re-thinking the machine pedal -- (s)

The Pandalon has been sitting for a while, as I've been absorbed in other projects and activities, but as usual, the delay has been productive: things have been "simmering" in the back of my mind, and I think I have some ideas on how to do various things better than I was planning before.

Most significantly, I think I am not going to attempt to build this complex two-stage, single-pedal, machine pedal system.  Instead, as originally planned, I will use the simpler method of controlling each of the two machine rails with a separate pedal.  So there will be three pedals total: machine-1, machine-2, and sustain.  And three bicycle cables, instead of two, but they are cheap and readily-available.  With the two machine pedals, it'll still be possible to get all the effects of the single two-stage pedal, plus of course more flexibility.  Since some of the stops are "buzzy" or harmonics-adding, and some are muffling in nature, there will be a general tendency to have two separate types of effect that may be wanted, two axes of sonic change which may not benefit by being forced onto the same continuum.  In any case, once the system is fitted for three pedals, it'll be easy enough to add the two-stage behaviour as an option, through modifications only to the pedal unit itself.

So that's all the explanation for why it "doesn't hurt" to go back to double machine pedals, but the actual *reason* is that the forces and frictions, at least the way my current stops action is designed, will work out much better this way.  The two-stage mechanism requires another layer of friction and spring force on top of the two "pulls" themselves.  The pulls already have tons of force, but I think they will operate smoothly and with a good feel, once directly connected to pedals with the right leverage.  I was having doubts about how clunky and noisy and possibly even squeaky the two-stage mechanism might be.  I had re-designed it several different ways, but nothing was quite sitting right about it, in my head.

The critical thing here is to have pedals which can be operated quickly and easily, without undue noise and without too "indirect" of a feel, even though the mechanism is totally indirect.  My model for the kind of pedal-use I want to support, is the virtuoso hammer-dulcimer players of the Eastern Block: they often have a damper pedal, which operates a simple pull-down damper on all the strings (i.e., normally-undamped).  They use this damper very quickly to accent and mute different parts of phrases, and there also seems to be a considerable linear range to the response of the damper, so they are able to obtain a range of tone colour from light damping (a change in harmonics but still with sustain) to very percussive highly-damped "plunks".  I want my buff stop to be able to do all of this, so that it can be used in exactly the same way.  I think the double pedals will enable this much more readily.

springs

After a number of attempts (resulting in a sizable "graveyard" of scraggly wire bits), I think I finally have the springs working, for the machine-pedal aspect of the stops mechanism.  It was difficult to find the right strength for the springs, to overcome the considerable (intentional) friction in the stops motions, without presenting so much force as to make pedal operation unpleasant -- or simply, so much force that my simple wooden mechanism would self-destruct.  And, it was difficult to fit the necessary springs into the cramped and irregular volume of space available for this part of the mechanism.  Here, then, is what I came up with...

First, I made models using thinner wire (#18 steel).  This is a crucial point; trying to work and re-work an already crooked and kinked-up piece of #12 steel wire, to snake around all the obstacles and fit into the right place, was an exercise in frustration which I repeated several times in previous attempts (I have already mentioned the "graveyard").  Once I realized I needed to fit a *pair* of springs into the same space, for each of the four ends of rails which needs a spring, I knew I had to get serious with the modeling.

Here are two of the four final pieces, in #12 steel wire.  You can see that I have marked left and right (one or two bands), because they are not interchangeable.  They're *roughly* mirror images, but each is shaped to fit into its exact situation.

Here you can see how the pairs of springs interlace, in-vitro.

