Is there is a lower frequency on vinyl or CD beneath which any content on the media could be considered noise and not part of the recording artists' conception or intent in making the recording?

Is it correct to assume that the master creation and fabrication process do not add anything to the supplied master recording at all? I am thinking back to the days of cutting LP masters on lathes and so on, where the cutter had to be skilled at ensuring a needle would not slip out of the groove on the final vinyl. I could imagine mechanical noise and resonances affecting the groove being cut but have no real idea whether any contamination of the source material would ever have happened in practice. And I simply don't know if the master recordings are subject to any filtering or other processing before a CD master is prepared. It would be interesting to know.

Presumably if there were any such low frequency or low amplitude content it would be stripped out by a lossy compressor but would be encoded by a lossless compressor to maintain "the purity of the source"?

Elsewhere I read that stripping out very low frequency content might improve compression for "lossless" compressors. I don't know whether there is low frequemcy noise to remove. Hence my questions.

I would like someone more knowledgeble to provide some information to help with the philosophical question here. Is the lossless compression of a "pure source" a warts and all thing or if low frequency noise could be identified as being a contaminant in the ripped "pure source", should it be heresy to consider removing it before, or as part. of a "lossless" compression?

(I personally see no harm in pre-processing a non-copyright audio source file for personal use, but I would guess majority opinion would be that anything to distribute to others ought to be warts and all unless one was the originating artist).

There is a great deal of sub-sonic noise coming off phonograph records. Between tracks, where there is no music signal, the general curve of a frequency analysis graph is lowest at the far right, the highest frequencies, and highest at the far left, the lowest frequencies. The highest level is often below 10Hz. Its level is often higher than many quieterl music passages.

This is supplied by many sources. Probably no LP has its hole is the exact geometric center, so there is a horizontal component added by the tonearm moving in an out because of that. Probably no LP is completely flat so there is a vertical component added by the tonearm moving up and down because of that. There are bearings for the platter and bearings for the motor and neither are never perfect. No doubt there are other sources. Some of them are certainly earlier in the production chain and thus part of the actual recording, the same factors I mentioned above are present at the cutting lathe. Rumble filters were once much more common on phono preamps because of all this but now have mostly gone out of fashion.

I'm not sure of the sources on all digital material on CDs, but excessive subsonics are generally observed there too. The levels are not as high as on recordings from LPs, but they do waste a significant amount of amplifier power.

My opinion is there is no down-side and definite benefits to filtering out the subsonics. I don't know if any such filtering is a lossy compression default, I haven't paid any attention, but my first guess would be 'no.'
Possibly many people will react against the suggestion that 28 to 30Hz is a generally good cutoff frequency, but my experience suggests it is. Quite aside from what I see with the many LP recordings I've worked with, my music playing experience suggest that cutting that high will rarely be a detriment. That said, a case by case check can't be a bad idea.

I think lossy codecs high pass the input signal. At least its hard to imagine most of them will devote a frequency band or transform or whatever they use to subsonic information. But I'm just guessing since I never thought to look into it.

I did try with a selection from an LP recording. I encoded with Lame preset -V 2 --vbr-new --noreplaygain. The decoded mp3 was a little bit different according to my frequency analysis, but there was no indication of filtering out, or reducing the level of, subsonics.

I applied my most commonly used rumble filter to the original and encoded again. The second mp3 was 3014 bytes smaller, not a lot relative to the more than six meg total (source is a little more than four minutes, 42.8 meg size).

Much the same should apply to tracks from CDs. At least to the reamdom few I've sampled, all of which had significant subsonic content.

Sorry for the long delay in thanking the kind people who replied to my original post - I've been somewhat distracted - sorry.

I now have to lose my internet connection for a while, but hey, it may give me some time to do some actual tests!

From earlier contributions it does seem that filtering out low frequencies won't make too much difference to the size of an mp3. I wonder whether the situation would be any different with a lossless encoder such as monkey's audio or flac? If a chunk of filesize can be removed without damage to the audio as intended by the recording artist then I personally think it is worth investigating. (It wasn't my idea but I didn't like the knee-jerk short shrift the proposer got in another forum).

I know the thought of doing anything as part of the lossless encoding process would mean it wasn't lossless, and that concept really bugs some pedants, so I am merely going to consider the concept of removing from recordings very low frequencies that were not deliberately put there by the recording artist as pre-processing, in the same way as removing clicks and pops from vinyl recordings before archiving. That way if I losslessly encode garbage then at least it will pure losslessly encoded garbage when it is encoded.

In the end it is down to personal taste and judgement I suppose. I just can't see why, if filtering out low frequency junk makes no audible difference, but could possibly make a difference in encoded file size, why lossless encoder developers couldn't add it as a "lossy" option to their encoder, where such a thing would be most conveniently located. Maybe it is a "sacred cow" topic, although it is hardly radical.

Lossless is lossless.

Talking about changes that might not affect transparency but do result in a decoded file being different from the original used for decoding belongs squarely in the category of LOSSY.

It's interesting that you should used the term "pedants". Replicating even the minutia is at the very heart of what lossless encoding is about.

I agree that it should not be part of a lossless encoding process. There are thousands of ways to process audio. Duplicating them (some of) in a specialized process whose purpose is clearly for one and only one particular thing seems like a poor idea that would interest very few people.

Pre-processing audio files to make them whatever you want them to be before encoding is a completely different thing, however. Since there are so many tools available that do a good job, no one who wants that result is denied anything.

If you are referring to vinyl LP, I could be wrong about this I am no expert in this area, I have read some books though. In order to reduce unwanted "rumbling" and other noises and compensate for higher frequencies a technique known as RIAA equalization would be applied to Vinyl recordings back in the day? I think that's what you are getting at? maybe




Don't quote me on this, but I think MP3 uses a FIR polyphase lowpass filter. That's how lowpass filtering is done per subband. Maybe that helps a tad bit.




Right the reason we say lossless is lossless, is because we had to show everyone how linear prediction coding works it would take a considerable amount of time and a lot of background knowledge.

Having recorded a fair share of LPs there is a lot of subsonic noise. I think it is generally considered good practice to try to filter out anything below ~15Hz for most systems in either the signal or the hardware.

I distinctly remember playing a CD I recorded from LP prior to enabling the high-pass filter which made the woofers on a DC coupled amplifier based system pump excessively. This can lead to higher distortion and even equipment damage at high volumes.

IIRC LAME does not high-pass filter signals and I think encoding low frequencies is not too tough for most lossy codecs (could be wrong though), so I would definitely do it during processing of the recording prior to encoding.

Every composition is different, at least to some extent, but for most LPs, filtering out everything below 30Hz would not make a difference -- as far as the music goes. A 40Hz cutoff would be ok in a significant percentage of cases. Same for most CDs.

"If a chunk of filesize can be removed without damage to the audio as intended by the recording artist then I personally think it is worth investigating."

= transparent lossy encoding