noise removal on a step respons of a battery: the step is also important

Hello all,

currently I am doing some battery tests to determine the numerical values of the components used in my battery model. What I do is: I discharge my battery with 10 amps and after a vertain amount of time, I set the discharge current to 0 so I can monitor the relaxation behaviour of the battery. Now it seems like the measurements suffer from some noise. Because the steps in voltage (see picture) are completely normal, and important to measure, I'd only need a filter which can filter the noise and not the step away. Anybody knows how to make that happen? Thanks in advance!

<<

>>

[Rick Rosson]

Most likely, you will need to use a low pass filter. The issue is that you first need to determine the appropriate cutoff frequency. To do that, you will want to look at the signal in the frequency domain.

1. Do you know what the sampling rate is for the signal?
2. Do you know how to transform a signal to the frequency domain?

[Image Analyst]

Linear low pass filters will blur out the step, while non-linear filters like median will not, at least not as much and not until the window size gets really big.

[Rick Rosson]

Fair enough. Good point.

[Jannick]

Indeed low pass filtering is not the best solution because of the reason given by Image Analyst. So, a non-linear filter is the best option in this case?

Answer 1

[Image Analyst]

The median filter, medfilt1(), is the usual recommendation for a first try at it.

[Jannick]

Thank you very much! this is already a pretty shar picture (whe I take the median over 30 samples):

[Image Analyst]

Usually you do an odd number of samples in the window so that you don't have a half sample shift along the x axis between the input and output.