Here’s the hard truth about analog modular synthesizers. As much as I’m attracted to them, they’re sorely limited in all sorts of ways. Also expensive.
For the last few days, while working with the Buchla Skylab, I’ve been contemplating what we might call process music — music in which tone production is sort of ongoing. Perhaps influenced by the composer/performer while it plays, or perhaps entirely automated, but that’s a less significant criterion.
This notion inspired me to put together a little piece called “Jungle Nights” in my computer using Csound:
I won’t claim this is a terrific piece of music. I’m not even sure it’s finished — I have some more thoughts about how to improve it. But it illustrates fairly well the idea of music as process.
The thing is, you couldn’t do this piece using the Skylab, not even if you cheated and added an outboard reverb and delay. The main tone and the drumming you couldn’t even come close to. The hissing noises and the little tone bursts could be approximated. And you certainly couldn’t do all four types of tones at once, let alone save your entire piece to the hard drive, load it up again next month or next year for further editing, and have it sound exactly the same as it does today. (The Skylab can store panel settings, but it has, quite obviously, no facility for memorizing how your patch cords are connected.)
The cost of a Skylab: close to $15,000. The cost of Csound: It’s free. Okay, I did have to buy the computer, but the cost of a swell computer, a MIDI keyboard with sliders, and an iPad so you can run a multi-touch sensing input would be around $3,000, and you could use the computer and iPad for a few other little things now and then. Also, the learning curve may be a little longer and steeper for Csound than for a Buchla system, but it’s not beyond human comprehension … and you can always buy a copy of my book (Csound Power) to get you started.
Learning curve aside, the advantages of a Csound-based system, as compared to the Skylab, are huge. They’re monumental. Unlimited numbers of oscillators and filters. Reverb and delay. Multi-segment envelopes. Audio recording and playback (multi-channel, if you like). Compatibility with OSC. Numerous types of synthesis. Ultra-precise control of pitch and timing. Complex mathematical processing of control signals. As many step sequencers as you’d like, with as many steps as you’d like. A broad choice of filter types. The list goes on.
Yeah, I love hardware with knobs and blinky lights. But maybe I can cure myself of that.
The substitution of a 285e (does these tones effortlessly, providing two such voices in tandem with a single 259e) along with the 210e for patch cable storage would go a long way here toward closing the gap. The Skylab is perhaps too limited in size, and this just underscores your point contrasting the cost. I guess my point is that a modular system, even the Skylab, is not a fixed system, so it is difficult to make fixed statements about it (perhaps in the case of the Skylab you could arbitrarily adhere to the recommend set of modules for the sake of discussion, but this would just be a rhetorical convenience). I might as well also add the well worn counterpoint that limitations are good. I don’t find anything to envy in a system that supplies an open ended number of oscillators at the touch of a button. Even a simple modular runs the risk of providing too many options.
This blog entry doesn’t even mention the most significant limitation of the Skylab — and of the entire 200e series. The pitch control over the oscillators is simply inadequate. These modules use a 16-bit A-to-D converter, but I’m pretty sure they aren’t even using all 16 bits of data, as 16 bits would give you about 0.33Hz resolution across the entire audible frequency range, or 0.16Hz resolution if you decide that the highest fundamental you want your oscillators to produce is 10kHz. In fact, the frequency resolution at the lower end is about 1Hz, which is simply not fine enough resolution for producing in-tune bass lines in 12-note-per-octave music. Since bass lines are one of the things that a lot of people want to use analog synthesizers for, and since Buchla systems are intentionally designed around the idea of 12-note-per-octave tuning (which is another limitation), the fact that you can’t produce in-tune bass lines in a $15,000 instrument is … well, I wouldn’t buy one. If you gave me one I’d never use it. The lack of frequency resolution has other consequences. If you want two oscillators to beat against one another at a very slow rate because they’re tuned, let’s say, 0.3Hz apart, you just plain can’t do it. In the melody range, you can choose dead-on unisons or 2Hz-apart beating. There’s nothing in between. Nor can you necessarily choose which you want. If you’re playing two oscillators from the touch surface, one key may give you dead-on perfect unisons while the key next to it produces beating, because the two A/D converters aren’t calibrated with one another. This is not a theoretical analysis; it’s based on my own experience with the machine.
