|
Post by Shreddie on Jan 16, 2010 22:48:52 GMT
Doing appropriate sustain-loops for Grand Piano samples is a challenge. Depends on the sample length you can grant for your venture. If you only have to loop the portion after let's say 5 seconds, then it is quite easy. But if you need to save memory you must consider a sustain-loop at about 2-3 seconds for the mid range and about 5-6 in the bass. The problem has a name: Decay. There is no chance to get a decent sustain-loop because the fall of the volume amplitude, so you will get hard jumps in the first place. You can try to equal that with a compressor, but this is the hard way. You can try to use loop crossfade, that is the soft way. This can be used to create a kind of moving of the harmonics as exactly this happens during the original piano tone decay. But it needs a bit experience, so be patient with yourself. That's why I still like Emulator X. If I am doing something like an acoustic guitar (much like a piano as far as dynamics go) I can select the portion of the sample after the high point of the attack then choose from a number of curves (linear, 3 exponential and 3 logarithmic) and the amount of gain (positive or negative) to level the waveform with the fade in/out function. It does work much like compression (though you do it manually and by eye) but with none of the pitfalls or artifacts if your compressor is not set up properly. Once the waveform is level it makes looping much easier as there will be no sudden jumps in volume. Once looped it's just a case of using envelopes to restore the natural dynamics. I think that's a much better option than compression, which can upset dynamics in the attack portion, or crossfading which can cause phasing at the loop point. Oh, and Klaus, I've been using EndlessWAV since you mentioned it a few months back... It's great so cheers for that! Psionic... Everyone will have their different ways of looping but for a long time I did it almost entirely by hand. When trying to find suitable loop points, the most important thing is to listen to the waveform very closely. With practice, you start to hear where matching parts of waveforms are... With acoustic instruments, harmonics vary over time and quite often have repeating cycles within them (much like a drum loop but far more subtle!) so it's important that the loop points you choose fit closely with these cycles otherwise you'll hear a noticeable change in timbre when you play back the looped wave. In other words, make sure those harmonics match as closely as you can get them! As for finding these points where they match... I do also look at the waveform and the position pointer when I'm playing as the waveform on the screen can (sometimes) give you clues as to where matching points are. And if you're wondering why I'm telling you this when you have automatic software like EndlessWAV... Well, even that bit of software likes to be given a rough idea of where to look! And a Klaus said, be patient... That is probably the most important skill/quality any samplist must have. It's tedious work and in the beginning, it's very hard as you're unlikely to have an ear for it. But I promise you, with time and practice it becomes much easier and faster.
|
|
|
Post by psionic11 on Jan 17, 2010 0:39:59 GMT
Aye, you hit the nail on the head with listening for the changing harmonic timbre. That's the first thing that I zone in on, and it's also the first thing that's a dead giveaway on a poor sampled waveform... no harmonic variance. I'm starting out with the bass note, C#0, and just listening to the fifth, then the tenth, and the upper 7th and other harmonics cycle in and out is very pleasing in itself. The fifth is the most dominant one, and ebbs in and out around a few Hz, so that's what I'd initially target as a loop point target. By the way, on a somewhat related tangent -- does the K5000 use pre-defined "harmonic constructs" in its additive synth engine? I imagine it has some "blocks" that emphasize or re-create some of the more dominant spectral signatures out there, for like brass or string, where the fifths and such are fairly strongly present compared to the root tones. Does it also allow you to determine the phase relationship, and the times each harmonic fades in and out? Reading up a little bit about it on the web only gives me hints, and I'd guess that you'd need the Wizoo book to really get a good grip on it. More to the point, as additive synthesis is really a super flexible but painstakingly laborious chore of building up the sine waves that characterize natural acoustics, can some kind of similar shortcut be done on the Fusion by creating "blocks" of pre-defined harmonic signatures, and then assembling these samples or multisamples m either via the multisample engine for up to 4 blocks, or using MIX mode for more dense and therefore realistic instruments...? Or is dynamic response via velocity and the instruments various registers more a factor, where active phasing becomes more of a defining sound that just a simple static spectral snapshot, so to speak? You also mentioned awhile back that your more native synth programming environment is FM rather than VA. I'd guess that there's a good deal of relationship between FM and additive synthesis, yes? I'm fiddling with FM8/FM7 trying to backwards engineer the sounds there onto the Fusion's FM engine, but oddly so far when I input the same operator relationships and the numbers, it doesn't sound close to what FM8 has. More specifically, I took a straightforward brass sound, which only has 3 ops, A modding B at around 45%, B modding itself around 45%, and then B modding C which is output as the brass sound. On the Fusion, op self-modulation anything much higher than 25% results in a noise-like quality, whereas on the FM8 varying self-modulation results in a wider variety of pleasant tonal characteristics. Perhaps the difference is linear vs exponential FM? Anyway, sorry for the lengthy derail. We now return you to your regularly scheduled programming...
