PCB Surface Finish for Switch Contact

[EPAIII]

I am working on a design where two PCBs will move in relation to each other in order to form a switch. This will be a change in distance apart only: there will be no sideways motion so no wiping action. Just contact and no contact. My goal here is to have a switch which will possess a high mechanical accuracy (+/- 0.0003" or better)(for the metric types here, that's +/- 0.0076 mm or better) over a long period of use. As far as I can find, there are no reasonably priced, commercially available switches with that degree of accuracy. But if you know of one for under US$ 1.00, please let me know.

I am considering other designs, but for this question the following applies. The contact between the boards would be a 3mm, stainless steel ball which will be constrained at one location relative to the PCBs by an insulator. The contact force will be as small as I can make it in order to prevent distortion of the PCB traces or the stainless steel ball. Hopefully this will be only a few grams. The stainless steel ball will also be constrained against any rotation by a friction fit in that insulator. I am considering stainless steel because of it's resistance to corrosion and the ready availability of high precision balls of that material at a very reasonable price.

My question is about the advisability of using an ENIG (Electroless Nickel/Immersion Gold) finish on the areas of the PCB where contact will be made. Of the finishes listed by most PCB fabrication houses, that one seems to be the best choice. But, will it work when used with stainless steel and in the long run?

The other ideas that I am looking at are the use of two of these stainless steel balls, one on each PCB. The big problem there is attaching stainless steel to a PCB pad. Acid flux is one choice but it would require extensive cleaning afterwards. Another I have looked at is a conductive adhesive. This would have more resistance, but my circuit can tolerate that. But I do not know about the long term reliability of such a bond. A third possibility would be a friction fit in a plated hole. But hole plating on a PCB is very thin and the specs for the finished diameter are not very tight. So getting a good contact that lasts for a long time (years) seems to be a very unlikely thing. Eyelets perhaps?

The positional accuracy of the balls in any of these alternatives would be ensured by a second PCB, with no traces, attached to the rear side of the primary PCB. The hole in the main PCB would be a bit larger than the ball size so it would rest on that back board while being installed.

I am open to any other suggestions.

[mikeselectricstuff]

ENIG is extremely thin - you probably need a hard gold plating as used for edge connectors.

[SeanB]

Make a larger hole, that fits the ball so it is at 60% of diameter, and use the conductive epoxy from the rear. Hole is plated through, and you would probably want to have a few vias to complement the big plated through hole anyway. Solder mask pulled well back at the rear, and a large plane, probably double the diameter of the ball, and the board under ENIG plated.  The vias untented, and 1mm diameter, so that the epoxy can also flow though them as well to provide extra mechanical anchor. Should hold the ball well, and the forces will tend to hold them in position. Prep for the balls would be a rough grind on the rear side, to provide a keying surface, and a clean in a solvent for both the ball and board before insertion with the rough side to the rear.

[moffy]

PCB laminate is not the right material to build precision switches, +/-0.0003" repeatability, because it warps and is not flat. When using CNC to try and mill traces, a probing operation is generally necessary first to create a height profile so that the milling depth would be the same over the whole board, the displacements can be quite significant.

[mikeselectricstuff]

PCB laminate is not the right material to build precision switches, +/-0.0003" repeatability, because it warps and is not flat. When using CNC to try and mill traces, a probing operation is generally necessary first to create a height profile so that the milling depth would be the same over the whole board, the displacements can be quite significant.

Aluminium PCB might be a better bet

[ajb]

Is initial accuracy important, or just repeatability?  That distinction is important.

There are solder-on contact pads meant to be used as more durable targets for spring contacts that could be useful.  Basically just little flat pads of plated metal that come on a reel.  Initial accuracy would depend on the soldering process of course. 

For making contact with the stainless ball, a PCB-mount spring contact could be a good option.  Lots of parts to choose from with different contact orientation, sizes.  Widely sold for battery contacts or to electrically connect a PCB to a housing for EMI control.  Should be possible to arrange two or more symmetrically to avoid spring pressure influencing position, if that's a concern. 

[EPAIII]

I think I have the warping covered. The boards will be relatively small, around 1.25"/35mm rectangles. The main and back boards will be combined with a good epoxy after roughing. This should produce a plywood like structure to resist warping. I think FR-4 may/will have some initial warp, but I don't think it will change over time. So once the device is assembled and calibrated it should be stable. Also the overall configuration of the device will help make any remaining warping in the swithch assembly irrelevant in the device's operation.