If you don’t find anything to envy in a system that has an open-ended number of oscillators, filters, envelope generators, effects processors, mixers, compressors, control signal logic processors, and voices of polyphony … I just don’t know how to respond to that, Doug. Any response I made would veer swiftly in the direction of insulting your intelligence, and I’d rather not go there.
Rough measurement shows the lowest frequency at about 27.5 Hz, and the highest at about 7kHz. Quantization appears to be just over 1 Hz as you say. I find the oscillators very usable, as do many others. There are plenty of recorded examples of the 200e available online, and it’s clear that the system is capable of tonal music. In any case, and supporting my original point, you can use an analog VCO if you can’t live with the quantization. You give up the tremendous sonic range of the Buchla oscillators, but many people go this route.
I don’t doubt that the oscillators are very usable for some sorts of music. I like the way they sound! But I’m also of the opinion that precise frequency adjustments DO matter. And of course, with Csound, I have both a tremendous sonic range AND precise frequency control. That was the point of my original post: Hardware modular synthesizers are utterly obsolete. If you like obsolete technology, then you may be very happy with a hardware modular. Hell, there are people today making digital instruments using the 8-bit SID chip from the Commodore-64 — and coming up with various arguments why this makes musical sense. If you like obsolete, don’t let me talk you out of it! But also, don’t try to convince me that it ain’t obsolete, because it is. As I demonstrated, I can run rings around the Skylab with a software system that costs $0. Rings the size of Saturn’s.
If precise frequencies matter, then use the Fine Tune knob on every oscillator. I will state in partial support of your struggle with tuning the beast that the 259e oscillator is harder to properly scale than the 261e. This is made more difficult as the 259e lacks the 261e’s scale trimming potentiometer behind the panel’s face.
I also think that one will have to define, or perhaps just be very liberal with the definition of the word “obsolete” in order to sustain the given argument. Is an ARP 2500 obsolete? No, it performs the same mighty job of making inspiring music. The same will hold true for your CSound-equipped computer if it survives 40 years or more. As a further arrow in this quiver: pianos don’t have pitchbend.
As a matter of being fair and forthright, I should let you know what kind of horse I have in this race. I own a 27 panel system. It certainly isn’t obsolete in being my primary musical outlet in a way that my Mac (loaded with purchased software) can not fulfill.
We currently have more and different tools. They coexist because each is potentially ‘better’, ‘more convenient’, ‘cheaper’, or ‘accessible’ than the other. For many, a laptop is a ‘better’ tool for the creation for music. For many, a laptop is the ‘worst’ or ‘wrong’ tool for making music.
Please take the opportunity to listen to BEMI’s “Etudes” CD for 200e before writing the Skylab manual. Also, check out Richard Lainhart’s “Oraison” and his other work on YouTube. That of Oxix52 and Alessandro Cortini (blindoldfreak) is also freely available. For some, the 200e is the opposite of obsolete. This sprinkling of videos would strongly support that argument.
The fine-tune knob doesn’t help. The voltage inputs from coarse tune, fine tune, and incoming CV (actually, CVs, as there’s one panel input and one direct input from the mod oscillator) are summed in the analog domain before being sent to the A-to-D converter. It’s the converter that’s the bottleneck. I did check this. Admittedly I did so on only one of the oscillators, not on the other one, but it would be very surprising if they were engineered in different ways, and I’m pretty sure Don would have mentioned it to me.
I agree that a Buchla system is a terrific resource for certain kinds of music! As is a harpsichord, and a harpsichord is obsolete technology by almost any definition you can think of. On the other hand, a software-based virtual harpsichord offers numerous advantages over the physical instrument, ranging from instant tuning to any scale (and no drifting out of tune) to the possibility of adding key velocity response if you want it. I would respectfully suggest that Csound offers the same kinds of advantages — and many others — over a hardware modular.
