Update on piano conversion

RIP The problem with that is that the soldering would be on the wrong side of the PCB, so the question for people who know more about that than I is if I could use the same mounting hole as a via, or if I need a separate via for that. As always, other solutions/ideas are well accepted and sought after.

"Mounting holes" aren't electrical, especially in Kicad. But the through holes used for through hole parts like pin headers are plated, just like vias. Things like pin headers can be soldered onto either side of the board, no problem. I do that on mine.

CyberGene My intuition tells me hammering the electronics directly is wrong. I can't explain why, maybe I'm wrong but I don't find that right, it will certainly stop working after a while.

The sensors themselves are potted in epoxy. Their failure is wildly unlikely. I'd expect the solder joints between them and the PCB to fail first. PCBs themselves have enough flex that I'd expect the traces to survive indefinitely.

Personally I've got little 3D printed protector things, so that the hammers strike those (and are kept _2mm away from the sensor surface), and the force is distributed across the whole PCB. I then put some very thick electrical tape on the back side to cushion the interface between the PCB and the rail.

RIP Now, back to my question….

If you're going for automated / professional assembly, that's probably a question for the assembler.

If you're doing it by hand, then a very gentle press fit is probably ideal.

Solder will bridge the gap of any sane drill size, but it'll take a lot more solder (mostly a problem for hand assembly), and you increase the risk of parts falling out,

JayKominek "Mounting holes" aren't electrical, especially in Kicad. But the through holes used for through hole parts like pin headers are plated, just like vias. Things like pin headers can be soldered onto either side of the board, no problem.

Awesome.

JayKominek If you're going for automated / professional assembly, that's probably a question for the assembler.

Sounds great, I'll ask then.

JayKominek If you're doing it by hand, then a very gentle press fit is probably ideal.

Thanks a lot for the tips!

CyberGene My intuition tells me hammering the electronics directly is wrong.

In my old Clav each key lever "hammers" upon its rubber contact.
And the rubber "hammers" upon the underlying PCB contacts.
Yes, they wear out eventually. But doesn't everything?

Digital piano actions are to a certain degree inconsistent in velocity detection because they rely on mechanical switches that will deteriorate from the constant banging. Optical contactless sensors can eliminate the wear. But if they are robust enough, they can take some beating.

I’d be also worried about vibrations in the rail causing neighboring sensors to oscillate longitudinally which can potentially reduce the precision of proximity detection but I guess that can be solved through using a very rigid rail that is fixed at multiple points.

CyberGene I’d be also worried about vibrations in the rail causing neighboring sensors to oscillate longitudinally which can potentially reduce the precision of proximity detection but I guess that can be solved through using a very rigid rail that is fixed at multiple points.

I was worried about that too. The flex computation from the specs did not help my "gut feeling". Holding that aluminum extrusion which you can see on the rendering I posted earlier today, did. That extrusion is really, really rigid. If you think the flex computation can help, the specs are on the misumi website.

CyberGene But if they are robust enough, they can take some beating.

JayKominek Personally I've got little 3D printed protector things, so that the hammers strike those

I could do the same thing Jay did, which you can see at https://github.com/jkominek/piano-conversion/tree/master/mechanical (I thought I saw a picture once, but can't find it right now). With the smaller sensors will actually be much easier to align and more effective since the "hole" would be much smaller and protective. I have just too many balls to juggle at the moment, but if anybody is interested in this design, and could do the 3D model for that, it'd be a nice plus. But it's just a plus, I'm confident it'll get years of service even without. The perfect is the enemy of the good, and if I try to do everything perfectly, I'll end up never finishing (which some days already feels like it)

RIP CyberGene But if they are robust enough, they can take some beating.

JayKominek Personally I've got little 3D printed protector things, so that the hammers strike those

I could do the same thing Jay did

And now I did (not 3D printed yet). Getting started was really tough, (see the commit messages for details). 3D manufacturing is not that easy, unless you can use just one product.

I also oh-so-slightly modified the board as follows:

• made it slightly longer, to be easier to attach to thinner extrusions (which happen to be larger, go figure MISUMI) for use as a key sensor
• being longer, I've been able to avoid one other bend in the trace of the sensors to the connector
• added silkscreen reference lines to aid alignment, especially for the key sensors which may not be super-accessible in all circumstances

And TADA 🎉 here is the result.

See description of what you are seeing at

https://github.com/davidedelvento/piano-conversion/tree/master/hardware/sensorboards/2040-centered-4x-QRE1113GR/ (not sure how to embed the README.md from github to here, sigh, it wants to show only by ugly face of more than a decade ago….)

