Hi Matt,
Thanks for the reply!
I am very skeptical that what I am observing is peculiar to the particular unit I have. Reading around online, I have come across lots of discussion, and even some videos, about similar behavior in other Arturia products, some of this going back quite a few years, especially in the Minilab. It seems that Arturia either consistently uses low quality encoders or doesn't design circuits to read their output in a way that properly controls noise.
First of all, one of the main issues I have is that in order to move through the range of output values from an encoder, I have to turn it nearly three full turns! This is true of all the encoders. And based on my reading elsewhere, this is true for all these devices. It isn't something wrong with this unit or my setup. And nothing you can do for me aside from installing different encoders with higher resolution would help this. One full turn of an encoder only gets me from 0 to about 47 out of the 0-127 range. The encoders seem to have a resolution of 48 increments per revolution. This means that in order to sweep through the range, I have to turn the encoder as far as I can, release my grip on it, regrip it, turn some more, release my grip, regrip, and turn some more. This is no good. I should be able to go through its entire range without releasing it. I can't imagine why anyone would want it to work this way.
Acceleration is not a very good solution at all. If I set it to "1:1", for example, which "fixes" this problem and makes it possible to sweep the full range in less than a full turn, the software merely multiplies each increment by four, which decreases the resolution of the sweep range by a factor of four. This means that it only outputs approximately 32 values rather than 128. Not good for audio work. You can hear the stepping. So these encoders are basically useless for modulating parameters of a synth while playing, unless you plan to smooth out the data after recording it. The fact is, the encoders have a fixed resolution. Nothing can be done about this in software. They only register 48 increments per revolution. The only way to get from 0 to 127 in less than a full turn is to multiply such that each increment now counts as four.
What I would like to know is if this low resolution encoder is something some people like, if such a design decision was made for a reason other than cost. Are lower-resolution encoders cheaper? Or is it actually useful somewhere to have encoders that require nearly three full turns to go from 0 to 127?
On top of that, the encoders produce noisy output, with values jumping around a bit, sometimes going backwards, while you turn the knob very carefully and smoothly in one direction. I've seen this described for Arturia hardware elsewhere. It isn't unique to my controller. And I did a little research on encoders. Apparently, they need to be "debounced" properly. Perhaps bouncing is the problem here. Obviously, hardware debouncing requires more circuitry, more components, and therefore involves higher cost. But software debouncing involves other problems such as introducing delays. Encoders with a lower or non-existent bounce rate are expensive.
I notice that if I put the encoders in relative mode, such that they only output two values, one for increment and another for decrement, if I turn the encoder at a slowish speed steadily in one direction, a fairly small percentage of the output values are for the wrong direction. This looks like the bouncing behavior typical of cheap encoders that I read about. But consider what happens in absolute mode when acceleration is turned on. If no acceleration is used, any bounce error only means that the value will go in the wrong direction by 1 step out of 0 to 127. Not too noticeable. But if we turn on acceleration, which just multiplies the increment in software, it also multiplies the magnitude of the error. So we get a jump of 4 in the wrong direction with 1:1 acceleration. With larger multiplications with high acceleration, the jumps in the wrong direction are magnified further. So while acceleration sort of "solves" one problem, it magnifies the other one.
I can show you the output if you like. The encoders on my NI Komplete Kontrol S25 by comparison give all 128 values very smoothly in less than a full turn. This is what one expects.
I am happy to report that the faders on the Keylab MKII perform nearly flawlessly. They give smooth, steady output, with no value-jumping. The mod wheel and pitch bend wheel usually skip every other value, seemingly as if the increments are being multiplied by two to get the full range, so they basically only have a 64 step resolution. No apparent backward motion or noisiness though, thankfully.
Channel aftertouch covers the range from 0-127 usually in less than 35 steps, fewer the faster you change pressure, sometimes as few as 4 steps. Pretty coarse. Not nearly as smooth as my NI S25. And the physical depth through which the aftertouch moves is far shallower than that of my NI S25.
As for the drum pads, I don't know what is typical for these as I have never had any before. How sensitive should they be, and how consistent should they be with lighter hits? With hard hits, they are pretty consistent. With lighter ones, some hits register and some don't and velocities vary. Is there a break-in period with these?
I recognize that these Arturia controllers pack more features in for the price than something like an NI unit with a similar number of keys. The Keylab MKII has drum pads, faders, outputs and inputs for modular synths, and so on. So I suspect that in order to put together a unit that has all these features at this price, costs had to be cut somewhere. I just have to decide if I want to accept these limitations or spend more and get fewer features elsewhere.
I'd actually appreciate if you could test another unit to see whether or not it produces behavior like I am describing. I bet you'll find that it does.
