Ken Elhardt - Strings bauen

[Moogulator]

ich kommentiere das hier mal nicht, es stammt aber von Ken Elhardt und das möchte erstmal erwähnen damit hier keiner denkt, ich hätte mir das ausgedacht..

BITTE DIES AUCH NIRGEND WO ANDERS POSTEN, ES DARF ABER DISKUTIERT WERDEN..

DAS SOLLTEN WIR KEN GEGENÜBER SO HALTEN.. danke schonmal..


Again from the SDIY list for any who are interested.
Sorry, but I'm still weeks behind and still haven't had time to catch up on
posts. Hope to do so by/over the weekend. Problem is I need this internet
computer downstairs, so I'm disconnected weeks at a time. Didn't want
mailbox to overflow so downloading mail now and figured I'd throw up some
recent string filter mods I'm working on. These obsolete the last ones I
did and these soon too will be obsolete too. The first demos used a generic
filter for all string sounds and there was no distinction between
instruments such as violin and cello. These ones I did a first pass on and
moved bands around trying to create peaks and dips where they would be on
the real instruments. Next pass I'll get compare to specific samples that
have certain characteristics I want and try to modify the string filters
even closer to the real thing. Following are some short demos ranging from
about 6 to 38 seconds. Should be listened to through good headphones (I
have not heard them through anything else).

This is a real viola sliding up in pitch on a string followed by the Nord
Modular doing the same thing. Even at this level, they are very close.

Filesize: 95 KB

String filter adjusted to violin.

Filesize: 568 KB

I'm playing the Nord violin with my right hand over a sampled string orch
preset from a Roland synth in the background. In the low range the nord
actually sounds better than just about all the violin samples I own,
noticable in particular near the end of the piece.

Filesize: 515 KB

Irish reel with nord violin doubled with jig flute sample just for the hell
of it.

Filesize: 140 KB

Best part about this piece are the cello parts which also end in a plucked
chord at the end (yes, they can do pizzicato too). Orginally the piece
wasn't too bright so I ran it through custom sharpening code I wrote to
brighten it up, but that really brought out the noise too.

Filesize: 573 KB

-Elhardt

 

[Rastkovic]

Wer weiss, seine MOTM Sachen fand ich auch schon sehr beeind

Wer weiss, seine MOTM Sachen fand ich auch schon sehr beeindruckend. Das hier allerdings klingt meinen Ohren schon sehr real, meinst Du es ist ein Fake?

Daniel

 

[Moogulator]

ich meine nix, ich poste es einfach..
bin noch im messe-

ich meine nix, ich poste es einfach..
bin noch im messe-nachwirkungs mode..

zzt bin ich nur zu 10% sichtbar

 

[Summa]

Bei der 2ten Sequenz des ersten Beispiels hoere ich Stufen i

Bei der 2ten Sequenz des ersten Beispiels hoere ich Stufen im Slide, das koennte natuerlich wirklich auf einen Synth hinweisen...
Schon Workstations hoeren sich imposant an, wenn man sie geschickt spielt, die Spielweise macht hier wirklich viel aus...
Also die Beatles Sequenz klingt nicht sehr beeindruckend, hat einen Touch von Harmonika und Akkordeon im Sound.
Gedoppelte Sachen lassen sich immer schwerer ausmachen...

 

[Moogulator]

hab von ken das go, ich werde noch was posten..
ist das all

hab von ken das go, ich werde noch was posten..
ist das alles an postings zum thema? *wi-wa-wunder*

dieser thread ist also legal

 

[Moogulator]

noch einen nachtrag, den ken mich bat auch zu posten:


P

noch einen nachtrag, den ken mich bat auch zu posten:


Per Mattsson writes:
Thanks Ken!, both for the synthesis research + putting this up online.<<

xavier tribolet writes:
Your explanations and examples are really interesting and incredible!<<

Howard Moscovitz writes:
Do you mind if I repost your message on the electro-music.com forum.<<

Thanks for the compliments. And by all means, if there are other message boards that you think want to know about this stuff, just cut and paste the messages there to. Anything that's posted on a message board is already out in the public so there is nothing to hide. Other people post my posts elsewhere too, and I find some of my AH posts ending up on a Hungarian site where I can't even read what they're saying.

Chet Singer writes:
I especially liked the Baroque ensemble. The fast bow scratches were just right: audible, but not synthetic.<<

Yeah, that takes some programming to get that to sound natural. I of course can crank it up to sound vulgar, or take it away all together.

