In the early 90’s, and after many years of programming synthesizers using either subtractive synthesis or FM synthesis, I only swore by samplers, having understood that what attracted me most in electronic musics was this work on sound experimentation, the research for incongruous sounds, including some very rich harmonic contents.
Still very expensive, high quality samplers were far beyond my financial means. I nevertheless managed to acquire an Ensoniq Mirage DSK, which however allowed me to do my first steps into the world of sampling. But I was dreaming to something better, and that’s where I saw this mythical machine the Roland W-30 was. Of course, it wasn’t a 16-bit sampler, 44.1khz (CD quality), but … nevertheless much more powerful than my modest DSK Mirage.
25 years later, I finally got my hands on this mysterious machine that had so often haunted my teenager nights ;-)
Let’s examine the « beast » more closely …
First of all, some bulk specificities: 16 notes polyphony / 28.8 seconds of sampling memory (@15khz) or 14.4s (@30khz) / built-in sequencer / 5 octave keyboard (responding to velocity and after-touch) / Large graphical LCD screen / floppy drive DD (720kb) / no built-in effects / loaded with 512kb of samples in rom (drumkits, synthesizers JX10, Juno, …).
I must point out that I come from the Akai world about sampling. Since my first Akai (S-950), passing by the S-2000 and then the S-5000. So I have a pretty good experience on the structure of memory, the organization of samples, programs, « multi mode »,… on the Akai samplers.
Also, I will only talk about the « sampling » part, because the W-30 is a workstation having an integrated sequencer I do not use, nor probably ever.
Let’s go for sampling...
Once in front of the W-30, all this must be questioned, because Roland chose for this sampler an amazing memory structure (as well as floppy disk structure) to say the least !
First of all, we are faced with a new terminology. We speak of « tone » (= a sample + parameters). They’re limited to 76 maximum which I’ll speak more in details further. We can already draw the following conclusion: The W-30 can only play up to 76 different samples. It will probably be enough in most cases, but it’s an astonishing limitation when you come from the Akai world.
Then, the memory (ram) unity is not expressed in kilobytes (or kb), but in « segments ». One segment = 0.4 seconds of recording time. This means that the sample length will be multiples of 0.4 sec (0.8 – 1.2 – 1.6 sec, etc …). And there is not just « one » memory, but … two ! Named « A » and « B ». When sampling the sound, it will be necessary to choose (in addition to its duration and recording quality: 15khz or 30khz) in which memory bank (A or B) we will place our sample.
These two memory banks (A and B) respectively contain 18 « segments », thus giving a maximum size of 2x 7.2 seconds = 14.4 seconds (why make it simple?) using the maximum sampling quality (30khz). With the reduced quality (15khz), the time length is doubled, and then reach 28.8 seconds.
Finally, the W-30 also has two « Rom » circuits (for a total of 512kb) containing a bunch of tones (samples). In fact, these samples use slots from the 33th to 76th. These tones are therefore permanently stored and can’t be modified. Our own samples will take place in the free slots left from 01 to 32. The rom samples are mostly dx basses, analog synths & leads (JX10, Juno, …) very well made, and depending on the circumstances, they can be opportunately used !
After this approach to memory management (to say the least exotic) comes the analysis of the quality of the sound sampled. I have to frankly say that it is positively surprising … Despite its mono input (thus all samples are mono), its resolution of 12 bits and the frequency of 30khz, I managed to sample sounds very rich in high frequencies with a fairly accurate rendition. It should be said that the converter of the W30 is a converter 16bit, which then compresses in 12bit for storage in memory. The Roland « D.I. » algorithm (Direct Interpolation) also allows a rendition with a minimum of aliasing and the result is really surprising in terms of quality !
Even when using the 15khz mode, I did a few tests on some « speeches » (human voices, interviews, …) and the result is frankly convincing !! If you want to get a lot of samples into memory, you should try to sample at 15khz, and if the result obtained is not convincing, then switch to 30khz mode and record the sample again. That’s probably the price to pay to get a large amount of good sounds into memory (no more than 32 anyway).
... And now, editing the samples !
Let’s move on to the possibilities of manipulating samples that are… quite poor. In the « Wave edit » menu, we’ve just the very minimum: we can copy a sample, we can « Truncate » (« Trim » in the Akai terminology), allowing us to cut the sample at the beginning and at the end. The function « Mix » merge two samples to make a single one. The function « Combine » (chaining two samples, to create one longer). We also have a « Digital Filter « (LPF / HPF), and finally « Loop » to loop the sample.
The loops fonctionalities are pretty rudimentary. There are only 3 modes: « one shot » (no-loop), « alter » (the sample is played back in both directions alternately), or the Foward/Reverse mode. On the other hand, we have a search function helpful to find the best « locators » to execute a loop, and another function to attenuate the phenomena of « loop clipping ». Thanks to the large graphic screen, we may use the graphic representation of a sample, using « zoom in » or « zoom out » to find the most suitable looping points more easily.
Unfortunately we don’t have some very practical functions like « Time Stretching », or « Normalize » however available on entry samplers of the time (as on the akai S950 for instance).
What is especially shameful is that the « Truncate » function, doesn’t allow to free unused memory, once « a segment » is not completly unused for a sample.
