Troll Xmas: Doubler named Hydra

[wlangfor@uoguelph.ca]

Originally, this was supposed to be a doubler like the waves one. But I'd found the extent of what it could do was provide extra amplification and digital clipping.

I wanted to create something that really doubled in a smooth way. I created this 5 band, fully functional version of Hydra in over 21 versions. It's a good example of Peak limiting, and some facets of stereo width treatment.

It has a math driven broadcast leveler which removes any essence of transition while editing the audio live.

CCBY, Le Attol, Martin Vicanek, Tor, MvdLee, Nubeat7, Dozius, MyCo, Wassaka





Download: Hydra - Revision21-60.fsm - RC Final, Final Pending Design Tweak (127.01 KB)

Manual: Download Manual Acrobat CS3+ Compatible (361Kb)



Previous Versions:
Download: Hydra - Revision21-31.fsm Release candidate (107Kb)

[wlangfor@uoguelph.ca]

So, I decided on the name Hydra7.

Enjoy this schematic; This version is a whole lot better than the Waves one, at least in the way of sounding like it's from a studio. The waves one has that odd digital gumminess though.
It'd be very hard to reproduce. At any rate this one sounds as if its from a studio. So, it has one bug only (that I know of); And that is that the left delay readout is reading backwards.

You'd have to look at the code; I think there is two instances of math going on that I forgot due to being rushed. At any rate, it's an easy fix. I'll update this afternoon as well.

Download on First Post

[wlangfor@uoguelph.ca]

Version 14.3 Rolls out, and its very stable. I'm happy because I can start mixing with it; I have a few projects to do.

However, its preset order will change meaning that if You make a track with it; It'll be lost, so.. Iono, You could wait for the final I suppose. I'll start with making a complete final of this one first maybe and then the port of Martin Vicanek's Feedback Network Delay reverb. I must say, I was happy with Martin Vicanek's FDN reverb and will undoubtedly use it in the years to come.

So, with this product I had to make a cross-fader switch. If You select 0, it fades out 2 and fades in 0, and the same for one and two. It's fairly intuitive and seems to supersede resulting feedback from intermittent connection feedback. I was thinking to save memory I could use some Boolean ands looking forward to keep the next signal open for business but I'll perfect the idea until it's compact and efficient
enough to rely on.
There's a lot of intuitive technologies in this; Including Linked knobs that allow dual mono or packed signal change. Pretty slick, and they even have a preset to remember the last input entry so when You turn link off, it remembers where things were at.

I was thinking, what if You were mixing live? That'd mean You wouldn't want volume to boost or drop significantly. I mean to do the same to the smaller ZDF EQ Knobs.

BTW, I'll make a smaller ZDF EQ next; Using the framework derived from this project. It sounds nice and I like just using knobs. Besides, only having two bands is handy assuming it works well. Due to the High shelf being strictly zero delay; it does not have extra math to support Higher shelves than roughly 11 or 10 K, but the peak will.

Anyways, I'm having a lot of fun with this series of products. It's nice to showcase what's really possible with Flowstone. I remember when I used to collect free plugins, one grainy sound after another and never any character. EQ's that add mud and sound thicker than dirt lol; I'm so glad I learned to do differently haha. Enjoy and make sure to thank Martin Vicanek from whom We're all getting free stuff if You're wanting to use this projects innards for something. That goes for the other listed contributors as well.

[wlangfor@uoguelph.ca]

Mhmm, so, the other night I was tired and didn't have time to update the delay rate; but now I did; And it should support up to 192k sample rate.

Works well, but I thought I would also add a "snap" module to avoid there being too much strain on the CPU when changing decibels and frequency. Maybe decimals rounded to 0.01 should do. Like using multiplication; It makes the math faster.

However, if ultimately the filter uses division technically that means due to the amount of hidden decimals being rounded it conceivably would take a loooottt longer. So let's hope zero delay filters really are that, lol. You know I was thinking of multiplying the other day and I just can't knock it.
I always wonder, why aren't these amazing audio nerds multiplying? Too bad really.
Hehe, so anyways; I perfected the delay times and added a de-noiser for the sample-based delay rate which should suffice; it has a hiccup or two; There's still a higher pitched phantom sound when shifting too fast (again, maybe I could use snap to solve that).

A few fixes to this product that I cannot remember right now and in regards to My ZDF EQ, I've been working on that and probably will try snap to lower the amount of CPU drain swells when dragging.

version 15.3 in first post. Will remove some superfluous components and switch fancy comment modules over to plain text boxes. Hopefully I can get the code down about 150k conceivably.

