Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: lowpass wav file to remove FM pilot signal (Read 4431 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

lowpass wav file to remove FM pilot signal

I have a digitised cassette tape recording (not my tape) in flac (44.1kHz / 16bit) that I've converted to wav for editing (remove non-program specific start and end audio).

It's a very good recording hitting the ~15kHz limit that the BBC imposes on FM transmissions, but more importantly sounds great for a cassette tape FM recording from 1996 with little hiss.

Looking at the spectrum I can see a pilot signal at ~19kHz (see here for a screeny).

My questions are:
1. should I bother to run the wav file through a lowpass filter to get rid of the pilot signal?
2. If so, how is it best to do this? I can do it in Audacity but thought that using a resampler like sox might be better as suggested here. If so, are those options correct? I'd use 0-16000 for the filter of course.

I will keeping two versions of this file:
(a) a lossless (flac) archived copy and
(b) a lossy (mp3) copy that I'll run a 16kHz lowpass filter on when creating it


Re: lowpass wav file to remove FM pilot signal

Reply #1 –
Doesn't Audacity use the SoX resampler by default now?

Re: lowpass wav file to remove FM pilot signal

Reply #2 –
That pilot tone is at -87dB. Good luck trying to head that above the normal signal.

You could also try a notch filter to filter out only the pilot tone.

Re: lowpass wav file to remove FM pilot signal

Reply #3 –
Since it's inaudible,  I'd just leave it.    Unless you wanted to try some general noise reduction for the tape FM/tape hiss, which should kill that 19kHz tone too, since it would be included in your noise fingerprint.

A notch filter would generally give you more reduction, but at -87dB and 16-bits, if you can knock it down by 10 dB you'll be down to dead-digital silence.    So a low-pass filter is probably equally as good.

Quote
but thought that using a resampler like sox...
Down-sampling requires   low-pass filtering so there's no advantage to resampling and  downsamping (except for a smaller file).    If you could downsample without low-pass filtering (which no resampler will do) the 19kHz "signal" would simply be "aliased-down"  to a lower frequency.

Re: lowpass wav file to remove FM pilot signal

Reply #4 –
Ok, thanks everyone. I think I'll just leave it.

Sometimes I overthink things and seeing this pilot signal I thought that maybe I could get rid of it for the lossless archival version, but as it's almost certain that I cannot hear it then why bother. I'll use the lowpass when making the mp3 though.

@EekWit: you're right, it does, cheers.

Re: lowpass wav file to remove FM pilot signal

Reply #5 –
Thought I throw my 2 cents in.  FM radio caps out around 15 KHz, so resampling to 32 KHz would remove the pilot tone and have no audible effects.

Re: lowpass wav file to remove FM pilot signal

Reply #6 –
Quote from: Chibisteven on
Thought I throw my 2 cents in.  FM radio caps out around 15 KHz, so resampling to 32 KHz would remove the pilot tone and have no audible effects.
Wouldn't that push dither noise further into the audible range?

Re: lowpass wav file to remove FM pilot signal

Reply #7 –
Quote from: pdq on
Quote from: Chibisteven on
Thought I throw my 2 cents in.  FM radio caps out around 15 KHz, so resampling to 32 KHz would remove the pilot tone and have no audible effects.
Wouldn't that push dither noise further into the audible range?

If the developer would simply resample without doing anti-alias filtering, then the aliases would become very audible.   If, by 'resampling', it is meant to 'anti-alias filter', then 'change the sample rate', then there is little additional benefit over and above just 'do the filter'.   So, I'd just suggest a brickwall filter at just below 15k with some large number of taps (if there is enough CPU.)   The 19k pilot tone would be effectively totally eliminated.   I would NOT suggest using a 2/4/6 pole IIR filter, because there is definitely an audible effect (which might or might not be desirable.)   If one just does a remez generated FIR filter with perhaps 1023 taps at 14.5kHz, then nothing above 15kHz will be allowed through.  (BTW, I sometimes use a 2pole IIR filter with a q of 1, at 15kHz to make certain kinds of musical sounds more 'sharp' sounding.)

