Theo Verelst Local Diary Page 81

I've ditched the usual header for the moment, I think it doesn't help much anyhow.
This page is copyrighted by me, and may be read and transferred by any means only as a whole and including the references to me. I guess that's normal, the writer can chose that of course, maybe Ill make some creative commons stuff one day, of course I have made Free and Open Source software and even hardware designs available!

Home (alt)   Tripod Home   Local Diary list   Tripod Diary List   Previous Diary

This page is under construction, so check back later, too.

July 16, 21:16, 2009

A forum about Kurzweil

Error in voice rendering what happens with my nine kayer sound
Posted 26 June 2009 - 12:21 AM

Hi all,

This nice sounding 9 layer soundfrom here
Called AnaFMSmpljoy2c because it combines two layer osc analog, 6 operator equiv FM and 2 layers of sampled EP, and it uses effects like narrow stereo chorus for the analog part. Playing it as a program at times gives me errors like here though:

see the 16 bit wav or mp3 examples.

Having the latest OS installed that's not good, or is there something the matter with the sound (or should I try to reinstall?)!

    Posted 06 July 2009 - 03:03 AM

    I yesterday made an experimental sound from the Horowitz piano sound of the presets, changing the tuning slightly from e.g. velocity, and I trimmed the filters and the envelopes, for purposes of preventing annoying harmonics in the high range. Don't worry if you don't really hear them: not many people are currently into trying out transient harmonic distortion from combined samples, but I try out the sound as good as I can, and prefer in this case (on stage without my system that is quite different, and that *is* what the PC3 is prepared for, too) a 'hifi' type of test, with my high quality monitoring, confirmed by headphones, various other speakers, indirect headphones, and with or without external DA converter.

I was after some more exp​ression freedom as far as the samples would permit that, coming from (in this case not to advanced chosen) chords played with notes with varying velocity. On of the original (Ray) Kurzweil issues I'm sure has been (when was it, I recall reading about it in 85 or so) to not just squeeze the samples in a small amount of memory but making them and the playback engine work good with all kinds of combinations of tones, velocities, possible crossfades and timbre crosses and such, interesting material, and a reason why software samplers with a boatload of samples can still sound sterile and hopelessly unharmonic, to the point of horrible I've regularly notices. Of course the Kurzweil is in the world top of piano sound reproducing machines, and I'm glad to play with one though naturally (I made and demonstrated a physical modeling guitar algorithm half decade ago) if I'd have money I'd want also to try the new Roland PM one, where the samples can't get in the way because there aren't any, really!

To add to the niceness and liveliness an warmth of the non-trivial kind I've tried to use some of the effects available like woodenizer and small cabinet. Don't forget like some others have said to add effects to your liking, there are many great sounding ones and when dosed rigth they work very well with the sounds, which is a compliment (because that's far from easy in the digital domain to achieve). A favorite of mine is to feel in another world by putting an open stage reverb on used on a extensive quadro setup, using the extra output pair of the PC3.

Well, just to make clear a few of the issues I punched on Rosegarden, used the pC3 'board' a little and recorded straight from digital out at 48kS/s (24 bit) possibly (hopefully) without any sample conversion and I put the master section and digital volume control on 'defeat'. Consequence was of course a sound at about -12 dB, so the mp3 sounds a bit unnice: (16 bits stereo, from 32 floats) (320 kb/s)

I guess one could comment on wether such a type of piano sound, which I'm sure isn't really in the presets, is more a musical type of sound that certain people might look. I am SURE it isn't more realistic in some important measures, no mistake about that. Oh I play piano from Gospel to Funk/Jazz Fusion since 80 or so, the piece for me is no 'piece the resistance', just something to try the sound out.

            Posted 14 July 2009 - 11:35 PM
    A thirty band analog type equalizer sure is cool, but I'd prefer a sound which is naturally good. I suppose still that in a certain measure (checking FFT or other spectrograms ?) the preset sounds a re accurate, I only made comments about sample cross modulation (it sound like to me) type of distortion in the higher range, and am aware of the difference between a live performance oriented sound and a horowith recording on Chesky or so.

