Best way to trace existing PCB (not for piracy!)

[daxliniere]

Hey all,
The membrane keypad on my garage sale Mitutoyo DRO has some breaks. Discontinued decades ago and replacement part (if available) will likely be 5-7x more than I paid for the whole DRO (which otherwise works).
I attempted several different micro-surgery methods to no avail - the tracks are held to the plastic film with wishes and unicorn tears.

So, I figured I would make a PCB with physical switches (low-profile tactile pushbuttons) to replace it. I've scanned the back of the membrane keypad and would love some advice as to what is the best way to clean it up and get it into a format for JLCPCB/PCBway/etc, please.

I've attached what I've done so far. (The PNG is sourced from am SVG file, which I had to 7-zip up to attach to this forum.)


Thank you in advance,

Dax Liniere.

[ataradov]

The easiest would be to measure the locations of all the buttons and recreate the board in a CAD. KiCad is free and would do the trick. There is no good way to get manufacturing files from a picture.

Don't bother with the film, make a regular PCB with a connector and use a regular ribbon cable.

[daxliniere]

Thanks, ataradov, but where there is a will, there is a way.

Yep, I'm not evening thinking of another membrane/flexible 'board'. 

[ataradov]

If you are familiar with any CAD tool, it would take a couple of hours at most to recreate this.

If you are not familiar - this is a good opportunity to learn. All you would need is create one button footprint. You won't even need to make the schematic, you can just layout the board.

[daxliniere]

Yes, this is exactly what I'm asking: What is the best tool to use to trace over this? I have no desire to draw up the schematic as it's irrelevant to manufacture.
Should I do this in CAD/F360 or should I be using a dedicated PCB design tool? I will eventually need Gerber files, I'm sure.

[ataradov]

And you can also attempt to add a real connector to this board. Details would depend on the mechanical design of the case. You can cut the existing cable, mount a connector (most likely using epoxy), and use thin wires to restore the traces. It might work.

[ataradov]

The best way is to just recreate traces by eye. Once you have locations of the buttons, routing is trivial to copy by simply looking at it. And locations of the buttons can just be entered as numeric values by hand. There is no need to do overlays or anything like this.

If you are not already familiar with any CAD tools, then use KiCad.

[daxliniere]

Hey Alex, thanks for your replies.

I am familiar with CAD, but KiCAD seems like as good a place as any to start with this process.

The original ribbon connector is unsolderable, even with very careful surgery, flux and a low soldering iron. I tried connecting in an IDC kinda way, but connection was sketchy at best. This is why I concluded the traces couldn't be repaired.
My plan was to see what connectors JLCPCB has available and source one, plus a double-ended ribbon cable to match.

As for fitting it into the case, the tact switches will be mounted directly behind the existing membrane switches, so it's effectively invisible from the outside. The only consideration is that the rows of 7-seg LEDs will sit a bit further back to accommodate the switch PCB.

I'm going to give KiCAD a try and see how far I get with it.

Thanks again and all the best,
Dax.

[ataradov]

It is not really clear from the pictures. It looks like the PCB just has the standard exposed pads for the rubber buttons and the actual buttons have a conductive rubber pad that shorts the terminals. Why not recreate the same? Why use buttons?

If the keyboard is actually membrane and contains the conductive part too, then I would still go with the membrane buttons, just the metal ones. They are readily available in different sizes https://www.aliexpress.com/item/3256801263748081.html They would have comparable thickness, so should not affect the mechanical fit.

I also would not bother with JLCPCB assembly. There is nothing to "assemble" here.

My idea with the connector was to mechanically fix it to the board, then  this part would be obviously solderable. Then run thin wires parallel to the existing tracks and use some conductive glue, no heat is necessary on that side. It is very hard to tell if this would work without looking at the PCB closely.

[andy3055]

Can you use conductive touch up paint and repair the broken traces?

[ataradov]

There is too much damage for the paint alone. Using thin wires to duplicate long sections of damage may help. But that paint in general is pretty hit or miss even on much less damaged sections.

[daxliniere]

Hey Alex. Actually, you've given me an idea. I tried and was successful at removing the flexible PCB from the front 'sticker' that has the conductive pads and metal dome 'snap' switches.

