PCB Thickness: What to Expect

[EPAIII]

Tried and succeeded? Well, NO.

I am working on it and it may be possible. But there are difficulties. I probably could solder ball bearing balls made of other steel alloys, but those would be prone to rust and I do not want that on switch contacts. So, stainless steel seem to be the way to go. The soldering fluxes that are in common use in electronics just will not work on stainless steel. You can use all the flux you want and all the heat and you are still left with a ball of solder that did not adhere to the stainless steel.

There is a flux that is said to work, but it is a strong acid. So, I would not want that anywhere near a PCB. Nor would I want it on the area of the SS ball that will be making contact with the opposite SS ball. So, how do I use that acid flux on about half of a 4mm SS ball to allow me to tin that half while preventing the acid flux to the other half. Oh, and I would not want any solder on the other half, just on the half that will bond to the PCB.

So, I am experimenting with techniques. So far, nothing that I can actually use. I purchased a solder pot and a couple pounds of solder. What I need is a way of coating 1/3 to 1/2 of the balls with something to keep the acid flux and solder off. Then dip in flux, followed by a dip in the solder pot, cleaning ALL the flux off, soldering them to the PCB (technique unknown), and finally removing that flux and solder resistant coating, perhaps with a solvent of some sort.

I am also looking at conductive epoxy. But I haven't purchased any to try yet. All around, it may be a much easier way to do this. And a few extra Ohms will not hurt my circuit.

I still can't get over the idea of trying to solder ball bearings.  You've already tried and succeeded, yes?

[amyk]

Show me just one switch where the manufacturer GUARANTEES activation AND de-activation (connect and disconnect) within a stated mechanical range that is under 0.001"/0.025mm and at a cost under, say 5 USD each (qty of 100). And that is guaranteed to maintain that same positional accuracy over thousands (preferably tens of thousands) of cycles and periods of months (preferably years). Heck, most switches deliberately incorporate some form of hysteresis. And I am trying to avoid any amount of that.

Actually, one thousandth of an inch is the absolute LARGEST range I hope to achieve. I really want ten-thousandths of an inch (microns) over those conditions, if at all possible. And these numbers come from the basic purpose of the device I am trying to make.

A digital caliper ASIC with some modifications to switch on specific readings is probably within the range of precision you're after, and 5 USD is definitely more than what some cost (in qty 1).

[Microdoser]

I still can't get over the idea of trying to solder ball bearings.  You've already tried and succeeded, yes?

Not only that, but to within those tolerances too. This would introduce yet another tolerance issue (and tolerance issues stack up rather than cancel out except in rare situations) of finding ball bearings with finer tolerances than you need, and soldering them in accurately enough that their final resting place is still within those tolerances.

I can't help but think this physical electrical engineering problem is one that might be better fixed by using a non-contact detection method and some tweaking in software.

[ajb]

There's no need for the positions (or sizes) of the balls to be so accurate.  The three-pins-on-three-pairs-of-balls arrangement is kinematic or critically constrained: there is exactly one position that will satisfy the system regardless of how accurately the contact components are positioned (as long as the errors aren't so egregious that they can't make contact at all).  Reasonable errors in positioning the balls or pins relative to each other will cause an error in the position of the stylus relative to the probe body, but again: that is what the adjustment screws are for.  The probe kinematics just need to put the stylus within a millimeter or so, and then you use the adjustment screws and a dial test indicator to adjust the runout (on a work probe) or perpendicularity (on a tool setter) of the stylus tip to as many decimal places as you want. 

I probably could solder ball bearing balls made of other steel alloys, but those would be prone to rust and I do not want that on switch contacts. So, stainless steel seem to be the way to go.

Carbide is typical.  Rigidity of the contact elements as well as the rest of the kinematic chain is more important to the performance of the probe than initial accuracy of the contact positions.  Imagine jacking up a car: the wheel doesn't immediately leave the ground with the first crank, because the weight of the car has compressed the suspension and flattened the bottom of the tire against the road.  So as you crank the jack up, you have to relieve that initial strain in the system -- the suspension has to fully expand, and the tire has to relax into its normal round form -- before it breaks contact.  Switch between front and rear wheels, and it will probably take a different number of cranks, because the weight distribution and suspension characteristics (and maybe even the tires) are different, so there is more or less strain in the tire/suspension to relieve before the wheel leaves the ground.  Whether the car is initially sitting on a flat surface matters a lot less (as long as it's not going to roll away on you...).   

The same thing happens in a probe like this, although at much smaller scales.  Under preload from the spring the mating surfaces of the pins and balls will conform to each other just like the tire conforms to the road, and the pins, probe stylus, body, etc all deflect a bit just like the suspension does.  If there is more strain at one of the contact sets than the others, because of differences in spring pressure, or nonuniform rigidity in the system, this will result in different actuation travels in different directions.  Ideally, the probe mechanism, not to mention the machine, would be perfectly rigid and any amount of deflection at the stylus tip would separate the probe contacts and be detected.  That's not possible in the real world, but a hard material like carbide that resists deformation under point contact is preferable for the balls/pins because it means less overall strain, and less variation in strain with load/position changes.  The rest of the system should be reasonably rigid as well, and most importantly it should be uniformly rigid so that the amount of deflection required at the stylus tip to part the contacts is consistent in all directions. 

It is possible to null out some residual errors via the chosen probing strategy or calibration method.  During a measurement, a longer actuation travel in one direction makes the stylus appear smaller in that direction, so a strategy that compensates for runout compensates for that as well.  A lot of that is baked into the standard probe/setter calibration routines from Renishaw et al, but getting the probe right to begin with so the error is consistent in all directions is of course better. 

I can't help but think this physical electrical engineering problem is one that might be better fixed by using a non-contact detection method and some tweaking in software.

Physical contact is the simplest way to achieve these levels of precision and repeatability in tool and work probing.  There are laser-based tool setters, but the optical and optoelectronic aspects required to get into the range of precision and repeatability that's required here are not trivial.  They have advantages for very small tools (<1mm), but there's a good reason that 99% of tool and work probes use the basic mechanical contact mechanism the OP is trying to replicate.  Work and CMM probes that can do surface profiling use strain gauges or capacitive sensors to provide an analog output, but they still use a stylus to physically contact the surface.  Anything else is too vulnerable to confounding factors external to the probe.  A work probe has to deal with all kinds of shapes and a wide range of materials -- steel, aluminum, plastic, composites, etc.  A tool setter can count on dealing with things that are roughly radially symmetric, but still a range of materials -- tool steel, carbide, bonded abrasive, diamond, etc -- and it needs to accurately detect the very outermost edge of a thin, sharp cutting surface, often with complex facets.  So all in all would be hard to come up something that would be simpler than mechanical contact unless you can constrain the material or shape quite a bit. 

[ajb]

Oh, and as far as soldering: if the balls are a suitably corrosion-resistant material to make for a decent switch contact with the pin, then why not use that to your advantage and remove the PCB from the mechanical equation entirely?  Fit the balls directly to a mechanical frame and use spring contacts on the PCB to make contact with them.  The frame could be something nonconductive (glass reinforced plastic, ceramic if you want to get fancy?) but a metal frame wouldn't be hard either: just use a nonconductive epoxy loaded with glass spheres (to enforce a minimum bond thickness) to bond them in.  Or heck, if the frame is aluminum, get it anodized.