You've programmed the software, now reap the benefits. Andy Honeybone gets into those silky 6 operator DX7 patches and converts them to good old 4-operator types.

At last, the time has come to transfer those six-operator voices to our four operator synths. To recap, DX7 type sounds can be faked on a DX21, FB-01 or TX81Z in dual mode provided that the voicing doesn't rely on fixed frequency operators, fancy multi-stage envelopes, algorithms with three modulation inputs to a carrier or multi-operator feedback. That's the theory, here's the practice.

For my first trick, the "BRASS 1" voicing from the International version DX7 mark I. First, check the algorithm — in this case #22. This splits neatly into two systems: one operator modulating a single carrier and another modulating three further carriers. This arrangement suits a combination of algorithms 1 and 6 on the DX21.

Starting with the easy two operator left-hand system, select algorithm 1 and turn off operators 3 and 4. Next, set the frequency ratios for both OP1 and OP2 to 0.5. Detuning is required and as the DX7 has twice the detune range of the DX21, the DX7 value is halved. The output levels can then be set along with the velocity sensitivity by using the DX7 values direct.

The envelope data is the most inconsistent between the two synthesisers, but before any fiddling is started, it's important to set the rate scaling parameter which again is half the DX7 value. Three issues ago, I featured the DX envelope generators in an article and you'd best refer back to it if you're confused by all those rates and levels. I tried to come up with a scheme for calculating values but gave up in favour of the good old 'lug oles'.

The part of the "BRASS 1" sound that we're working on is a 16 foot "wah" (J Arthur column here we come) and it's fairly obvious that the carrier envelope is going to plateau after a slow attack whereas the modulator will share the attack but drop down to give the desired timbral modulation. Look for similarities in the envelope data and make use of the EG copy facility as much as possible.

A word of warning about many DX7 voice charts. Seldom are patches created from the INIT VOICE state and so many junk values are to be found in listings. Prime examples are pitch envelope rate data when all pitch levels are equivalent and level scaling data when right and left depths are zero.

Fortunately, LFO data is remarkably similar between all machines in range if not in effect. If an LFO rate seems wrong, don't live with it just because the voice chart says so. The most obvious difference is the lack of sine and downward saw waveforms. The triangle wave is a perfectly good substitute for the sine and the saw is no great loss.

(Click image for higher resolution version)

Having edited this voice 'partial' as those smug D-50 owners might say, give it a name and store it in the 'A' bank or equivalent. There may be some of you who will object to leaving two operators unused, but we're here to imitate rather than go one better.

The other half of the voice is constructed from algorithm 6 and is an 8 foot pitch "wah" with a slight beat. The editing procedure is exactly the same. The DX21's single level scaling parameter is used to prevent the tone getting too bright at the top end. Editing all done, the voice is named and stored in bank 'B'. Selecting dual mode and both voice halves will give a sound with a richness of which you probably never knew your DX21 was capable.

Right, that was easy but what about some of the DX7 mark II voices? OK, let's look at "PianoBells". The algorithm is #5 which is three systems of singly modulated carriers readily faked by a combination of DX21 algorithms 1 and 5. Starting with the left-most system we hit a snag. The frequency ratios required are 20.00 and 4.00 and the DX21 can't provide the first. No panic, simply choose 10.00 and 2.00 and transpose later. Continue editing and store the first bell fragment.

The remainder of the voice is a Rhodes timbre and a duplicate bell sound severely detuned and brightened with feedback. Because we're halving the frequency ratios, we now need a ratio of 1.50 for the piano and the DX21 won't oblige. Here we resort to psycho-acoustic trickery. The attractive part of the voice is the detuned bells and the piano lurking somewhere beneath takes second place. It's more important for the piano to be at the correct octave than it is for it to be of an identical timbre. The solution is then to set the piano modulator to 3.00 and say no more.

(Click image for higher resolution version)

After the "PickGuitar", the next most talked about DX7 II voice is the "VibraPhone". Can it be translated? Yep. On analysis, the sound is three operators worth of vibes and three operators used to produce a click. Algorithm 27 can be simulated by 5 and 4 in combination on the DX21. The lefthand three operators produce the vibes sound and the only point of interest is the unmodulated carrier providing the fundamental frequency.

The right-hand three operators are arranged as two modulators and a carrier. The DX7 calls for a frequency ratio of 31.00 but, although outside the DX21's range, experimentation shows that the ratio is not critical and that the DX21 top value of 25.95 is quite acceptable. Storing the 'click' as a separate voice enables its use in combination with other timbres such as the 'bank A brass' described above to give ear-catching hybrids.

The faked 'VibraPhone' stands up well to the DX7 original but obviously can't compete with the subtleties of random pitch and aftertouch pitch bias. The DX7 version is also better controlled at the lower end thanks to its comprehensive level scaling — a feature in which the FB-01 shows some improvement over the DX21.

Erratum: Apologies all round to those who have suffered personal embarrassment resulting from the frequency ratio, detune and amplitude modulation enable parameters of operators 2 and 3 on last month's DX21 voice chart programme being transposed. Define the following —

:swap2/3 (n-n) DUP 2=IF DROP 3 ELSE DUP 3=IF DROP 2 THEN THEN;

— and add 'swap2/3' as the first statement in the definitions 'detune' and 'osc freq' and also after the 'DUP' in m.sens.