I would like to use PCBs for some of the parts due to their ready availability and relatively low price. I am trying to work around their undesirable characteristics. I will also be using Send-Cut-Send for some of the flat metal parts. Again they will be easy to acquire and relatively inexpensive.

In any case this is experimental. I will build several prototypes and use at least some of them for long term testing. I will see how well it works.

PCB laminate is not the right material to build precision switches, +/-0.0003" repeatability, because it warps and is not flat. When using CNC to try and mill traces, a probing operation is generally necessary first to create a height profile so that the milling depth would be the same over the whole board, the displacements can be quite significant.

[EPAIII]

Just repeatability. The initial accuracy will be established with a calibration procedure.

I have considered the spring type contacts. And I still am. But I am looking into several options.

I am not aware of the contact pads. They sound interesting. Can you provide any further details: who makes them, what is a good search term for them, etc?

Is initial accuracy important, or just repeatability?  That distinction is important.

There are solder-on contact pads meant to be used as more durable targets for spring contacts that could be useful.  Basically just little flat pads of plated metal that come on a reel.  Initial accuracy would depend on the soldering process of course. 

For making contact with the stainless ball, a PCB-mount spring contact could be a good option.  Lots of parts to choose from with different contact orientation, sizes.  Widely sold for battery contacts or to electrically connect a PCB to a housing for EMI control.  Should be possible to arrange two or more symmetrically to avoid spring pressure influencing position, if that's a concern.

[MarkT]

Not sure stainless is the right choice, it won't solder easily to say the least.

Why not extract the contacts from a cheap relay, they come conveniently mounted on brass or beryllium copper strip which will solder...

[EPAIII]

I do intend to price aluminum PCB as one option. I have not purchased any aluminum boards and have no idea of how they compare in price to FR-4. It seems to me that extra steps would be needed in their fabrication and that would mean a higher cost. Also the relative volume of the two types would come into the equation.

On top of all that, I am hoping to panel-ize the PCBs for this project in order to keep the price down. But one of them should be a heat insulator. Aluminum does not seem to meet that requirement. I would guess that using aluminum for some of the PCBs and FR-4 for the others would be an additional increase in the cost.

Again it is the repeatability that really matters, not any initial error that can be removed in the initial calibration. As I said, I am trying to work with the characteristics of the materials. And I am considering other designs. I am only asking about one aspect of one of those possible designs in this thread.

PCB laminate is not the right material to build precision switches, +/-0.0003" repeatability, because it warps and is not flat. When using CNC to try and mill traces, a probing operation is generally necessary first to create a height profile so that the milling depth would be the same over the whole board, the displacements can be quite significant.

Aluminium PCB might be a better bet

[EPAIII]

It is difficult to solder SS to copper traces! Wow, you think.

I knew that from square one and, in spite of that, have already already invested dozens of hours reading about it and well over $100 in experimenting with it. Yes, it can be done, but it is not easy. And an acid flux is probably absolutely necessary. I have made a preliminary judgment that it would take an inordinate amount of time and would probably leave some acid residue on the PCB. I probably will not be going that route.

As for salvaging relay contacts, I have thought of that. But here's the thing, relays are designed to close with a small wiping action of the contacts. That wiping is there to remove any corrosion on them. It would be difficult to install them in my device while keeping that wiping action.

Not sure stainless is the right choice, it won't solder easily to say the least.

Why not extract the contacts from a cheap relay, they come conveniently mounted on brass or beryllium copper strip which will solder...

[moffy]

I don't know what your constraints are but I would prefer a non contact method, though that adds to the complexity, specifically capacitive, which can be quite sensitive as its sensitivity increases as 1/d.
https://www.mtwmag.com/innovative-capacitive-sensors-for-industrial-use/
Not that you use their sensors, but the electronics is straightforward. It's just an example of resolution but the approach might be counter to your search for a simple low cost solution.

[EPAIII]

I had not mentioned it, but one of the biggest constraints is battery powered. With a low current LED as the indicator, I hope to use a single, 3V, coin cell, probably a CR2023. And I would like it to last for a long time, a year or more, between changes. I know that does not completely eliminate an active sensor, but it does make it difficult. It might even put in the realm of a custom IC.

I am going to check out your link.

