How do I remove vibrato?

• Oct 18, 2020 - 13:02

This mscz

Flute.mscz

of a flute playing "A" whole notes tied to each other, has vibrato. How do I remove the vibrato?


Comments

Vibrato is recorded into the sound font so MuseScore does not remove that. I won't claim it's impossible but it would be quite difficult at best. There are probably some Baroque sound fonts that don't have vibrato recorded in them.

In reply to by mike320

That is unfortunate. I am using a spectrum analyzer to extract the harmonics for the different instruments - to help students understand how harmonics results in different instrument sounds.

The vibrato adds confusion. Will a future revision allow us to remove the vibrato?

In reply to by mike320

Rather than attempting to remove, the initial recording could be without vibrato - unless the recordings are from an actual instrument.

If the recordings were made, without an actual instrument, who has the root note with appropriate volume harmonics? This could save me considerable time!!!

In reply to by bhs67

If the initial samples were not from real instruments, then doing a spectrum analysis to understand characteristics of those instruments would be worthless, wouldn't it? So you need to continue to use sample of real players and real instruments. You just need to find someone who has recorded one with as little vibrato as they can muster.

In reply to by Marc Sabatella

Once the instrument sounds have been characterized (harmonics / amplitudes) then we should be able to replicate them!

This is what I have extracted from the MuseScore flute (440 Hz, with vibrato):

Flute.jpg

If we could capture sounds from a flute without vibrato, it should sound like a flute without vibrato! (I think that vibrato distorts the spectrum analysis a small amount).

And once we've captured these sounds, we should be able to add as much vibrato as we want (pitch changes and rate changes).

For additional complication, the harmonics created could change for higher and lower pitches. That is why I'm selecting 440 Hz for now.

I'd like students to understand that the reason a trumpet sounds different than a flute is harmonics. Ditto for singing. Harmonics is key for everything that we hear!

In reply to by bhs67

I think this FFT looks fine. Because the important thing is the peaks. Since the harmonic sequence is already in a certain order, knowing the decibel of the peaks (I think the ones on the graph are voltage value in percent) allows you to evaluate this graph.

Disadvantages of analyzing instrument sound with FFT and re-creating with sine-waves :

  1. That changes in time are not reflected in the instrument. Yes, we have x-axis and y-axis, but no z-axis.
    It's like taking the center frame from an animation (for example a bouncing ball) and showing it. Or as if you got the average score of the players who scored in a team).

  2. That the main frequency (root) dominates all others. Because phase-changes are ignored.

  3. Due to the limited number of harmonics used, High-fidelity instrument sounds cannot be obtained.

Flute and clarinet is the simplest sounds (and may be oboe too). Try a steel-string guitar, trumpet or violin. You will find that it is very difficult to get the same timbre. You will always get unnatural, synthesized (plastic) sounds.

I have created many instruments with the help of FFT, in a sound font (which I did not publish) that I named Plastique (because this is how I hear the instrument sounds created with this method). PS: One of the most difficult one was the Timpani instrument.

In reply to by Ziya Mete Demircan

I converted the dB into voltage percentages - easier for people to comprehend!

I understand the limitations. :-)

To me, it is important for students have a solid foundation. Understanding the basics makes everything else easier to understand!

Memorizing notes, key signatures, etc. is easy. Comprehending can be more challenging!

Using C# programming, I decided to create sine waves, with zero harmonics - can adjust the volume for every frequency. The sine waves use the same zero crossing - intentionally, as this more accurately represents actual instruments.

This will help me analyze the accuracy of the FFT results - and let me attempt to emulate different instruments.

In reply to by bhs67

MuseScore is notation software which can be a tool used in learning music. And whilst MuseScore 4 will definitely focus on improving interaction with better sound options (such as easier VST integration) it by no means will suddenly become a soundscaping/analyzing tool.

The main method of internal playback will still be driven by the synthesizer reading samples from soundfonts. Whichever those samples are.

In reply to by jeetee

Just saw your response. I understand.

I don't know much about the synthesizer, but typically, from an electronic perspective, it involves converting digital numbers into analog signals - which we perceive as sounds. The question is, how are those digital numbers created.

Many years ago, Kurzweil, created a piano that did this exact same thing! It was an interesting approach!

In reply to by bhs67

The MuseScore internal synthesizer is not the same as some hardware synthesizers. It is a synthesizer in the sense that it takes sound samples (and not synthesizes those) and applies some limited manipulation to them to "synthesize" the audio.
So yes, it goes from digital numbers (the samples and score data) to other digital numbers (the digital audio signal) which gets then processed by the "synthesizer" of the playback device (such as your soundcard) into actual analog data.

In reply to by jeetee

Not sure what that means. A high speed microprocessor / microcontroller, with a high speed Analog to Digital Convertor, can capture all of the music sounds (the sample rate must be more than twice the highest frequency).

If no changes are made by the micro, the conversion with a Digital to Analog Convertor, will produce the identical sound.

The accuracy depends on oversampling / undersampling, the quality of the microphone, and the quality of the speaker ... and the soundcard! :-)

In reply to by bhs67

There are a plethora of hardware synthesizers that don't start from recorded samples, but use actual oscillators as their starting point. That was the main point I was trying to clarify, MuseScore's synthesizer does not work that way, it starts from recorded samples.

In reply to by jeetee

I understand. What I described starts from recorded samples! :-)

The steps I mentioned are used to reproduce what was recorded and stored. Sampling at more than twice the frequency is necessary (Nyquist).

Reproducing the sound requires converting to a frequency spectrum, adding zeros between the samples (often three zeros), then converting back => which results in an original image (at least close to what it was).

Without the conversion and adding zeros, would look like a jagged signal.

The signal could have been sampled at a four times faster rate, but consumes extra storage space. Sampling, etc. reduces storage space => results in a good signal at the end!

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