I need to sleep on it, and hope that both @"CyberGene" and @"JayKominek" (and possibly others) give a look at both the latest revision of the board and of the protector, and now I am really ready to order both. I plan to do it at https://www.pcbway.com/ which I like better than JLCPCB and 4PCB: one-stop-shopping (even 3D printing, so one order for everything) and especially a decent submit form rather than the unusable cr** [1] the other two offer.

[1]: for a newbie like myself, their staying in business implies that other people, likely experts, don't cry as much as I do when using them

How are you going to attach the plastic coverings to the PCB? Plastic may cause a lot of clacking noises if large enough and hit by a hammer.

CyberGene How are you going to attach the plastic coverings to the PCB?

Yes. As almost everything else, Jay had the initial idea, so if it's a good one, it's his, but if it isn't working, I designed it myself, so it's my fault 😐

Plastic may cause a lot of clacking noises if large enough and hit by a hammer.

You are never satisfied, @CyberGene are you? 🙂
You are probably spending too much time with @porco_rosso these days 🤣 (no offense intended Eddie, if anything that is actually a compliment for being a perfectionist -- as I am in fact myself)

I thought of that already. If necessary I can and will use thin fabric/felt stuff (already have lots of samples) in between

• metal and PCB (in the bass),
• metal and plastic spacer (in the treble, to correct for the different height due to overstringing)
• plastic spacer and PCB (ditto)
• PCB and plastic protector (the thing you are talking about)
• and even plastic protector and the hammer

In theory you could do digital analysis of these noises, but that's really above and beyond what I am willing to make. Instead, trial and error should be manageable enough (besides that the necessary thickness of the ones mentioned in the last two bullet points might make the sensors buried "too deeply" inside, but I doubt we'll reach that point -- the other ones can be as thick as I want and easy to adjust for given the extrusion attachment design I have).

Now, for something which I have only done in a cursory way, please go check if my tolerances are enough for the manufacturing/assembly process while still not leaving too big of a gap for the hammers to be caught into 😃

Plastic spacer mentioned above:

https://www.amazon.com/gp/product/B01LXGL3HG/

I just had to say something 😀 My way of thinking is always about seeking for what may go wrong, apologies if it sounds like I’m critical. Ask my wife, she will complain about me always thinking only about the negative outcomes of everything 🤣

Other than that it looks like you’re making great progress and you said you were gonna freeze it, so that’s even more good news that you managed to find time to work on that project! I’m looking forward to the final result.

As to PCB, I don’t feel I’m competent enough to comment on that since what I did on Cybrid was probably very amateurish and I haven’t read or researched anything about good PCB design. My understanding is that’s not very important unless you create very high-frequency schematics and/or analog type of electronics. But I might be wrong.

CyberGene apologies if it sounds like I’m critical.

You don't. That's the right way to do the design if you want to have a minimal chance to succeed.

CyberGene Ask my wife

Too bad we leave so far, you and I would be best friends, and our wives too 🤣

CyberGene As to PCB, I don’t feel I’m competent enough to comment on that since what I did on Cybrid was probably very amateurish

The comments I was looking for are very basics: KiCad does not give errors (but gives too many warnings because it does not like the way I made the silkscreen-over-other-things-including-itself). But are the default minima ok? Am I coming too close to the adjacent jumper connector since I arrive at 45 degrees to reduce the number of turns? Very simple things like that, both these listed here and other I might not have thought about. But I was asking that part to mostly Jay. For you, @CyberGene I was thinking of the sensor protector, since I banged my head on the keyboard too much [1] to trust myself that I got it right. If the holes are in the wrong spot, I'm just getting shipped a bunch of just plastic from China….

[1]: See here

RIP I plan to do it at https://www.pcbway.com/pcbway.comhttps://www.pcbway.com/No preview could be generated for this link which I like better than JLCPCB and 4PCB

Forgot to mention:

1. their price structure is not "raw", which I like. With the others, ordering one board or two costed the same money. Ordering 10 or 20 (or any doubling up to thousands) costed 5-10%. I totally get why, but that put psychological pressure on finding asap lots of people interested in the project and order together. With these, even with the volume discount which is there, the price is not so massively biased toward huge numbers. As such, I can order what I need for myself, plenty of spares (but not a full second instrument) and be content. If anybody wants to order theirs….
2. they offer a one-click option in which if I want I can make my project available for others to order. So, once I'm settled, I don't have to explain on github what to do and how, I can just say "click here, pay and the correct parts will show up at your door". Given how painful it's been exploring the other two PCB manufacturing and assembly companies, for me this is a huge plus.

