mono songs saved as 1-channel audio files

Out of curiosity, I picked apart some lossless stereo files of some mono songs from the early 1970s. I was curious if there was any benefit of saving them as 1-channel (true mono) lossless files, rather than 2-channel (stereo) lossless files where the channels have pretty much the same content.

I'll give you the answer first: For a mono song, I found a benefit to saving one track of the stereo file as a 1-channel ("mono") audio file.

Then, methodology: To look at each track, I perform an out-of-phase-sum test, or OOPS.

The process is straightforward in Audacity. Import a stereo audio file, use "Split Stereo to Mono" in the little rectangular menu at the left edge of the track, invert the right channel (by selecting the now-mono track corresponding to the right channel, then Effect > Invert), and hit play. If there's a volume imbalance between left and right channels, you can compensate for it by adjusting the volume one of the two now-mono tracks. If the song is mono, you'll get some nice cancellation between the left and right channels, which should leave behind some non-musical artifacts (like dithering noise).   

(To make sure that Audacity doesn't introduce dithering artifacts of its own, go under Edit > Preferences > Quality, set the Default Sample Format to 16-bit, and set the High-quality Conversion Dither to None. Do this before you start.)

Dithering noise is noise that is intentionally added in software to randomize the smallest bit(s) of an audio signal. In general, it improves the sound at very low volumes (for CDs, that can be from around -90 dB to around -70 dB). Audio editing programs usually add it when they perform some kind of volume adjustment. For most recording/mixing/editing applications, you don't need to pay attention to it; programs like Audacity automatically add in dithering noise as a default setting. Wikipedia has extremely detailed entries for "Dither" and "Noise shaping" - it will surely tell you more than you need to know.

I found that the vast majority (maybe 80%) of good-sounding CDs with mono tracks, as well as just about all homemade two-channel audio tracks made by some superb audio people on this board, have left and right channels that don't sum to zero, but instead sum to dithering noise. Put another way, the left and right channels aren't digitally identical, but differ by an amount that represents dithering noise. The dithering noise can vary from track to track, from about -90 dB to about -40 dB.

Weird! I would have guessed that mono songs (with good mastering) would have left and right channels that would be digitally identical, but most don't. I did find some that were digitally identical (like the Grass Roots' The Complete Original Dunhill ABC Hit Singles CD from 2014), but this was a tiny minority, maybe 10% of everything I looked at.

To see what effect the dithering noise had on file size, I took a CD-ripped track and saved it as a stereo (2-channel) FLAC file. Then, in Audacity, I saved only the left channel as a true mono (1-channel) FLAC file. (Technically, I saved both as WAV, then used the same software and the same settings to data-compress to FLAC.) I compared file size (or, equivalently, data rate) for the two-channel and one-channel files. For tracks that sum to dithering noise, I found that the one-channel FLAC files are about 20% to 25% smaller than the two-channel FLAC files. Put another way, the FLAC encoding spends 20-25% of its effort to properly encode the dithering noise.

I also found that for a track where the left and right channels are digitally identical, saving just the left channel of audio as a 1-channel FLAC file has exactly the same file size and bitrate as the original 2-channel FLAC file. That's a good sanity check.

Interesting.

And it got me thinking: For most applications, like playing a CD, no one cares that there's dithering noise between the two channels. But for some applications that have a limited bandwidth, such as a streaming radio station that streams at 128 kbps, there could be some negative effects from the dithering noise. Basically, 20-25% of the bandwidth is taken up by encoding dithering noise, which nobody really cares about for mono songs. Imagine if we freed up that 20-25% to better encode just the audio we do care about. Doing so would likely improve sound quality. We can do just that by using just one channel of the song (I chose the left channel), and saving it as a 1-channel audio file.

Using just one channel from a mono song also eliminates artifacts that arise if the left and right channels are out of synch. I found one example - the mono song "Hearsay" by Soul Children, from the Time-Life CD Rhythm & Blues Vol. 20 1972 (1996), also released as Solid Gold Soul Vol. 7 1972 with the same mastering. "Hearsay" has the left and right channels out of synch by about two samples and a small volume difference between left and right channels. Here, saving just the left channel as a 1-channel FLAC file cut the file size roughly in half, meaning that the FLAC encoding devoted about half of its bandwidth to ensure that the left and right channels are reproduced properly (meaning, out of synch!) For synchronization errors, saving just the left channel as a 1-channel FLAC file really improves the sound in a tangible way, by eliminating the desynchronization problem.

