Samples, Synthesis And The Return Of The Filter

We're in the second half of the 1980's, the warmth of Analogue synthesis is almost a distant memory, replaced by the cold and to be honest, thin sound of FM synthesis, which was now heard just about anywhere where music was being played. One strength the DX series had was deep and round bass sounds, but for those daring to experiment with the mass of options hiding behind the 2 line LCD, the results tended to be harsh metallic clangs, another thing FM synthesis was good at(!) There was another competitor who's synthesis was very similar, that was Casio with the CZ range of synthesizes using Phase Distortion synthesis, but somehow Casio had managed to inject a little warmth into the sound, choosing to use harmonically rich waveforms, rather than sine waves as used by Yamaha. Alas, the CZ range was also quite difficult to get to grips with for many, with more complex envelope generators as well. They were popular, but the DX7 got there first, and still ruled the synthesizer roost.

Then the slumbering synthesizer giant Roland awoke, and once again, the synthesizer world was to change with the launch of the D50.

 

Roland-D50
The D50 had a deep rich sound that could swirl around the stereo spectrum, and sounded much more realistic than any other synthesizer that had gone before, thanks to the revolution that was Linear Arithmetic (LA) synthesis, the first mass market Sample + Synthesis sound generation system. Suddenly, the warmth of synthesizers of old was back, not quite the same, but was back, and the D50 became the new wonder synthesizer of the 1980's.

Linear Arithmetic worked in a different way to all digital synthesizers, and used a simple perception trick to make it sound as realistic as it did. But the thing that got many people excited was that it had a proper synthesis system hidden within, and also heralded the return of the filter, not quite the same as in days of old, but still a welcome addition to the synthesis world.

So how did it work and what was the simple perception trick?

The human brain does most of its perception of what a sound is, during the initial attack stages of the sound.  If the attack portion sounds realistic enough, the remaining section will be perceived to sound real so long as it sounds similar to what the brain expects to hear. The D50 had a number of ways to construct a sound, including samples of real instruments, samples of older synthesizers, actual synthesis waveform generators and of course Time Variant Filters so shape the harmonic content of the sound. Now, remember that this is around 1986, and memory is phenomenally expensive compared to today. The sample ROM on the D50 would not be big enough to hold the average PC BIOS on today's PC's, being around 8MB(!) in size, so to compress the many sounds that were available, only the attack portion of the sounds were sampled and stored for percussive and piano type sounds. Only string sounds were long enough to have sample loop points, but even then the samples were kept short to save space.

So, the typical piano sound on the D50 used the sampled attack of a real piano, and constructed the rest of the sound with it's synthesis engine. This was then passed to the filter section, which added the final shaping of the sound to give the sample to synthesis sections a smoother transition. Because the brain heard the piano at the beginning, the rest of the sound must be a piano, even though in truth, it was not. However, the D50 was not really known for it's piano's, it was known for it's swirling pad sounds and searing lead sounds that brought back memories of classic Tangerine Dream tracks and complemented the sound of groups like Fleetwood Mac and many others. So, just how did this system work?

The D50 had at least two sound sources, known as Partials; these were either a sample from the sample library (Either onboard or on expansion cards), or a synthesis waveform from the on-chip oscillator. This was then passed to the Time Variant Filter (TVF), which could change the harmonic content of the sound over time, thanks to the rather complex envelope generator. From here, the sound is sent to the Time Variant Amplifier, which also has a similar complex envelope generator, and then on to the effects unit... The what? Yes, an effects unit, more on this later.