I wonder if I can use the flex PCB as a UV mask to etch my own board?
One potential issue is that the flex PCB has a kind of conductive paint applied to the contact fingers.

[ataradov]

There is no way to use that as a mask. Just design a PCB with the same routing, order it in 0.8 mm without solder mask and you will get the same functionality with a minor increase of thickness. But the domes would make contact with the PCB at the same level, so functionality would not be affected.

[ataradov]

Another option I would consider is hand wiring a PCB using low profile buttons (like this https://www.aliexpress.com/item/3256801553660538.html) and either a prototype PCB or just a plain copper clad board. This would require some effort, so I'm not sure it is worth it.

For the board of this size with gold finish PCBWay quotes  $100 (5 pcs). and with a HASL finish it is around $60.

[daxliniere]

There is no way to use that as a mask.

Hold my beer...
Seriously though, let me teach you something really cool:

Another option I would consider is hand wiring a PCB using low profile buttons (like this https://www.aliexpress.com/item/3256801553660538.html)...

Dude, that's what I said I was going to do in my very first post.

So, I figured I would make a PCB with physical switches (low-profile tactile pushbuttons) to replace it.

[ataradov]

Seriously though, let me teach you something really cool

I've started making PCBs when ordering them for $5 was not a thing. I'm aware how PCBs are made. If you think you can do - go ahead. I'm telling you, it is not as easy as it seems. Also, you would spend more money on chemicals and equipment than just ordering a professionally made PCB.

There is also no need for any of this UV stuff. This board is simple enough that it can be made using printer and iron method. Way easier, way cheaper.

Dude, that's what I said I was going to do in my very first post.

From your first post is was not clear how that keyboard is designed. With the most recent pictures it is much more clear how it is actually designed.

And again, I would only go for the switches if you are making it on a protoboard with hand wiring. If you are ordering a PCB, go with the original pattern and use the original buttons. If you want to get fancy, make the button mesh footprint be compatible with the physical button.

With those switches I would also worry about plastic wear. The cap on the switch is pretty small, and over time it would imprint on the plastic and potentially break through. So, I would also add some sort of hard baking to the plastic domes if you go that route.

[tooki]

There is no way to use that as a mask.

Hold my beer...
Seriously though, let me teach you something really cool:

Yyyyeah... I'm pretty sure he's WELL aware of how DIY etching works, bro.

Another option I would consider is hand wiring a PCB using low profile buttons (like this https://www.aliexpress.com/item/3256801553660538.html)...

Dude, that's what I said I was going to do in my very first post.

So, I figured I would make a PCB with physical switches (low-profile tactile pushbuttons) to replace it.

No, he doesn't mean the same thing. You're talking about having a custom PCB made, and he clearly understood that based on his many replies before that one. When he offers the alternative of "hand wiring", he means NOT using a custom PCB, but rather generic perfboard or similar, and actually manually running actual wires to the buttons.

[SiliconWizard]

If you really want to make a PCB out of an image and it's single layer, and you're not too keen on recreating from scratch, wouldn't the following tool help?

https://www.eevblog.com/forum/eda/i-made-an-svg-vector-graphics-to-gerber-converter/

Not tried nor have any opinion on this, but you can contact the author.

[tooki]

Yes, this is exactly what I'm asking: What is the best tool to use to trace over this? I have no desire to draw up the schematic as it's irrelevant to manufacture.

Unless you have zero interest in ever designing a PCB now or in the future, then you absolutely should draw the schematic first. Schematic->PCB layout->Gerber->Manufacturing is THE process for making PCBs, so a) it's the process the software is designed around, and b) doing the schematic means you CAN'T wire the PCB wrong. I cannot underscore enough how important (b) is! So if you have even the slightest interest in electronics, learn the process, you'll be doing yourself a favor.

P.S. I'm not talking hypothetically about the advantages. At work, a fellow apprentice was tasked with copying some PCB layouts (for practice boards) . I didn't see the result till much later when we needed to change something. Oh. My. God.  Turns out, he'd done it without first creating schematics, and thus had to "creatively" undermine the layout software every step of the way to get the job done. All of the mechanisms in the software that are there to prevent errors simply don't want you connecting "random" copper, which is exactly what it is without a schematic. It would have been far better for his learning, as well as for future project maintenance, to do a schematic first.

[daxliniere]

doing the schematic means you CAN'T wire the PCB wrong.

