the various variables? I do not completely understand 'WE" and 'AUX'.

Do you have the actual NOx values? (Is this a calibration test?)

I do not understand the technology of the sensor, and this may be necessary to devise a way of correcting its readings. Do you have a published way of correcting the sensor values for the humidity? Is there a specific way that humidity interferes with the sensor (does it 'poison' it, physically block it, or something else)?

EDIT -- (24 Oct 2025 at 00:45)

See if the sensors dexcribed in HUMIDITY AND TEMPERATURE CORRECTION FACTORS FOR NOX EMISSIONS FROM DIESEL ENGINES matches your sensor. If so, we can probably use one of these correction approaches. Those appear to be relatively straightforward, however it will be necessary to state the units of your data, and make any necessary unit conversions to work with these equations. Specifically, Equation (4) would be easy to code.

.

[Dharmesh]

@Star Strider

Sorry. , the sensor outputs two signals: WE and AUX. The difference between them is that WE includes the gas concentration, while AUX does not. The purpose of this is so that AUX can be used to identify whether any noise or environmental factors have affected the sensor.

However, both signals are influenced by temperature and humidity. Temperature is relatively simple to correct you ensure that the gas concentration is zero (or very close to zero) and then create a baseline by measuring at various temperature steps.

In this case, we can assume that the gas concentration is close to zero. At this stage, we do not need to calibrate the concentration.

The signal we need to focus on is the corrected WE signal, shown in red. Primarily, you can see humidity transients, which simply affect the current and therefore the final voltage.

There is some information on these sensors in the application note, please see the last page:

aan_110_app-note_v0.pdf

Sensor data sheet for NO2. All other sensors would in principle be similar

alphasense_no2-b43f_datasheet_en_4.pdf