CyberGene How are you going to attach the plastic coverings to the PCB?

The machine screws securing the PCB to the rail go through the plastic. It's under compression, and there's no perceptible play in the assembly. You could add some glue if you could find a glue which would adhere to the surface of a PCB, as well as the PLA plastic. Hyde glue is too brittle (tried it), and superglue doesn't cure between those two (tried it, too). Then I just decided it wasn't necessary any time soon and stopped worrying.

RIP But are the default minima ok?

The default minimum track spacing is conservative and pretty safe for this project. Especially for the sensor boards, where the signals don't have frequency content much above 20kHz. Keeping things tight, but abiding by those spacings, is (I think) a reasonable compromise.

(Life is trickier with digital signals, if you recall circuit theory 201, square waves have "infinite" frequency content, So there's speedier stuff to deal with. The ADC board is the sensitive one, from a layout standpoint.)

JayKominek The default minimum track spacing is conservative and pretty safe for this project.

I was actually thinking of my modifications, for example this one:

And I guess the majority of my insecurity comes from not fully knowing notation and the process here. What are those thin rings around the thick rings? Should have been both tracks for pad 5 and 4 more to the right to a) stay closer and b) avoid that notch on the pad 4 connection?
Another thing that confused me is the warnings of silkscreen being "erased" by the copper tracks. Aren't the copper tracks covered by an insulating material on top of which stuff can be written? And uncovered only in the areas where soldering is expected? It seems so in the raytracing 3D view, but the warning seems to indicate otherwise.

Also, when ordering or requesting for an estimate, I need to select the accuracy I want (sigh, can't they figure it out from the file I'm uploading???) and I could use 8mil which is slightly more than the 0.2mm selected in KiCad, so I selected 6mil, which costs (marginally) more but more importantly prevents the use of the 2oz Cu density (which may be overkill, but it's almost free at 8mil accuracy)…. Probably none of these matters but still it takes a toll on my decision making….

@RIP, is that a diy avatar or stock? Either way, a very good likeness! 😀

RIP What are those thin rings around the thick rings?

The clearance that needs to be maintained by unrelated nets. You might note that the track for Net-(U2-Pad3) makes a tiny little jog around pad 7's ring.

RIP Should have been both tracks for pad 5 and 4 more to the right

The track to pad 5 (net E3), looks fine in that picture. The track to pad 4 has some wasted copper, where it looks like the track tried to go to the left and then you brought it back, but it won't have any negative consequences.

RIP Another thing that confused me is the warnings of silkscreen being "erased" by the copper tracks.

Are you sure it is referring to tracks, or to the copper of the pads? You've got silkscreen going over pads 8 and 7 in that picture, which, as it warns, will have to be omitted.

You're right that copper which isn't intended to be soldered to will be covered by solder mask, and is a legitimate place to put silkscreen. That just doesn't apply to all of your copper.

RIP I need to select the accuracy I want (sigh, can't they figure it out from the file I'm uploading???)

Perhaps you want even tighter tolerances than what appears in the file. How would they figure that out?

If you specified 8, they might catch it and warn you when they do a manual review. Not all of these checks boardhouses have are fit for generating automatic feedback to customers.

RIP I selected 6mil, which costs (marginally) more but more importantly prevents the use of the 2oz Cu density (which may be overkill, but it's almost free at 8mil accuracy)….

Absolutely no need for 2oz copper on the sensorboards. The only value is for carrying higher currents.

I'd pay for the service which matches the tolerances you used, the risk/reward/frustration tradeoff isn't great for trying to save those pennies.

Drew-r @Del Vento, is that a diy avatar or stock? Either way, a very good likeness! 😀

https://avatarmaker.com/

JayKominek The clearance that needs to be maintained by unrelated nets. You might note that the track for Net-(U2-Pad3) makes a tiny little jog around pad 7's ring.

Ah! I thought that was my unstable hand when making the track! Should I further need to tweak the board will take that out.

JayKominek The track to pad 5 (net E3), looks fine in that picture. The track to pad 4 has some wasted copper, where it looks like the track tried to go to the left and then you brought it back, but it won't have any negative consequences.

Thanks so much, I appreciate the clarification. Should I further need to tweak the board will take that out.