Do you know what else guarantees you can't wire it wrong? Using the original transparent flex PCB as the photo-positive mask!

[ataradov]

Do you know what else guarantees you can't wire it wrong? Using the original transparent flex PCB as the photo-positive mask!

You can't use it directly because of the bridges. You need to remove those at least. Plus you need to check that this green material is UV transparent. It is highly likely that it is not.

But if you think you can do that, try it. It would be a good learning experience.

But the first issue you would need to solve is find a copper clad material in that size with pre-applied photoresist. This alone would likely be more expensive than a manufactured PCB. And chemicals involved in this process are highly corrosive and need to be handled and disposed with care. It is a huge headache, especially if you don't plan on making more PCBs yourself.

[tooki]

doing the schematic means you CAN'T wire the PCB wrong.

Do you know what else guarantees you can't wire it wrong? Using the original transparent flex PCB as the photo-positive mask!

Are you here to get answers to your question, or to be a smartass and try to lecture people more knowledgeable than you?

Your original question was:

So, I figured I would make a PCB with physical switches (low-profile tactile pushbuttons) to replace it. I've scanned the back of the membrane keypad and would love some advice as to what is the best way to clean it up and get it into a format for JLCPCB/PCBway/etc, please.

Making a replacement PCB for tactile switches rules out reusing the old membrane layout to begin with. So even if you could make an optical copy of the membrane traces onto a blank PCB, it wouldn't get you to your stated goal.

[daxliniere]

The stated goal was to get the keypad of the DRO operational again. The solution evolved throughout this thread as all good ideas do. (The design 'journey' has been documented here.)

Alex. Where on earth did you get the idea that it would not be UV transparent? Sure, some aren't. You seem to be set on your negative train-of-thought about this.
And who cares about the bridges? If you apply a moment of thought, you would know that they could easily be removed (and replaced with a back-of-board jumper wire).
And, and, and why does it need to be pre-applied photoresist film? (as if that's hard to find anyway...  )

In the end, it would take me a few hours to redraw the flex PCB. Having said that, by the time I drive to RS to get the positive-photoresist PCB and other stuff, it might be the same time as redrawing it and cutting on my CNC mill.

See? Evolution of the solution.

[ataradov]

Sure, go ahead. I don't really care what you do.

If you have a CNC mill, then it would probably work as well.

[nctnico]

Or just buy a brand new replacement for less than $100:
https://www.ebay.com/itm/363797941713

Otherwise I'd design a regular PCB (create schematic and route the whole thing) but it will be hard to beat the price of the readily made keypad from the Ebay listing. And don't bother with making your own PCBs. Last week I got rid of all my UV exposure and related stuff. Too much hassle and poor results.

[julianhigginson]

well that thread sure was exhausting....

[Doctorandus_P]

Another vote for KiCad.

Quite recently a new feature has been added to KiCad-nightly. You can now load an image in the PCB editor and use that as a template to guide footprint and track placement.

https://forum.kicad.info/t/post-v6-new-features-and-development-news/32633/58

Do note though that it's a new feature and that the nightly version is likely to have some bugs. PCB's made in the nightly versions can also not be opened in the stable KiCad version. It is however a policy of KiCad to release a mayor version once a year, and KiCad V7 is therefore expected late this year or early next year.

[JohanH]

I have recreated PCBs in KiCAD by importing the image as a background in KiCAD PCB tool, then drawing new traces according to the background. Of course, first a schematic should be created.

[PlainName]

PCB's made in the nightly versions can also not be opened in the stable KiCad version

Why is that?

[pcprogrammer]

Why not use EasyEDA https://easyeda.com/editor

No need to install software and it is possible to order from JLCPCB directly or create gerber files. It is not difficult to work with it. You can also print the design 1:1 on your printer (If you have one of course) and check if things fit behind the existing panel.

Dual layer is not very expensive and makes the layout job easy.

Way easier than at home messing with chemicals to make your own PCB.

[JohanH]

PCB's made in the nightly versions can also not be opened in the stable KiCad version

Why is that?

Importing projects made by newer versions of software into older software versions is much harder to do than the other way around and is many times impossible. In order to make it possible, you would have to patch the old version so that it is able to import new features etc.