I edited the above sound more, and changed the effect chain, and think the result is improved to start sounding like an actual grand in a space, depending on amplification and volume. It's for download on my Wiki for those who want to try.

I must compliment on the Kurzweil sample-path because it allows what I've done with the sound, which is *very* far form trivial...

Posted 12 July 2009 - 01:16 AM

View PostJohnKeeL, on 11 July 2009 - 11:31 PM, said:

Ok here´s a quick question, i have the following interface: MOTU 828Mk3 + MOTU 8Pre
What´s the best way to connect the PC3 to it for optimal recording quality?

A quick answer: I think the question is important because norms change some ways since decades ago but it isn't always clear how signals go.

In short I'd think you probably would want to use a digital input from your unit. I don;t own a Motu, I use A Lexicon Omega unit already for years and using quite extensive quality monitoring I'd say in these times of the return of high signal quality (the 'normal' HiFi adagium as in neutral is best) the extra DA conversion in the PC3 followed by an extra AD conversion in your interface is not very beneficial. Probably noise can be excellent (I know I can easy work over hundred dB and more) so that's great, especially when you use balanced cables, and the overall signal will probably still be pretty and good, but those converters, and their assumptions (filters, sampling type) and the fact they'll be not running at exactly the same sampling frequency can cause hard to define (altough) measurable signal deviations, which probably are nicer prevented by using the good digital output from the Kurzweil, setting it to the right frequency on the master page.

I get the best results doing this, even better then taking the analog output of the PC3 and pluggin directly in my (very high q) preamp: a tos cable to the lexicon digital in. For recording you may want to check the interface you use follows the incoming digital audio signal sample for sample, so that you don't do a (possibly only small) sample rate adaptation, because that is also going to sound less optimal. That also holds for the computer software you use o record with: preferably make it slurp the incoming samples without changing them (sync source to the tos input).

In the context of plugging in and playing: take a normal mono jack cord, plug into some at least half-decent keyboard amp, and the sound will be more than fine. No misunderstanding, I'm talking about trying to get the best (preferably neutral) result, not getting a very good result, than most decent things will work fine.

Adding a preamp when you use analog outputs will possibly stop your AD converter input from blowing up ( a possible studio reason) add of course the character of the preamp to the signal (analog warmth is not likely but possibly the result), create some distortion or even clipping (probably not desirable), be an impedance matter (can the output drive the inputs with no preamp (most probably the pC3 can).

Because the PC3 of course has a DA converter in its analog output path some more things will happen when you add or do not add preamp buffers. It could be the preamp acts as a sampling residu filter. Otherwise it can be the high frequencies present in the output will upset the preamp and create annaying types of distortion: trying out I guess.

Balanced has of course the major advantage of (besides from creating a little more headroom) preventing hum and other disturbances, so it's great the Kurz has 4 balanced outputs. In these days of switched supplies and DA converted sounds, there are all kinds of high frequency issues with signal grounds, so in extreme cases, when the grounding circuit in the balancer can't keep up, you can try using unbalanced signals if you have strange 50/60/100/120 HZ related transient distortions or worse (like tones in the signal radio type things). An unbalanced inoput will mostly have at least one internal stage less than it's balanced counterpart, which may prevent distortion (just try, no need to bother much).

For the people having made remarks about the sounds: it seems to me that it is somewhat comparable as with serious quality bluray players: the built in high number of bits converters are great. Period. But for the perfectionists probably things can be said, which could simply be there are imperfection in them, possibly also numerically. I don't recall (or maybe they aren't there) the exact specs of the Kurzes' output, but it may be that on very good audio systems this is an subject. Trying the digital out is interesting in that sense.

Since there are probably 'things about sampling' and of course the on this forum mentioned 'preparation for the live stage' things going on with the sound, it could in principle be that the extra transients from you preamp will sound wonderfull to your ears, or that this so and so mixer preamp is actually what was used for sound making monitoring: so by all means use the same :)

Posted 03 July 2009 - 09:33 PM


Euh, like I communicated last week slightly about with SoundTower:
considering the PC3 (Maybe the prophet 8 for instance requires a serial number for the software to work/download? I got to buy 'my' PC after I could play around with the ST SW, actually) in this case is free as in free of charge what is the reason to keep (maybe some parts) it Closed Source, which even in Windows-World nowadays seems hardly a must.