[wlangfor@uoguelph.ca]

So, I wanted to turn in early and I updated this doubler named Hydra with a nice slow analyzer readout Originally made by KG_Is_Back (and Maybe others) which is modded by Martin Vicanek. I made sure to update the Info section so as to make the contribution attributed.

I must admit, now I'm really happy with this product; I'd like though to add something which mutes the weird warping sound when You change delay time. I was saying to Nubeat7 about this but he had no idea how to fix it, guess it's out of his Niche; But maybe someone else has a fix or ideas?

I know delays Built For LFO can alter for float decimals as well, rather than just samples but is it the same? At any rate, to be fair; It may be that this delay I have in there now is faster.

Image and Download on first post.

I tested this on FL Studio 12, and was surprised when I was reminded of the Adam Szabo Track I Wanna Be with Johnny Norberg. Tested it with a doubler added and it sounded even better!

[Halon]

Great stuff. Thanks

[wlangfor@uoguelph.ca]

Great stuff. Thanks

Yes, I agree - it turned out pretty, but; conclusively based upon My field testing of it; It's unstable. The big knobs and other features of it are too much draw on resources and I intend to slim it down. I decided that I would make it multiband instead. Doublers in My Opinion are only good for a little piece of sound, whereas chorus usually only good for treble.

Therefore; I'll put together an intuitive four band doubler/chorus/width application that has stereo functionality. It's cool that these knobs came out so well, but that's the nature of a prototype.
I was using very recently the MSW2 by soundspot and I was pleased with the way it allows width to be adjusted on a taper. I had done something earlier when I worked on a crossover equalizer named the Bronislav 752 and I'd added the choice of chorus or doubling for each of four bands. Which BTW uses Martin Vicanek's own HP and LP filters to affect crossover.

I was able to re-delineate these low and high pass filters so that there is no use of either division or high level mathematics. I've tested parallel crossover routed to a transfer function readout; and it seems at unity. I'm pleased and it would seem that now they are even faster than Zero Delay Filters. But they are first order of course, not second. Their grade is also not as steep but nonetheless, nice. I intend to use these low and high pass filters for the new version of Hydra.

I suppose this solves a dilemma that I have had in the past; and that is that I like to use Parallel processing for equalizers so that I can affect width. But with this new Hydra I can affect, width, doubling, and at under 0.8% CPU . BTW I've optimized the ZDF Filters whose construction was coordinated by Martin Vicanek and Adam Szabo. I was able to optimize them and they will also be added to the next release of Hydra and also to the great ZDF EQ. As far as I can tell, these optimizations almost negate CPU spikes entirely. Which is fabulous.

For those who'd wondered Why, It would seem that replacement of long division by Adam Szabo did most of the work to that regard, so that He would have a float variable defining simple division (1/variable) so that definition could be utilized as a multiplier. It's a trick which avoids some of the cost of long division. After inspecting this and though optimizing it to a further extent I must admit that I've developed a great respect for Adam Szabo's abilities. Many of His methods seem to mirror the logic and rationale which I employ.

I'll be back Thursday and probably with a new ZDF EQ, Hydra, FDN Reverb, and with something called a Mic Boom switch. It's a bass addition which You can add after the fact for Vocals. It's perfect for instances when You have filtered away bass to a large extent, and wish to hear the bassy sound without distortion.

[wlangfor@uoguelph.ca]

So, these tweaked ZDF Filters may actually be the best variant for driving the impulse. I'd wanted to understand the ZDF and attempted to make it from primitives. It loops back on itself and a few of its core pieces are re-iterated in on itself leading to the solved filter computation. And that's undoubtedly hard to mimic on the primitive level.

I'd replaced some math that divided larger sums than a static 1. And static 1's are a good idea because there is less computing, but in the ZDF's case for actually processing audio; It would seem there is a finer balance. So, My best guess is that the tweaked versions I produced might be the best for driving an impulse plot graph. And the reason I felt that was the case was by examining the processed cycles. The Cycles in Adam Szabo's variant was less than Martin Vicanek's but the cycle processing remains at a constant hum, BUT it has spikes (due to the lack of hierarchy in the called instances [probably intentional])

like so:

,,,,,,,/,,,,,,,,


/ = spike

But mine does this:

, , , , , ,

because the math of mine is so efficient; it ignores the fact that the computation is not dependent on hierarchy and just efficiently pushes out an intermittent amount of activity. But that's exactly what a ZDF is not supposed to do or else it takes a tick to catch up. So, in this case, a failure means a useful variant probably better for producing the graph.

