Jump to content

Brainstorming A Diy Concertina Midi Controller


Recommended Posts

It has occurred to me that this project will come to naught if we don't have some really good sound samples.

I don't yet know enough about what is involved in creating and managing sound fonts, but I think that we ought to be trying capture the best quality samples from the best instruments that we can lay our hands on.

 

Agreed. T'would be a shame to "spoil the ship for a ha'panny's worth of tar".

 

 

I don't yet know enough about what is involved in creating and managing sound fonts, but I think that we ought to be trying capture the best quality samples from the best instruments that we can lay our hands on.

 

I expect to post more about this later as I think that this would be an important project in general, but is this an area where your expertise could play?

 

Urk! Moi? (Hastily looks over shoulder to see who might be standing behind him...)

 

Quite possibly, but there may be better-resourced candidates. I'm not familiar with how MIDI samples are created either (part of my interest in this project is to fill in some big gaps in my understandings). But clearly they start with making very careful recordings of real instruments, and I'm reasonably well equipped for that. Why don't we start by setting out the requirements of the recording aspect of the project and see who's best placed to do it?

 

I'm assuming we need close, clean recordings of a number of notes, with minimum reverb and background noise? (You can add reverb later, but very hard to remove it). Correct?

 

Does it matter if they come from an Anglo? I do hope to also have an English in a week or so. Would that be preferable, or doesn't it matter?

 

How many notes do we need to cover the range? I believe quality piano emulations involve several samples per octave, transposing the rest. But, as I say, I'm a fine example of blissful ignorance.

 

How do we handle dynamics? Do we need separate recordings of the sample notes at a number of dynamic levels? How many?

 

What recording standard needed for the sample notes? CD standard (16 bit, 44,100Hz)? Wav file? Or can we go straight to the sampled state without passing through a normal sound file format?

 

What else?

 

I imagine, once we can answer the "what should it look like" questions above, we could ask anyone who feels they are well placed to supply the notes to put their hands up.

 

Terry

Link to comment
Share on other sites

  • Replies 92
  • Created
  • Last Reply

Top Posters In This Topic

You will need a good many of different samples covering the dynamic range, not leaving too much to extrapolation, I'd guess... As to me, the dynamic curve is of high significance for the concertina sound.

Edited by blue eyed sailor
Link to comment
Share on other sites

I'm very interested in collaborating on the concertina Soundfont project, providing the result is licensed such that it can be included with a future commercial product I'm designing. Something like Creative Commons Attribution ShareAlike would be ideal. I haven't made a sound font before but am very keen to learn. It seems to me that a really good sound font is going to be crucial to the success of any electronic concertina. I've already done a little bit of reading about how they are created and the software that is available to do it, but I lack either a high-end instrument or professional quality recording equipment to record the samples.

Link to comment
Share on other sites

What recording standard needed for the sample notes? CD standard (16 bit, 44,100Hz)? Wav file? Or can we go straight to the sampled state without passing through a normal sound file format?

AIUI you need a sound file (typically wav) per sample, though the person recording the instrument need not be the one to spend hours clipping the recording session up into individual samples. I would suggest recording and working with the samples at 24 bits if we can. Most SoundFont editors and the latest version of the SoundFont file format support 24 bit samples, which gives much greater dynamic range (whether your synthesiser can do anything useful with the extra 8 bits is another question!). I'm not so sure about the optimal sample rate, though I would say CD quality at a minimum.

Link to comment
Share on other sites

I'm assuming we need close, clean recordings of a number of notes, with minimum reverb and background noise? (You can add reverb later, but very hard to remove it). Correct?

 

Does it matter if they come from an Anglo? I do hope to also have an English in a week or so. Would that be preferable, or doesn't it matter?

Not sure why anglo vs. English should matter, especially if the first instrument it'll be tried on will likely be a Hayden duet. Besides, each individual instrument has its own sound, and there are even significant variations in attack and dynamic response. I for one would very much like to have a variety of "real concertina" sound fonts to choose from. Metal ended vs. wooden, ebony vs. amboyna vs. rosewood, brass vs. steel reeds, Jeffries vs. Linota, Edeophone vs. pinhole Aeola, etc. And of course, for the English there are treble, baritone, bass, and even piccolo. (Anglos and even duets also have octave shifted baritone and piccolo versions, though they're much rarer.) But their waveforms aren't simply octave shifted versions of the waveforms from the normal range. We'd need separate recordings.