I don't know what your constraints are but I would prefer a non contact method, though that adds to the complexity, specifically capacitive, which can be quite sensitive as its sensitivity increases as 1/d.
https://www.mtwmag.com/innovative-capacitive-sensors-for-industrial-use/
Not that you use their sensors, but the electronics is straightforward. It's just an example of resolution but the approach might be counter to your search for a simple low cost solution.

[EPAIII]

Well, I went down that rabbit hole. Never did find a single Micro-Epsilon product where it could be purchased so could not get even the slightest idea of cost. But their stuff looks really nifty ($ $ $ $ $ $).

Did a search at DigiKey for capacitive sensors. The least expensive one was $12.90. That's more than ten times what I want to spend. I looked at the DigiKey listing but could not find either an accuracy spec or a data sheet. I guess you buy one and play with it until it does something. Sounds like a super high price for what is probably a metal plate surrounded with a ground ring. Perhaps it has some built in electronics. I don't know. Oh, but it IS ROHS3 Compliant. Whoopee!

Their next sensor, by price costs $52. That's about what I would like to sell my entire device for. NOT MAKE IT FOR, SELL IT FOR! Perhaps they are flexible on the price. Or maybe they have coupons.

And don't tell me I am crazy. The Chinese make a micrometer that can measure to 0.00005" / 0.0013mm, ship it across half the world, and it still sells for around $40 with a profit for both the Chinese factory that made it and the importer who sells it here. They don't do that by buying parts from others at $12.90 each, allowing those other companies to make a tremendous profit. I am looking for the <= $1 "solution" (I hate that word but it fits).

I tried Mouser but they seemed fixated on turning the room lights on and off.

[moffy]

I had not mentioned it, but one of the biggest constraints is battery powered. With a low current LED as the indicator, I hope to use a single, 3V, coin cell, probably a CR2023. And I would like it to last for a long time, a year or more, between changes. I know that does not completely eliminate an active sensor, but it does make it difficult. It might even put in the realm of a custom IC.

I am going to check out your link.

I don't know what your constraints are but I would prefer a non contact method, though that adds to the complexity, specifically capacitive, which can be quite sensitive as its sensitivity increases as 1/d.
https://www.mtwmag.com/innovative-capacitive-sensors-for-industrial-use/
Not that you use their sensors, but the electronics is straightforward. It's just an example of resolution but the approach might be counter to your search for a simple low cost solution.

Unless you can accept a really low duty cycle, which doesn't make sense for such a high resolution sensor, I can't see an active sensor lasting a year on a CR2023 either.

[mikeselectricstuff]

I had not mentioned it, but one of the biggest constraints is battery powered. With a low current LED as the indicator, I hope to use a single, 3V, coin cell, probably a CR2023. And I would like it to last for a long time, a year or more, between changes. I know that does not completely eliminate an active sensor, but it does make it difficult. It might even put in the realm of a custom IC.

I am going to check out your link.

I don't know what your constraints are but I would prefer a non contact method, though that adds to the complexity, specifically capacitive, which can be quite sensitive as its sensitivity increases as 1/d.
https://www.mtwmag.com/innovative-capacitive-sensors-for-industrial-use/
Not that you use their sensors, but the electronics is straightforward. It's just an example of resolution but the approach might be counter to your search for a simple low cost solution.

Unless you can accept a really low duty cycle, which doesn't make sense for such a high resolution sensor, I can't see an active sensor lasting a year on a CR2023 either.

Depends entirely on duty cycle - one advantage of a physical contact sensor is you can have near-zero standby current if you only need to wake on a change of state, and near-instant (few microseconds) wakeup.

Even something that periodically wakes to do sensing is entirely doable on a 2032 if you don't need low latency - the biggest factor is how long you need to stay awake. There are  plenty of MCUs with suitably low standby current.

[mikeselectricstuff]

Might help  if we had more info about the application - I'm guessing a touch probe for cnc?
 
Inductive proximity sensing might be an option - inverse-square law helps here, and this can be done with extremely low power draw - ping a tuned circuit and count the number of rings. I used this many years ago on a PIC for sensing metal vanes in a water meter, this drew something like 5uA avarage waking a few times a second

May be doable with off-the-shelf SMD inductors.

If you can use the contact event as opposed to proximity, peizo may be worth a look. or an accelerometer.