Thanks for the tips on music. I’ll check them out.
Jim maybe you could try this on CSound. http://soundcloud.com/dougcl/buchlajazz
It would be dead easy, but not worth the effort.
I doubt it, but that’s not the point. It should at least be clear from this clip that the 200e is not limited to an aleatoric sideshow. You didn’t say that it was, but it could easily be inferred from your statements by someone who didn’t know any better, and god forbid, respected your opinions.
“Aleatoric sideshow” — I like the phrase. And no, I would never have characterized the Skylab, or any other hardware modular, as being limited in that manner.
With respect to the question of whether anyone respects my opinions, or ought to, it’s not my place to be the judge of that, but on the whole I don’t think I have anything to apologize for. I’m sometimes wrong about music technology, but I’m seldom wrong in dumb or obvious ways, and when I learn that I’m wrong, I admit it. I was totally wrong about Native Instruments Reaktor 1.0, for example. When Reaktor 3.0 was released, I admitted as much in my follow-up review. (This was many years ago.) If I have said anything that was factually incorrect about Buchla gear, please — let me know!
It appears that the 1Hz quantization is across the entire range, but I could be wrong as I tired out at 350Hz or so. It gets harder and harder to dial in 1Hz changes as you get into the higher frequencies (we’re probably dealing with an exponential input response upstream of the ADC.)
Worst case, a chromatic semitone falls exactly midway between two quantized frequencies, ie. a 1/2Hz error.
Since the number of quantized frequencies per octave increases exponentially, the 1/2Hz error in higher octaves is not perceptible.
My three 259e’s performed exactly the same. All quantized frequencies were identical. Your statement that the ADCs have varying calibration is either false, or I am lucky.
Since all three oscillators are the same, dialing in the exact same pitch on all three is easy. Further, the CV resolution of the touch keyboard far exceeds the 1Hz limit rendering it capable of selecting any of the quantized frequencies. Your claim that neighboring keys deliver different results is untrue, even in polyphonic mode with several oscillators involved.
Regarding beat frequencies, the oscillator frequencies may differ in 1Hz increments, not two as you state.
So there is little in your facts to amend, and I’ve already addressed your opinions.
First, Doug, I need to thank you for providing a link to one of your Buchla tracks on Soundcloud. After listening to seven or eight of your tracks, I have a much better understanding of what you’re aiming at musically, and I want to compliment you on the sonic variety and expressiveness. You’re doing good work.
I was also being too optimistic when I said it would be easy to reproduce that track. Several non-trivial programming techniques would be required. I’m working on a “re-imagining” of the track in Csound, not trying in any way to steal your music but simply asking myself, if I had an impulse to do something similar, how would I do it? It’s an interesting project, and I may have some results in a week or so. It won’t be “the same” in any way, but I think the comparison is bound to be enlightening.
I’ve been asked to ship the Skylab back, so I’m no longer able to verify technical questions. I can only go by my memory of what I observed. I did, however, encounter situations (evidently due to slight variations in the calibration of the A-to-D converters) where the touch keyboard would produce a unison on one key and beating on an adjacent key. You may be right about the 1Hz increments of the beating; I was listening to tones that had a number of overtones, and I may have misinterpreted what I was hearing.
I asked Buchla why you got hired to do the manual when you seem to have nothing but contempt for their instruments.
I think I need to defend myself a little here. It is grossly inaccurate to suggest that I “have nothing but contempt for their instruments.” I’m a big fan of hardware modular synthesis, in fact. Initially I was happy about the idea of writing the manual in exchange for a Skylab. However, as I spent time with the Skylab, I became aware of certain limitations that I had not known about initially. These limitations, which seemed to me to be very significant musically, affected my opinion of the instrument.