You can see how limited the clearance is, between the nameboard and the wrestplank (it's about 3").  I've had to substantially refine the design to fit into such a space: the original idea would have required more like 12" of clearance behind the panel (so it obviously would have had to be placed in a very different way on the instrument, not on the nameboard).

various refinements to stops mechanism; major redesign of damper system

As you can see, I added front guides to the "lower units", as I've taken to calling them.  I had front guides in all my drawings of this mechanism as I developed it, over several years; but as I built it this first time, I ended up not fitting front guides at first, and then I thought I'd see if it could survive without them: simpler.  But no.  The spring force from the rollers below tends to drive the lower units upward until they rub against the lower machine rail.  I had to grind away some of the inner surfaces of the machine rails, so they wouldn't rub against the shafts of the stop levers (overall, this mechanism should be built with a tad more clearance, next time); but grinding wouldn't help here, the spring force would just push the lower units higher until they contacted the new surface.  Nothing else in the mechanism provides the necessary stiffness and force to resist this.  Fortunately, though unplanned, it works out that exactly two thicknesses of my plywood (i.e., the guide itself and then a plywood mounting block) is just the right depth, so that the end of the lower unit dowel is just retained in its full travel, but the nearby fourth actuator-wire hole is not blocked by the guide.  Thus, the dowel-holes of the front guide do not have to be notched, like the ones on the back guide, to clear the actuator wires.  So, they have more "guiding ability" -- which is important because the most-crucial top of the circle is necessarily notched anyway: there's not clearance for a fully-closed circle on top, it has to be cut off, again because of my too-tight clearances.  Anyway, the guides, such as they are, seem to help adequately.

The other change, to not-yet-implemented aspects, is that I have decided to put 2:1 levers in between the machine rails and the sliding trapwork with the inclined plane, which moves with the pedal.  The force on the rails will be quite large, and therefore very strong return springs will be needed.  Some mechanical advantage from the levers will help the sliding planes, which I call the "hills", do their job more smoothly.

---

Meanwhile, I've benefitted from this long break in working on the keyboard action itself.  As often happens, my mind churned over a problem that was bothering me, and somehow it arrived at a solution "in the background", as my conscious focus was on other things (the stops).  The problem was, too much friction in the damper system (with the springs and monofilament going over "pulleys"): I had assumed that monofilament making a close 90-degree bend over a smooth metal surface, was close to frictionless, but far from it.  I could possibly have mitigated the issue by using actual, moving pulleys, but then things just got 98 pulleys more complicated, where I thought a straight piece of wire was going to do.  And, 98 pulleys of unknown design, from an unknown source.  Me, fabricate?  All this bothered me.

But then I realized a better solution, going back to something I had discarded before, is to use levers.  There is not room for damper levers above the keylevers, as in some early designs of mine; but there is room below.  The change works out to be smoother than one might think.  Same damper guides in the upper touch rail.  Same monofilament adjustment screws in the tops of the keylevers.  But instead of springs and pulleys, the monofilament directly lifts up the front end of the damper lever.  The back end goes down, by a smaller amount (about 2:1), taking the damper wire down with it.  The weight imbalance of the asymmetrical lever is, hopefully, enough to return the mechanism by itself, without added weights.

Not only is this likely to move with less friction felt at the keyboard, but also it provides a nice low-force way to implement "sustain", by lifting the front ends of all 49 levers at once.  I have been worried that the force on the stops mechanism from pulling down 49 damper springs, was going to be too much and would make the machine pedal interface unworkable.  There are still many worries there, but this big issue is at least significantly mitigated with these gravity-levers.

starting on the stops mechanism

Quite a bit of fabricating of parts for this mechanism; it's certainly more complicated than what traditional harpsichords have had for stops control (a simple lever at best, or just "reach in and push this part over a little").  But for the flexibility it provides, and particularly given the relatively small space the mechanism had to fit into, I am proud of how simple the design is, actually.  I've had this on paper for several years now, plan to use it on all my keyboard instruments, but this is the first actual test.

The one slightly non-trivial part was making the slots in the machine-rails; these are of poplar.  I was able to "drill-route" them using a sharp drill bit on the drill press, by drilling holes with diameters that intersect, then grinding away the rough sides that remain using the same drill bit, carefully working until a channel is produced that the drill can slide freely through from end to end.  Results look quite crude and home-made, but functionally it works fine, and perhaps the rough sides give a better grip on the screw eyes or something.  (These rails could also have been built out of strips and blocks glued together, like the way some harpsichord registers are made, if one wanted to avoid my non-Kosher sideways drill use and still do it with minimal tools.)