I'd imagine the ownership of the software ideas and probably the control of the software and users dynasty are at stake, to make an honest start of a discussion about the subject, as far as people are interested, and of course not in the "Free Beer" sense primarily, let's say I'd not even mind paying money and signing a non-disclosure agreement to get a Midi-def. But it's not like ST needs a patent from the sw as it is, I think it is almost a direct programming of the same info avialable from the PC3 screen, except some saving and of course a large overview of the parameters and a midi link, which isn't always the best idea (try removing some layers from a many layer sound with the sw, the PC dsoes it much faster, read: in interaction time, in my case). Maybe the internals of the Kurzweils need protection from vandalism or simply from copying, I don't know. The samples, the ideas in the DSP signal path and so on: could be.

I mean, unix was first open source and half commercial and later on often commercial and now again in Linux incarnation Open Source. I don't claim it in my perspective currently is great to make open source software, I guess I's like to make money, but those concepts are in the free west not intertwined in such fashion. And I feel much better working on Open Source ideas, regularly, and feel soooo much better on my Fedora with Xwindow variations and all kinds of legal open software variations!

Long the the brave and the free or is it...

Feel free to kick the idea in some other forum of course,
but with a link please...

Posted 14 July 2009 - 02:59 AM

View PostStefan Lindmark, on 06 July 2009 - 11:14 PM, said:

The download link is at the bottom of the page. No need to register to download if there's a server problem with the form.

Yeah, thanks. As I indicated somewhere, I had the software already, it actually was fellow reason for the instrument purchase, so I knew that, and must say I was very glad about. I just thought maybe something is wrong and thought I´d enter the data: it´s being asked for so maybe you guys keep statistics or something. Or maybe it´s a Linux Firefox thing that may need attention or just a spur of the moment, I don´t know.

For a hobby research project along the lines of
this (see first subchapter) mathematical formula rendering , also mentioned here:
Creating arrays of wave formulas with BWise I´d like to do some MIDI patch programming with the PC3.

For an impression: my server allows for making math sounds all automatically, see this example page: Needless to say I´d like to transfer (or ´map´) math sounds (especially to begin with complex additive synthesis sounds like above) to VAST, but preferably automatically (by hand is a lot of work and my software (Like Bwise ) can vary rapidly create graphs and formulas from them and allow changes too).

And an Example song with sampled drum (Made on Linux with FOS), 1 FM piano and the mathematical sounds targeted for the PC3 (I might try to VASTify it by hand but if I don´t belief I can do automatic MIDI programming in the overseeable future my motivation is lower) panned hard left and right (melody and bass): h.mp3 The melody is the solution formula of a differential equation (see
for wav version).

I would find it great to take a wave, map it to an number of in this case VAST operators of good quality, and automatically play such sounds on the Kurzweil synth, and be able to play with it with much polyphony and like with samples and modulations.

When I in general asked for more (self) programming over MIDI a little while ago (in fact to map a general graph automatically to a number of cascaded layers), Mark from Soundtower suggested:
"If you have an idea for the HD V.A.S.T editing you can share it with me: bitmap would be the best. I check it and if it is very good we may tailor it in the editor. It will have your signature on it but it will require a fraction of your time...."

Clearly that isn´t so easy in the case I´m interested in.

Well, to whoever will either state official company policy ("Njet") or hopefully understand I want to built in like my own stereo system in a nice car I bought and would like to know the wiring: I don´t mind risking a reboot or so, any form of info (a small part of source code, or a scan of napkin with the main parameters, or is DSP code being programmed?) to be able to set actual parameters myself over midi, like on the pages on the display of the machine from a program or script I make, is welcome. I mean somebody somewhere has something of a list of those things or soundtower couldn´t have made software, unless maybe the Soundtower editor and the internal sound editor contains a DSP assembler, but hey that´d be cool too, unless of course that is secret.