[JohanH]

Why not use EasyEDA https://easyeda.com/editor

It's not bad advice, but consider what you need. Personally, I like to archive old projects and be able to open them again, import them to newer software versions and create a new revision if needed. Now with a cloud service, can you be sure that you can do this in the future?

With KiCAD, in ten years the software could change so much so that a new version can't open/import a really old project, but you should be able to fire up a virtual machine, install an old version and at least be able to open the project.

[mikeselectricstuff]

Yes, this is exactly what I'm asking: What is the best tool to use to trace over this? I have no desire to draw up the schematic as it's irrelevant to manufacture.

Unless you have zero interest in ever designing a PCB now or in the future, then you absolutely should draw the schematic first. Schematic->PCB layout->Gerber->Manufacturing is THE process for making PCBs, so a) it's the process the software is designed around, and b) doing the schematic means you CAN'T wire the PCB wrong. I cannot underscore enough how important (b) is! So if you have even the slightest interest in electronics, learn the process, you'll be doing yourself a favor.

Not necessarily. There is no right or wrong way in engineering, only more or less appropriate for a particular situation.
Doing the schematic first can impose unnecessary constraints as choices are nailed down too early in the process. In many cases there is a lot of scope to swap pins & devices ( MCU I/Os etc.) and the optimal arrangement can be far from obvious until you start looking at the physical layout.
I never start from a schematic, instead I create nets as I go along during placement, and to a lesser extent, routing - put some or all of the parts down, create nets for all the things that are fixed, then see what the best options are.
There is still a netlist, it's just created during the layout process. (I use PCAD2006, which allows nets to be created during layout - I don't know how common this is in other PCB software).

For a job like this, where you already have a layout, starting with a schematic would be a waste of time unless you intend to make substantial changes.
As mentioned above it's not worth the effort to convert from a scan, and much quicker to re-draw it from scratch in a PCB package.
 

[pcprogrammer]

Why not use EasyEDA https://easyeda.com/editor

It's not bad advice, but consider what you need. Personally, I like to archive old projects and be able to open them again, import them to newer software versions and create a new revision if needed. Now with a cloud service, can you be sure that you can do this in the future?

With KiCAD, in ten years the software could change so much so that a new version can't open/import a really old project, but you should be able to fire up a virtual machine, install an old version and at least be able to open the project.

There is an offline version of the software available, so store that package with your designs, which you can also archive locally, and you are good to go, as long as the hardware will support it. Because old software like for instance OrCAD needs either an emulator or old hardware to run. Might be a bad example when a new version of the suite is still available today and old designs can be loaded into it, but he.


About drawing a schematic first and then the PCB, I think that depends on what you are used to, but I see it as an iterative process. I mostly draw up a schematic first, then start to shuffle the components round on the PCB design and when needed I do pin or component swapping in the schematic and annotate the PCB until fully satisfied with the result.

[PlainName]

PCB's made in the nightly versions can also not be opened in the stable KiCad version

Why is that?

Importing projects made by newer versions of software into older software versions is much harder to do than the other way around and is many times impossible. In order to make it possible, you would have to patch the old version so that it is able to import new features etc.

Ah! Of course - the nightlys are going to be newer. Duh

Edit: to clarify I was thinking it meant if you used the nightlys then you couldn't open the files with the next full version (that is, the stable when it's caught up). But you can read it either way and obviously I saw the wrong one.

[tooki]

Why not use EasyEDA https://easyeda.com/editor

It's not bad advice, but consider what you need. Personally, I like to archive old projects and be able to open them again, import them to newer software versions and create a new revision if needed. Now with a cloud service, can you be sure that you can do this in the future?

With KiCAD, in ten years the software could change so much so that a new version can't open/import a really old project, but you should be able to fire up a virtual machine, install an old version and at least be able to open the project.

FYI, the way EasyEDA works right now is that it runs the version of the software the project was last saved in. If a new version is available, it asks whether you want to upgrade the project or not. So it seems that at least now, it actually has multiple software versions on the server to run as needed.

[daxliniere]

For a job like this, where you already have a layout, starting with a schematic would be a waste of time unless you intend to make substantial changes.

Finally! The voice of reason!

Yes, I am trying to create a 1:1 replica of the original membrane PCB, there is no need to interpret the wiring.

[ataradov]

Finally! The voice of reason!

