Real Time Low Pass Filtering

Hello,

I am trying to apply the low pass filter to my data. My samples are like square wave and I want them to be smoother. In order to do that, I am using low pass filter. This works fine for offline operations, I mean, I have all the data and time vector, so it is easy to filter the data. But I want to do it real-time with minimum delay, like 1 sample delay. Here is my .mat files and graphs that I got.

clc

clear

close all

temperature = load("C:\Users\User\Desktop\Samples.mat");

% Samples.mat have temperature data.

filtered_temperature = lowpass(temperature, 1, 1000);

plot(filtered_temperature)

hold on

plot(temperature)

% temperature.mat and filtered_temperature.mat attached.

Thanks for your help.

Answer 1

[Benjamin Thompson]

All digital filters working in real time must have some delay, this is a law of filtering in general. The amount of delay depends on the type of filter, how fast the passband falls off, etc. You can use the "fir" option of lowpass to ensure an FIR (finite impulse response) type filter which has constant delay in time. Then if you have a data file, your problem is not "real time", right? The delay of an FIR filter is equal to the order or number of coefficients in the filter. So just adjust the time values of the output so they line up with the original time values.

[tinkyminky93]

Actually I haven't got data file always. It is sampling the temperature data and sends them to me without .mat format. I merged them manually so we can say that it is real time. For example it is sampling with 1000 Hertz until I power off my device. I need to do it online, not offline with 500 datas. Other things that you mentioned are right, it is a law of filtering. Can we do it without filter, any advice about it sir? Thanks.