I´d probably share it and possibly in Open Source form as it is, if that is a valid community oriented consideration.

Long live free Midi!


Theo Verelst

Posted Yesterday, 03:42 PM

View PostGildabass, on 15 July 2009 - 11:20 AM, said:

wow ! theo ! i didn't get half of what you do, but it is incredible ! is it music ? is it physics ? I really love the concept of creating melody with mathematical formulas…

It's clearly music from I suppose a singular Swedish original source, being made by using a MIDI song file I found on the internet ("Hasta Manana midi" in Google), which sounded cool, played by rosegarden on Linux with a soundcanvas in Fluidsynth (or Qsynth) for drums, a Hexter DX-7 simulation program connected for the piano chords, and two programs of mine being mixed in for the bass and drum sounds. In fact those programs contained a waveform from adding sine waves with exponential decay, but different time constant for each sine wave, and in accurate mathematical ways and 64 bit sample computations, and the other a 'symbolically' solved differential equation with almost a sine wave as a solution, also sample-approximated very accurately, and I might have mixed in a bit of LADSPA reverb over the mix, I would have to look that up (the 'diary' or more like 'reporting' page I mentioned probably has some more info).

Those mathematics for the bass and solo can directly be correlated with physical resonances of respectively the electronics of a number of tuned but different damped oscillations, and a simple system which resonates in a particular way which leads to the used differential equation, so your right: it's close to physically makable systems. Samples and DSP easily make waves which are not relatable to actual physical systems, because the resonating strings or air columns in the sampled instruments cannot be detuned (even a bit) without losing their sense of correct dimensions, it is as if the sounding board and harp of the piano for instance are shrunk or expanded in size. And when using seperate samples per key and velocity, the non-linearity of the resonating boards and string influencing each other (especially for interesting complicated Jazz-chords) and the non-sampleability (for the mathematically oriented: Impulse responses demand the same starting conditions for each impulse convolution to be correct, as soon as a wave 'stands' or bounces in the space, that condition is already theoretically violated) of the resonances of the air in the sounding space makes the result per definition also somewhat until quite far away from the real thing, even on a good monitoring system.

I suppose the melody could be considered a very basic but fundamental solution of the Physical Modeling kind, but pretty accurately so. I did some real PM program and computations too ( see here and demo sound file ), live demonstrated on a large european Open Source conference, but that's harder, and the mathematics of such simulations are very (to do good: extremely) hard.

I'd like to try making the type of simulation work on my PC3, but I don't know if that is possible with the dynamic V.A.S.T. architecture, I'd have to make coupled harmonic oscillators, of at least 64 or so on a row, and probably even coupling them in the cascade direction and then around is hard, but maybe some of it can work already, I need to try. It takes a lot of computations, but the Kurzweil has some pretty strong DSP going on.

So the above is like a compromise: mathematical waves can often be mapped on the VAST arch, but I cannot at the moment do that automatically because the MIDI exclusive (or what is it) protocol isn't 'available'. I don't know if that will turn out to be official company policy (would be a disappointment, I mean even most consumer HiFi systems used to come with diagrams) we'll have to wait and see. Maybe they don't want a whole miriad of open source or worse private close source software things messing with the machine (would create more hard-to solve support problems, possibly).

I'm glad you like the type of sound, I think that's why I made mention of it!
Both the bass and melody sound are probably directly from my mathematical function sound example page and are on my mental list to implement on the PC3, which should work.

Mathematical sounds ported to the PC3

Making songs with a new instrument

I tried my self made / adapted sounds in a song, with global multiband compression, made of four tracks, recorded digitally over TOS 44.1kHz/24 bits, in like an hour.

    Sheet_1-Untitled [.wav] (12.5 MB , CD-Q)    kurznewtv1 [.mp3] (2.3 MB, 320kbps)

    kurznewtv1_1 [.mp3] (1.4 MB)     kurznewtv1_2 [.mp3] (1.4 MB)  256 kb/s