 

I suspect that if we manage to compile one sound font, the next will be easier and subsequent ones easier still... up to a point of diminishing returns. So if we manage one, I hope we can keep going. I would even be interested in sound fonts from some less than "perfect" instruments. They have their uses, too.

 

As for adding reverb, I wouldn't want that at all. Add that to a sound font and it's no longer a "real" concertina sound. One can always add it as an "effect" when playing, though I wouldn't. In fact, one of my pet peeves is sound engineers who add reverb to everything without even first listening to the uncorruptedaltered sound. And if they add reverb to something that already has reverb in it... I'm not sure I'd want to be in the same room. (Might be interesting, though. What's the audio equivalent of a moiré pattern?)

Link to comment
Share on other sites

Maybe we should build a digital version of the Horniman museum: high quality sound samples of the best vintage concertinas.

 

Wow, I really like that idea. I could envisage a project that would serve a lot of the museum's needs and the concertina world's needs. It might bring up some interesting philosophical issues for the museum, so it would need to be entered into without an immediate deadline. On the flip side, it's quite likely that many of those concertinas won't play well, and that would pose practical and conservation issues. Perhaps drop it in the "what do we do after this project?" box. But certainly don't let that idea fade away.

 

Jim's point about having a range of voices available makes excellent sense, and counters my concern that other people might well be better placed than say I am. All we need for the moment is anyone to make some recordings, others and betters will then follow when we have the hardware to realise them.

 

Alex, recording in 24 bit should present no problem, if indeed it is of any benefit. I imagine that sound fonts are pretty compressed, like mp3s, to minimise storage space and computational overheads. I usually record in 24 bit so that I have plenty of room to fiddle around in, and still end up with 16 bits for CD quality.

 

Have you (or anybody else) concluded which sound font editor looks the most promising? (Free, or close to it, would be nice of course!) Perhaps we should start there and see what sort of input it likes?

 

Terry

Link to comment
Share on other sites

Terry:

 

I guess that I did not communicate too well, I did not mean the concertinas that are in the Horniman Museum. I suspect (don't know) that few of them are in good tuning. I suspect that they are mostly exhibition pieces rather than playable instruments.

 

No, I meant a sort of virtual audio Horniman: a collection of the best digitized sounds from the best concertinas that are in good tune, which is really what Jim is saying.

 

As far as taking audio samples is concerned:

 

It is my understanding (I stand to be corrected by someone who knows this stuff better than my cursory consultation with Mr. Google) that samples should be taken from (ideally) each note at a variety of amplitudes with no over-driving of the sound levels. Ideally the instrument should be in very good tune. Once you have a set of samples then somebody has to assemble them into a sound font file, this does not have to be the same person as the sampler. Any out of tune notes would have to be digitally brought into tune. The more samples the better, the more amplitudes the better. I think the idea is that, although it is possible to digitally transpose for any missing notes and amplitudes at playback time, if you can minimize the amount of digital modifications and artefacts that take place then the more faithful will be the playback. Whatever is wrong in the basic audio samples (distortion, missing harmonics, ...) will also be manifest and unfixable in the final files, so good audio equipment (studio quality) should be used to take the samples.

 

This process can lead to very large sound font files. In the early days, the number of samples used was minimal with no amplitude samples. In those days computer or synthesizer memory was a scarce resource. Result: terrible sounding midi sampled instruments - midi only sounded good for synthetic instruments not for sampled instruments. If you have a Windows machine then the built-in sound fonts on that sound awful, the font file is small and MS have never upgraded it. Listen to a Yamaha digital piano (I don't mean the cheap arranger keyboards) and I am pretty sure that you will notice the difference in the sound emanating from its sound fonts. I think that they are using the same basic process of creating sound fonts - along with some top flight sound engineers and musicians.

 

Nowadays memory is cheap and plentiful so it is feasible to dedicate the hundreds of megabytes a big sound font needs. One nice set of free general sound fonts is the Fluid R3 general sound font, it is 143mb resident in memory. There is a harmonica instrument in Fluid R3that I use a proxy for a concertina - there is no concertina.