Given your talents, you might want to consider the purchase of a G2. I'd
guess you could fit your entire filter array into a single G2 patch. A
monophonic audio-processing G2 patch can span up to two DSPs, one in the
VA area and the second in the FX area. You could create different
"instrument bodies" by detuning the entire array up or down slightly with
a single control. And by assigning the filter amplitudes to the G2 front
panel knobs, you might be able to eliminate the post-processing EQ.<<

I had originally planned on buying a G2. But they still don't have all the modules implimented. It's not as much as an improvement as I thought. It's another big chunk of cash to spend when I'm not working right now. Plus the Creamware Modular costs less and does more, so I'd almost rather go that way. But I also have Reaktor, and I was alway planning on moving it all over there, as there are a number of things the Nord just can't do realistically G2 or not, such as convincing violin portamento. Also, my main driver patch is at 98% usage, and all my three string filter slots are at 100% usage, so unless the G2 is a minimum of twice the speed, It probably won't fit in a single G2 patch. The post processing EQ won't be needed after I make many filter adjustments.

Also shifting the filters up and down is a no-no. Just shifting one semitone up or down and the overall sound can be dramatically different and quite unattractive. Tiny changes in resonance and the effect can disappear fast. I'll be doing custom filters for each type of instrument and they will be quite unrelated to each other. To change from a violin to cello with require loading in a new filter set. This was later done in my "adjusting to real instrument" demos, though I still need to make one of two more passes beyond those.

Ian Sayer writes:
This is exceptional work. I know you must have invested considerable time in this, but I for one, would love to see the string filter patch.<<

It takes days to try different filter configurations to even get something that sounds somewhat like what I want. Then it's another long miserable period of time analysing real string samples, taking FFT's (sometimes animated ones of instruments sliding up a string), trying to determine all formant peaks and dips in the harmonic spectrum, and trying to set up filters that mimic them, and sweeping sine waves through my filter bank to see what it looks like and why some many unpredictable things seem to happen when moving the filters around which I have to fight with. And all the pain in the ass to get the high freqs to sound unsynthetic (another big pain). It's hell, and as with most of my sounds and want them for my use, as opposed to flying around the net and showing up in other people's recordings and them getting the credit.
--------------------------
Posts related to second set of demos:

Wan Kemper writes:
These are stunning sounds again. Each time i think it can't be any better, and you just do that. Never ever heard such realistic violin sounds from an electronic device.<<
Rodge writes:
Thank you for posting these, they are absolutley astounding. It makes me want a G2 just for this one sound. Incredible!!<<

Thanks again for the compliments. Remember that these were done on the original Nord Modular, not the G2.

Ville Gustafsson writes:
He also posted some string recordings on the nord modular list a couple of years ago which he claimed was done on the NM1but in reality FAKE, ie processed acoustic recordings. Unless Mr. Elhardt cares to post his string patches for us all to try out, I think we can safely assume these are FAKE FAKE FAKE.<<

This is all news to me. I'd like to know what those recordings were, and why it is necessary to process acoustic recordings that already have all the proper body resonances with additions cavity formants, which would sound awful.

Chet Singer writes:
I've had a listen, and these are just fantastic. I'm especially impressed with the bow noise and bow-scraping attack, such as the viola slide and some of the other hard-attacked notes. Can you reveal how that's done?<<

My slot 1 string generator patch uses two oscillators. You're always hearing a hardsync'd oscillator, but since most of the time it's in perfect sync with the master oscillator (that you don't hear) it just generates a sawtooth wave. But on the attacks, I hit that sync'd oscillator with a burst of filtered noise which causes it to bounce around, yet at the same to try to sync to the master oscillator. This is kind of what a string does as the bow starts to pull it, but it randomly slips prematurely before its cycle is fully complete generating smaller shorter waveforms, but still overall kind of syncing to the overall pitch of the master osc. The freq of the attack noise, the length, and the depth can be adjusted to taste. I believe the depth of it is under velocity sensitive control.

As for additional bow noise, to get a coarser sound overall, I'm using a wavewrapper, and modulating the CV input of that with some noise. This also adds brightness to the sound like an exciter and gives a more complex and moving higher freq content to hide the potential synthetic sounding high freqs. That and the sweeping delay lines animate the higher harmonics to act like a real instrument up there. If both of these were taken out, the upper end would sound too sawtooth/synthetic, or if my filters are too resonant up there, it sounds too digital, like a digital wavetable oscillator. But the wavewrapper's sound is very dependent on what's fed into it, and that takes fooling around with so as not to sound brittle and digital. In my patch, it comes after two LPF's (combined for 48dB cutoff). All this is very sensitive, as changing one thing can affect others. For instance, if my sweeping delay lines are too fast, it diffuses my vibrato, and I can't get a strong vibrat!
o. The driver patch is kind of balancing on a razor's edge. (Note that my string patch also uses kybd velocity and an EV-5 pedal to control dynamics).

