# Variable Window Size of Blackman FIR filter

I am trying to implement an FIR filter with a Blackman window function. The solution I got for my signal filtering problem is to have different window sizes for certain characteristics of the input signal. My current solution uses a set of filters working in parallel and the output switches between them based on the input signal characteristics. However, some discontinuities may occur in the output signal. My question: Is there an efficient way to implement an FIR filter with variable (or adaptive) window size or this whole idea is unreliable?

• What is your signal filtering problem? Do you need to restrict this to be FIR filters with Blackman windows? Please tell us the bigger picture of your problem in case there is a much simpler solution. Commented Aug 29, 2020 at 11:44
• It can be FIR filters with any window function that has a triangular shape. The problem is similar to regular signal smoothing. Different amplitudes of input signal require corresponding different window sizes for better performance. Commented Aug 29, 2020 at 15:02
• "better performance": uh, the solution you describe doesn't seem to be true to the input signal (i.e. no, you would not change the smoothing filter when your input amplitude is different; usual filters are linear). So, you have some special requirements that you really should explain. What kind of signal is this? What are you planning to do with it afterwards? I think the solution to your problem is very different to the answer to your question. Commented Aug 29, 2020 at 15:17
• This is a virtual simplified system of an industrial power converter. I will try to explain this problem in my own words as I don't have much experience in power electronics. This power converter has two modes of operation: 1. A single step that can have different amplitudes. 2. A sequence of pulses with random amplitudes. The system objective is to smooth the input step as an s-like signal with certain rates. So that the internals of capacitors and inductors have some required transient response. Commented Aug 29, 2020 at 18:54
• The best I got for now is to have a Blackman FIR filter with different window sizes. The shape of the Blackman window achieved the best transient response for the power converter. However, when I switch to the second mode of operation lots of spikes occur as I don't have an efficient way to switch the FIR filters outputs. I am also not sure this is the best way. I was told that this a more of a control problem than DSP. But, the system is implemented with much optimized hardware and software for FIR that I need to get advantage of it. Commented Aug 29, 2020 at 19:02