|
|
|
Post by Shreddie on Jan 17, 2010 3:36:50 GMT
By the way, on a somewhat related tangent -- does the K5000 use pre-defined "harmonic constructs" in its additive synth engine? I imagine it has some "blocks" that emphasize or re-create some of the more dominant spectral signatures out there, for like brass or string, where the fifths and such are fairly strongly present compared to the root tones. That's where the K5000 is fairly good. It has options where you can adjust meaningful sections of it's 64 harmonics in one block if you wish... Odd, Even, Fifths, Octaves, Bright (upper half), Dark (lower half) etc. Of course, you can also adjust the level of each harmonic individually. That's the one failing of Kawai's K series additive synths. It only allows you to use synchronised sine waves. Real world sounds when stripped apart to their harmonics use both sine and co-sine waves, often with slight pitch shifts to some individual harmonics. The K5000 can't do that. As an example... A triangle wave actually has an identical harmonic signature as a pulse wave at 33% (at least I think it's 33%) but as we all know, they sound very different and have different waveforms. This is because the pulse wave has co-sine as well as sine waves. Those co-sines are very important indeed if you want to emulate real world sounds... Something the K5000 isn't very good at most of the time though some convincing sounds can be had... But as you will see in the answer to you next question, the K5000 can be extremely expressive regardless of that. Yes, each harmonic has it's own fully specified (velocity/control sensitive) ADSR. These can be adjusted in the harmonic blocks mentioned earlier or individually. It might have individual LFOs too but I've not got that far yet! It's those individual envelopes (and their sensitivity) that give the K5000 it's incredible expressiveness. Thanks to those it can do things in ways that no subtractive synth can. Allowing certain harmonics to keep ringing on where a filter would have shut them out, developing some harmonics slower than others in the attack etc... And you can do that with extreme precision to create patches that are probably more expressive than any other electronic instrument. In that respect the K5000 is in a league of it's own. I think that's also why it's a synth that confuses me a bit... I've always seen synths as tools, separate from my body and lacking in expression. My bass on the other hand becomes an extension of my body due to it's expressiveness. The K5000 seems to blur the line between other synths and acoustic instruments... It may be electronic but it really can react to your playing like an acoustic instrument. Well, from what I have read, most people would say that the wizoo book is a must have but for some reason additive synthesis seems to work the way my mind does so I'm finding it incredibly easy! In theory, it would be possible to get somewhere near to what the K5000 does using samples of sections of harmonics (samples from my K5000s perhaps? Hmmm!) but it would be lacking in some respects. Using individual samples of sine waves and spreading them across multiple patches in mix mode would indeed be possible with the Fusion and it may well be capable of far more that the K5000s but it would be terribly labourious just to get one patch together... I know... I've tried it! Also doing that, you could well run into some polyphony issues with the Fusion. The K5000s had a massive amount of processing for it's time and still has quite alot today! But that processing is optimised for Additive synthesis... To give you an idea... A K5000 series patch may contain up to 6 ADD (additive) 'sources', each of those sources effectively has 64 oscillators each with it's own envelope... That's 384 oscillators and envelopes! More than the Fusions total polyphony! However, you can combine up to four 6 source patches into a multi for a staggering 1536 sine waves! And that's before the subtractive synth section and powerful 128 band formant filter come into play! However... Combining 4 6 source patches will reduce the K5000 to a monosynth! Additive really does take a massive amount of processing. If you want a real look at how good additive can be when the processing is available, I suggest you take a look at Camel Audio Alchemy VST, it has up to 600 partials, sines and