For physical contact, gold on both sides is probably the  only  reliable option. all the cheap ball-based motion sensors I've looked at use gold contacts and a gold-plated ball

Also think about whether you can be clever about power saving - e.g. in the case of a CNC touch probe, you can use motion ( accelerometer) to go to sleep when it hasn't moved for a few mins.

[moffy]

One last thought, if you had the contact point as a piece of piezo material you can get reasonable voltage output for low levels of deflection. That way you wouldn't need an electrical contact, you would need a solid stop to prevent damage to the piezo material.

[mikeselectricstuff]

Looking back at the first post, .0076mm is getting into the world of surface electrostatic effects, thermal expansion etc. I can't see physical electrical contact working for this except maybe in combination with some kind of lever to amplify the movement.

Simple contact with a piezo may have issues with hysteresis ang ageing, but maybe a piezo which is actively driven at its resonant frequency, and sense damping when it makes contact.

[ajb]

I am not aware of the contact pads. They sound interesting. Can you provide any further details: who makes them, what is a good search term for them, etc?

They're often listed alongside pogo pins and other spring-loaded contacts, since they're meant to be used together. Hopefully this search link will work: https://www.digikey.com/en/products/filter/contacts/contacts-spring-loaded-pogo-pins-and-pressure/311?s=N4IgjCBcoKxgTFUBjKAzAhgGwM4FMAaEAeygG0QBma%2BANgAYQBdIgBwBcoQBldgJwCWAOwDmIAL6SgA

Pairing one of those with a solid domed pin (example) may provide a better-defined contact point vs two flat pads, at the expense of more deflection of the contact due to higher contact pressure. 

Looking back at the first post, .0076mm is getting into the world of surface electrostatic effects, thermal expansion etc. I can't see physical electrical contact working for this except maybe in combination with some kind of lever to amplify the movement.

 

Those and several other challenges have to be factored into the mechanism, materials, and manufacturing choices for sure.  OP had a previous thread about a CNC toolsetter where some of this was discussed.  But it is doable with a well-designed and well-made mechanism -- commercially made CNC probe systems (eg Renishaw) claim repeatability down to <1μm for mechanical contact systems. 

[Conrad Hoffman]

Non-wiping switch contacts seem like a reliability problem unless both contacts are gold plated. You can make a very accurate hall effect sensor with (duh!) a hall effect device and two neo magnets. Place the magnets side-by-side with the fields opposing. This creates a very narrow transition region so the hall effect device switches suddenly. Little neo magnets are cheap but I haven't priced simple hall effect devices lately.

[mikeselectricstuff]

commercially made CNC probe systems (eg Renishaw) claim repeatability down to <1μm for mechanical contact systems.

How do they work? Actual contacts or something else like piezo or inductive  ( shuffles off to ebay to look for a broken one to take apart....)

[ajb]

commercially made CNC probe systems (eg Renishaw) claim repeatability down to <1μm for mechanical contact systems.

How do they work? Actual contacts or something else like piezo or inductive  ( shuffles off to ebay to look for a broken one to take apart....)

There are probes used for surface scanning that use capacitive or strain gauge mechanism for analog outputs (more useful on CMMs than CNC machines), but general purpose toolsetter and work probes are almost always mechanical contacts. This sort of mechanism seems to be what everyone does, but there may be other solutions out there: https://youtu.be/vJdcbhGa7Jk https://youtu.be/LXLAGprt3Is

[thm_w]

So OP are you building a tool setter? Why not link the previous thread for context.

There are tons of discussions and testing related to sensing endstops:
https://www.cnczone.com/forums/mechanical-calculations-engineering-design/149985-forum.html
youtube.com/watch?v=il9bNWn66BY
youtube.com/watch?v=2o8woJfTEVc
youtube.com/watch?v=XLa5PICnxpg

If you want to do better than that, as mentioned, you'd want some kind of mechanical advantage linking into a switch or sensor (opto, hall, whatever).
You don't need to buy a $12 part from digikey, you can use a 50c microcontroller to do capacitive or inductive sensing.

[moffy]

According to the second video thm_w posted:
at the 11:43 time mark a simple inexpensive microswitch had a standard deviation for repeatability of about 1um, problem solved.

[EPAIII]

I had not mentioned it, but one of the biggest constraints is battery powered. With a low current LED as the indicator, I hope to use a single, 3V, coin cell, probably a CR2023. And I would like it to last for a long time, a year or more, between changes. I know that does not completely eliminate an active sensor, but it does make it difficult. It might even put in the realm of a custom IC.

