mono LPs, signal and noise

[L-R to Mid-Side preset]

I've been living under the illusion that recording of a mono LP in stereo, then converting to mono in software (after declicking), resulted in a 6dB improvement in signal to noise, 3dB from lowering the noise, relative to the ‘stereo' version, and 3dB from increasing the signal level.

Recently noticing that there did not seem to be any increase in signal level from converting to mono, I made some measurements on the album I was transferring.

Before converting to mono, the two channels differed by 0.74dB, in both Peak Amplitude and Average RMS (Analyze/Statistics). I therefore used the average of the two channels to compare to mono results.

The mono version, whether I concocted it with the Channel Mixer L-R to Mid-Side preset, the Channel Mixer Average preset, or through Convert Sample Type, was equal to the stereo version's averages in these two measurements -- no increase in signal level.

I marked music free sections in six between-track segments. The decrease in measured noise varied quite a bit. The average for the six selections was: peak amplitude down 4.86dB and RMS average down 5.12dB.

This does approach 6dB improvement, but it is all from lowered background noise, none from increased signal level. Is this indeed what should be expected?

1dB difference between the two channels is probably not at all bad considering most phono cartridges. The phono amp and soundcard might also contribute. Do other people who record in stereo, then convert to mono (when both channels are in about the same condition), just process it as it gets recorded to your hard drive, or do you first amplify the lower level channel to make both equal in amplitude, then convert to mono?

Well, you don't get a signal level increase when you play a record in mono instead of stereo, because the sum total of energy extracted from the groove hasn't increased in the slightest!

The more interesting consideration is the noise. If the noise from each side of the groove happens, unfortunately, to correlate then you won't get any improvement in SNR at all either. It isn't usually quite as bad as this - the less it correlates, the more it moves out to the sides of the stereo image. You have to note at this point that similarly to the signal, the total amount of noise hasn't actually decreased or increased - it's the distribution of it across the channels that has altered, and that's what you have to take advantage of to decrease the noise in the end result.

If you do a L-R to M-S conversion of your two-channel mono recording, you should end up with the uncorrelated noise appearing as the S (Side = difference) signal, and the wanted signal as the sum. And sometimes you can use the S-channel noise signal as a noise reduction profile to some advantage, but obviously results from this will vary according to the amount of groove correlation.

While the total energy has not been increased by converting to mono, I though the process involved adding the signal from each channel to get the single mono channel. I see that is not happening with what I’m doing, but it seems to make sense. It also seems to be what I thought I read in the past.

Noise that is in, part of, the original recording, should completely not-correlate (in theory). Noise that was added after the LP disk was stamped (or in the stamping process?) would not necessarily be congenial to our desires to remove it in this fashion; it would be purely a matter of the luck of the draw and how that noise came to be (I find that quite a bit of the energy of many clicks ends up being removed by converting to mono).

There is some intrinsic noise added by cutting the recording into the master disk and by the physical playback process. Does the out of phase part of this normally come to 3dB, 6dB, or something else?

The SNR can be increased even if the signal is not because the noise is reduced.  If you sum the two channels
of a mono vinyl recording, it depends how you sum it.  If you use the Channel Mixer Average Mix preset, then
the resulting signal strength will be about the same (50% + 50% = 100%).  If you use the Full Mix preset,
it will be 200% (+6.02 dB).

Summing the two channels will include all clicks from both the right and left channels, but at 50% amplitude.
Using one channel only will include the clicks from only one channel, but at full 100% amplitude.

As SteveG said, the noise reduced by summing the channels depends on how the record was cut.  Some
records have the low noise (up to about 500 Hz, decreasing as you go higher) cut out of phase.  This is
probably done so that when you send the bass of both channels to a single speaker, the low noise cancels.
Other records may have the noise of one channel rotated 45 degrees with respect to the other.  Recently, I took
three copies of a 1st pressing of a record, with the same dead wax engravings, and found that the noise
correlated perfectly between two of them, but not the third!