co-sines and may be configured almost any way you wish. A fairly beefy PC is required to run it though. No! On the one hand the FM on the Fusion or my FS1R can be set up with no FM taking place... Effectively using the 6 or 8 (respectively) operators as individual harmonics generators when set up appropriately. But all you're really doing there is something along the lines of a souped up tonewheel organ which is a basic additive synth. True additive starts at about 32 harmonics according to some literature that I have read. As for how similar proper FM is to additive... They're totally different beasts. FM basically works by distorting waves in a precisely controlled manner to create sidebands which gives you timbral change. Additive builds a waveform from the ground up, it's far more precise and predictable. Yes, it is... They're different forms of FM... Working on the same principal but with different maths. Ah, but before we do, a little more on the K5000 series. From the beginning, Kawai had intended the K5000 series to be giant killers. They filled them with a truly massive amount of processing for the time which included a superb (even by todays standards) effects section with two bus effects (chorus and reverb) and four freely configurable (serial, parallel or any combination thereof) multi effects processors. The 'R' and 'S' models only have four part multitimbrality and 32 part polyphony whereas the 'W' version has 64 part poly (split into 32 for ADD and 32 for PCM) and 32 part multitimbrality though much of it's multi timbrality was down to it's PCM sounds, I believe it was the first synth to have 32 part multitimbrality way back in 1996. The 'W' also has a well specified 32 track sequencer but lacks the useful 'macro controls' of the 'S' model. A 76 note K5000X which combined all the features of the 'W' with the macro controls of the 'S' model was previewed at NAMM many years ago but due to poor sales of the K5000 series, it was dropped. The K5000W first retailed at £2000 in the UK but after a couple of years you could find them being cleared for £550 (an 'S' was £500)... The K5000 series nearly killed Kawai and they never made another true synth, instead concentrating on their core business of pianos and stage pianos. Even though the K5000 series was dropped from Kawai's line up in '97 or '98 they continued to develop for it (OS and Banks) until 2002 I believe. The latest OS for the K5000s being version 4.04. All previous versions and patch banks are still avalilable on the Kawai website.
|
|
|
Post by kpr on Jan 17, 2010 14:24:51 GMT
Shreddie, thanks for the cool reminder regarding the Emulator X functions. I did not much with that, but I will look into this more closely now. Perhaps it is another helpful tool I haven't realized so far.
Cheers
Klaus
|
|
|
Post by psionic11 on Jan 17, 2010 15:12:32 GMT
Sweet, thanks for that detailed review of the K5000 series. It answers a lot of my questions, and more. Pity about the synchronized waveforms, not allowing user or machine control of the phase relationships. The comparison of the 33% pulse and triangle waves was quite illuminating.
Odd thing is, I got up early with the Fusion on my mind (somehow dreamed that there was a hidden mechanical feature on the rear which allowed you to plug a VGA monitor to it!). I was sorting thru the various drumsounds, thinking of a way to create a sort of mini-drum machine accessible while in Program mode (will just have to use the Mix mode instead)... and lo and behold I came upon an overlooked bank I had: the K4. Although it's just 147 PCM waveforms, there are a few additive sine waves from the harmonic series, 1-9. Not very high up in the series, but enough to get a taste for additive.
Last thought, bringing it back to the Fusion's multisample engine. I'm fairly certain, but just looking for confirmation on the OSC start point parameter. When you choose use, say a 5% sample start time, what you're telling the Fusion to do is to just play the sample not at its beginning, but at 5% into it... useful for softening transients, since they're at the beginning... but you're not really changing the resulting phase of what we hear, correct? In other words, if you loaded the same sample twice, but have one have its Start Point later than the other one, then the two waves won't be out of phase now, will they?