JayKominek Del Vento Another thing that confused me is the warnings of silkscreen being "erased" by the copper tracks.

Are you sure it is referring to tracks, or to the copper of the pads? You've got silkscreen going over pads 8 and 7 in that picture, which, as it warns, will have to be omitted.

You're right that copper which isn't intended to be soldered to will be covered by solder mask, and is a legitimate place to put silkscreen. That just doesn't apply to all of your copper.

You are right I was being sloppy with terms. What it was saying is "Silkscreen clipped by soldermask". Now, most of that I understood (places where there is actual soldermask, such as the sensor pads). Some others were weirdly termed, but still understandable (got out of the board by a fraction of a mm, so it should have said "clipped by edge cut"). Others, were probably not real warning, but just display things, such as the "plus" sign (to indicate polarity for the power supply) "disappearing" under the track when I placed it to the left of the jumper where it seemed to belong better. But on close inspection, there was no arrow pointing to that as a problem, so that was probably just a display thing of PCBEditor and would instead appeared correctly in the final board? In fact it does so on the 3D viewer…

JayKominek I'd pay for the service which matches the tolerances you used, the risk/reward/frustration tradeoff isn't great for trying to save those pennies.

In fact that's what I requested in the estimate. The only concern was of the limitation it introduced to the 2oz thickness, but…

JayKominek Absolutely no need for 2oz copper on the sensorboards.

If that is the case, then it sounds everything is fine and if I am satisfied (and they are too), I can proceed with the order.

JayKominek The only value is for carrying higher currents.

Yes, the value for higher currents I (obviously (?? 🤣 ) knew. Since it was cheaper for the less accurate process, I threw it in anyway in an earlier estimate whose order I did not complete.

I'll keep y'all posted! Hopefully this is the last or very close to last iteration!

I ordered the sensorboards (PCBs, parts and assembly/soldering of both SMT and TH) as well as the 3D protectors at PCBway as described early. Should be easy to reproduce for others, but not as easy as I hoped (I had to "recreate" the order rather than saying "let others order exactly what I did" -- link on the repo). About $400 for everything, which includes:

• manufacturing of 60 PCB boards, in 2.4mm thickness
• purchase of 60x4 QRE1113GR sensors and of 60 connectors
• surface soldering of the 240 sensors on the 60 boards
• through-hole soldering of the 60 connectors, 30 on one side and 30 on the other (as shown in the picture posted earlier in this thread)
• manufacturing of 30 3D-printed protectors
• shipping
• paypal fee
• coupon discount

Excluding the latter three items, the price is $7.26 per per sensorboard, or less than 2 dollars per sensor (the sensor itself is about half a dollar). In case you are wondering why 240 sensors rather than 176, it comes to:

1. The register division of the keyboard in sections forces me to have a few unused sensors in each section, or order a few 2-sensor and 1-sensor boards. The latter is much more expensive because they don't have volume discount

2. Ditto for the 3 pedals, which may require one sensorboard each (I might use something else, TBD)

3. I need/want a number of spares in case any is damaged in transit, assembly, when wired or just if there is "infant mortality", not to mention the possibility to run tests outside the piano when that is finally assembled and I won't want to touch it.

   All of these together, with price considerations, made 240 the "best" choice

It was easier to deal with this company than with the other two I tried earlier. We'll see the quality when stuff arrives: @JayKominek I plan to give you a couple of these sensorboards, and if I will be ready a couple of cables as we discussed at on github so you can test both. If you have a spare ADC board (even rev0 and even unpopulated besides the connector), that will help me planning cable routing in practice rather than on paper. I plan on waiting for a little bit before orderding the subsequent parts, after assessing how satisfied I am with these.

Please check the attached pictures of the assembled boards, especially the polarity or orientation of the components and tell me if there is anything wrong.

By the way, as your requirement to solder 30pcs J1 in top and 30pcs in bottom side, but this is a 1 layer board, the solder paste cannot be added in bottom side, we can only solder in top side, pls advise.

After your approval we will move to next steps.

Additional attention: Soldering for some through hole parts is not completed because we want you to confirm if there is anything wrong with the orientation. If not, solder paste will be added soon.

Alos some dirties on the board will be cleaned before shipment as well.

Obviously I replied

Thanks for your message. As far as I understand Gerber files and PCB manufacturing, the through holes for the J1 connector should be plated, just like vias. As such, soldering on the back side, even though it would not be connected "anywhere" on the back side itself, should be electrically connected on the front side. Isn't that the case with the way through holes manufactured with your process?