Davepnord writes:
But the fact that you went to such extremes to convince people it (the Sage) actually existed makes any of your sub<a href="https://www.sequencer.de/specials/sequencer.html">Sequencer</a>uent claims suspect. What did you expect?<<

That's because that was a hoax, and to complete a hoax, one has to fess up to doing it. There's no point in faking Nord patches. The nord patches are already fake, because they are electronic, not acoustic. I don't think one can fake a fake sound. And by me showing what the raw unprocessed sound vs the process sound, sounds like, it's plenty clear it's all a synth. Besides, I've posted patches to other realistic sounds I've done such as my Piano Crash and Recorder sounds.

Wout Blommers writes:
Today I get the feeling things like the Sage story work against you, which isn't a nice thing. It takes away the real value of the achievements you made.<<

Actually, I don't think it works against me. I know people are just bringing it up to try to force me to post patches. I can see right through it. But perhaps next time I'm at my Nord computer, I'll do a screen dump of all four slots and post a very small picture for you to see. Also, if I do give in
and do a Wendy Carlos style "Secrets of Synthesis" CD, I will give up the patches on just about everything I've done. Just have to decide if there is enough market for such a thing.

BTW creating a violin sound isn't that difficult, although not so good as yours <<

Anybody can patch up a sawtooth wave with vibrato and call it a violin sound, but it's not. Creating a realistic violin sound is very difficult and requires filtering almost beyond what can be done in analog, not to mention all the subtleties in it's dynamics and the driver portion (mine is about 100 modules total). I'm trying to create string sounds real enough that they can be played right within a real acoustic ensemble or replace a real instrument in an ensemble, and virtually nobody could tell it's fake. For instance, take the real violins out of Kansas' song "Dust In The Wind" and replace them with mine, and mine would sound like real violin replacements.

-Elhardt

 

[Moogulator]

Since the subject of multiband bandpass filterbanks for use

Since the subject of multiband bandpass filterbanks for use in string
filtering came up on the Synth DIY list and I went through all the trouble
to respond, might as well post it here where I know there are some who are
interested in some of my latest techniques and how they sound. So below is
the post I made to that list.

----------------------------------


Jürgen Haible writes:

(Seriously: Ken Elhardt's string sounds are more convincing than my demos, and he used a handful of readily available FX boxes. Maybe he likes to talk about that)<<

Yikes, I'm weeks behind on catching up on e-mail and posts so I'm just seeing this now. I realize I probably appear as if ignoring people but I'm just busy and really tired of the internet in general. I'll be catching up in a few days I hope, but I can tell you everything you need to know about string filtering and have a few new mp3 demos of my Nord 40 bank string filter in action.

Too time consuming to cut and paste questions, I'll just answer them. Using equalizers for the job? They don't work. I also have that Behringer 24 band feedback surpressor. The sound will still sound synthetic, just filtered differently. No EQ has enough bands anyway. You need about 4 to 6 peaks and 4 to 6 nothches per octave. A 31 band EQ only has 3 bands per octave, which only allows 1 peak and 2 notches or vice versa. Nor are they as steep and drastic as needed. JH could have mentioned some of the stuff I use as it's no secret, such as delay lines and phase shifters as comb filters. But I'm trying to get away from them because I have no control over how the peaks and notches are placed within something like a 40 stage phaser. And the metalic sound that delay lines create makes me have to use many in parallel to randomize the frequency response, and again I have little control over where those peaks are.

I've actually done some high count multiband filtering on my Nord before. One demo I put up is the following showing the effect on an the unprocessed sound. Actually it's part multiband filterbank and phasershifter. I don't think something like the Moog string filter gives such a dramatic change to the sound.

Filesize 200,622 bytes:

As for my latest experiments below, I'm using my Nord Modular as sound source and to also process its own sound (I'm plugging the output of the Nord back into itself to process itself since I need the power of three DSPs just for the filter bank). I've tried many different configurations, such as 6 bands per octave (every other semitone), 4 per octave, alternating spacing, random spacing, and so on. The best overall sound was alternating between 2 and 3 semitones for about the first 24 filters and then compressing down to 2 semitone spacing higher up. This produced the most wood-like tone, while not sounding too hollow, too synthetic, or all kinds of other problems I've come across. I'm then running that through a Behringer 31 band True Curve EQ to shape the overall sound for a more refined sound.

