My girl turned 1 year yesterday, and amidst the hustle and bustle of the holidays, it was impossible to have a proper birthday party until today. Since "over the river and through the woods to grandma's house we go" is a 2 1/2 hour drive, I thought it a fine opportunity to do some work. I thought now might be a good time to share some specifics on my work. I'm sure to more experienced Reason users what I'm doing is nothing to really celebrate, but it's a big deal to ME! Besides, if there are other users on my same level reading this, it's bound to be helpful.
Here I am working my way through the DX7 factory patches. I decided today I'd recreate the Clav 1 preset. To start, I loaded up my Default FM Decay patch--a very simple 3-osc. FM patch (only the first osc. is active) with an ADR amp envelope. My main problem is that I'm working on the old iBook and can barely see the display for all the sunlight streaming in. I proceed to work over 2 hours trying to figure out how to make one stinking feedback loop and simulate the effect of a pair of carrier waves with two modulators stacked on each in serial with 3 hard-wired FM pairs. Generally speaking, there are a couple of simple ways to approach this.
The #1, absolutely MUST do this step in making a sound is analysis. I load the DX patch in FM8 and switch to spectrum view. To make a long story short, the important thing is switching operators on and off to figure out what effect each operator has on the composite sound. From here, the sound designer has a choice: I can approximate the sound of the combined serial operators by using parallel FM oscillators arranged in pairs. While I've gotten good results from this in the past, certain kinds of instruments are difficult to assemble this way, ESPECIALLY bright, distorted, brassy kinds of sounds. I feel that the results here are still too sinusoidal for what I'm trying to do. DX7 FM sounds have an almost fat, analog quality and evoke the sonic behaviors of acoustic instruments. This happens mainly as a result of how the DX7 uses feedback loops to get waveforms other than sine. Thor allows no such variation. There simply are no means of creating simpler waveforms that make up the basic components of this clavi sound. This clavi is a very dry, pinched sound, maybe a little nasal and definitely a bit on the thin side. I'm thinking a combination of HP and LP filters can make this work (possibly, I don't know yet). Now, another very useful analysis tool is Absynth. It's a good testing ground to better test what happens in FM synthesis by using Transform in the wave window. I only have Absynth on the MBP, so all I have at the moment is FM8 and my ears.
The other choice I have is to use the modulation matrix to do some fancy routing to create the necessary feedback loops and thus some necessary timbre variation (getting closer to that thin nasal sound I mentioned). I decide to go this route.
For roughly two hours, I experiment with a plethora of various mod routings to get the same feedback effect as the DX7. Eventually I had to give up and start over. I was disappointed because I felt I wasted a perfectly good drive when I could have had a great Clav patch. It seems I'd forgotten one of my own guiding principals: This collection is INSPIRED by the DX7, not a direct replication of it. I start to wake up somewhere between Carthage and Yazoo City on the return trip. It was here I figured out what I needed was a compromise between fancy routings vs. keeping it simple and staying true to the original DX7 preset.
Here's what I came up with: Serial arrangements of FM operators are essentially like performing multiplication (it's more complicated than that, I know, but the logic makes sense when you take this to Thor). The ratio of A:B:C is roughly 8:0.5:2. In this case, the first 2 numbers can be multiplied to get this: 4:2 using two operators. Thor does whole numbers only, so perhaps a different way of looking at this arrangement is 16:1:4. Let's say I ignore the first modulator, I get the ratio 1:4. I'd previously experimented with 8 and 16 both as modulators and I found that it sounded to much like emphasizing a high partial completely uncharacteristic of the Clav 1 sound. Now, the 1:4 mod:carrier ratio isn't all that great either, but I'm limited on choices here. I already know I want to modulate FM amount using the filter EG, so I set this up in the matrix. I can't get a true feedback loop, and I certainly don't want the extreme effects of self-modulating a modulator. Instead, I set up Osc1 to modulate its own pitch and set a fairly high amount of FM. The result is not a bad almost-clavi sound. But it's still too much of a filtered analog saw wave to really be useful.
I'm going back to the FM8 patch and having another look at the second set of operators. The ratio is 6:0.5:0.5. I'm thinking we can do something similar as with Osc1, but maybe with two pairs of oscillators this time. To compensate, I'm changing the ratio to 12:1:1 and bringing Osc2 and Osc3 down an octave. Osc2 ratio is 12:1, FM amt 24. I deactivate the output of Osc2 going into the LPF--we won't actually hear this one. I leave the Osc3 ratio 1:1. I route Osc2 in the matrix to Osc3FM frequency and scale it using the filter EG. Getting better! Not quite the DX7 patch. Not even close, honestly. But still a usable clav patch.
Here's where inspiration took over: Listening to a real Hohner D6, you'll notice there's typically an audible click when the damper return to the string. This is a necessary component of the clavi sound that the DX7 is trying to imitate. However, this click isn't present at all in the DX7 patch. What we need here are two separate volume controls via EGs.
First of all, I tweaked the amp EG so that decay is 2.61 sec and release is 122 ms. The release has to be fairly quick, and JUST long enough to make that click audible. Now to create the click effect, the filter EG will have to be inverted. Vel and Kbd amounts will have to be all the way down, envelope amount all the way up, and cutoff fairly high at 8.84 kHz. A typical ADSR setting in the filter EG won't work anymore. Sustain and release times are all the way down, while A=1.02 sec and decay is all the way up. On a basic level, this will work well with your filter, but the filter EG is affecting more than just cutoff. It's affecting also the relationships between oscillator FM amounts and oscillator pitch. Where filter env is controlling Osc1FMamt, I program -47 (remember, we have to invert our envelope to account for that clicking sound on release). Where is scales Osc3FM I set -87. Perfect!
Actually, it sounds more like a cheap imitation of a harpsichord than a true D6. It doesn't even really represent the DX7 preset very accurately. Part of this is due to the fact I'm using rounded-off settings for FM frequencies while the original DX7 is slightly offset. I can compensate by using lfo to vary pitch slightly or I can detune the oscillators (or use chorus). There's definitely a loss of authenticity here between my imitation and the original DX. But it's still a very good clavi sound as well as an open door for other kinds of related timbres.
In conclusion, I'm including an image of this patch so you can get an idea of how it looks.
No comments:
Post a Comment