For my own library (hand-picked tracks for every version of every song), I'm going through the folders and manually creating 1-channel FLAC files for every mono song, using the left channel from the 2-channel CD rip or homemade file. I estimate that I have about 1,000 files left to work through.

(If you're going to attempt this, don't just do a mass file conversion to "mono"; I think doing so will sum the left and right channels, which will hard-bake any desynchronization errors into the 1-channel file.)

I'm curious if anyone else has dug into this, or attempted to use 1-channel FLAC files for mono songs.

Edited by crapfromthepast

The vast majority of dither on ordinary CDs prior to around the 2000s was just a normal artifact of playing a mono tape through stereo gear. That is, the "dither" in those recordings was not generated noise that's added during digital processing, but rather, just the random noise that's inherent to the tape format, as well as in the electrical circuitry in the chain of audio gear in between the tape and the ADC. That's why there's rarely a CD where a mono recording manifests as bit-for-bit identical left & right channels. Also, as Ron noticed, sometimes there's a sync difference of 1 or 2 samples; that was probably introduced by the digitizing gear (I think the classic Sony PCM "adaptors" in the 1980s did this)...it's not enough to hear, but it's not perfectly mono.

As mastering tools became more sophisticated in the mid-1990s and beyond, more mastering was done in higher bit depths, which was then converted down to 16-bit for CD with the addition of digital dither. On mono recordings, sometimes this dither is in stereo, sometimes in mono. It's very inconsistent. Any random noise, be it tape hiss, electrical circuit noise, or dither, is very hard to compress losslessly. The more intense and stereophonic (different in each channel) it is, the more it will bloat the bitrate.

Dither is mainly for reducing bit depth; it eliminates quantization noise and allows for what is effectively a lossy version of the very quietest audio to exist "in the noise". However, it's still noise and is not ideal in all situations. I avoid using it if all I'm doing is cutting and pasting. As Ron indicates, Audacity's default settings add dither when saving, and it's better (IMHO) to turn this off and only use it when you know you want it.

As for the method of generating pure 1-channel mono from a recording that's mono-with-stereo-hiss, that's kind of a can of worms. When you sum the channels, you effectively reduce the hiss and increase the SNR by 3 dB. But if the signal is better in one channel, or there's a sync issue, as Ron says, it's best to just pick one channel. You have to decide track by track.

There's also an issue which has come up for me: some music was recorded with imperfect mono, where particular instruments or the background hiss associated with just those tracks on a multitrack (or channels in a mixer) might have a slight affinity for the left or right, and ultimately the result is a very narrow stereo mix. Accidental as it may be, it just doesn't sound the same when forced into pure mono. Should such music be treated as mono at all?

Now for the million-dollar questions: Do you really want to go to the trouble to permanently process an entire lossless collection in this way, forcing pure mono on the imperfect-mono tracks, just to save some space, even though disk space is cheap? And will you be keeping the originals just in case? And is this really the best way to go about it, or could you instead just tag the files to indicate what processing you want upon playback? (e.g. using foobar2000 with the Dynamic DSP component.)

For me, after experimenting 7 years ago with making some 1-channel mono versions of some rips, and running into these issues and philosophical questions, I decided to just live with the imperfect mono for now. I'm more interested in curating my collection and getting the best-sounding masterings. Worrying about the ideal processing of mono is much lower priority.

But to answer the general question, yeah, I've thought about this topic quite a bit. :)

Edited by mjb50

Thank you to all in this thread for explaining this to me.

This topic has always been interesting to me (even though
I've always been somewhat uninitiated to do any
research...)

Andy

Absolutely fascinating! I never tried this and would never have guessed the dither would increase the file size that much.

I've converted 2-channel "mono" files to both = left or both = right but in saving with dither again I may have just restored the excess data I was trying to eliminate.

Dither isn't specifically a stereo tool as I understand it, so it's interesting that a stereo 2-track would create a different dither pattern in supposedly identical channels. A mono channel should have dither, too.

The only thing I can think is that dither is supposed to be random, so maybe that's why they're different?

Gene - I was surprised to see that the dither is in stereo. I could try to rationalize a reason for it, but I don't fully understand why it's in stereo either.

Hykker - The steps you're taking are perfect. I think if you just save your final mix as a mono file (1-channel), rather than a stereo file (2-channel), you'll get the benefits of all the steps you're taking. (I think Audition has a "mix to mono" option, or something like that. Cool Edit Pro had such an option.)