If that DOES result in phase differences, esp if it's two sine waves but you happen to start the second sine wave multisample 90 degrees into the wave, then haven't you just effectively made a co-sine out of that sine? Seems to me then you could use the Fusion -- with K5000 multisamples -- to manipulate the phase relationships meaningfully by using the MOD MATRIX to control the OSC startpoints in realtime via velocity, right?
And bringing this all back to the OP, if I were to use mono samples of the Yamaha C7, I could use my ear and my hunches about phase relationships to add back that subtle but critical phasing that the 3 natural piano strings per note have in the actual acoustic world...
|
|
|
Post by psionic11 on Jan 17, 2010 15:16:16 GMT
Oh yeah, I gave Endless WAV a quick going over, but it didn't match my workstyle very quickly (especially since it didn't find my soundcard automatically, and trying to dig into it for help files or preference settings wasn't easily found).
I have downloaded Audacity though, and I got fairly productive rather quickly. Looks like this will be my tool of choice for the Yamaha C7 multisample work. Thanks for the suggestions... now back to the studio work/play for now.
|
|
|
Post by Shreddie on Jan 17, 2010 19:58:50 GMT
Sweet, thanks for that detailed review of the K5000 series. It answers a lot of my questions, and more. Anytime... If there's any more you want to know, just ask! I found it very interesting and a bit of a surprise when I first read it! I think I read it somewhere amongst the Synth Secrets series... That by the way is one of the most interesting and useful series I have ever read on synthesis. If you haven't already read it, I strongly suggest you make a start. Steve's Lost Art of Sampling series is also an invaluable read for the budding samplist! Cool! Right then... If you are using just pure sine waves then that would be the case... But If you were using an additive derived waveform, you would run into problems. If you set it up so that the root was 90 o out of phase, the second harmonic, being at three times the frequency, will have moved 360 o out of phase relative to the non-phase shifted wave... In other words, it will be back in phase with itself and bahaving as a sine, not a co-sine. I think it would be the second harmonic anyway. Other intermediate and higher harmonics will also have shifted in their relative phases. However... While it wouldn't be additive synthesis in the conventional sense, it could be a fairly funky form of synthesis in it's own right so it would be worth trying out! One important thing I forgot to mention (I was tired last night, sorry!) is that real world sounds use sines, co-sines and reverse phased sines and co-sines as well as sines at various other degrees out of phase with the root. That's what makes additive synthesis so difficult.
|
|
|
Post by Shreddie on Jan 17, 2010 20:19:25 GMT
Shreddie, thanks for the cool reminder regarding the Emulator X functions. I did not much with that, but I will look into this more closely now. Perhaps it is another helpful tool I haven't realized so far. Anytime Klaus! I do like Emulator X alot. Its sample capture and editing functions are rather comprehensive so it's what I use for much of that. On the down side, it's synth engine is a little slow to use due to its complexity. But for sampling, it has lots of useful features that you may find handy... It's sample editing section is reasonably straight forward to use too, I would guess that you could get to grips with it in a couple of hours quite easily... I certainly did! If it's any good to you, I could perhaps start another thread or send you a message regarding Emulator X to give you a good overview of all its sampling features in a few days, I won't go into any details about its synth engine though. I should point out that I only use the first version of Emulator X. The last version was Emulator X3 which will differ from mine a little. Also, E-Mu have confirmed that X3 will be the last ever version of Emulator X and will cease support for it at some point in the future. On the up side though, as it is end-of-line, it does mean that it's much cheaper than it was!
|
|
|
Post by wigworld on Jan 18, 2010 19:11:38 GMT
Anuone know if these Yamaha C7 samples will work with the free version of Kontakt Player?
|
|
|
Post by psionic11 on Jan 19, 2010 6:13:07 GMT
Just for you, to answer this question, I went ahead and downloaded and installed the free Kontakt 4 player. I didn't adjust the latencies or anything, but yes, it does play in there.
I guess I'll use that to test drive the slimmed down version destined for the Fusion. Hope I don't get too lost exploring the all the samples included with it.
