[sdiy] Synth Keybards and Number of Keys

[Edward King]

----- Original Message ----- 
From: "John Luciani" <jluciani at gmail.com>
To: "Edward King" <edwardcking2001 at yahoo.co.uk>
Cc: "Synth-DIY" <synth-diy at dropmix.xs4all.nl>
Sent: Friday, March 02, 2007 5:15 AM
Subject: Re: [sdiy] Synth Keybards and Number of Keys


> On 3/1/07, Edward King <edwardcking2001 at yahoo.co.uk> wrote:
>>
>> I am building a couple of two octave pedal boards that use slotted
>> optical switches
>> (PCBs arrive tomorrow!). I am thinking of doing some keyboards that use
>> slotted
>> optos in an analog mode.
>>
>>
>> -- if you encounter any problems, let me know and I'll do what I can to
>> help.
>>
>> -- Conversely, I would like to build pedalboards, so perhaps you can 
>> share
>> your stuff with me too!
>
> I purchased a couple of Wersi pedal boards that were surplus (although
> it wouldn't
> be hard to build these from scratch). On the back of the pedal boards
> there are a spots
> for four circuit boards. Each circuit board monitors a group of either 
> five,
> six or seven pedals. The boards are designed with jumpers so that they can 
> be
> cut. Readback is done using four 8 bit parallel to serial latches (one
> per board).
> The boards are daisychained and a single header 4x2 header connects to a
> uC board.
>
> There is a fair amount of side-to-side slop in the pedals. Probably too 
> much for
> the slot size of the opto. I am probably going to use some steel dowel 
> pins
> underneath the pedals to remove the slop.
>


I had this problem too (although to a much lesser degree I should imagine).

Although the key pivot rod (hardened steel) and bushes are a pretty good 
fit, there is a just a pinch of wobble. Given that the slot is only 3mm 
thick, I had to put guides on the front of the chassis to guide the linkage 
arm and reduce this. The guides are basically nylon blocks with a notch cut 
into them, so although they have continuous contact with the linkage arm, 
they dont cause any noticeable friction and they are very cheap and easily 
replaced.

I also committed a cardinal sin.....(no, nothing to do with a goat, I 
promise)....I rounded the inner upper part of the opto switch slot casing so 
that even if the vane does wobble a bit over time and through wear and tear, 
the vane still goes into the slot regardless.
Of course, the operator will realise that this is happening  if / when it 
does through the feel of the key travel...and hopefully have the sense to 
replace the nylon guide before it mullas the opto switch.
I committed another cardinal sin (god, you're going to think Im such an 
amateur now) by putting rubber grommets as shock absorbers inbetween the top 
of the pcb onto which the opto switches are mounted and the screws that hold 
them in place so that rather than cracking the pcb upon a "vane -vs- switch 
position mismatch" the whole thing moves just enough to absorb the shock....
Before you say anything, I know its the wrong approach.

Fact is that using opto slotted switches is totally the wrong approach 
anyway and on future projects I will use LVDT's (I think they're called) 
which are discreet, compact linear position sensors which use fields to 
measure position. The ones I have in mind come with a controller configured 
and callibrated for linear output.
Opto slotted switches are the cheap way of doing displacement sensing, but 
not the best way by a long way IMO. Its worked well so far, but I wouldnt 
like to vouch for the reliability until Ive thrown this thing into the back 
of a van and gigged with it a few times. Im almost itching for the setup to 
fail so that I can rip them all out and start again with the LVDT thingies.

LVDTs can be mounted a little like shocks. the ones I was looking at, you 
mount one end like you do a shock absorber and then attach the actuator to 
the key bottom. This arrangement isnt quite linear (because a key moves in 
an arc, not linearly) but its near enough to not worry about it too much 
unless you're perfectionist. If you are a perfectionist then you can make 
the thing even more linear by linking the key to another lever to the 
actuator, thereby flattening out the arc somewhat.

Other approaches I looked into included:

linear position sensors using that carbon impregnated stuff (the name of 
which escapes me now) but essentially a slightly more rugged slider. They 
are not cost effective at £35 per sensor (even in bulk) and the MTBF is only 
50,000 operations at best anyway.
Industrial ones are available which improve on this, but they are more 
costly by a factor of 10-100 depending on type and are physically huge.

optically encoded discs
impractical because significant gearing up would be required to get 
reasonable resolution. Would therefore also dramatically confine the 
keyboard feel to a certain set of dynamics (it makes the keys feel like one 
of those gyro toys). I tried this method despite being advised against it 
and wasnt surprised that it felt horrible.

