Understanding the DX7 (Part 4)
The fourth part of Jay Chapman's guide to Yamaha FM. This month, he broaches the delicate subject of programming sounds from scratch.
Having completed his analysis of what goes into Yamaha's factory preset DX voices, Jay Chapman moves on this month to programming some basic sounds from scratch.
It would seem that, after an exhaustive survey of DX series preset voices, the time has come to start programming a couple of voices of our own. Both the following voices have been kept simple so that we can build on what we talked about in the earlier articles in this series. DX9 owners will be pleased to hear that they will be able to program the same voices on their instruments by 'interpreting' the instructions given below (unfortunately, I don't have a DX9 so I can't tell you exactly which keypads to press!).
Only two Operators are used to create the voices. They are combined in the 'modulator/carrier' configuration already explored in previous issues. In this way simplicity is assured but, as you can imagine, the voices are not exactly going to shine in terms of quality and depth. More complex programming will follow in later articles - it's best to learn to walk before you can run...
There are in fact two ways of programming your own voices on the DX series - either you modify an existing voice or you start from scratch. Both the voices discussed below were programmed from scratch, so the first thing you have to do in each case is a 'VOICE INIT'. Press the brown FUNCTION keypad then the green VOICE INIT keypad and finally the green YES keypad twice. You now have the DX7 in about as basic a state as it can be. Only Operator 1 is contributing to the sound you hear, producing a sine wave whose pitch is governed by the keyboard. In terms of envelope shape, we're presented with the simple 'gate' effect of key down is on, key up is off, as on an electric organ.
The first voice to come under discussion is actually the first piece of DX7 programming I attempted, having devoured the manual and tried some of the experiments talked about in earlier articles. This may sound a bit silly to you, but I thought it very important to be sure that I had understood enough to make my new DX7 sound like the cheap, tatty, monophonic, kit-built, one-oscillator analogue excuse for a synthesiser that it was replacing.
Before we actually get to work on the DX7, let's have a look at what we are trying to do. Figure 1 shows the state of affairs that I wanted to imitate - some of you may even be able to work out which cheap synth I had. The idea of this patch is that a harmonically rich waveform for the oscillator (a sawtooth, perhaps) is fed through a Voltage Controlled Filter (VCF) and a Voltage Controlled Amplifier (VCA). The ADSR-type Envelope Generator controlling the VCA is set to give practically any envelope you like, though I tend to set a long release time to try and make up for my having slow fingers!
The essential character of the voice is given by setting the AD Envelope Generator controlling the VCF frequency to give moderately slow attack and decay times. This opens and then closes the filter, and provided the resonance control is set suitably you can obtain a 'wah' effect as the higher frequencies come in and then fade away. I'm fairly confident you will have heard sounds very like this one on quite a few keyboard albums in the past.
The first component I needed to create was the 'harmonically rich waveform'. Having tried the 'modulator/carrier' experiments I knew that I could produce some sort of complex waveform by messing about with the modulating Operator's frequency and output level. This I proceeded to do (without a great deal of scientific method) until the sound produced seemed to fit the bill. To follow in my somewhat inglorious footsteps you should continue on from the VOICE INIT by turning off Operators 3 through 6 and then press the green OUTPUT LEVEL keypad and select Operator 2 if it is not already selected. Your display should look like:
We are now in a position to mess about with the modulation. The difficulty here is that it's preferable to alter both the output level and the frequency of the modulating Operator at the same time, since it's the interaction of these two parameters that determines the overall modulation effect. As we have only a one-dimensional data entry control rather than something like a joystick (which is 'two-dimensional' and can therefore handle two parameters at once) we have to keep leaping between the output level and the various FREQUENCY keypads (stick to FREQUENCY COARSE for the moment), changing the parameter values a little at a time and listening to the results.
As an example, I found that if Operator 2's FREQUENCY COARSE was set to 4.00 and its output level was set to 75, the result sounded quite reasonable. Getting to this point was a somewhat hit and miss affair, I must admit: some information that could be of use at this point is to be found in Dave Bristow's Getting Started booklet which I understand is now supplied with all DXs. On page 23 you'll find a number of Operator configuration diagrams (ie. partial algorithms) which relate to the production of waveforms you may recognise from your pre-FM days. In particular, the diagram reproduced in Figure 2 will give a 'sawtooth' output wave. You might like to try the sawtooth wave instead of the sound we've arrived at so far. You could select algorithm 9, for example, and set the parameters for Operators 1 and 2 as specified in Figure 2. At least then you'll know exactly where you're starting from...
