[solved] Multiple instruments in unison end up "eating each other up"

• Aug 20, 2022 - 17:17

Having already listened through bunch of classic music MuseScore scores that I was able to find and download from the public domain, I must state that I constantly bump into one problem - multiple instruments playing same lines, either in perfect unison or in perfect octaves, often sound as one instrument to me. Many times I've been told that several instruments, for example, several woodwinds playing in unison, is for the sound enrichment. But I don't seem to hear that enrichment when I play back those scores in MuseScore. All I get is bigger loudness and something that comes out as one sound. Is it my hearing or my computer problem? Or, perhaps, I have some wrong settings? Look, for example, at the last measure of the following excerpt:

Brahms_-_Hungarian_Dance_No._5_exrpt.mscz

Those two notes at best sound as only something like two trumpets in octave to me.


Comments

I'm not quite sure what you expect. Obviously, computer sound generation without a huge effort of fine-tuning (as it is done e.g. for computer-generated film music) will always deliver so-la-la sound. Brahms wrote this for a real orchestra - expecting that a medium-quality(**) and, more important, heavily interpolating(*) sound software would deliver what a real orchestra does, is in my opinion (and experience) hopelessly "computer-naive". Especially brass sound is, as far as I know, one of the hardest to emulate; and the "chorus problem" = how to make multiple voices sound like they are played by "a few instruments" is a special bummer.

(*) Soundfonts contain discrete sampling points for different notes, various loudness, maybe different playing styles (simple ones like legato, staccato, pizzicato, ...), some chorus (multiplicities). A synthesizer software will, when rendering the sound, select some loudness, playing style, chorus etc. "in between" - and then have to interpolate the sounds to its chosen "playing vector". Some virtual instruments try a different approach, namely to mathematically simulate the instruments - either with a nearly physical model, or with a statistical model that "somehow" replicates the measured sound production of the original instrument. Obviously, these models will deviate from the real instruments; and so in some (or many) situations sound differently.

(**) There is a reason that VSL sells for a few thousand $/€.

H.M.

In reply to by hmmueller

@hmmueller: Thank you for your reply. It is kind of discouraging and encouraging at the same time. On the one hand, you say that I am computer-naïve - that's, of course, discouraging; but, on the other hand, you are also mentioning computer-generated film music, which, although not without a huge effort of fine-tuning, can somewhat reach the desired results. That's, of course, encouraging.

"and the "chorus problem" = how to make multiple voices sound like they are played by "a few instruments" is a special bummer."

-- I am a bit confused here. Don't multiple voices already sound like a few instruments? What my original problem in my original post was is that multiple instruments playing a same voice (in unison or in octaves) usually don't sound like a few instruments - they sound as only one instrument instead.

What happens when you play two identical instruments unison: two identical samples overlap each other. This results in a doubling of the volume.
If these were two real-life instruments, neither waveforms nor intonations would be exactly alike.
Nothing can be done to waveforms because the same sample is reused (except using a different version of the same instrument from another soundfont).
But it can be done with intonation. Selecting the measurement in which one of the instruments is unison and changing the tuning from the Inspector window between 7 and 15 will work.

Attachment Size
Unison-tuning-trick-z01.mscz 17.9 KB

In reply to by Ziya Mete Demircan

Another thing to be aware of - while in theory two identical samples overlapping perfectly would result in a uniformly louder sound, the reality is, it seldom works out that way, because the samples aren't quite overlapping perfectly. The way computers work means the sounds are more likely offset ever so slightly in time, just enough to make the sound not just louder, but also somehow "weirder" - more like a "flanger" effect is being applied.

And to be clear - none of this is due to any quirk of MuseScore, it's inherent in the physic of sound. Doesn't happen with real instruments played live because there is no chance the waveforms will be identical. But it's happens with any synthesizer based on sample playback, which is to say, almost all modern synthesizers that are trying to reproduce physical instruments as opposed to producing purely electronic effects.

In reply to by Marc Sabatella

@Marc Sabatella: "And to be clear - none of this is due to any quirk of MuseScore, it's inherent in the physic of sound. Doesn't happen with real instruments played live because there is no chance the waveforms will be identical"

-- Why does then MuseScore not use several different samples for one instrument?

In reply to by Marc Sabatella

@Marc Sabatella: "...but SDZ and VST instruments are, as is Muse Sounds"

-- I see. Thank you! Can you, please, provide any clue on how I could get to all of that (SDZ, VST instruments, Muse Sounds)? In that other response of yours you used the term "round robins". Is that a kind of umbrella term to refer to SDZ, VST instruments and Muse Sounds?

In reply to by innerthought

Round robin is the term for the technology use by these formats (typo on my part, I meant SFZ, not SDZ) to include multiple samples of a single note so the synthesizer can choose between them. Normally this is used to do things like make tremolo not sound ridiculous, but in theory, it could also be use for simultaneous notes to avoid this artifact.

VST instruments are supported by MsueScore 4 (currently in beta testing), as is Muse Sounds. So see the announcement of the beta in the Announcements forum for more info. SFZ soundfonts can be loaded in MuseScore 3, but again, I have no idea if any particular SFZ soundfont or MuseScore's SFZ player (Zerberus) happens to implement this in a way that solves the issue.

In reply to by Ziya Mete Demircan

@Ziya Mete Demircan: "But it can be done with intonation. Selecting the measurement in which one of the instruments is unison and changing the tuning from the Inspector window between 7 and 15 will work"

-- This one is a very interesting technique. And it really works! Thanks for sharing it with me.

"Nothing can be done to waveforms because the same sample is reused (except using a different version of the same instrument from another soundfont)"

-- I wonder what stops MuseScore from using several different samples for one instrument?

The other problem you have here is that everything is panned center. So of course things can cancel each other out. Or sound too loud. Try panning instruments the way they would be placed on a stage. Experiment with dynamics and volume sliders. FF for a flute can't be as loud as FF for a trumpet. F for a violin is not the same as F for a trombone. Players in an orchestra listen to and blend with each other. On the computer, you have to do that.

Do you still have an unanswered question? Please log in first to post your question.