3

Normally, if you merge stereo to mono, the center component gets a +3dB boost because it is doubled. Is there a way to merge stereo to mono while keeping the balance between the center and sides?

5
  • 4
    Which DAW? Most can change the pan law applied. This is Cubase - i.stack.imgur.com/JuyXs.png
    – Tetsujin
    Oct 26, 2023 at 15:31
  • 1
    Please don't cross-post to multiple SE sites. Pick one & stick to it, or flag for migration. Is cross-posting a question on multiple Stack Exchange sites permitted if the question is on-topic for each site? - sound.stackexchange.com/questions/52529/…
    – Tetsujin
    Oct 26, 2023 at 17:04
  • @Tetsujin pan law has no impact on merging already mixed stereo track into mono. Oct 29, 2023 at 21:45
  • Or to be more precise: has no effect on relative levels in such case, which seem to be the subject of the OP question. Oct 29, 2023 at 22:50
  • Can you explain why you think the "center component ... is doubled"? By "center component", do you mean the elements panned to the center? When an element is panned to the center, that means half the level of the element is in the left channel and half the level is in the right channel. It's the same total level across both channels, whether those channels are summed or not should have no effect on the total energy of any element panned to any position in the stereo field. Nov 1, 2023 at 14:46

4 Answers 4

2

If you want to control the relative levels of mid versus side information, a clear way to go is first run the stereo signal through a mid/side processor, then run that output through whatever process is collapsing to mono. If the mono collapse seems to change the balance of mid vs. side, you can compensate by adjusting the levels in the mid-side processor.

9
  • Unfortunately, this won't work. Side channel includes L and R in opposite phases. Mixing it with Mid will cancel either L or R. Oct 27, 2023 at 16:44
  • @user1079505 That’s not how I understand M/S working, and I’ve used both M/S miking and processing techniques extensively. Oct 27, 2023 at 17:24
  • 1
    I think you should clarify what you mean by "mid/side processor". If you simply transform L & R to M & S and then fold down by summing M + S, then @user1079505 is right: You end up with just the Left channel. I think you're talking about a more complex processor that transforms L & R to M & S, performs some processing, and then transforms back to L & R to put onto the bus. In that case you will have a more useful result when folding down the L & R bus to mono.
    – Theodore
    Oct 27, 2023 at 19:06
  • 1
    @Theodore Right. Probably some phase transformation of either M or S component could prevent that unwanted cancellation. This is hinted by user95217 , but that answer needs some more explanation. Oct 27, 2023 at 19:09
  • Wow what are we even talking about in these comments? I can’t figure out how one would think that I'm talking about treating L as mid and R as side. How would that possibly fit any definition of “mid/side processor”? Oct 27, 2023 at 19:31
2

What is the question about?

When one mixes down a stereo recording to mono, the tracks which were originally closer to the center sound louder than the ones originally mixed on the sides, in comparison to the sound of the original stereo recording played on stereo speakers.

Why does it happen?

Let us imagine three different audio tracks e.g. three recordings of three vocal lines. For simplicity, let us assume they have the same average amplitude and loudness, and letters A, B and C represent the waveforms of these tracks.

Let us mix a stereo track in a way that the track B is in the center, track A is hard-panned to the left, and track C is hard-panned to the right. Let's do it without touching the mixer levels, and without pan-law compensation. The resulting waveforms of the left and right channels are:

L = A + B , (1)

R = B + C . (2)

Let us play this on stereo speakers. What will be the relative volumes of the three tracks A, B and C? A and C are played by a single speaker each, so they will have the same volume.

Track B is played by both speakers, with the same amplitude, so it will sound louder. By how much? The signal is 100% correlated in both speakers, which may result in a boost of up to 6 dB. However, sound coming from two displaced speakers and also possible reflections will result in a pattern of constructive and destructive interference in the listening space. For now, let us assume the track B sounds 3 dB louder than the tracks A and C, though let's keep in mind that the exact value may depend on the speaker placement, and acoustic properties of the listening space. Let us also note that this is not the boost mentioned by the OP.

What happens in mono mixdown?

Mixing down to mono involves adding two channels of the stereo track together, and dividing the result by 2 to avoid clipping:

mono = (L + R)/2 = A/2 + B + C/2 . (3)

What are the relative levels of track A, B and C now? A still has the same loudness as C. B has now twice larger amplitude than A and C, which corresponds to 6dB difference. In stereo the difference was 3dB. Therefore, the relative level difference between the track B and the tracks A and C increased by another 3dB in comparison to listening in stereo. This is the 3dB boost mentioned by the OP. Please also note that this value is approximate, just as the difference between levels in stereo we assumed is approximate too.

Why the pan law is not a solution?

Because it is irrelevant in the context of a single stereo track.

Pan law allows to automatically reduce the levels of tracks panned in center. The mixing engineer could have used it in the example above to reduce the level of the track B to prevent it from sounding louder than the tracks A and C in stereo. That would be equivalent to adding an additional factor (e.g. 1/sqrt(2) for 3dB attenuation) in front of the component B in the Eqs. (1) and (2) when producing the stereo mix.

However, pan law is not applicable to the subsequent process of summing the stereo signal to mono. It doesn't have any effect on the additional relative boost introduced in Eq. (3).

Why mid/side processing is not a solution?