A simple case: I sampled a sound and I had expected its duration to 2 seconds. After removing the « blank » at the start and at the end, the new sample length is now about 1.7 seconds.
So 0.3s is freed which could be allocated for other samples let’s say. Sadly no … because a whole segment is 0.4sec. So my 2s sample length needs 5 segments. After truncate the sample (2s -> 1.7s), I still have 5 segments because the 5th segment is still in use for storing my sample data. This shouldn’t be a big issue only for a single sample. Now what about 10 samples of the same case (2s -> 1.7s)… So 10 x 0.3 sec = 3 seconds of unused memory but therefore impossible to free… that makes a huge non-sense. With such a system, imposed by the W-30, these 3 seconds of unused sampling memory will never be freed and reassigned to the main memory … this is just « lost ». It’s a shame, especially because the amount of sample memory is far from beeing infinite on the W-30.
Let’s finish this visit with the manipulation of the filter and the LFO (a single LFO). Again, surprise: the LFO offers only two waveforms (Sin and P / H). But on the other hand, there are other useful parameters and the filter (TVF) is pretty good. It’s a resonant filter, which can be affected by the LFO, or the keyboard velocity, or the very complex envelope (EG). The filters are really very good ! We could almost perfectly simulate the behavior of purely analog synths.
Thanks to the « Tone edit » section, we can modify how the sample behaves (filter, volume, envelope, lfo, play mode, velocity, …).
In this section you will find everything you need to have our original sample, once you have went through « Wave edit » and then the « Tone edit », to finally become a « Tone » with the desired behavior!
So finally … all that left to do is to assign our tones to a « Patch » (« Program » in Akai terminology). The W-30 allows up to 16 patches (= tone arrangement) to be handled.
« Patch edit » is the place that will mainly consist of placing our tones into a keyboard zone (keyboard mapping), and control the behavior of these (volume, which output to send the sound to, …).
Important to know: a sample has a maximum transposition range of 2 octaves (a sample recorded on C3, can’t be played at C#5 for instance).
Here too, Roland uses a special machanism : « Patch Split » (which as its name does not indicate at all) which is the place where we will do the « keyboard mapping » using our « Tones ».
The principle is to map the keyboard areas with the desired tone (s). The W-30 allows you to play two tones at once when you press a key on the keyboard (same system used on EMU Emax or Emulator). It can be the same tone, or two different ones. Depending on the keyboard velocity, tone # 1 will be played or tone # 2. It is also possible to use this system to emulate a stereo field (two identical tones, distributed on two separate outputs, with an LFO varying the volume of each, simulating a stereo pan varying from left to right alternatively).
It will therefore be necessary to press all the keys of the keyboard, to have an overview of all the assignments made… this being shown on a graph representing a keyboard with tiny dots showing the « zone » assigned to a tone … you really need very good eyes, and an excellent memory to remember everything. This is unpractical and a badly conceived interface design, to my opinion, to manage the pach assignements.
For the rest, the patch mode allows to define the volumes, the outputs (there are 8 individual mono outputs on the W30), the settings of the pitch bend, the keyboard setting … it is quite complete I must say.
To summarize, and as a conclusion concerning the sampling part, I would say that the W-30 is a sampler with a very good sound quality (despite the limitations I mentioned), has excellent resonant filters, and a complete synthesis engine (subtractive) that could turn it into a real synthesizer! The keyboard is very good, well balanced and accurate, as well as general construction which is quite solid. The two weak points we can find on this kind of vintage machines (more than 25 years old) are ofen :
- The floppy drive (type « chinon »), operating on « MF2-DD » (720kb, double-density) disk.
- The backlight of the large LCD screen which tend to dim
The W-30 pros :
- Despite audio limitations (12bits / 30khz), the samples obtained are of very good quality
- High quality resonant filters
- Complex envelopes (one for TVF, another for TVA)
- Easy operation thanks to the large LCD screen and intelligently structured menu
- Good keyboard quality (responding to velocity and after-touch)
- Robust construction (despite a fairly light weight, which is 9.8 kg approx. )
- Interesting sounds residing in Rom memory
- 8 individual outputs
- Ability to upgrade the machine, using a SCSI kit, with an outside scsi socket already provided for this purpose.
The W-30 cons :
- No built-in effects
- Limited number of digital processing (no time-stretching, or normalize functions for instance).
- Pretty rudimentary loops functionalities.
- Very special ram memory structure, generating quite a lot of waste of unused memory
- Limited amount of memory (28.8 sec maximum, using the 15khz mode).
- The Operating System operates in « Overlay » mode (like the EMAX) -> the system disk must often be left in the machine because the W-30 often load some data
Let’s conclude this overview by analyzing the electronic guts of this very pretty machine :
The heart of a sampler uses three essential elements: The microprocessor / The memory / The filters & outputs.
Here are some key components of this machine :
- 1x Intel N8097BH (16bits micro-controler, same family as 8x9x)
- 4x HM65256BSP-12 (16-bit pseudo static-RAM / system & patches data)
- 5x MB81C4256P-12 (12-bit dynamic RAM / samples data)
- 1x MD6209 (16-bit D/A converter)
- MB654419U (gate array / filter)