Guides for the rails, and several other structural supports, are glued to the back of the nameboard, which forms the basic chassis for this whole mechanism.  The guides, 1/4" dowels, are glued into shallow drilled holes.  The rest of the blocks are simply glued to the surface.  Although I tried to mask-out some areas where I thought I'd be gluing, turns out in every case I had to sand and scrape away paint and finish to place each new block as I went, and the masked areas never got used.  Oh well!  (This is the cost of pre-finishing parts as early as possible, on an evolving prototype design, rather than delaying finishing as long as possible.)

The stop knobs have to be threaded through first, and their screw-eyes installed, ahead of everything else, or else they can't be reached conveniently.

The rails slide into place over their guide-dowels...

When a stop knob is turned to the left or right, its screw-eye engages a slot on one of the rails.  The rails are moved by the machine pedal.  When the stop knob is pointing straight up, its screw-eye misses both rails, and thus the stop is in "manual" mode.

The backing plywood, and all the additional parts of the mechanism, are held with sheetrock screws, so everything can be disassembled at will.

The stops knobs were initially hanging down in a position past their normal full-out travel position, with the screw-eyes resting against the nameboard.  Here I have pushed them in slightly and turned them to engage with one of the machine rails, which are themselves resting in the full-out position with gravity.  Thus, the back ends of the stops dowels are now visible at the surface of the plywood guide holes, in their correct full-forward position (where I can reach them to attach the lower portions).

The lower dowel portions, I don't have a good technical name for.  The upper dowel turns with the knob, this part is rotationally stationary but it slides in and out with the knob.  Anyway, this portion, each one I have drilled with four holes so that the order of the stops can be re-arranged.  (Although I have already chosen the actual order of the knobs, now determined by the names painted on the nameboard, the ordering of some of the back-end mechanism may still change.)

The spring-force of the wire actuator arms on the rollers, along with the springiness of the thin 1/4" dowels themselves, gives each stop a gentle but positive "click" action, where it wants to be fully-drawn or fully-pushed-in, but not somewhere in the middle.

Here the four left stops are pushed in...

...Here they are pulled out.  The rollers turn through a small angle, which will be translated to pull-force by wooden arms at the outside ends of the rollers (not constructed yet).  The length of the arm will determine leverage of course; in most cases, the motion will be reduced 1/2x or 1/4x from the knob travel (which is about 1/2").

Here I've added the brackets which carry dowels for the machine-pedal carriage mechanism, not yet constructed.  The dowels slide back and forth with the motion of the pedal (transferred by bicycle cable).  Elements attached to the dowels carry inclined planes (the "carriage"), which along with rollers cause the rails to be pushed forward in sequence at the proper point in the pedal travel: first the top rail, then it is joined by the bottom rail.  Strong spring force pulls back the rails, and also the pedal carriage.

Here you can see that I have used the re-arrangeability of the rollers, to put stop #8 on the back roller, out of the sequence followed by the other seven stops.  The 8th stop is "sust", and it needs to reach downwards to control the damper-lift, which is more convenient to do from the back roller; most of the other stops will pull cords or levers to the back and above their corresponding rollers.

With all the knobs turned to the right, all stops are engaged with the lower rail.  Here, they are all in the pushed-in position...

...When I manually push the lower rail forward, all the stops move forward to the drawn position.  The difference is highlighted by the unpainted-but-finished section of the dowels, which becomes visible.  I may paint this region red, or something.

nameboard and stops design

Well, now it's official, I have to finish this instrument by the end of this year!

