Saturday 29 May 2021

Project Proposal seeks Student... (Keyboard Noises)

You know when you have a good idea, but don't have the time or resources to actually do it? Well, here's an example... Cue: picture...

Photo by Kelly Sikkema on Unsplash

You have probably noticed that just about every electronic musical instrument makes a different noise when you press and release the keys on the keyboard. Some are quiet, and you barely notice them. Some are noisy and intrusive in some circumstances. Some make springy sounds. Some make a sort of clacking sound. Some are so jarring that you have to use headphones or turn the amplifier up. Some are an intrinsic part of the instrument (acoustic piano, clavinet, harpsichord...). Some defy description. Everyone has their own preferences for what is an ideal 'noise', or lack of it.

But what is difficult to determine is what an instrument will sound like - in advance. In general, and especially in these 'online purchase' times, it is only when it arrives that you actually get to hear what the keyboard actually sounds like, especially long-term. Even a few minutes in a music shop may not give you a very good idea of how it will sound in your acoustic environment.

So I gathered some keyboard noises from a few instruments: high velocity key-down, and high velocity key release. This is what I got (time waveforms at the top of each set, spectrograms underneath in colour):

SY99, Montage 7, and CLP-930 key noises

These weren't perfect recordings. But what is intriguing are that they key-up is sometimes louder than the key-down (SY99), and that there are wide variations, particularly the Clavinova (echoes from the cabinet, or scrapes from the action?). You will eventually be able to get edited versions of my recordings here.

VI-Control Forum

This topic came up recently on the VI-Control Forum, a place where composers, musicians and technologists with an interest in virtual instruments (Kontakt et al) discuss a variety of topics. In this case it was specific to MIDI Controller keyboards, but the subject of keyboard noise is really a generic one across all electronic musical instruments with keyboards. I proposed that a crowd-sourced approach to gathering the sounds might be a way to get some useful material (by asking VI-Control Forum members to record their keyboards), although I acknowledged that there were some problems to overcome. There wasn't much feedback, and this often happens in forums - they aren't really places for marshalling big campaigns involving lots of people, unless there's a very strong reason that will  motivate people.

The Alternative

Since then, I've been thinking about it a bit more, and I decided that an alternative way of gathering the data could be an interesting experiment. I am sure that it is a good idea, and that the results will be useful, but I don't have the time to do it myself.

So I have written up some more of my thoughts and the background research that I did around the idea, and have produced a rough Project Proposal. That's what follows. My hope is that a university or college student on a music technology course, or maybe an intern at a manufacturer, sees it, and decides to do the research and then publish it. If you know someone who might be interested, then feel free to give them this URL or the project proposal below!

- cut here - - - - - 

Project Proposal

Title: Noises in the Keybed.

Author and Licence: This project proposal is written by Martin Russ aka Synthesizerwriter, and this work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Abstract: The keys on the music keyboards (keybeds) on electronic musical instruments make noises when they are pressed and released. Data on the Sound Pressure Levels (SPL) of these noises, as well as subjective descriptions, and recordings of the sounds that are produced, are not widely available. As a result, it is difficult to make informed decisions when deciding which instrument to use for a particular application. This project aims to provide information on how relevant data can be captured consistently and reliably, and subsequently made available to anyone who would find it useful.

Procedure: There are several stages required to capture relevant data of the noises made when the keys on a music keyboard (on an electronic musical instrument or MIDI Controller keyboard):

1. Preparation of the keyboard, the recording/measuring equipment, and the performer. 

2. Key-down noise (pressing the key on the keybed, from the default 'rest' position, to the full depressed position, where it is stopped by the physical keybed)

3. Key-up noise (releasing the key on the keybed, so that it returns to its default position)

4. Processing of the captured audio files and SPL levels

5. Processing of data from the processed audio files


1. Preparation.

The capture can be carried out by individuals acting as part of a crowd-sourced team, where the results are collated by an individual or a team. Alternatively, the capture can be carried out by a single individual or a team.

The keyboard should be placed on a solid surface, ideally one that is free of rattles, creaks and other noises. The 'performer' should checking for any variation of noises across the range of the keybed, and then choose the noisiest key as the one to use for SPL measurement. If the noise is consistent across the keybed, then a key in the middle of the range of the keyboard should be chosen. Whichever key is chosen, it should be marked with a removable sticker, and this will then be the only key used for the capture. 

If a microphone is used then it should be placed 1 meter vertically above the keybed. This makes positioning easier than close mic'ing. Because the source of the sound is typically inside the keybed, then positioning the mic above the keybed (and with the microphone) aimed at the keybed, should capture the sounds produced by the keyboard, and should not interfere with the pressing and releasing of the keys on the keybed by the 'performer'. This positioning of the microphone should be possible with most mic stands.

If an SPL meter is used then it should be placed in the same way as the microphone: 1 meter vertically above the keybed. 

Low-cost SPL Meters. You 'can' get cheap (about 20 USD) SPL meters from Amazon et al.., but their supplied calibration is difficult to determine, there is anecdotal evidence that indicates that measurements may be several dB different relative to a properly calibrated SPL measuring device, and proper calibration devices cost about 6x that (about 120 USD). This approach seems to be a recipe for making the results worthless because of lack of calibration - but may be the only apporach that is feasible when using crowd-sourcing.

Higher-cost SPL Meters. Properly calibrated and calibratable SPL measuring devices will typically cost over 120 USD, and will require the use of a calibrated sound source (about 120 USD). The same SPL measurement device should ideally be used for the whole of a session, and ideally for as many captures as possible. 