I am going to check out your link.

I don't know what your constraints are but I would prefer a non contact method, though that adds to the complexity, specifically capacitive, which can be quite sensitive as its sensitivity increases as 1/d.
https://www.mtwmag.com/innovative-capacitive-sensors-for-industrial-use/
Not that you use their sensors, but the electronics is straightforward. It's just an example of resolution but the approach might be counter to your search for a simple low cost solution.

Unless you can accept a really low duty cycle, which doesn't make sense for such a high resolution sensor, I can't see an active sensor lasting a year on a CR2023 either.

Well, yes it will be a low duty cycle. But with a current draw of under 2mA the CR2023 should last well.

[EPAIII]

Can't keep any secrets here. You guys are too smart. So yes, I am trying to do a type of probe. It only needs to indicate in one dimension. If fact, it should be strictly limited to one dimension.

Some of you are concerned about my accuracy numbers. I don't know what is possible. I am trying to find that out. Naturally I want the best accuracy possible. And the REPEATABILITY of the switch mechanism is a major factor in that. Thermal expansion is definitely in the mix. Surface electrostatic effects? I must look that one up. Is it a fancy way of saying capacitance? If what I have is just a DC circuit, does that matter?

As for all the talk about active sensors and sleep and wake-up times, this is a FLASHLIGHT circuit: battery, switch, and LED. That's it. I am not even sure I need a current limiting resistor as the internal resistance of the CR2023 seems to limit the current flowing in the low current LEDs I have found. And when the switch is OFF, zero current flows. I guess you could call that "sleep mode". I am not worried about those details. At the present I just need to build a repeatable switch. And one with a long life, if possible.

In spite of all that, I seem to have stumbled into a way of building that circuit that allows it to operate in multiple ways or modes while other, similar devices are limited to just one. That should provide a competitive advantage over the others, at least for a time.

Those contact pads look interesting. I had hoped they might come with a dome shape, but it looks like they are all dead flat. So I am back to the stainless steel balls. The pogo pins would be good except the radius at the tip (0.5mm) is relatively short. So, after a number of cycles against a gold pad, the gold on the pad would be indented quickly and I don't know what effect that would have. This seems to be validated by the fact that the gold plating on the pins is twice as thick as that on the pads.

As for those Renishaw probes, that is a WHOLE 'NOTHER LEAGUE. And their prices definitely reflect it. But then, in about 15 minutes of looking, I could not find even a single specification on them. If I had a product that was worth those prices, I think I would be shouting the specs from the rooftops.

Anyway I do appreciate the help. And perhaps I/we can come up with something that fills a different nitch.

I am not aware of the contact pads. They sound interesting. Can you provide any further details: who makes them, what is a good search term for them, etc?

They're often listed alongside pogo pins and other spring-loaded contacts, since they're meant to be used together. Hopefully this search link will work: https://www.digikey.com/en/products/filter/contacts/contacts-spring-loaded-pogo-pins-and-pressure/311?s=N4IgjCBcoKxgTFUBjKAzAhgGwM4FMAaEAeygG0QBma%2BANgAYQBdIgBwBcoQBldgJwCWAOwDmIAL6SgA

Pairing one of those with a solid domed pin (example) may provide a better-defined contact point vs two flat pads, at the expense of more deflection of the contact due to higher contact pressure. 

Looking back at the first post, .0076mm is getting into the world of surface electrostatic effects, thermal expansion etc. I can't see physical electrical contact working for this except maybe in combination with some kind of lever to amplify the movement.

 

Those and several other challenges have to be factored into the mechanism, materials, and manufacturing choices for sure.  OP had a previous thread about a CNC toolsetter where some of this was discussed.  But it is doable with a well-designed and well-made mechanism -- commercially made CNC probe systems (eg Renishaw) claim repeatability down to <1μm for mechanical contact systems.

[EPAIII]

While 3D printers are certainly ONE application, their layer thickness between 0.1 and 0.3 mm is hardly high precision. Yes, I would love to have a device that is useful for 3D printing.

But there are many other applications where MORE precision is an absolute must. Just look at those Renishaw probes mentioned above. And many of the applications are somewhere between the coarse resolution of a 3d printer and the devices that absolutely require something like the Renishaw probes.

