Band pass filters, from theory to practice

[ikalogic]

Hello i am strating this thread to get some help about Active band pass filters using op-amps, as i tried to do many schematics i found on the net, and it never worked.. and i never knew how to fix it because i don't fully understand the process. So.. here we are with my first 'stupid' question!

1-If the input of a band pass filter is an oscilating signal (pulses) how is the output signal (in case the input frequency is not rejected?) is it a smoth DC signal? is it the same input signal but amplified?

thx

 

[evandude]

1-If the input of a band pass filter is an oscilating signal (pulses) how is the output signal (in case the input frequency is not rejected?) is it a smoth DC signal? is it the same input signal but amplified?

a band pass filter will always output an AC signal. To really understand what's going to come out of it, you have to understand the concept of the frequency spectrum of a signal - any signal can be represented as a sum of sinusoidal waves of different frequencies, and the response of a filter to the actual signal is the same as the sum of the responses to all those sinusoidal waves (superposition) It probably isn't a concept you can just immediately understand, usually it's a sizeable chunk of an EE course in college.

Check out this page, which has a few (hopefully) helpful diagrams:
https://mathworld.wolfram.com/FourierSeries.html

You can find the response of the filter to any sinusoidal wave since you know the filter's frequency response (ie - the gain vs. frequency) Usually (though not necessarily), a bandpass filter will have close to unity gain in the passband and drop off outside the passband. The output of a bandpass filter to a sinusoidal signal of a given frequency is then going to just be the same sinusoidal signal, scaled by the gain of the filter at that frequency. If it's within the passband, the output should be pretty close to the input, if it's outside the passband it will be attenuated.

If you want to visualize the response due to a more complex signal, such as a square wave or series of pulses, you would probably want to take the fourier transform of the signal, which converts it to the frequency domain so you can see its frequency content, and see how that lines up with the filter response. However, just for an easier "getting a feel for it" analysis, designing a simple filter in some circuit simulation software and running some signals through it to look at the output is probably the best bet. For example, what you should expect if you put a square wave or a series of pulses with a certain frequency through a bandpass filter, is an AC signal that looks roughly like a sine wave at that same frequency, with a degree of distortion that depends on how good the filter is. If the filter were a perfect bandpass (completely rejected anything outside its passband, completely passed anything inside it... impossible to build one but useful to think about) then you would get a perfect sine wave, and if it's not a very good filter then it would probably look more like a smoother version of the input signal, with more rounded edges.

Edit: I grabbed a slide from an old project presentation of mine and attached it. The BLUE signal on the left is the input of the bandpass filter, and the signal on the right is the filtered output. The input signal is a bit more complicated than you're asking about, but the response of a square wave (such as the green signal which I've drawn on the left) would look similar.

 

[stevez]

This may sound critical but I don't mean it that way - have you tried a library? Maybe you don't have access to a good one so the internet is all you have. While the internet does provide access to a lot of stuff you're missing an awful lot of good publications. I've seen quite a few references to good books on bandpass filters and it's likely that one of them has all you need. Seems to me that a number of magazines I've scanned over the past few years cover this subject quite well. I subscribe to QST (ham radio) and the June 2006 issue has an article on an audio filter for CW as well as some advice on how to work thru the formulas.

 

[dknguyen]

Ikalogic, do you know about FOurier analysis? How a signal is made up of many superimposed sinusoid signals of different amplitudes, frequencies, and phase shifts? A filter works by supressing a select range of those sinusoid signals (based on frequency). If you don't know that, then there;s quite a bit of research you have to do.

TO answer your question, if the oscillating signal is a sinusoid, it will only be DC if the filter supresses all frequencies larger than 0Hz. THis is a low pass filter. Of course, real filters never supress frequencies perfectly and don't have sharp cutoff points for the frequencies they do supress, so it won't be perfect DC.

If the filter supresses signals higher than the frequency of the oscillating sinusoid signal, output and input will be the same (ideally anyways). THis is a high pass filter.

A band pass it a combination of the two. If the oscillating signal's frequency falls in betwee the minimum and maximum pass frequencies, then it will be supressed. If not, output = input (idealy).

 

[Roff]

Ikalogic, do you know about FOurier analysis? How a signal is made up of many superimposed sinusoid signals of different amplitudes, frequencies, and phase shifts? A filter works by supressing a select range of those sinusoid signals (based on frequency). If you don't know that, then there;s quite a bit of research you have to do.

TO answer your question, if the oscillating signal is a sinusoid, it will only be DC if the filter supresses all frequencies larger than 0Hz. THis is a low pass filter. Of course, real filters never supress frequencies perfectly and don't have sharp cutoff points for the frequencies they do supress, so it won't be perfect DC.

If the filter supresses signals higher than the frequency of the oscillating sinusoid signal, output and input will be the same (ideally anyways). THis is a high pass filter.

A band pass it a combination of the two. If the oscillating signal's frequency falls in betwee the minimum and maximum pass frequencies, then it will be supressed. If not, output = input (idealy).

Corrected:

A band pass it a combination of the two. If the oscillating signal's frequency falls outside the minimum and maximum pass frequencies, then it will be suppressed. If not, output = input (ideally).

Of course, if the input signal contains some frequencies within the passband, and others which fall outside the passband, then only those within the passband will be passed.

 

[dknguyen]

Ah, thanks. PASS-BAND, duh. hehe

 

[ikalogic]

Okay i'll be back here after some readings and research about Fourier.. maybe with a bunch of new questions..

thanks for all.

 

[Sceadwian]

The stuff you can do using Fourier methods in the digital domain is pretty impressive though, cutoff's can be almost instantanious and as close to complete rejection as you can ever get. Pretty processing intensive though.

 

[dknguyen]

What are wavelets? I don't see a course in them at the university but I've sort of heard the word tossed around. Maybe it's a graduate thing. Im not sure if it's even used practically or if its still in research.

 

[ikalogic]

Also, by the way, are there ready Band-pas filter ICs to start with?
Can a 556 tone decoder be used? (i guess PPL is also a way to detect/reject frequencies right? )

Just to let you know guys, i am in Mechatronics Engineering.. we don't do lot of electronics, nor electronics related math, that's why it isn't that easy for me..

 

[ericgibbs]

hi ikalogic,
Have a look at www.analog.com site. Lots of applications, try to get
yourself on their tech mailing list, well written info.

Also they are helpful with free samples.

Regards
EricG