...or at least it is not a solution in its simplest form

Let's start with explaining how mid/side processing works. Stereo left/right signal can be converted to mid/side channels, as follows:

M = (L + R)/2 = A/2 + C/2 + B, (4)

S = (L – R)/2 = A/2 – C/2 . (5)

one can then perform some processing on these channels, resulting in modified waveforms M' and S'. Then, we can convert the signals back to left/right:

L' = M' + S' , (6)

R' = M' – S' . (7)

Let us mix this down to mono:

mono' = (L' + R')/2 = M' . (8)

One might be tempted to boost the S channel, in order to compensate for the boost of the tracks in the center during mono mixdown. However, as shown in equation (8), side channel does not contribute to the mono mixdown. Therefore, any kind of processing of the side channel has no impact on the mono mixdown, and in particular, cannot compensate for the boost.

How can shifting phase help?

Mixing a signal with its copy with phases rotated by 90° results in sqrt(2) increase in amplitude, that is 3dB boost. Let's rotate the right channel from Eq. (2) by 90°. We get:

R'' = B'' + C'' , (9)

where R'', B'' and C'' are signals R, B and C transformed with the 90° phase shift. Let us mix it down to mono

mono'' = (L + R'')/2 = A/2 + C''/2 + (B+B'')/2 = A/2 + C''/2 + sqrt(2)·B/2 . (10)

What happened with amplitudes? Shifting phase of C'' does not (or should not) affect its sound. Thus, A and C still have the same amplitude. The amplitude of B is now sqrt(2) times larger than A and C, or 3dB more. This is the same difference as it was in stereo. Therefore the phase shift trick helped to compensate for the mono mixdown boost. A phase shift different than 90° will result in a different level of compensation.

8
  • I don't find your first sentence to always be the case. This is why Tetsujin and I have mentioned pan laws, because depending on the pan law in use, the panning can compensate for this possibility. Oct 31, 2023 at 13:42
  • @ToddWilcox 1) by first sentence, do you mean the opening paragraph? Later in my answer I explain why the actual amount of the boost may vary depending of the listening conditions, therefore yes, it might be that the boost is very small or even none. 2) In order for me to answer, please write explicitly panning of what. Oct 31, 2023 at 14:45
  • Can you explain the difference between your first sentence and your opening paragraph? Seems to me like your opening paragraph is exactly one sentence long and therefore those two phrases refer to exactly the same thing. Nov 1, 2023 at 10:55
  • @ToddWilcox correct, I was making sure we are talking about the same thing. Note, the full explanation of the issue stated by the OP takes two more consecutive sections. Could you still please comment about how exactly would you want to address the issue by panning? Preferably, using the examples and naming I introduced in my answer? By the way, here is a really nice video by Dan Worrall demonstrating various issues with mono mixdown: youtube.com/watch?v=spaqBr-cCFw Nov 1, 2023 at 15:46
  • Is it possible to phaseshift a complex signal like the right channel of a whole song over the entire frequency spectrum? Or does this method only apply to simple signals or ones with harmonics?
    – Kefas
    Nov 9, 2023 at 15:40
0

You could put a Hilbert transform on one side signal before the addition. I would suspect that you might get more useful responses in the DSP stackexchange or possibly the Sound Design Stackexchange.

1
  • 4
    This doesn't really tell us how or why this would work, or even how to do it. For those of us who are just simple old sound engineers rather than mathematicians en.wikipedia.org/wiki/Hilbert_transform didn't help my comprehension one bit. May as well be in Martian.
    – Tetsujin
    Oct 26, 2023 at 15:35
-1

Folding down to mono by summing Left and Right doesn't merely double the Mid† component, it discards the Side component entirely:

The transform from Mid-Side to Left-Right encoding is as follows:

L = (M + S) / 2

R = (M- S) / 2

When folding down to mono by summing, we get:

mono = L + R

which is:

mono = [(M + S) / 2] + [(M - S) / 2] = M

...in other words, just the Mid component.

This is why testing for mono compatibility of a mix is important.

Even if you don't fold down to mono electronically (or digitally), the result is essentially the same when listening to a binaural recording when the distance from you to the speakers is substantially greater than the distance between them.


† I am using the more common "Mid" term instead of "Center".

7
  • 3
    But the op is not folding down something that has been converted to mid side (or from a mid side recording), it’s a normal mix, where the left and the right are usually relatively different. They collapse down to mono just fine with no audio discarded, as they are not out of phase versions of each other to begin with.
    – OwenM
    Oct 27, 2023 at 19:21
  • 1
    @OwenM The conversions from mid-side to left-right and back are transform codings that lose no information. Unless the Left and Right are equal, by definition, folding down from two channels to one loses some information. The mid and side information are always there in the left and right signals whether we think of them that way or not.
    – Theodore
    Oct 27, 2023 at 19:34
  • Is it common to have the same signal in opposite phase on left and right?
    – ojs
    Oct 27, 2023 at 22:06
  • It is not, so that is my confusion with this answer. If you have an instrument hard panned left in a stereo file that is then converted to mono, you will get that instrument present in the mono file at the same amplitude as it was in the stereo file. I agree that folding it down from mid-side behaves as explained, but that is not what the op is talking about doing here, I would posit that the op's question has nothing to do with mid side processing, unless the recording they are hoping to convert has some mid-side encoded content, or they wish to use the method to manipulate it in some way.
    – OwenM
    Oct 27, 2023 at 22:38
  • 1
    Regarding “mid” vs “center”, AFAIK they are two different things. I’d suggest not using one to mean the other in either direction. Oct 28, 2023 at 2:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.