And there are other considerations. Like the inconvenience of having wires running everywhere. Or the relatively large size of some of the devices. Some shops will want a device that can be moved from one place to another as needed. Many people do not want the trouble of constantly replacing batteries. Provision for over-travel: even my Chinese mill has switches with provision for over-travel. If you have a device with sub-thousandth accuracy, there is no faster way of destroying it's accuracy than running into a HARD stop. Ask me how I know.

I am not trying to make a device for ONE application. I want to make one for MANY, MANY applications. A Z axis, Swiss Army Knife. I want it to have ALL the features. And I want it to be extremely affordable. Oh, and of course dead nuts accurate.

Yes, I know I am asking for a lot. But if it was easy any of those guys would do it.

As for the previous thread, I have been on and off this project and I had forgotten that it was here. I searched and perhaps this is the one:

https://www.eevblog.com/forum/projects/pcb-thickness-what-to-expect/

I have done a lot of work on it since then and also had some week long absences from it. And my memory has never been my strongest suit. Ask any of my teachers, well any who are still around and able to answer.

I was concerned about hole sizes as related to the diameter of the SS balls and the board thickness. I was still thinking about soldering the SS balls at that point. Some experiments have turned me away from that. I have gone through a lot of work on just how to construct these switches since then. About a week or so ago I had a sudden thought that perhaps the SS balls did not need to be mechanically and electrically attached to the PCBs. They could float between two PCBs and not make a closed circuit until the set point was reached. But I am not an expert on the various finishes for PCBs so here I am back here asking for help. For knowledge that I don't have. And I appreciate the answers; all of them.

I am puzzled about the use of stainless steel. It seems like it should make a good, long term material for electrical contacts. Yet, I have seen few uses for it that way. Why? Perhaps it just falls in some middle ground where it is not best for high current/high power circuits and also not the best for low signal level circuits. Lighting an LED seems to be between those two extremes and they can't make switches and relays that are optimized for everything. But then, isn't SS used in things like flashlights and toasters and many other non-critical things? Perhaps it is a good choice.

So OP are you building a tool setter? Why not link the previous thread for context.

There are tons of discussions and testing related to sensing endstops:
https://www.cnczone.com/forums/mechanical-calculations-engineering-design/149985-forum.html
youtube.com/watch?v=il9bNWn66BY
youtube.com/watch?v=2o8woJfTEVc
youtube.com/watch?v=XLa5PICnxpg

If you want to do better than that, as mentioned, you'd want some kind of mechanical advantage linking into a switch or sensor (opto, hall, whatever).
You don't need to buy a $12 part from digikey, you can use a 50c microcontroller to do capacitive or inductive sensing.

[thm_w]

While 3D printers are certainly ONE application, their layer thickness between 0.1 and 0.3 mm is hardly high precision. Yes, I would love to have a device that is useful for 3D printing.

I don't see how that is relevant, did you read moffys comment?

You can find the Renishaw specs at the end of their technical documents, eg: https://www.renishaw.com/cmmsupport/knowledgebase/media/pdf/en/04c7251ac8c14277b20fbd540cc0da55.pdf

PH6M autojoint repeatability: 1um
TP6 uni-directional repeatability: 0.35, TP1: 0.5um

[donlisms]

I can't see any reasonable alternative to the normally-closed switch approach. The one I saw used three balls in a ring, each insulated from the others, with a spring holding the ring down against a contact plate.  A lever attached to the ring would be pushed sideways, letting at least one ball switch open at the instant of contact.  No choice but to be precise.  No gaps while a switch closes. No calibration. I guess the error would be determined entirely by how stiff all the parts were, and how much tension was required. 

I guess it would require a rethink for one axis.  And it requires an inverter of some kind, to close when one of the switches opens, but that doesn't seem difficult, especially in light of all the advantages.

[wizard69]

I am working on a design where two PCBs will move in relation to each other in order to form a switch. This will be a change in distance apart only: there will be no sideways motion so no wiping action. Just contact and no contact. My goal here is to have a switch which will possess a high mechanical accuracy (+/- 0.0003" or better)(for the metric types here, that's +/- 0.0076 mm or better) over a long period of use. As far as I can find, there are no reasonably priced, commercially available switches with that degree of accuracy. But if you know of one for under US$ 1.00, please let me know.

You are asking for a lot.   There are industrial switches that make use of Sapphire or Ruby tips but they are not cheap. well not $1 but cheap relative to some industrial switches.   I've used such to get less that 1 micron repeatability on some CNC machines decades ago.   They where a good solution at the time.

