Classic ZynAddSubFX VST download

[ishkabbible]

Agreed - I would really like to see some "standard" scaling with a musical meaning, like the old 1V/ocvtave scaling that made analog synths somewhat interoperable with each other. A big part of the problem internally is that things seem to get converted to exponential at very random points in the signal chain - may have made sense when exp() was a very expensive operation, but on modern processors it is actually relatively cheap - IMHO the "right" way would be to run linear signals everywhere, and do the exp() only at the final point-of-use (perhaps optionally, to allow linear control as well). The controls are only accessed once per buffer anyway, so the overhead would be small.

[fundamental]

The problem there is you potentially double the load and save time. Saving is probably not too much of a problem, but loading needs to happen as quickly as possible if you want to change voices on the fly..

The numbers disagree with you here for the most part. With a cold filesystem cache, theses are the numbers I get.

The 96th-100th percentile should get some profiling to bring those numbers down, however the rest of the loads take a pretty insignificant amount of time.

[fundamental]

A big part of the problem internally is that things seem to get converted to exponential at very random points in the signal chain - may have made sense when exp() was a very expensive operation, but on modern processors it is actually relatively cheap

Eh, expf/powf can still be quite expensive relative to other operations at least according to some semi-recent profiling. That said, the parameter space exposed to the user shouldn't be distorted by this performance requirement.

[Marginal Ray]

Ishkabbible, you perfectly sum up the situation (the 3 windows, etc). It's a bit difficult for me, with my bad English (but Google translation is generally worth!).

I'm happy this fantastic synth is still "alive", even as VST. Paul Nasca desserves recognition from the “synthesizer community”, as a kind of Robert Moog of software!

The synth is important, but what sound designers do with it is as important. I must admit that the patches made by Paul Nasca, and the additional patches I found on Internet sound better that most of the commercial stuff. I'd be curious to know what Paul Nasca is doing now (professionnaly, of course).

What concerns interface, I found that: http://budislavtvp.deviantart.com/art/Z ... -455890191.

[folderol]

... and another thing

I wonder how much precision we need for saved values. For example Volume works out to approximately 0.4 dB per step, yet you would have to listen extremely carefully to pick up 1dB difference! Frequency is more critical. People can usually detect quite minor pitch changes, especially trained musicians. However, even there we have the normal scaling + very precise microtonal settings already available.

I accept that things get a lot more complicated when you start involving some of the more esoteric filters, but it would be interesting to know what actual difference to the sound there was with finer divisions of the saved values, as opposed to the dynamically changing ones (which are already done as floats).

Has anyone actually done tests on this?

[ishkabbible]

The biggest place I find I need better saved values is in the LFO frequency. By their behavior, I almost think they are exponential. I meant to go check- thanks for reminding me. I would consider that a bug If you listen to the LFO sawtooth upsweep sound that happens around 1:10 in "Ishka-Who?" you can hear why 7 bits isn't enough. I'm sweeping the LFO frequency very slowly through the MIDI-learn feature. Down around 1 - 2Hz where it starts, you can't hear any quantization. But by the end of the sound at around 1:22 - 1:23, we are almost a factor of 2 per step.
So I would argue that saving LFO frequency as a float makes a BIG difference at "higher" (50 - 100Hz) LFO frequencies.
ANd fundamental beat me to it - Loading and saving are quite efficient. I keep the zipper turned off on mine just because I sometimes like to edit the .xiz files, and the largest one of my patches, which is actually a kit of 4 patches that all fire off on each note, is 85K. Probably 5K with level != 0.
I haven't done any *nix work since I was running BSD 4.2 on a PDP-11, so I don't know if this is an issue in linux, but the windows file system starts getting noticeably ugly with just a few thousand files in one directory, so I assume that the per-file cost exceeds the per-size cost. Without running the benchmarks, my gut feeling is that doubling the number of entries in the XML costs less than having an extra file in the folder that winders has to touch for some obscure reason whenever you open ANY file in that folder.

[folderol]

@fundamental
I hadn't realised filesystem times were improving that much - must be getting old

@ishkabbible
Good point about the LFO, I'd forgotten that one.

All told, I can see that a combined file would be a good transition.

I'd much prefer this to a one-way upgrade. For whatever reason, other musicians may not be able to upgrade their software, and this would deny them access to recent patches. Added to which, for such a one way upgrade you would have to change the extension - or face a barrage of complaints that files were broken.

[fundamental]

@ishkabbible

Yeah, 7 bits can lead to some stepping effects. With the midi-learn system I've setup anything with a floating point parameter can have 14 bits accessible via a coarse/fine midi CC, though this still does require the change to floating point.
Per the linux filesystem, they do tend to be somewhat optimized for large numbers of small files, but I don't know how well ext3 compares to ntfs.

@folderol

Per the hard drive speeds, my disk is pretty darn slow, so I'd imagine better numbers could be seen with more recent hardware.

Per the mixed files, having the forwards/backwards compatibility does have noteworthy advantages for the user, but I can't justify the additional development cost (in terms of time).
For this sort of change, I'm the only one currently doing this scale of work on the original codebase and I'm slow enough as is.

[ishkabbible]

@fundamental - I am already working on it A lot of the stuff I did in "Buchlava in C++" required real analog control, so I have already switched a few controls over. I did it in a very brain-dead way - range is still 0-127, just in floating point.