Summing the two channels of a mono record is a tricky business, though.  If the high frequencies are not
aligned perfectly (due to imperfect cartridge setup so that the timing between left and right is off), then
summing will weaken the highs because they will partially cancel.  Also, not only are the left and right typically
at different volumes, but the frequency response will not match, either, so just amplifying one channel to
match the amplitude of the other is not enough.  If the timing were perfect, then summing the two channels
would produce a result averaging the frequency curves of the two channels.

In my view, to do it right, you need two tools:  1)  a frequency matcher (Audition doesn't have one), and
2) a Center Channel Extractor tuned for vinyl (Audition's isn't, as far as I can tell).  Audition can provide
other functions helpful for the task, including Channel Mixing, and timing correction. 

Hehe - Long time no hear from!

If you sum the two channels of a mono vinyl recording, it depends how you sum it.  If you use the Channel Mixer Average Mix preset, then the resulting signal strength will be about the same (50% + 50% = 100%).  If you use the Full Mix preset, it will be 200% (+6.02 dB).

It's not quite as simple as that, I'm afraid. Certainly not when it comes to arriving at the total amplitude you get from summing the output of a cartridge to mono, anyway. This is because it is an energy sum - which means that if you use the full mix preset, you will end up with a result that sounds louder - ie, it's not the same as just joining the two signals together. The problem is that stereo signals contain direction information as well as amplitude, so simply considering the signals alone doesn't provide all of the information necessary to arrive at the correct amplitude of the centre component. To get the correct answer, we have to consider a dB addition of the absolute power that each channel represents. And that means that the final result as a centre channel pressure would be expressed as

p= sqrt (p₁² + p₂²)

and that would give us 3dB more signal at the centre, not 6dB. If it wasn't like this, we would never have got away with supposedly mono-stereo compatible records at all. It would be more accurate in some ways if we used 70.7% as a mix figure rather than 50% for this sort of application - that would give us a closer approximation to a realistic result  - but it wouldn't alter the SNR one jot; just the max level of the track.

Some records have the low noise (up to about 500 Hz, decreasing as you go higher) cut out of phase.  This is probably done so that when you send the bass of both channels to a single speaker, the low noise cancels.

If you find a record like that, it was cut by somebody grossly incompetent. If you get the bass information out of phase on a record, the first thing that happens is that the stylus mistracks badly, because the track would try to throw it clear out of the groove! The idea is, and has always been, that the bass information is in phase, and therefore always converted to the same vector sum - which is vertical movement. So, true stereo compatible mono cartridges have always had to have the same amount of vertical compliance as stereo-only ones. This is why bass instruments and kick drums have always tended to be panned to the centre in a mix intended for vinyl release.

And they are generally limited in level very carefully - this was always a major part of the skill of preparing a master for cutting. If you don't limit the bass excursions, then grooves crash into each other - the level of bass still determines the overall width of the groove, after all. So in a cutting room, you always have some sort of frequency-dependent stereo width control - that's the ultimate fix for the problem, after all.

Hi SteveG and AndyH,

In the summing thing, I was referring only to signals from mono records, or any two-channel mono signal.  You can test it:
Do a Full Mix on any two-channel wave. Get stats.  Do a Full Mix. Get stats, which will be 6 dB greater in power and amplitude.

In the record cutting, I was referring only to low "cutting" noise, not the lows on the source tape.  If you look at the wave
of the noise (between tracks) recorded from vinyl, for many records, when the left waveform goes up, the right goes down,
and vice versa.  If you play it with bass going to a single speaker (such as many computer speaker systems), the bass in the
noise disappears, but you can hear it clearly when playing either of only one channel.

I keep a 70.71% mix preset around, plus a script for properly folding down stereo to mono 

In the summing thing, I was referring only to signals from mono records, or any two-channel mono signal.  You can test it:
Do a Full Mix on any two-channel wave. Get stats.  Do a Full Mix. Get stats, which will be 6 dB greater in power and amplitude.