I am considering other designs, but for this question the following applies. The contact between the boards would be a 3mm, stainless steel ball which will be constrained at one location relative to the PCBs by an insulator. The contact force will be as small as I can make it in order to prevent distortion of the PCB traces or the stainless steel ball. Hopefully this will be only a few grams. The stainless steel ball will also be constrained against any rotation by a friction fit in that insulator. I am considering stainless steel because of it's resistance to corrosion and the ready availability of high precision balls of that material at a very reasonable price.

One of the biggest problems we had back then was the mechanical arraignment.  I couldn't possibly say if this will work and give you the accuracy needed.   I'd suggest though that highly polished and hard surfaces may be a big help here.   Also you need to protect from mechanical over travel.

My question is about the advisability of using an ENIG (Electroless Nickel/Immersion Gold) finish on the areas of the PCB where contact will be made. Of the finishes listed by most PCB fabrication houses, that one seems to be the best choice. But, will it work when used with stainless steel and in the long run?

The other ideas that I am looking at are the use of two of these stainless steel balls, one on each PCB. The big problem there is attaching stainless steel to a PCB pad. Acid flux is one choice but it would require extensive cleaning afterwards. Another I have looked at is a conductive adhesive. This would have more resistance, but my circuit can tolerate that. But I do not know about the long term reliability of such a bond. A third possibility would be a friction fit in a plated hole. But hole plating on a PCB is very thin and the specs for the finished diameter are not very tight. So getting a good contact that lasts for a long time (years) seems to be a very unlikely thing. Eyelets perhaps?

The positional accuracy of the balls in any of these alternatives would be ensured by a second PCB, with no traces, attached to the rear side of the primary PCB. The hole in the main PCB would be a bit larger than the ball size so it would rest on that back board while being installed.

I am open to any other suggestions.

I'm not sure if having anything attached to PCB will work in your favor.   Unless you have production level goals for this, I'd just go out and buy a sensor that can already do what you want.   You will still struggle with the mechanical systems to get the repeatability you want.

[wizard69]

I think I have the warping covered. The boards will be relatively small, around 1.25"/35mm rectangles. The main and back boards will be combined with a good epoxy after roughing. This should produce a plywood like structure to resist warping. I think FR-4 may/will have some initial warp, but I don't think it will change over time. So once the device is assembled and calibrated it should be stable. Also the overall configuration of the device will help make any remaining warping in the swithch assembly irrelevant in the device's operation.

I would like to use PCBs for some of the parts due to their ready availability and relatively low price. I am trying to work around their undesirable characteristics. I will also be using Send-Cut-Send for some of the flat metal parts. Again they will be easy to acquire and relatively inexpensive.

In any case this is experimental. I will build several prototypes and use at least some of them for long term testing. I will see how well it works.

PCB laminate is not the right material to build precision switches, +/-0.0003" repeatability, because it warps and is not flat. When using CNC to try and mill traces, a probing operation is generally necessary first to create a height profile so that the milling depth would be the same over the whole board, the displacements can be quite significant.

It is more than just warping.   When you get into mircron and sub micron range everything impacts performance.   We had to pay special attention to air conditioning ducting and that was on a machine made largely of steel and cast iron.   You would think that the mass would make rapid changes impossible but if you do the math it doesn't take much to be thermally impacted.   Then you need to consider things like mechanical vibration that can impact repeatability.

[EPAIII]

I don't think I ever mentioned or dreamed about micron or sub-micron resolution. 0.0001" is about the best I dare dream about and that is about 2.5 microns. And I am aware that even that would be prone to temperature changes and perhaps other effects. When metal parts are to be machined to 0.001"/0.025mm or better, the heat of cutting them will lead to distortion. It is necessary to allow them to cool to room or a specified temperature before final measurements and cuts are made.

I have struggled to make parts accurate to +/-0.0001" and it can be a challenge. I am not just blowing smoke.

I think I have the warping covered. The boards will be relatively small, around 1.25"/35mm rectangles. The main and back boards will be combined with a good epoxy after roughing. This should produce a plywood like structure to resist warping. I think FR-4 may/will have some initial warp, but I don't think it will change over time. So once the device is assembled and calibrated it should be stable. Also the overall configuration of the device will help make any remaining warping in the swithch assembly irrelevant in the device's operation.

