How come when I run a high pass/low pass filter...

[Steelflight]

How come when I run a low pass and/or high pass filter on my recordings in Audacity (High pass-250, Low pass-600), these frequencies still appear when I load these edited wav files in Praat? I really like Praat's spectrograms, sonograms, and the various analyses you can do with them.

I filtered out a recording with a high pass of 250 Hz and a low pass of 600, but I still get a pickup of 1xx Hz in my spectrograms in Praat. Anyone know what the deal is? I thought the filter would have "attenuated" such a low frequency...

Thanks!

[ermine]

I thought the filter would have "attenuated" such a low frequency...

It has. But you have to ask yourself by how much. Your spectrogram shows a typical range of about 60dB, and to ask a 250Hz filter to have attenuated an octave down by more than 60dB is asking a lot - how steep did you specify your filter, and if you are filtering that hard, how do you know you are not corrupting the signal? A filter blends from the passband to the stopband, and typically steeper filters adversely affect the tonal quality.

You ideally want to get a bit closer, but you do have a signal that is visible so you shouldn't need to filter that hard. Try expanding the LF section or use a log frequency scale

[Steelflight]

How "hard" would you filter it then? Just do the high pass of 300 HZ like you suggested earlier?

[Steelflight]

And what do you mean "expand the LF section"? I've been looking at it and can't quite figure this out!

Thanks.

[ermine]

Take this recording of a scops owl for example. I love the loneliness of this owl signal coming from about half of a mile away, but let's face it I'm not going to win awards with it as it is weak and swamped by crickets. I don't generally do filtering if I want a spectrogram, so let's see what this does.

scops1.png is what I get if I take this into Raven straight. Pretty rough though the owl signal is visible. Let's see what we can do. First thing to note is I am looking over a long period of time (2.5 secs) so I am not placing a premium on temporal resolution. So let's see if some frequency resolution helps - owl songs nend to be long notes held for a while, as opposed to say the twittering of a wren. So I rack up the resolution bar on Raven (the right-hand slider) all the way up to 975, which gets me scops1_res975.png

I then observe that all the action I want is down low, and i can get rid of those pesky crickets if I zoom in to the frequency range using the + sign on the vertical slider in Raven to give me a closeup of the range 0 - 3.8kHz. I don't need all the HF rubbish. I observe the scops starts at 1.6kHz rapidly falling to 1.3kHz which is held pretty much steady with a faitn waver at the end. If I were looking for harmonics I would look for the second harmnic at 1.6kHz x 2 - 3.2kHz but there isn't any above the noise floor so I guess this is a pretty pure call.

This was recorded with a parabolic dish at night and you really shouldn't try and record things half a mile away but Raven can still pull something from the mess. You don't need any filtering if you are simply using the sound file for spectrograms - it doesn't help make things more visible. It can help tmake things more listenable - the owl would be clearer if I filtered out some to the HF from the crickets, though I would damage the soundscape, such as it is.