Also the source sound you create to run through the filter bank is important. Just using a sawtooth oscillator with vibrato is too organ-like. I'm using several delay lines slowly sweeping around kind of as subtle choruses (also used this on my older string sounds), and they act as harmonics animators. If you've ever looped a single cycle of an acoustic instrument, it just sounds like an electronic tone because there is no motion. This also makes the vibrato sound sweeter and more string-like. You also want if possible, randomization of note pitch as fretless violin notes are always somewhat out of tune. You need human-like vibrato, changing in speed and depth. You need control of the string loudness and tone after the key is pressed so you can phrase it the way a string player would. I'm also using a Nord waveshaper module that adds some high freq harmonics for a brighter more scratchy sound which really helps bring out the buzzy top end of the cello, and I can modu!
late it if I want bownoise. If you want to play fast notes, you'll want some kind of bow slipping/scraping sound on the attack. And when I move my string source sound to Reaktor, I'll add the little portamento slides that really help (can't get to sound real on the Nord for some reason). In other words you have to really losen up and get an animated sound. Organ type sound in, organ type sound out.

Filter plot of white noise running through my string filter:

http://home.att.net/~elhardt4/String_Filter_Plot.jpg

(Note that these demos should be heard through good quality highly detailed headphones to hear all the subtleties. I'm often shocked by how the beautiful tone and highend detail can disappear when played through other means.)


Here are some new demos. This one alternates between my warm and fuzzy newage Nord violin and the real one in this Ackerman recording. I had to splice in solo guitar parts from the beginning of the piece, thus some obvious sudden edits. The timbre of the two is quite close just using my default filterbank band placement, so much so that with just a little effort in playing/phrasing and timbre adjustment I could probably replace the real violin in the entire piece and few would notice the change. Following is the timing of what is real vs Nord in the recording. Also threw in a Nord cello toward the end, really showing off the wood-like quality of the filterbank as the virtual bow scrapes across the last cello note.

0:00 to 0:22 Nord
22:00 to 0:38 Real
39:00 to 0:58 Nord
1:00 to 1:16 Real & Nord playing together
1:12 Nord cello enters, all Nord through to the end

Filesize 1,801,370 bytes:

Here is a snippet from the Brahms violin/cello concerto. First take is with the real cello, second take is the Nord. Because my bands aren't in same places as the real thing there is a timbral difference (if I can correct, will be a near perfect match), but in the low end you will see that the cello-like characteristics of the Nord sound almost every bit as much like a cello as the real one, just with a different body timbre. Needs more work in the higher range to get that sad/whiney/stringy cello type sound.

Filesize 541,641 bytes:

One pitfall that I've encountered is when using the same string filter for all the instruments in an ensemble. I did one recording I didn't even place online because the entire orchestra sounded like it was playing through a tin can even though the individual instruments sounded pretty good by themselves. When I thought about it I realized that they were all emphasizing each other. Where as in a real orchestra, different instruments have different frequency responses and they smooth and average each other out. This is just one more reason to do it digitally where you can change filter settings unlike a fixed analog filterbank. None-the-less I tried to do a number of things differently to minimize that bad effect in my following Bach performance of a small up-close chamber group (with sampled harpsichord). Still it's a quite resonance sounding and badly distorts and sounds like total crap in many other ways through computer sound systems. Needs to be played through som!
ething that can handle severe resonances popping out all over the place, such as quality headphones. Even there, there were some problems, such as the popular Sony MDR-7506 which sounded ear-piercingly nasal in the 2K to 4K range. If using those, or others with the same problem, listen to the second file labeled (Sony). If using Sennheiser HD-280, AKG K240, Fostex T50RP (needs treble boost if possible), or most others, try the first one labeled (Senn).

Filesize 809,553 bytes:

---------

Just a couple of demos using my older methods for those who don't know what I was talking about. This string orch uses delay lines and a phaser.

Filesize 439,382 bytes:

All the sounds in this one are from analog or virtual analog synths. Violin and cello use delay lines, the Andromeda guitar sounds use a phaser for nice body resonance. BTW, filter banks are good for other things too. Just in my fooling around I've gotten complex timbres that sound like pianos, harps, sitars, bagpipes, and big organ pipes, that don't sound like anything you'd normally get from an analog synth.

Filesize 881,478 bytes:

-Elhardt