Capture measurements should be of SPL and the sound simultaneously, and so a microphone and SPL meter should both be positioned 1 meter vertically above the keybed. They should not touch. One possible alternative to using separate microphone and SPL measuring device (meter) many be a calibrated USB microphone. One example is: These cost about 120 USD, but can be used to capture both the audio and a calibrated level for the audio (it may be possible to also derive the SPL).

The recording environment should be as quiet as is practical. A 30 second recording of the background noise in the recording environment should be made at the beginning and end of each capture session, so that the recording can have noise reduction applied, if this is required. The ambient temperature in the recording environment should also be noted at the beginning and end of each capture session.

The instrument, of which the keyboard is a part, should be set so that it does not make a sound. For electronic musical instruments this generally means that it should not be connected to an amplifer, or the internal loudspeaker amplifier should be set to minimum volume. For other instruments, then damping may need to be applied to any vibrating part that is not directly a component of the key mechanism. 

The performer, whose role is to press the key on the keybed at five second intervals during the capture session, should wear cotton clothing, a sleeveless tee-shirt, no jewelry, no rings, no bracelets, no ear-rings, etc. Ideally, they should wear no metal or other material that could make any extra noise. The only noise that should be recorded is the sound of the key on the keybed. The performer should breathe quietly and moderately during the capture session.

It may also be possible to find data on the SPL of keybed noises on the InterWeb, although the author has not found any such data. A search for this would be good practice before committing to a full project.

 2. Key-down noise 

The performer's task is to press the defined 'middle' key on the keybed, moving it from from the default 'rest' position, to the full depressed position, where it is stopped by the physical keybed. The start of recording should be indicated to the performer, who then waits 10 (ten) seconds, and then presses the key. Once pressed, the performer then waits for 10 (ten) seconds, and then moves to the next stage ('Key-up')

A range of velocities or strengths, of pressing the key should be used. The minimum set should be High (as fast and hard as possible), Low (as slowly and lightly as possible), and Mid (mid-way between the previous values). If MIDI is available, then MIDI velocity can be used to achieve consistency. Max would have the MIDI velocity value of 127, Mid 64 and Low should be below 30, but this depends on depending on the playing ability of the performer. 

Notes should be taken during the capture session as a subjective record of the 'sound' of the key noises. Words used can include: clicks, snaps, clunks, thumps, spring-buzzes, slides, sizzles, etc.

The three MIDI values:127, 64 and 10-30 should capture most of the variation of 'sound vs velocity', and a single key is not onerous to record: less than 30 seconds of WAV, and probably less than a quarter of an hour to accomplish. If it is discovered that there is a particular velocity or strength of pressing that produces a different opr louder noise, then this should be captured and noted.

Three key-down captures should be recorded.

3. Key-up noise 

The performer's task is to release the key on the keybed, so that it returns to its default position. Based on the techniques used to record acoustic pianos, then there are two extremes that should be captured:

a. Starting from the key being held down (the ending position of Stage 2), then the performer moves their finger backwards so that the key is allowed to rise up to its default position on its own. The performer should strive to avoid making any noise as the key is released.

b. Starting from the key being held down (the ending position of Stage 2), then the performer moves their finger upwards and off the key as quickly as possible, allowing the key to rise up to its default position on its own. The performer should strive to avoid making any noise as the key is released.

Once the key has been released, the performer should wait 10 (ten) seconds before the next Stage 1 key-down, or if this is the third capture, then the performer should wait 10 seconds and the capture will stop.

Three key-up captures should be recorded in total per chosen key on the instrument. If one key has been chosen, then three captures. If two keys have been chosen (noisiest and quietest, where there is a noticeable range of noise levels), then two sets of three captures should be recorded in total.

4. Processing of the captured audio files and SPL levels

The recordings of the key-down and key-up events stages (2 and 3) will be in pairs (down then up), and there will be three contiguous sets of the pairs. Each stage should last for ten seconds, with an extra 10 seconds at the start.

The audio files can be trimmed so that there are 10 seconds of near-silence before the first stage 1 capture, and 10 seconds of near-silence after the third stage 2 capture. The audio files should be stored in a linear, uncompressed format. WAV files, either 24 bit or 32 bit floating point, ideally at 48 kHz sampling rate. 44.1 kHz can be used if it consistent across all captures. Mixtures of 44.1 kHz and 48 kHz sampling rates should be avoided if possible.

If an SPL meter is used, then the peak measurement during the three captures should be noted.

Files should be saved with names that include the date, instrument name, a capture session ID, the sample rate, and the bit depth (24 or 32float).

5. Processing of data from the processed audio files

A spreadsheet should be used to hold the metadata noted during the capture session. Columns should include the date, instrument name, a capture session ID, the sample rate, the bit depth (24 or 32float), the measured peak SPL, the name of the performer, the name of the recording engineer or co-ordinator, any notes (quiet, noisy, clicky, thumpy, etc.) and any other comments. 

A cloud-based spreadsheet is recommended. Airtable is a sharable spreadsheet with advanced capabilities that offers a free service for small numbers of rows in the spreadsheet.

Stages 4 and 5 are just spreadsheets, and file management, and will appeal to a very particular type of person. 

The results should be shared as widely as possible. The VI-Control Forum is one possible place where a post could be placed.

Conclusions: I hope that this project proposal is useful to someone. Please feel free to share the URL or a printout of it to any researcher, student or intern that might find it useful or inspiring. This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License, so that it is available to all under clear usage terms.


If you find my writing helpful, informative or entertaining, then please consider visiting this link:

Synthesizerwriter's Store (New 'Modular thinking' designs now available!)

Buy me a coffeeBuy me a coffee (Encourage me to write more posts like this one!)

No comments:

Post a Comment