I would like to use PCBs for some of the parts due to their ready availability and relatively low price. I am trying to work around their undesirable characteristics. I will also be using Send-Cut-Send for some of the flat metal parts. Again they will be easy to acquire and relatively inexpensive.

In any case this is experimental. I will build several prototypes and use at least some of them for long term testing. I will see how well it works.

PCB laminate is not the right material to build precision switches, +/-0.0003" repeatability, because it warps and is not flat. When using CNC to try and mill traces, a probing operation is generally necessary first to create a height profile so that the milling depth would be the same over the whole board, the displacements can be quite significant.

It is more than just warping.   When you get into mircron and sub micron range everything impacts performance.   We had to pay special attention to air conditioning ducting and that was on a machine made largely of steel and cast iron.   You would think that the mass would make rapid changes impossible but if you do the math it doesn't take much to be thermally impacted.   Then you need to consider things like mechanical vibration that can impact repeatability.

[EPAIII]

How that is relevant? I both read moffys comments and watched the video referenced in them. The guy who made the video was talking about 3D printing. He was in a room with reels of filament hanging on the walls. The video was ALL ABOUT 3D printing. So the accuracy needed for 3D printing is the BASE LINE in his thinking. And I am sorry, but the layers put down by any 3D printer that I know of are not measured in microns. They are measured in tenths of a mm (~= 0.0039"). That's about 100 times larger. He even states that 50 microns is more than good enough. I am sure that is true for 3d printers, but there ARE other uses.

As for the repeatability of micro switches, they were one of the first things I thought of and I tested their repeatability months ago. In the video ONE sensor of each type was tested and that includes ONE micro switch. I tested FIVE, name brand micro switches. While two of them were accurate to the 2.5 micron level, the other three were not. But lets be generous and say 50% were. I need not one, but THREE switches in my device and a quick price check shows prices start around $1.95 each. But I would need to buy 6 in order to find 3 good ones so 6 x $1.95 = $11.70. While not a deal-breaker, it is something to consider. Ant then there is the size. That $1.95 one is 10mm x 16mm x 28mm and that does not include the terminals which stick out an unspecified distance further, probably making it 38mm long. Three of them would more than double the size of my device in two directions and increase it by about 1.5X vertically. There are smaller ones, but they will be more expensive, probably around $3 to $5 each or $24+/- for six.

And one more bad characteristic of almost any switch that is commercially available: toggle action which produces a dead zone. The ON point and the OFF point will have a distance between them. So they do not change state at the same point for travel in opposite directions. And this is another characteristic that almost no switch manufacturer provides in the specifications.

What I don't see is how a micro switch could possibly be used in my device.

While 3D printers are certainly ONE application, their layer thickness between 0.1 and 0.3 mm is hardly high precision. Yes, I would love to have a device that is useful for 3D printing.

I don't see how that is relevant, did you read moffys comment?

You can find the Renishaw specs at the end of their technical documents, eg: https://www.renishaw.com/cmmsupport/knowledgebase/media/pdf/en/04c7251ac8c14277b20fbd540cc0da55.pdf

PH6M autojoint repeatability: 1um
TP6 uni-directional repeatability: 0.35, TP1: 0.5um

[EPAIII]

Three points of contact define a plane. So, that configuration is almost mandatory. In my case the three balls would not be in contact but when all three of them come together, the circuit closes and the contact plate would be parallel to the base of the device.

And an inverter is not needed.

I can't see any reasonable alternative to the normally-closed switch approach. The one I saw used three balls in a ring, each insulated from the others, with a spring holding the ring down against a contact plate.  A lever attached to the ring would be pushed sideways, letting at least one ball switch open at the instant of contact.  No choice but to be precise.  No gaps while a switch closes. No calibration. I guess the error would be determined entirely by how stiff all the parts were, and how much tension was required. 

I guess it would require a rethink for one axis.  And it requires an inverter of some kind, to close when one of the switches opens, but that doesn't seem difficult, especially in light of all the advantages.

[EPAIII]

I would love to get back to my original question.

I have looked up the contact pads suggested above and they seem to have the same or a similar surface finish as the hard gold finish of PCBs. It may be a question of which of those two is less expensive.

But I would still like to know if there are any other finishes that I could consider. What worries me about a gold finish, even a very thin one, is that my SS balls could create a depression in it. Then if the ball assembly shifted sideways, the ball could make contact on the wall of that depression instead of fully seated in it.