Rotary pots
Dont go there.

standard 40 or 60mm faders
Dont go there either. Cheap ones have MTBF of 20 or 30,000 and expensive 
ones are too expensive for a less than great solution.

liquid and chemical based
My wife (chemist and microbiologist) ran off some liquid solutions for me 
which would last a reasonable amount of time, offered reasonably linear 
resistances for a given travel of contacts and were non-toxic and non 
corrosive (ish).  At first she thought I was mad considering putting liquids 
inside an electronic device, but I assured her that if it was good enough 
for the pc market with the liquid cooling systems, it was good enough for 
me....
they tested well, but the packaging was the main problem and leakage would 
have in fact been a major risk.

simple LDR's
inside a shielded channel with the light source at the bottom of the key in 
a shielded channel. Light pollution, configuration and calibration proved to 
be no good. very innaccurate. I was hoping that someone else had tried this 
and could "shed some light" :o))
but I didnt find anyone whod been able to make it work well.

Open Hall effect switches
Completely inappropriate.

QTC pressure sensitive pills
i experimented using a lever with minimal travel, but a variety of pressure. 
This actually worked well, but mechanical manufacturing inconsistences 
proved to be a nightmare with callibration and mapping. I think that this 
would be a nice way to go (given my new-found love of these wonderful little 
devices which I am over-using like crazy) but some of the manufacturing 
inconsistences (my bad) would have to be minimised or at least quantified 
and so the calibration wouldnt be that bad. Its nice to have a little 
"randomness" in there at the hardware level because the resulting sounds 
will ultimately sound more organic, but there is a big difference between 
randomness and rubbishness!
A potentially better solution is to use a spring attached to the bottom of 
the key, not far from the pivot point, with a medium density rubber stopper 
on the end, which comes into contact with the QTC pill. Choice of spring and 
mounting point are quite important though as the QTC pills have an upper 
limit (measurable, not physical) of 100 Newtons. I daresay they can tolerate 
more, but theres no point if your measurements top out halfway through the 
key travel or only use half or less of the scope of the thing.

Resistor wire coil + scissor type contraption
Basically, a resistor wire coil laid along the chassis (front to back) and 
another coil attached to the underside of the key, but to one side so that 
the two came into contact side by side more, the further the key came down.
Horribly innaccurate and you can feel weird dynamics at the key, almost like 
a pair of scissors.
Not sure why I even tried this, but it gave me the idea to try....

innards from a Scaletrix controller.
Say no more.

Force and Tension guages
These actually worked surprisingly well both on the back of the keys (beyond 
the pivot point where the return springs are mounted) and underneath the 
keys, at the front of the keyboard chassis. I got good resolution with the 
right amplifiers, but the solution is cost-prohibitive, the control 
electronics add excessive complexity to the machine and most of the sensors 
I looked at were either too large or too heavy. Futek manufacture some 
amazing ones (including very compact devices with built in canbus, usb, 
serial and other interface technologies) but the prices are not for the 
faint of heart.

Resistive wheel
I got inspiration for this idea from a simulator design in the 90's
Basically, a small wheel is attached to the bottom of each key through which 
is passing a small voltage. the wheel is made from a conducting material (of 
course) and is springloaded. It runs down a resistive material (the insides 
of a cheap fader are good for this) which is set at an angle (slope) which 
is set to maximise contact, whilst minimising wear and tear, errors and 
noise.
Contact between the supply and the wheel can be improved by using a sort of 
mini out-of-phase split ring type setup.
I never tried this method with this application, but it worked well on the 
sim. That one used a much larger wheel though and to my mind, this may prove 
fiddly to implement in this context.

Accelerometers
The two that I tried were very effective, but one of them topped out way 
before the maximum key velocity and didnt register on very slow movement of 
the key. This was my fault as Id chosen the wrong device for the range of 
accelerations. The two I bought cost £8 and £20-something respectively.
£8 per key isnt bad really when you think that the devices in question are 
exceptionally reliable and very accurate generally, contain most of the 
electronics you need and are small enough to be mounted on the underside (or 
even inside) of the keys with little hassle.
The output is not quite linear for the single axis models though so a 
two-axis model with the right logic is required if you want truly linear 
representation. Also, the one with the higher range goes a bit mad when the 
key hits the bottom of its travel for harder velocities (as you would 
expect) but this is a good thing in that you can use this to identify the 
bottom of the key travel coinciding with the activation of the aftertouch 
sensor.
I like this solution, but it boils down to this....you're measuring the 
movement, not the displacement. That is to say, you're only getting the 
displacement as a result of the movement, but if the movement is too slow 
for the sensor to detect, you dont get the displacement.
I like the idea of using this solution instead of the traditional bus bar 
velocity sensing setup and using someting else for the displacement because 
I am only using velocity sensing as well for completness and for sounds 
which dont need displacement sensing, I can save processor time.
I may chuck another keyboard together to see how this works out in practice, 
but at £700 for just the sensors, it would be an expensive experiment.






		
___________________________________________________________ 
Does your mail provider give you FREE antivirus protection? 
Get Yahoo! Mail http://uk.mail.yahoo.com