Figure 3 shows the connections made for the sawtooth wave (ignore the pitch 'EG' part which relates to the 'Syndrum' voice to be described later). Note the feedback loop from the output of Operator 2 going back to its input. The feedback parameter controls how much of the output is fed back to the input and thus acts in much the same way as the output level parameter of a modulating Operator. Of course, the more output you feed back the more complex the output becomes, which means that the fed back signal is also more complex, which in turn means that the self-modulation is more complex, which again means that the output is more complex, which... fun isn't it?! We'll consider this subject in more detail in a later article: in the meantime you might well be able to see why such feedback is used to create a noise source on the DX (consider the definition of white noise).
So far all we have is the equivalent of a VCO producing a sawtooth waveform. The 'wah' effect - due to the VCF and its AD EG shown in Figure 1 - can be produced by setting the parameters of the EG in the modulating Operator, to give the required slowish attack/decay curve. It's important to realise that the envelope should not start from and descend to a level of zero as it does in the AD envelope in Figure 1. If it did we would only have the more complex harmonic content of the sawtooth during the 'wah' and we would be left with the simple unmodulated sine wave of the carrier operator for the rest of the time!
To obtain an envelope for the modulating Operator (number 2) as shown in Figure 3, I came up with the following parameters for Operator 2's envelope generator:
I'll leave you to set up your own amplitude EG ADSR curve, which will mean you deciding on and setting up the parameters for Operator 1's EG, of course.
While the voice we've just produced is not going to win any prizes (understatement of the year!) it shows a very important principle of FM synthesis in action. It's not too difficult to see that with several Operators connected in a fancy algorithm with various modulations (possibly of modulations!) going on, we can produce very complex timbres. However, of far greater importance is the timbral movement we have created in programming the 'wah' in this 'cheap synth' voice.
Because at each point of modulation in our fancy algorithm we could set about forcing such timbral movement. Also worth considering is the fact that at each such point we can apply not just the simple AD envelope, but rather the eight-parameter envelope available in each Operator on the DX series. This is one of the facilities that allows DX synthesisers to imitate acoustic instruments so well, since the change in the relative levels of fundamental and several other harmonics during the sounding of a note can be synthesised accurately.
The second voice is, I admit, a pretty weak attempt at a syndrum. Again, the simple approach - with just two Operators being used - means that the quality of the imitation is fairly poor. Things will improve in later articles, rest assured.
As with the 'cheap synth', the two Operators are used in modulator/carrier configuration. The sound I wanted here was something bordering on the metallic, and this was obtained easily using the same method (I think 'mucking about' would be a suitable technical term!) as that described for the cheap synth voice. This time feedback was not required. The values I used for Operator 2's frequency coarse and output level were 2.00 and 70 respectively. Operator 1 is already set up sensibly after the VOICE INIT, of course.
To get the characteristic pitch rise or fall associated with a syndrum the pitch envelope generator is brought into use. The pitch information fed to the Operators is a combination of the keyboard pitch and the pitch EG envelope (this combination takes place logically at the point indicated by the dotted arrow in Figure 3). As you can see, the pitch envelope starts high and settles to a medium value (actually 50): this will give a falling pitch.
When the pitch EG is not in use, all of its level parameters are set to 50, which corresponds to the keyboard playing at the pitch it thinks it is tuned to - if you take my meaning. The pitch EG has the same parameters as the EGs in the Operators.
Since we want to start high we must set level 4 high. Note that the envelope starts and finishes at level 4 - the pitch EG diagram in Figure 3 shows the return to level 4, represented by the dotted line across the top of the envelope diagram at its extreme right. We don't actually hear this pitch change because it occurs only after key release, when the carrier's (amplitude) EG curve is at zero anyway.
The fall in pitch must take some time so we need to set up a suitable rate parameter. Since the pitch change occurs from level 4 to level 1, it's rate 1 that needs altering. You can see this easily if you look at the EG diagram on the front panel of your synth. The pitch EG parameter values that I settled on are as follows:
The last component we have to set up is the amplitude envelope of the syndrum. This involves choosing suitable parameters for the carrier Operator's EG. The sound should have a fairly percussive start and then fade away over about half a second. Envelope generation on the DX instruments is essentially no different from that on any other synthesiser, so I don't see this side of things causing too many problems.
The percussive start to the envelope is obtained by setting both rate 1 and rate 2 to high (ie. fast) values. Level 1 is left full on at 99, which should always be the case for the main (or only!) carrier Operator, in order to avoid unnecessary loss of level. Rate 3 is then set fairly slow, to give the half-second fade out. Rate 4 is left alone since effects after key release are ignored for simplicity's sake. Note that I rarely look at the parameter values changing in the display when I'm setting up envelopes - instead, I tend to experiment by moving the data slider about until it sounds right. Not very scientific, but it works.
I found the following values effective for Operator 1's EG:
|Operator 1 EG||1||2||3||4|
So there you have it - 16 syndrums at the same time over 5 octaves. Will wonders never cease?
Gear in this article:
Feature by Jay Chapman
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!