Some of the limitations, notably the absence of general-purpose voltage attenuators and inverters, would naturally disappear in a larger system. (For the record, I should perhaps mention that the ARP 2600 had this functionality, at considerably less cost, more than 30 years ago.) One limitation, however, the lack of fine resolution in oscillator pitch, cannot be remedied.
If fine pitch resolution is not important to you, and if you’re ready to buy a slightly larger system in order to get CV processing, then a Buchla system might be exactly what you want. Go for it!
Also, there is a lot more to “music” than playing western scales in tune… I think your rants miss the whole point.
Certainly, there’s more to music than playing in tune. (I make no distinction between Western and non-Western or non-traditional scales.) But this assertion misses the point. I could, with equal validity, state that there’s more to music than producing a good, satisfying tone with your instrument. There’s more to music than being able to play precise rhythms and count rests in an error-free manner. There’s more to music than being able to balance the relative volume levels of the instruments in an ensemble. There’s more to music than studying theories of composition. There’s more to music than the ability to improvise fluently.
All of those statements are true. However, the fact that there’s more to music than item A or item B does not imply for a moment that item A or item B is irrelevant to fine music-making. Some people have no need to improvise, granted — but almost every other item in that list is very specifically necessary to fine music-making. You have to be able to play in tune. You have to be able to produce a good tone. You have to be able to play rhythms and count. You have to know when to play loud and when to play soft. If you’re going to compose original music, you have to understand all sorts of stuff about composition.
So … what’s your point, Evan?
Jim, if you find the attenuation and inversion facilities less than adequate in the Skylab, swap out a module for the 256e and/or 210e. I don’t think the Skylab module complement is optimal either. Because it is a modular though, you have the ability to tweak it to your needs. The module complement is just a *suggestion* from B&A. You know this. So it’s disingenuous of you to make statements about the Skylab lacking in these areas. Actually, I don’t know what your beef is. You had a 200e sitting in your lap, in trade for some work that you enjoy doing. It seems like you had some kind of crisis about the frequency quantization in the oscillators. I can understand this. Too bad you didn’t calm down after this initial reaction and spend time exploring the system a little more. In the end, if you really couldn’t live with the oscillators, you could simply replace them with analog oscillators. Again it is a modular, tweak it to your taste. I find it hard to believe that you would find yourself too constrained by the system to enjoy using it and make music with it.
I considered swapping for a different module to get the voltage processing. But yes, the frequency quantization of the oscillators was a deal-breaker for me. To the best of my knowledge, Buchla does not build true analog oscillators, so there would be no easy way to replace the digital modules with analog ones. Using true analog oscillators from Doepfer, MOTM, or Sound Transform in a Skylab system wouldn’t be a piece of cake, because, quite aside from the configuration of the modules within the case, Buchla (for reasons that are buried deep in the sands of time) does not adhere to the one-volt-per-octave standard.
There were some other issues going on at the time, having to do with the business arrangement. The upshot was that they changed their minds about having me write the manual. I didn’t refuse to do the job; what I said to them was, “I can’t do this job in exchange for the hardware. That deal is not going to work for me. I would like to be paid instead.” If true analog oscillators had been an option, I might very well have felt differently about it.
As an independent expert, I feel that it’s my duty to make publicly available whatever technical information I may discover about a music system. I also feel that it’s appropriate for me to offer cost comparisons. Anyone is entitled to do whatever they like with the information. You’re certainly free to ignore it — but if I make a technical statement that’s incorrect, I certainly hope you’ll let me know so I can correct it!
I’m sure Buchla systems are wonderful resources for certain types of music-making. They’re also very expensive. I don’t think either of those points can be disputed. Personally, I’m a big fan of modular hardware instruments — my very first synth, back in 1981, was a four-panel Serge, and prior to that I had learned synthesis on an ARP 2600. I don’t think I have ever made a statement that could be interpreted as a blanket condemnation of Buchla instruments … unless you think “they’re very expensive” qualifies as a blanket condemnation. If I had a big pile of money, I might buy a large Buchla system. Or a large Doepfer system. In the end, I would probably go for the Doepfer because of the true analog oscillators. Smooth control of oscillator frequency DOES matter.