Post #3:

You won't even need to make the schematic, you can just layout the board.

But given that replacements are readily available, they are probably the best way to go. Ordering the board that size would cost about the same anyway.

Milling or any other methods trade in quality and time. It all depends, may be worth trying. You will always have an option to buy the replacement panel.

[nctnico]

For a job like this, where you already have a layout, starting with a schematic would be a waste of time unless you intend to make substantial changes.

Finally! The voice of reason!

Yes, I am trying to create a 1:1 replica of the original membrane PCB, there is no need to interpret the wiring.

You may think that but the whole goal of a schematic is to be able to check whether everything is connected in a sensible way. The keypad will typically consist of a matrix but the PCB layout can severely obscure that. It would be a hard lesson when it turns out the wiring is wrong after spending a lot of money on having a PCB made.

And why aren't you ordering a replacement panel? I doubt going the DIY route will end up being cheaper.

[ataradov]

You will still have the layout file to check against if necessary. There is absolute no need for a schematic here. Plus you can make an error in recreating the schematic anyway. It is just a waste of time.

[tooki]

Yes, this is exactly what I'm asking: What is the best tool to use to trace over this? I have no desire to draw up the schematic as it's irrelevant to manufacture.

Unless you have zero interest in ever designing a PCB now or in the future, then you absolutely should draw the schematic first. Schematic->PCB layout->Gerber->Manufacturing is THE process for making PCBs, so a) it's the process the software is designed around, and b) doing the schematic means you CAN'T wire the PCB wrong. I cannot underscore enough how important (b) is! So if you have even the slightest interest in electronics, learn the process, you'll be doing yourself a favor.

Not necessarily. There is no right or wrong way in engineering, only more or less appropriate for a particular situation.
Doing the schematic first can impose unnecessary constraints as choices are nailed down too early in the process. In many cases there is a lot of scope to swap pins & devices ( MCU I/Os etc.) and the optimal arrangement can be far from obvious until you start looking at the physical layout.
I never start from a schematic, instead I create nets as I go along during placement, and to a lesser extent, routing - put some or all of the parts down, create nets for all the things that are fixed, then see what the best options are.
There is still a netlist, it's just created during the layout process. (I use PCAD2006, which allows nets to be created during layout - I don't know how common this is in other PCB software).

For a job like this, where you already have a layout, starting with a schematic would be a waste of time unless you intend to make substantial changes.
As mentioned above it's not worth the effort to convert from a scan, and much quicker to re-draw it from scratch in a PCB package.

My point is this: starting from a schematic is what most EDA software is designed around. Hence why I made my suggestion with the caveat of not bothering if PCB layout is something they never intend to do again.

As for pin swapping: Yep, that’s an important thing, which is why good software also helps you swap pins and components around in the layout and push the changes back to the schematic.

For a job like this, where you already have a layout, starting with a schematic would be a waste of time unless you intend to make substantial changes.

Finally! The voice of reason!

Yes, I am trying to create a 1:1 replica of the original membrane PCB, there is no need to interpret the wiring.

I wasn’t being unreasonable. You clearly didn’t bother seeing the point of my comment: IF you plan on doing electronics in the future and making more PCB layouts, then learning the standard way is a good learning experience. (And reverse engineering a schematic is a great way to have a schematic to practice layout with without having to design a circuit from scratch.)

If you don’t plan on ever laying out a PCB again, then skip it. Just know that the software may fight you every step of the way.

You may think that but the whole goal of a schematic is to be able to check whether everything is connected in a sensible way. The keypad will typically consist of a matrix but the PCB layout can severely obscure that. It would be a hard lesson when it turns out the wiring is wrong after spending a lot of money on having a PCB made.

Exactly.

You will still have the layout file to check against if necessary. There is absolute no need for a schematic here. Plus you can make an error in recreating the schematic anyway. It is just a waste of time.

For a one-off thing, yes. But if learning electronics is also a goal, then it could be time well spent.

[PlainName]

Plus you can make an error in recreating the schematic anyway.

Indeed, but the point of the schematic is that it's easier to read than a layout, so mistakes should be more obvious. Further, you can decide your layout is not going to make it for some reason and do it completely differently, but the schematic will ensure that it's still electrically correct.

[ataradov]

This topic is about recreating an existing layout, and a very simple one at that. What is not going to make it here?