Hmm... despite what it says, Audition can't measure power directly at all - in fact all references to power should be removed - as I've said for years. Yes, if you do a 100% mix you will get 6dB more amplitude in the resultant file, but if you put this mono signal into a loudspeaker in the centre of your stereo spread, it would be too loud. You don't want to believe everything that Audition stats tell you - the labels are wrong, and there are factors that they don't take account of.

Adding the two signals that arrive at speakers or earphones that are separated gives you a resultant signal that doesn't represent the signal level that would have been presented to a centre speaker at all - that's why we need panpots that follow a -3dB (which is also arguably incorrect as well, but for a different reason) law for panning a mono signal across a soundfield - which is what the Audition panpots do.

If you want to know why the -3dB figure is arguably wrong too, it's a slightly different, and longer explanation. But however you look at it, a 100% mix is quite inappropriate for creating a mono track. Pictures actually make this a lot easier to explain, but barring some sort of miracle, I don't have time for a few weeks.

Hello Steve,

If you get the bass information out of phase on a record, the first thing that happens is that the stylus mistracks badly, because the track would try to throw it clear out of the groove! The idea is, and has always been, that the bass information is in phase, and therefore always converted to the same vector sum - which is vertical movement.

I believe you have that the wrong way round!

If you take a mono record, it will have been cut with lateral movement only, ignoring any unwanted incidental vertical movement that may have occurred. When played on a stereo system, this will give a central sound image, which represents in-phase signals.

See this link:

http://www.aardvarkmastering.com/riaa.htm

Paragraph B, Item 4.

RECORD INDUSTRY ASSOCIATION OF AMERICA, INC.
BULLETIN E 3
STANDARDS FOR STEREOPHONIC DISC RECORDS
October 16, 1963

B. The Stereophonic Groove

4. When equal in-phase signals are recorded in the two channels they shall result in lateral modulation of the stereophonic groove shall produce equal in-phase acoustical signals.

On a stereo record, out-of-phase signals will, therefore, give vertical movement of the groove/stylus.

So, true stereo compatible mono cartridges have always had to have the same amount of vertical compliance as stereo-only ones.

It has always been my understanding that the reason for compatible cartridges having sufficient vertical compliance was to avoid, or at least minimise any damage to the vertical motion of the record's groove.

This is why bass instruments and kick drums have always tended to be panned to the centre in a mix intended for vinyl release.

Surely, the reason for those instruments to be panned to the centre of the mix is to ensure that they result in only lateral movement of the groove/stylus? I would imagine that if it resulted in vertical movement that would be more likely to 'throw' the cartridge out of the groove!

And they are generally limited in level very carefully - this was always a major part of the skill of preparing a master for cutting. If you don't limit the bass excursions, then grooves crash into each other...

Isn't that more likely to occur with lateral movement than vertical?

... - the level of bass still determines the overall width of the groove, after all.

Not so - it is the vertical movement (i.e. the out-of-phase signals) that determines the width of the groove. It is the lateral excursion of the groove that is determined by the level of bass.

Regards,

Dave.

[If you take a mono record, it will have been cut with lateral movement only, ignoring any unwanted incidental vertical movement that may have occurred.]

If you take a mono record, it will have been cut with lateral movement only, ignoring any unwanted incidental vertical movement that may have occurred.

Well, unless of course it's a Daimond Disc.

This is a little off topic, but I listened to a recording of a conference in England a while back geared toward audio restoration research, and the featured speaker was from a company that had figured out a way (you knew somebody would do it eventually) of taking snapshots of phonograph records and then converting those to music files. Since there's no needle in a groove the noise level is cut way down like this. It more or less ignored scratches jumping right past them, and if they were bad enough it would attempt to fill in the blanks and did so quite well. It was astounding to my ears.

I believe you have that the wrong way round!

Yup - so do I... 

Most of the arguments are the same, just the directions are different. Usually I check first - I don't normally trust my memory, and there's why!