 

If you want to have a listen to some different sound fonts then try Musescore, an open souce music notation program available on Win/Lin/Mac. You can change the sound font file from its default not very nice font to Fluid R3. See this page for some more info about sound fonts in Musescore.

 

http://musescore.org/en/handbook/soundfont

 

Note: While Musescore does not know abc format natively, there is a plug-in that allows you to import an abc file into Musescore and then play it back. If you have the Fluid R3 font loaded then it will sound much better in Musescore than in EasyABC or ABCExplorer.

 

Here is a concertina sound font that I found on the web:

https://www.dropbox.com/sh/kx6ude5uwzoan15/0pMQSd11c4

 

It is a sampled baritone concertina made by Phil Taylor (no relation). I don't know what it is about this font, but I do not find it particularly satisfying.

 

Michael Eskin has made some samples from an Edgely and a Carroll:

 

http://www.youtube.com/watch?v=A4quFimqQEE

 

 

 

Don.

Link to comment
Share on other sites

Alex, recording in 24 bit should present no problem, if indeed it is of any benefit. I imagine that sound fonts are pretty compressed, like mp3s, to minimise storage space and computational overheads. I usually record in 24 bit so that I have plenty of room to fiddle around in, and still end up with 16 bits for CD quality.

Exactly, if the samples need tweaking a bit before use then 24 bits gives us more headroom to do so without losing too much quality. I suppose recording at higher sample rates would have the same benefit, particularly if we have to change the pitch of any out-of-tune samples, even if we decide to downsample them to CD quality before putting them into the SoundFont. My understanding is that the samples in SoundFonts aren't compressed at all - they are just PCM data. Not all synthesisers support 24 bit samples, but 16 bit synths should just ignore the extra data.

 

Have you (or anybody else) concluded which sound font editor looks the most promising? (Free, or close to it, would be nice of course!) Perhaps we should start there and see what sort of input it likes?

My ideal would be to use an Open Source editor that supports Mac/Linux/Windows. There are a couple that look promising, Polyphone and Swami Project. I've yet to get either working on my Mac yet though - I'll have another play this evening after I get back from work.

Link to comment
Share on other sites

Great summary, Don. The default SoundFont that comes with Quicktime on the Mac isn't great either. What I've been doing to get better sound out of something that I've composed in EasyABC is to export it to a MIDI file and then either play it back in VLC (somewhere in the preferences you can tell it what SoundFont to use - under the hood it uses the Open Source FluidSynth synthesiser) or using FluidSynth on the command line to render it to a Wav file then Lame to compress it to mp3.

Link to comment
Share on other sites

I guess that I did not communicate too well, I did not mean the concertinas that are in the Horniman Museum. I suspect (don't know) that few of them are in good tuning. I suspect that they are mostly exhibition pieces rather than playable instruments.

The sad thing is that I believe very few of them are even exhibited. They're mostly "vault pieces".

 

No, I meant a sort of virtual audio Horniman: a collection of the best digitized sounds from the best concertinas that are in good tune, which is really what Jim is saying.

Part of what Jim is saying. I'm actually advocating also compiling some sound fonts from concertinas that aren't "the best". Maybe even one or more with "gasps", "chokes", slow response, and/or out of tune notes. Do you think every honky-tonk piano piece should sound like it's being played on an expensively maintained concert grand? If not, then why the equivalent for concertinas?

 

One great thing about doing it that way is that with only one instrument you could switch between playing a "junker" on a boozy drinking song and an "amboyna Edeophone" on something baroque, as well as switching between the sweet amboyna and a honkin' Jeffries sound.

 

If that's a new idea in the MIDI world, I hereby declare that this post constitutes "publication" and that the concept is now in the public domain. B)

Link to comment
Share on other sites

Going back quite a few posts there was a comment about the ease of playing and the responsiveness of Dean's S -Wave. It was like I had new responsive and accurate fingers when he brought it for me to try (we make his button covers ). I will never be a "good " player but I really did feel that my capabilities had gone up quite a few notches. I am sure that there would be no problem in replicating the many and varied tones, wheezes and bum notes from a wide variety if machines. Talk to Dean

Link to comment
Share on other sites

Part of what Jim is saying. I'm actually advocating also compiling some sound fonts from concertinas that aren't "the best". Maybe even one or more with "gasps", "chokes", slow response, and/or out of tune notes. Do you think every honky-tonk piano piece should sound like it's being played on an expensively maintained concert grand? If not, then why the equivalent for concertinas?