Oh, and by the way … thanks for continuing the dialog!
” Buchla (for reasons that are buried deep in the sands of time) does not adhere to the one-volt-per-octave standard.”
1.2V/octave = 100mV/semitone (100 cents).
1V/oct 16-bit resolution for a 10V range = 0.183 cents
1.2V/oct, 16-bit resolution = 0.153-cent resolution range
Noober, I’m sorry to have to break it to you, but unless I’ve missed something (and I don’t think I have), the implication you’re making, which is that Buchla systems have better pitch resolution because they use a 1.2V/oct standard, is simply wrong.
If we use a 16-bit converter to translate an incoming control voltage across a 10-octave range, then we will have 65,536 pitch increments within that range. Each pitch increment will be 0.183 cents, exactly as you stated. But that will be true whether the A-to-D converter is calibrated to deal with an incoming analog CV signal of 1V/oct, 10V/oct, 3.14159V/oct, or whatever. In each case, if we want the oscillator to have a 10-octave range and we’re using a 16-bit converter, we end up with exactly the same pitch resolution.
To get your result, we have to retain the maximum voltage value of 10V for the analog CV signal, and couple it with a 1.2V/oct standard. This will narrow the available pitch range of the oscillators. Our oscillators will now have a pitch range of only 8.333 octaves — 8 octaves plus a major third. This is 100 scale steps in 12ET, rather than 120 scale steps, so the pitch increments produced by the A-to-D converter will indeed be 0.153 cents. But in order to get this value, you’ve cheated. You’ve narrowed the allowable pitch range of the oscillator. Exactly the same result could be obtained by using a 1V/oct standard and calibrating the A-to-D converter to max out at 8.333 volts.
The reason I said “buried deep in the sands of time” is that I suspect Buchla has continued to use the 1.2V/oct value because that was what he chose for his first instruments in 1965, and he wants his instruments to be backward-compatible. Backward compatibility is a good thing, but the same result could be achieved by building 1V/oct oscillators (which would be compatible with the rest of the known universe) and putting a little switch on them that would select 1.2V/oct inputs if desired.
In any case, a 16-bit A-to-D converter is not adequate for control of pitch. 24-bit converters should be used, at the very least.
Maybe the most amusing thing about this exchange is the idea of obsolescence. This is made even more entertaining by the insistence that a programming language developed smack dab in the middle of the 1980’s is somehow less obsolete than a modular synthesizer first developed only ~20 years prior. I’m sure it could be argued that Csound became obsolete with the development of sophisticated and reliable real-time software synthesis environments (and this cool thing called interactivity). Or that the text-based paradigm for coding became obsolete when graphical programming environments paved new ground. I wouldn’t argue that because, clearly, people still use Csound in creative, new, and developing ways.
I just don’t get why the benefits of one thing make another completely obsolete. Would it make any sense to say, “Certainly all music techniques, styles, and methods prior to the year 1999 (or any arbitrary point in time) must clearly be obsolete because they cannot do what newer music can.”? Maybe the main issue is your insistence that obsolescence is universal, not just a perspective held by an individual.
I play the cello, and that’s an obsolete instrument. As I think I pointed out a couple of weeks ago, in an earlier comment, the fact that some piece of equipment is obsolete does NOT mean you can’t or shouldn’t make music on it.
But if you want to talk about non-obsolete modular synthesis software, we could talk about either Reaktor or Reason. Reason is much easier to use, of course, and has a multi-track recording environment (which you’d have to build for yourself in Reaktor), but Reason is more limited in certain other ways. You pretty much can’t play microtonal music with Reason, it doesn’t like sending control signals to other devices, and it has a fixed set of synthesis methods, but if you’re okay with all that, you can profitably think of it as just one giant modular synth, especially now that the new Rack Extensions provide CV processing.