As you can see, I have started fabricating the front panel (nameboard) and other elements related to the stops mechanism.  In order to do the lettering, I had to finally decide on the order of the stops:

8va sitar bsoon buff mod mute1 mute2 sust

...which is somewhat determined by the mechanical constraints, but turns out to be a good ordering from a functional point of view as well.  This gives something of a tonal continuum, left to right, from "buzzy" to "muffled".   Also, all of the stops which modify the tone after initial excitation, i.e., the "mutation" stops, are to the left.  On the right side are the controls which do other things than tone mutation: the moderator affects the initial production of tone, the mutes bring on silence.  Originally I had planned to make these mute stops "pull for sound", but now they will instead be "pull for silence".  This is mainly to make them work in a more-useful manner with the machine pedal (the left pedal): as the pedal is pushed down, adding sonic effects (which, generally-speaking, tend to reduce the total volume of sound), it is more likely that one will want to remove an active string from the pair, rather than add one.  Especially given that some of the effects (sitar, 8va) only affect one of the two strings in a pair: the first pedal position (part-way down) can introduce the effect while leaving both strings active, and then the second pedal position (all the way down) can mute the other string which doesn't have the effect, thus causing the effect to become much more prominent.

The stops mechanism will be attached to the back of the nameboard, fitting into the 4 inch space between the nameboard and the front of the wrestplank (usually nameboards are mounted right up against the wrestplank).  The entire assembly will be held in place with two screws, not glue, so that it can be easily removed, analogous to the keyboard and action.

Pulling the knobs will activate organ-type rollers which transfer the motion to the sides (four to the left, four to the right), where attached actuator arms will pull control wires forward through a distance of about 1/2" (3:1 leverage from the knobs).  Each stop mechanism will have a low-force spring return (low enough force as to not override the stronger spring-induced "click" which seats the stop knob into either the fully-pushed-in or the fully-pulled-out position.

The stop knobs can be rotated as well as pulled.  When the knob is pointing straight up, in the vertical position, the stop is in manual mode, and it can be pushed or pulled freely.  Turning the knob to the left engages the stop with pedal position 1, and to the right engages with position 2.  When stops are engaged to the pedal, they can no longer be moved by hand; instead, as the pedal is pushed down, first all the stops engaged to position 1 are moved outwards; as the pedal is pushed further, the stops engaged to position 2 also move outwards.  Releasing the pedal retracts the stop knobs, in the corresponding reverse order.  (Stops can be disengaged from the pedal, by rotating the knobs back to vertical, at any time, whether the pedal is down or up; the stops stay in whichever state they were in at the time they were disengaged.)

One issue yet to be figured out, is how to manage the spring force inevitably associated with the sustain mechanism (damper-lift).  Unlike all the other stops, which don't have an intrinsic spring force and which will have specific return springs added, the sustain stop has a relatively large amount of spring force to act against: it has to push all the damper springs down simultaneously, thus lowering the dampers away from the strings (in my unusual, dampers-from-underneath design).  This force is no problem to handle with the leverage of the dedicated sustain pedal (the right pedal).  However, in keeping with my gadget-centric design sensibility, I want to also treat the sustain function as "just another stop", susceptible to the handstop and machine-pedal mechanism like all the rest.  (Initially I avoided this conclusion, but at the very least I wanted to have a handstop which allows the dampers to be permanently disabled without having to hold the pedal down: this, after all, is the primary mode associated with historical pandalons (more often spelled "pantalons"), as well as with the hammered dulcimer itself, the underlying inspiration for this type of instrument.)

Having the sustain stop susceptible to the machine pedal, would seem to imply that the "click" force of the sustain handstop will have to be strong enough to hold the stop in the out position, against all the damper spring force.  I don't yet have a clear vision of how this will work.  I suspect the "click" will have to be too strong.  Using a mechanism involving an inclined plane which "plateaus" can eliminate the back-pull of the springs when the stop is fully out, but still something will have to provide all the force to move the knob in the first place.  Anyway, no clever solution here yet, but I'm seeking one...

keylevers...