If there were no readily available version, if I were to recreate this, I would also add through holes from the buttons to the pads on the other side. Even if I completely mess up the layout, it would not matter. I can always wire the boards however I want.

This is not some mass production situation, you are just making one. Stop overthinking it and just make it and be done with it.

[daxliniere]

Progress.
(Laminated MDF test plate)

[fourfathom]

Are there actual traces in those interdigitated fingers?  It looks like some of them are completely milled away.  And your un-milled groundplane will also be connected when the conductive switches are depressed -- will that affect the switch decoding?

[Warpspeed]

Its obviously a keyboard layout.
Those interdigital fingers get shorted together when a button is pressed.
Pretty simple, no electronis, just dumb wires and connectors.

Its probably the only example I can think of, where a circuit diagram is not going to be at all useful.

[daxliniere]

Are there actual traces in those interdigitated fingers?  It looks like some of them are completely milled away.

Yes, you're correct, they were eroded away. This was just a test as this would be the finest thing I've done on my CNC mill. Now that I've seen it works, I just need to adjust the radial offset to account for the V-shaped cutting bit. Also, this is a piece of MDF laminate flooring, so not nearly as rigid as copper-coated FR4.

And your un-milled groundplane will also be connected when the conductive switches are depressed -- will that affect the switch decoding?

Ahh yes, that's an excellent point. I'll add some cutaway rings around the interdigitated* fingers to avoid shorts. (*by the way, I LOVE that word.) Cheers, fourfathom.

[fourfathom]

Yeah, I've done a bit of PCB milling on my Shapeoko.  Keeping the PCB flat is critical, as it doesn't take much variation in V-bit depth to go from "just a scratch on the copper" to "no trace left" (especially with narrow traces).

[daxliniere]

Hmm.. seems like 0.8mm endmill might be the better option, then. I have a 0.5mm, but I've been too afraid to use it - it looks like it'll snap if I look at it the wrong way! 

[PlainName]

Hmm.. seems like 0.8mm endmill might be the better option, then. I have a 0.5mm, but I've been too afraid to use it - it looks like it'll snap if I look at it the wrong way! 

For a PCB the way to go is with a V-cutter and a big angle - 90 degrees is good! Obviously, whilst the tip is much less fragile, the depth of cut is really important so your PCB must be flat and square. I use carpet tape to stick the PCB onto the mill's bed, which ensures it's flat and only varies by the PCB thickness tolerance.

Also, 3-in-1 or other cutting oil spread over the surface both keeps the bit from going blunt and stops the PCB dust from going everywhere.

Apols if you already do or knew that. Originally I tried small bits, shallow V-cutters, etc., so ISTM you may be at in the first leg of your PCB making journey.

[daxliniere]

Apols if you already do or knew that. Originally I tried small bits, shallow V-cutters, etc., so ISTM you may be at in the first leg of your PCB making journey.

No, no, this is all great to hear. Thank you for sharing your experience. I'm sure it's saved me from a few mistakes.

Dust is not such a problem for me as I have my CNC in a cabinet, but that's interesting about the blunting. Is that a big concern when making such little cuts in FR4?

[PlainName]

FR4 seems to be pretty hard on sharp edges - try cutting or shaving it with a knife a few times. A better indication is the copper tracks which, without oil, I find start nice and well defined but quickly get rough and require a tool change to get a good edge again.

[nctnico]

FR4 is glass and epoxy resin. HSS tools are useless for FR4. Really go for tungsten (or other very hard metal) bits if you want to have some lifespan.

[SiliconWizard]

KiCad allows importing on any layer from vector graphics (SVG and DXF currently.)
Not directly bitmap images, but there's a separate tool for converting bitmap images to footprints, that you could then just add on the copper layer.
Alternatively you can use Inkscape or some other tool to convert your bitmap to SVG (vectorization will be better using Inkscape I think) and then import that in KiCad directly. That's pretty straightforward.

Do not expect not to have to do manual touch-ups though. And of course you need to finely adjust the dimensions.

[PlainName]

Really go for tungsten (or other very hard metal) bits if you want to have some lifespan

Indeed - I use carbide for drills and they last an age. But hard metal is very brittle and snaps easily when not going in dead straight. Not so hot for tiny milling bits.