John Dyson  

Re: lowpass wav file to remove FM pilot signal

Reply #8 –
Quote from: pdq on
Quote from: Chibisteven on
Thought I throw my 2 cents in.  FM radio caps out around 15 KHz, so resampling to 32 KHz would remove the pilot tone and have no audible effects.
Wouldn't that push dither noise further into the audible range?

One could use a flat dither at the bottom bit since you're only dealing with 2 channels here and already dynamically compressed material when dealing with lower sample rates that could push shaped dither into the audible range, if you're paranoid.  I question it's audibility because FM hiss is a lot louder than the noise floor of a 16-bit recording.  Now if you had something with 20 channels and the recording is low amplitude, then I be much more worried about how things add together.

FM radio and audio cassettes are so noisy that they practically self-dither when you record them.  Random noise in FM radio has a triangular effect so it's more concentrated towards the higher frequencies and because the stereo signal is modulated in the higher frequencies it tends to be more noisier when listening in stereo for this reason.

Re: lowpass wav file to remove FM pilot signal

Reply #9 –
Quote from: Chibisteven on
Thought I throw my 2 cents in.  FM radio caps out around 15 KHz, so resampling to 32 KHz would remove the pilot tone and have no audible effects.

32KHz sampling rate is not the way to go. You are using a semi non standard samplingrate to overcome an issue that is not an issue in the first place.

15 years ago 32KHz was quite common in the days of DAT long play mode tapes. These days more and more soundcards and amplifiers do not understand the 32KHz sampling rate anymore, requiring you to upsample again.

In that case you're better off with a lowpass filter, or no filter at all.

Re: lowpass wav file to remove FM pilot signal

Reply #10 –
With that super low amplitude I would just do nothing and treat them as typical 16/44 files.

Re: lowpass wav file to remove FM pilot signal

Reply #11 –
The "-87" should be compared to the overall level, which seems to peak at -27 dB or so.
Still, a -60 dB tone at 19 kHz should not be much? Likely it won't even be useful as perceptual noise substitution? ;-)

(By the way, a cassette deck capturing 19 kHz - that sounds like a decent one, doesn't it? Or was the pilot tone just so much louder on the air that the -87/-60 is after significant roll-off?)

Re: lowpass wav file to remove FM pilot signal

Reply #12 –
The major ways that 19kHz subcarrier leakage might hurt is if you send the signal to old analog circuitry.   Even then, -60dB is pretty low.  However, if you are seeing higher levels, then it could be a problem on electronics that has bad HF nonlinearities.   Old transistor designs (e.g. simple, three transistor gain circuits without appropriate management of HF range in the design -- and other such situations) can happily and every effectively produce audible intermodulation distortion (not necessarily as bad as having birdies, but simply an obvious edge to the sound.   Also, old tape gear didn't like strong levels of HF (again, -60dB is okay, -20dB might start causing problems.)
In the digital world, it is VERY easy to get rid of the 19kHz signal, and frankly, if it bothers you for some appropriate reason, then just nuke it with a nice filter.  If you use Sox (I live in mostly a command line world, since I am an EE/software person, not a music professional), then just add the option 'sinc -15k' or 'sinc -14.5k'.   These commands will do a sharp filter without adding terrible phase or peaking issues.  (Sox is free and even though it has a fairly ugly command line, it generally works fairly well.)   So, if you are using sox, then the following command might be useful:

sox infile.wav outfile.wav sinc -15k

I use sox on Windows and on Linux, and is a nice little swiss army knife that doesn't require firing up a big GUI to do a simple task.

John Dyson

Re: lowpass wav file to remove FM pilot signal

Reply #13 –
I ran some 96k RMAA tests in the studio I previously worked for.
https://hydrogenaud.io/index.php/topic,96560.msg806718.html#msg806718

Test signal playback and redigitize device:
https://www.rme-audio.de/en/products/multiface_2.php

Cassette recorder under test, using a type I tape with Dolby B enabled:
http://tascam.com/product/cd-a750/

I attached the flac of the sweeping IMD test, notice the 1kHz spike, it is the IMD! Also notice the imperfection of the phase analysis and the nice high frequency roll off.

The subjective sound quality of actual music playback is not problematic at all, it just sounded like what a cassette deck should sound.

It would be possible to improve the result by using hi-end decks and tapes, but that 19kHz pilot tone is just too low to worry about. Just think about the amount of distortion in the whole audible spectrum.