Now, one thing to bear in mind is that what I have described above is a sound, but when you play on the keyboard, you are playing a performance, confused? Let me explain. A performance is made up of one or two sounds, these can be created together with the performance in mind, or any two sounds that you think sound good together. The performance sets how the sound will respond over the keyboard, the tuning, the overall volume, what the pitch bend and modulation controllers will affect if they are used, and what effects will be used on the sound (There's those effects again, be patient, they will be covered soon). If you create a performance that uses two sounds, then you now have four partials (Sound sources or the D50 equivalent to oscillators) to make your ultimate swirly sounds with. Now, bare in mind that the D50 can play a maximum of sixteen notes at any one time, each one of these notes will be one sound, so if you play a performance that uses two sounds, then the maximum number of notes you can play is halved. Even though this is mentioned in the manual (Quite a few times actually, each time it is explained in such a way that it never sinks in), many people felt duped, no wonder as all the marketing material pushed sixteen note polyphony as a strength over the DX7 and CZ 1000.

With a performance that used 4 partials, all using synthesis or synthesizer waveforms, you had something that could stand up against the analogue synthesizers of old, or if you had something that used synthesis and samples in a creative way and were cleaver at manipulating the delay parameters and envelope generators, you could create swirling pads that evolved over time and stayed interesting for as long as you held the key down. Compared to the cold harsh sounds of the DX7, this was a lush vibrant synthesizer. But as it wasn't made up entirely of resisters, capacitors, valves and transistors, how did it create such a vibrant sound? No one really asked that question, but the answer has already been mentioned (Can you guess what it is yet?).

The effects section of the D50 was not really a first, analogue synthesizers has used effects for years, mainly chorus and simple delay effects, with the odd spring reverb in some of the more exotic, and some esoteric models, but with digital effects, the D50 could take it to a new level, allowing the effect to become part of the sound sculpting process, indeed, for the D50 it was integral to the sound generation process, and nearly all of the sounds that shipped with the synthesizer out of the factory depended on the effects for their ability to impress the listener. This was the D50's strength, and it's Achilles' heel, pressing a single button, or selecting a single option in the performance settings, and the D50's sound became as brittle and weak as the other digital synthesizers that it was up against. The warmth and depth of the sound was the synthesis equivalent of a magicians smoke and mirrors.

This did not stop it selling by the truckload, and it did not stop Roland capitalizing on the popularity of the D50's sound and desirability. Soon after, Roland launched the MT32, and advertised it as a multitimbral (Able to play many different sounds at once over MIDI without requiring additional synthesizers) version of the D50, which in reality, it wasn't. The MT32 had a cut down version if the D50's synthesis engine, and a crippled effects unit in comparison. Also, there was no way to edit the MT32 from the front panel (Though it could be edited over MIDI from a computer) and there was no battery backup to keep user programmed sounds in memory once the power was turned off (Even though there is a space on the circuit board with connectors marked as backup battery!). These were possible on the D50, and the D50 also had an optional editing surface, which allowed you to tweak sounds as you played, just as you could with the old analogue synthesizers, but this was expensive, and these days rare to find. Next to the D50, the MT32 sounded weak, but on it's own, it could fool the uninitiated that it was a fully-fledged D50 filling the music with ambiance and character.

Not long after this, smaller synthesizers were released based on the MT32 sound engine, namely the D10 and D110 sound module, and later the D20 and D5. In the LA synthesizer range, the D5 was unique as it did not have any onboard effects, meaning that out of the entire Roland range, this was the weaker and more exposed synthesizer, and to be honest, did not really sell by the bucket load (The D10 was better value and second hand, was cheaper then a new D5 at the time). The march of Roland and its LA synthesis was unstoppable, and samples were embedded into synthesis. In fact, Roland were so pleased with the way LA worked, they took the sound generation out, and built a series of samplers with the remainder of the synthesis engine. The S50, S20 and S10 samplers were launched and considered to be the best sounding samplers around and very easy to use, as there was an option to attach an external monitor and control the system with a mouse. However, Akai had beaten Roland to the sampler market with their easy to use S700 and S900 samplers, and had the market sewn up. Roland at best were a bit player with arguably better hardware, but at least they had the D series synthesizers to rely on, then Korg returned to the market and things changed once again...