You are correct that it will be quite a job. The hardest part for me was NOT yielding to the temptation to convert the controls to useful units in the process - I actually tried, and pretty quickly ran aground on the problem of where the exp() happens But I developed a pretty good idea of how to do it, and I didn't break it in the process. I really need to get git working so I can release some of these mods.

But fixing the XML load/save with parallel FP fields is trivial: forward compatibility is a requirement - new code MUST read old files, so you already need a class to do the conversion, making it convert in both directions is a small delta at that point.

@folderol - you really should upgrade that paper tape reader

[fundamental]

@ishkabbible

I guess that's one way to do it that maintains compatibility quite cheaply, but that sure is ugly. I really would love for the interface to deal with real units s.t. it's not knob twiddling until it feels about right, but changing the range from 0-127 to 0.0f-127.0f is the quick path.

I have worked on the task of converting to sensible units and it's not too bad on a small scale. A lot of thought just needs to go into the final decisions, and getting an effective bijection makes things quite messy translating back to the 0-127 domain (though not impossible).

For some things I'd certainly consider the 0.0-127.0 hack, though it is just a relatively ugly hack.
In regards to the LFO behavior, it's already somewhat broken at those frequencies anyhow with the massive amounts of aliasing that occur at higher LFO rates.

[ishkabbible]

<hangs head in shame>
But hey, it worked, so we'll call it a "proof of concept"

[Marginal Ray]

I’m a happy member of the zyns fan club, but I realize that this forum is not the best place to convince anyone how good this synthesizer is!

Nevertheless, I noticed that Ishkabbibles’ music is mainly electronic, a domain where Zyns’ superiority is not evident, and were the lack of a real arpeggiator can be problematic (but there are several good free midi arps. to download). But his electronic sounds are so organic, so analogue like, that listening to Ishkabbibles music sometimes reminds me Kraftwerk.

For me, Zyn shines as alternative to real instruments. Here’s a demo:

I rapidly composed a embryonic kind of rock “song” (no vocals), not nice, not very structured, and only suited for this test purpose. It was directly edited in Reaper with the mouse (not played on a keyboard). Please, don’t care about the musical (un)interest of this work. The topic is Zyn.

There are 4 guit tracks, mostly saturated and a bass track. Drums are from EZdrummer is both versions.

In the first version, bass and guitars are “real”: Ibanez Bass samples with a Amplitube bass preset. Guitars comes from Amplesound: over 4 giga high quality samples recorded from a PRS custom. Each track has its specific amp & cab chain, mostly from Lepou (free virtual amps that sound better than commercial products – like Zyn!) :
https://soundcloud.com/marginalray/elec ... al/s-idLg3

******************************************
The second version is 100 % zyn, except drums. No additional plugin, just original Zyn presets, with no changes applied to anyone.
The “guitar” presets change during the song, to show what Zyn can.
https://soundcloud.com/marginalray/elec ... yn/s-fFUXF

****************************************
I tried to achieve the same result with other synths, like the acclaimed Synthmaster (the player), or the demos of Diva (excellent for subtractive stuff) or Zebra, both from Uhe. Alchemy (Camel Audio) was excluded, because most guitar-like sounds are based on samples. It's simply impossible to achieve a rock feeling, because it always sounds very electronic (good for Krafwerk, not for rock’n’roll).

I could make the same kind demo with plucked sounds, brass, Rhodes, organs, reeds, etc. Again, the goal is not to mimic the real thing, but get the same kind of sensation, of emotion, vibrations when listening.

[ishkabbible]

@Marginal Ray - LOVE the highly-distorted ZYN guitar! Would love to have that patch in MY library. Is that a patch you can share? It would make a nice counterpoint to the one I developed for part 2 of "Panda" (I used the Arturia ARP-2600 for that one, recreating a patch I used to use on my real 2600 back in the 80s). My original plan was to have one of my guitarist friends play one of the lines while I played the other, but she got bogged down in another project and never had time to record her part, so in the end I just expanded on the placeholder I put in for her to grok what I was looking for.

sound better than commercial products – like Zyn!

No doubt! There is a LOT of incredible freeware out there. ZASF is an amazing synth - It spans the entire spectrum from organic to electronic quite gracefully. I think in a large part just due to its sheer size - where else are you going to find a synth with 16 oscillators, each of which is a complete synth in its own right (2 oscillators, 3 LFOs, VCF, 2 VCAs, and 4 EGs). And that is just the ADDsynth part. It is just an awesome all-around synth - I have yet to imagine a sound that I haven't been able to make in ZASF.

[Marginal Ray]

Hi ishkabbible,

I only used presets from the conventional Zyn guitar bank. Nothing special.

Reaktor from Native Instruments is surely more powerful than Zyn or any other synth, hard or soft. But I don't have time, competence and motivation to create my own patches. So I will not create my own synthesizers, what Reaktor is actually made for.

Just in case, I just discovered this interesting free virtual drummer : http://www.powerdrumkit.com/download.htm.

[folderol]

@Marginal Ray
Pretty good
I can understand people not wanting to create their own sounds but would definitely recommend looking in KVRs list of patches to complement the presets you get.

@ishkabibile
At last year's LAC I actually had someone say there were too many controls! I told him he didn't have to use them all if he didn't want to

One thing that surprises me is that hardly anyone seems to use kits to build up multi-layer voices.