I have started fabricating the keylevers and associated parts, which involves much repetition at rates of 49 or multiples thereof.

Unlike in my three-key prototype, on the real keys, I have decided to use little blocks to hold the striker spring wires.  On the prototype, these wires made a 90-degree bend and were seated into shallow holes in the keylever itself, but this has proven unsatisfactory: a firmer seating is needed, and I don't like drilling deeply into the already precariously thin and light keylevers.

The blocks have a horizontal hole drilled through, end to end.  The spring wire will seat into the back end of this hole, facing the striker.  The hole on the front face may be used for an optional "secondary touch" mechanism which I have designed; I won't be sure whether the mechanism is needed until I play the instrument for a while.

I made this little jig to hold the blocks in position for drilling.

I have gradually developed a technique for obtaining reasonably-accurate hole positions, despite the remarkable tendency of small drill bits to wander, and despite what seems like a large amount of play in my drill press bearings (are they all like that, or only the dirt-cheap ones from China?).  The key is to have a well-placed starter mark, which I make with an awl; and then I position the drill (with the power off) with its tip centered in the starter mark, and *then* turn on the power.  Because I often need one hand to help hold the piece, and the other hand to hold the drill press lever, I installed a wire-tie hanging off the end of the power switch, so that I can pull it with my teeth and thus turn on the power when my hands are occupied.  Potentially dangerous, but effective!  (Given how much easier it now is for something sweeping past to snag the wire tie and turn on the drill, I'm now much more careful about pulling out the little plastic "ignition key" which disables the switch, when I'm not using it.)

I thought I'd have to rig a complicated and tedious clamping system to glue the blocks onto the keylevers.  However, the Titebond II sets up so fast on these dry, porous, absorbent basswood joints,  that I found I could just hold the joint firmly by hand for less than a minute, and the parts would already be firmly stuck together.  Then I'd just let them sit for an hour or so, and good to go.  Indeed, one block I noticed I wanted to move, after only about two minutes of drying, and I just *barely* got the pieces apart in time; it took quite some force to break the bond.  Very much longer, and it would have been the wood that was breaking.

Here are the 49 keylevers in the frame.  I will next attach the plywood keys (but not their ceramic and glass coverings, yet).  Then I will do the painting and finishing I have planned for the levers: black paint on the keys, clearcoat on the rest of the levers.  Then I will finally attach the guide-wires, spring-wires, and felt pads for the strikers (thus these items will not be exposed to the clearcoat).

I have found rectangular glass tiles with a white background, which are the perfect shape and surface quality for the sharp keys.  However, I'm still searching for the right solution for the natural key coverings.  I am convinced that ceramics *may* be a good key covering; but I've been unable to find pre-made tiles which fit the bill, despite many "almosts".  So I am now planning to try making my natural keys out of clay, myself -- for the second time.  I already tried making the naturals from "Della Robbia" oven-bake clay, but this type of clay will clearly not be durable enough for this application.  So now I will try "real" clay, fired in an actual kiln at 2000-something degrees F, with glossy black glaze.

framed for posterity

Done with the basic keyboard frame, which gives a good chance to view it and contemplate its basic design, before I add a whole lot of other parts and mechanisms (such as the dampers) which will tend to obscure the basic simplicity of the striking mechanism.

As you can see here, there are three crucial rails, plus the decorative/protective black front rail.  The fulcrum rail has little blocks of wood which support the keylever axle wire, and also it has a "comb" of steel nails, which guide the keylevers (these unusual details, instead of the usual center "balance" pins underneath the keylevers).  The lower touch rail, farther back, has the other "comb" of nails, completing the side-to-side guide system for the keylevers.  The upper touch rail has no nails, but it has holes drilled through which will guide the damper wires (as seen on my 3-key prototype).

Here the frame is attached to the instrument (which is resting sideways, on its spine edge).

Next up, a whole lotta keylever fabrication, using the lessons I've learned from the prototype...