 

One great thing about doing it that way is that with only one instrument you could switch between playing a "junker" on a boozy drinking song and an "amboyna Edeophone" on something baroque, as well as switching between the sweet amboyna and a honkin' Jeffries sound.

Good point. There is such variation in tone and other factors (e.g. action noise) between different concertinas that it would be nice (eventually) to have a selection of different fonts to choose from. I think we should aim for the best quality samples we can manage though, regardless of the quality of the instrument being sampled.

 

I wonder if SoundFont velocity layers are sufficient to accurately represent concertina dynamics. My understanding is that their intended function is to play different samples depending on how hard you hit a key. What happens with a MIDI concertina if you press a button when the bellows are still (so the velocity is 0 and the synthesiser chooses the sample for the lowest velocity available if it does anything at all), then while holding the button down you gradually increase the bellows pressure up to the maximum level and the concertina generates a series of channel pressure messages telling the synthesiser about the increasing pressure. Can software synthesisers switch between velocity layers in response to channel pressure messages or do they simply control the amplitude?

Edited by alex_holden
Link to comment
Share on other sites

 

I wonder if SoundFont velocity layers are sufficient to accurately represent concertina dynamics. My understanding is that their intended function is to play different samples depending on how hard you hit a key. What happens with a MIDI concertina if you press a button when the bellows are still (so the velocity is 0 and the synthesiser chooses the sample for the lowest velocity available if it does anything at all), then while holding the button down you gradually increase the bellows pressure up to the maximum level and the concertina generates a series of channel pressure messages telling the synthesiser about the increasing pressure. Can software synthesisers switch between velocity layers in response to channel pressure messages or do they simply control the amplitude?

 

 

This sounds like support for after-touch, but I'm not 100% sure how often after-touch is mapped to note velocity in a way that switches samples; Horns and woodwinds do this, so I'd think it's likely that a good sound font engine would support it. I don't think that answering this question has much bearing on the construction of the sound font itself, but rather is a requirement on the sound engine software.

Link to comment
Share on other sites

Thinking about the logistics of collecting sound samples.

 

In an earlier post I glibly said that we needed studio quality sound samples. This leads to a problem: studios are not that portable and the concertinas are all over the world. Terry is in Australia and I imagine that the biggest collections of concertinas is in Ireland and the UK, which would require some very long microphone leads...

 

@Terry, and any other sound engineers/enthusiasts out there:

 

Is it possible to create a microphone set-up (two mikes, I guess) that is small enough and robust enough to be sent through the mail, that could be set up by anybody reasonably able to read and follow instructions(!!!), can record faithfully and with enough response to capture all of the harmonics, and that can be plugged into almost any computer system to capture the data?

 

Oh, and price is an issue as well...

 

Maybe the microphones that some performers directly attach to their concertinas? What would a recording studio use to record from a solo concertina?

 

I am wondering if we can come up with a package that we can send to concertina owners along with some detailed instructions so that the owners can record the sound of their concertina. The recording gear would then get sent on to the next person volunteering the sound of their 'tina. The owners would upload the sound samples for somebody else to tinker with and assemble into a sound font file.

Link to comment
Share on other sites

Going back quite a few posts there was a comment about the ease of playing and the responsiveness of Dean's S -Wave. It was like I had new responsive and accurate fingers when he brought it for me to try (we make his button covers ). I will never be a "good " player but I really did feel that my capabilities had gone up quite a few notches. I am sure that there would be no problem in replicating the many and varied tones, wheezes and bum notes from a wide variety if machines. Talk to Dean

@Old Nick

 

Thanks for bringing this up again as I am considering the button 'architecture' for a cheap DIY jobbie.

 

The easiest and cheapest thing to use would be a button array of off-the-shelf miniature electronic push switches. Is this what the S-wave uses?

 

Alternatives involve keeping (or building anew) the existing mechanical button system on a concertina and sensing the movement of the pads with hall-effect sensors or optos. The trouble with this, for me anyway, is that is one of the parts of a Chinese junker that should be junked and building anew would take a lot of time and layout design.

 

I am hoping that you are going to say that the S-wave uses simple push switches (no springs, levers and pads) and that what you are saying about improving your playing style applies.

 

Don.

 

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×
×
  • Create New...