Uh-oh, I'm going more and more virtual. I haven't used Reaper in months, but within this last month I've used Kontakt 4, Komplexer, Soundforum Synth, Epsilon/Tranzilon, Audacity, MIDI-OX, BC/FCB managers, BCR edit, EX5 factory, Awave, Fusion Converter, Line6 groovebox...
|
|
|
Post by wigworld on Jan 19, 2010 10:11:10 GMT
Thanks Psionic! I didn't fancy a massive download if it wouldn't work with my software. Looking forward to a Fusion version - thanks to all those working on it.
|
|
|
Post by Shreddie on Jan 19, 2010 16:22:20 GMT
Hey Psionic... I'm going off topic again! For a bit of additive fun, have a look at 'sscsss' down near the bottom of the EXTech member downloads section. I've not looked at it myself yet but I think it could be rather interesting!
|
|
|
Post by psionic11 on Jan 19, 2010 16:49:41 GMT
Hey Psionic... I'm going off topic again! For a bit of additive fun, have a look at 'sscsss' down near the bottom of the EXTech member downloads section. I've not looked at it myself yet but I think it could be rather interesting! Very interesting! I will be definitely be looking at this later after work. Thanks.
|
|
|
Post by psionic11 on Jan 22, 2010 4:54:56 GMT
I took a look at the EX5 sscsss application. It does look intriguing, and I managed to run thru the mini-tutorial to create the sample .S1m file. Being able to visually inspect waveforms as you create them would be super-handy, but unfortunately the GnuPlot program that can look at the created samplesets is just too complicated for me to install.
So I spent a little more time on the C7 conversion project. Working with just one note and all of its velocity samples, I've trimmed it down to 18% of its original size. If I keep that ratio with all the other notes, then I can create a 65MB version (@ 44.1kHz). So I'll experiment a bit more with this same note and its velocities trying to squeeze it a bit smaller, most likely by cutting down the sustain lengths by around 10%. I'd really like to try using 1-shot samples with the tail of the sample faded out, rather than looping, as that, to me, would sound more natural and also require less tedious work for a beginner at this.
Also, I'm not sure how I did it before, but now when I convert the sample rate to anything other than 44.1kHz, it changes the pitch of the sample. That makes total sense, but I wonder how I was able to do the A/B comparisons I mentioned earlier without pitch changes... perhaps there are different methods for re-sampling, and the program I was using before to try different sample rates used an algorithm different from Audacity....? Being able to re-sample to 32kHz while keeping pitch and quality intact could really trim down on the final sizes.
Oh yeah, and I take it that I should keep things at a 24 bit rate? That's the bitrate that the Fusion uses for its .wav's, right? I seem to recall that the higher the bitrate, then the higher the dynamic headroom...or is it higher = more detailed quality...
|
|
|
Post by Shreddie on Jan 22, 2010 13:03:42 GMT
I took a look at the EX5 sscsss application. It does look intriguing, and I managed to run thru the mini-tutorial to create the sample .S1m file. Being able to visually inspect waveforms as you create them would be super-handy, but unfortunately the GnuPlot program that can look at the created samplesets is just too complicated for me to install. That's a shame... I've not looked at that program yet, I don't have time at the moment. I think you'll find that the vast majority of Fusion samples are 16bit. If I were you I'd dither the samples down to 16 bit rather than resample to 32KHz. 24bit samples do indeed have higher headroom but as for quality, most people can't tell the difference, or much of a difference, between 16 and 24 bit samples. The quality of peoples monitoring equipment has more of an impact on perceived sound quality than bit depth above 16bit... Drop the bit depth to 12 or 14 bits however and the differences will start to become more obvious. Many more people however will be able to tell the difference between 44.1 and 32KHz samples... Dithering down to 16bit will reduce file sizes more than resampling to 32Khz too. As for me, I record samples at 24 bit and any presets for my computer created from those will usually remain at 24 bit but those are my main backups. Otherwise, I almost always use 16bit samples. I always dither down to 16bit for my hardware as it saves me 33% memory over 24bit files without having much of an impact on sound quality. Oh and another tip for you... When looping, use a decent pair of headphones. Good headphones will reveal details and artifacts that speakers will mask.
|
|