I intend on a re-try with ZDF's so as to optimize the math to a lesser extent; but not too much smoother. Only to the extent that there is less spikes.

[wlangfor@uoguelph.ca]

Can one go to far in a revamp of a product?

Yes, lol - And I just did. It's actually something like NASA grade techno-cool. So, I wanted a doubler as You know. And I wanted to have five bands, but what irked Me was that there was a roughness to the effect. I wanted the doubling, granted but not the grit.

I wanted thickness, but not the mud. So I made an AGC Floor and Ceiling controlled Equalizer to drive this thing. It smoothes out volume issues to avoid there being dissonance.
Pretty exciting, as I have never heard of that being applied in things except maybe Autotune and Melodyne, Waves Live Tune and Graillon, Pitcher.
You're welcome to use any feature within this Alpha towards Your own work. Most of what's in here is original. The only exception is the NEDF's. They're a low and high pass that's a derivative of Martin Vicanek's work. I was able to improve the grade of the cut so that it is a bit more linear and there is less swell. Yet, the swell still exists to a small extent.

But I think they were designed that way, just to work. And to further that end I simplified the math. New version on first post and all images removed like in the new version of Spectre. Keeping up with Mac users and Mac FL Studio users obviously becomes something to think about. If they're using FL Studio they can probably run a native plugin which is pretty bad; Only, images from PC wouldn't work.

Thankfully, Trogluddite's Ruby Image draw thing was available and I used that to create My signature and a Canadian flag.

[wlangfor@uoguelph.ca]

So, I'm taking this project into a new spectrum of respectability, I wanted to create something that's much better than Waves. I like that they make simple plugins (which I use all of the time) but quite often there is readily available features within flowstone that can achieve a bit more.

So, I'll break down what is different in this new version:

• Automatic Gain detection and Unification:
  It's not enough these days just to say that You're adding gain. You have to finitely say how much gain. And You cannot always use a linear calculation to achieve this. Especially in the light of affecting a segmented piece via crossover. I was able to do this with a system that utilizes ASM peak detection which is then sent to a mono float. the result is then changed into resulting decibels.
  
  Ultimately, I made a few features which allow normalization and speed of the attack. I often hear doubled stuff which seems a bit grainy. Almost like a grain effect. And that's not what I wanted, so I utilized this method.
  
  
• Automatic Stereo Width detection:
  I created an algorithm utilizing tor's stereo width code [from audiotekknik.net] (which I ASM optimized), it's able to test the width and decide if the stereo width is already too wide. I was using loopmasters ultimate vocals samples to benchmark its effect.
  
  I also used some drum samples and an old Roland Edirol VSTI to assess functionality. Ultimately, because I was in a rush; there might be a missed knob quantity. I might update the file. Attack should be 15 to 50 and release 0.1 to 0.001. You could always set the minimum of attack to 5 though. It seems to operate at 5 well enough though it says the number is beyond spec.
  
  So ultimately, because My application tests for both stereo width and for volume it's well beyond most things that You can find. Even AGC apps and Youlean's stuff. (Youlean, You're welcome to the width detection algorithm, but please attribute the fine people who were sourced). And I'm glad of this, because I just wasn't happy with anything on the market.
  
  
• Improved Delay infrastructure:
  Martin Vicanek and his simple delay ASM code utilizing simple packing and SSE was a great saver of CPU swell and it leaves Me at 1.3% and it peaks at roughly 2.1-2.2% overall CPU load.
  
  
• New knobs and style:
  I was able to look over some vector style knobs by both pall and VanderLee; both of which came in handy. However, they use a lot of kB in the schematic which makes Me want to default My own LA school type of knob code for the sliders (I did the same for the sliders in the T-verb and FDN Reverb). I hope to get back to roughly 70kB.

Let me know what you think of the knobs and stylization. I notice the "colour" and "sound" is not up to My standard, but it will be. it will take a lot trial and error, but I'll achieve the right smoothness and colour eventually.
Hopefully this is a wakeup call to real programmers out there. I've heard it said that a real "programmer" wouldn't use flowstone, but in most respects; It's faster than plain C++. You should give it a try; because armed with logic and coding know-how and with keeping in mind of processor limitation, the abilities are greater as well.