Techniques and strategies for prototyping electronics circuits (2020) | Hacker News Reader

Techniques and strategies for prototyping electronics circuits (2020)

youtube.com

79 pointsbrudgers26 comments

We do zero prototyping like this anymore. Actually, we almost never carved copper plates anyway, we would just get perf boards with plated holes and wire everything together.

Everything gets a pc board made now. If you can't buy a dev board from the mfg or a 3rd party, then you make your own.

Surface mount mostly killed it off. But I don't miss spending hours looking for a problem that turns out to be just some breadboard issue that broke when you jiggled it wrong.

I've been waiting for the 3d printer conversion that deals with all the drilling, positioning, etching, gluing together multilayers, and eventual pick and place with optical inspection.

Someone has to do this, eventually electronics hardware prototyping has to stop lagging behind several decades.

Why would that make economic sense when the PCB makers already have the necessary machines and I can get a small 2-layer board delivered straight to my desk for a few tens of dollars? Electronic prototypers are not waiting around for some hoped-for long-tail revolution, because it's already possible to click a button in your EDA program and get hardware via fedex almost instantly. And we've had this capability at monotonically decreasing prices for over 20 years.

Because tomorrow is not almost instantly, except in the land of circuit boards.

Your argument doesn’t pass a self-integrity check. You can get boards for a few tens of dollars. You won’t get them today.

You can get boards tomorrow, or if you happen to be close to a fab and submit your order before lunch maybe same-day. This doesn’t cost tens of dollars.

I think the main thing is it’s really rare to just have one project in the works anymore. So you buy the 5day turn and work on something else in the interim. If somehow you really blew it in the project planning stage, and you simply must have it now, then I guess it’s kinda a toss up. In those emergencies rarely does cost matter but definitely your geographic location will limit options.

I guess this is the point — the workflow is adapted to the lack of immediate feedback.

A soft goods shop I know bought everything they need to do prototypes in house. Now they can crush everyone else on delivery schedules and are fully self-determined and therefore also more reliable on those timelines than anyone else. Those two things together draw very lucrative projects.

Imagine if you could do the same thing with electronics, moving prototyping entirely in-house and modifying the workflow to match. It would create an entirely new class of shop, and I would bet a lot of interesting changes would flow from that.

Yeah but neither will a multi-axis robot that can etch and drill and pick and place and solder. It's going to cost an absolute fortune even with fantasy leaps of technology.

> Yeah but neither will a multi-axis robot that can etch and drill and pick and place and solder.

So simplify: mill instead of etching, which also handles drilling. This needs 3-axis movement, which is all you need to drop solder paste on pads and do pick and place too, you just need a tool changer for each task. Not a trivial project either, but it's been done and certainly not for an absolute fortune.

Final step is reflow soldering, and this can be done using just a heated plate [1]. This all seems doable for most hobby sized PCBs.

[1] https://www.solderreflowplate.co.uk/

Yes; milling is fiddly but because the forces are so low (and you can reduce the mechanical accuracy needed with depth probing), it's possible to do reliable pcb milling for similar to cost of decent fdm printer.

People are doing this today but everything is jank, nobody's repackaged it with prusa-like reliability at hobbyist prices AFAIK.

Ideally if you could get v thin copperclad with a substrate that doesn't produce toxic dust, with the right software and jigs you could produce "basic" multilayer boards(1) with 1 "pin aligned" flip and a few manual tool changes for drilling etc.

It all seems possible today but has never been packaged nicely and made cheap, I suspect because the market is not big enough.

(1) one big issue... the boards produced would be vastly inferior to "real" multilayer boards, would mainly be usable at low frequency, you could get complex layout / routing but they'd suck at high frequency and in several other ways. You would need to relayout / redesign to target a "real" process, which is somewhat undesirable for prototyping.

Someday you will start with a sheet of copper with green soldermask. A laser will cut out the traces then take off just the masking where the component leads need to be soldered.

They can pretty much do that now. But nobody has a nice solution for thru-holes yet. And single sided boards are pretty boring.

This wouldn’t work because you have to somehow remove most of the copper under the soldermask without removing the soldermask.

It's just lines of bare fiberglass showing. That doesn't hurt anything.

but it's all about the managed lag.

how else would all those billions in R&D would be worthwhile for investment institutions?

as I see it (and stretching my reasoning), the lag is also part of what maintains the prestige of many academic and research organizations.

the billions in R&D are not all about the outcomes, a lot of them are spent making sure it's really damn hard for any rivals to catch up. how exactly? I cannot know but I can infer it's got a lot to do with having nobody able to see the whole picture, anybody can only know either how to design the chips, xor how to build them.

if everybody is as good as MIT, then MIT is no longer MIT. somebody has got to make sure some of those 3rd party (and far away) institutions stay there, in the back.

if everybody could do "2nm" process (whatever that means), then TSMC wouldn't be ahead of Intel, and so on...

Some of the prototypes he shows in his videos are really beautiful, or I'd just be echoing the "just lay out a board" chorus. There are people you will never get away from this kind of prototype/one-off construction and that's OK. You'll still see it occasionally in PCB RF design, too. I.e. if you want to do a bunch of experiments (antenna designs, etc) today and don't have a board mill.

But if you are looking for advice on learning to do this stuff, it's a backwards approach for 2021.

If you have the materials on hand, making one of these boards will be quite a bit faster and cheaper than laying something out and getting it quick-turned.

The downside is that the quality won’t be as high, you’ll only get a single copy (or however many you make by hand), and you can’t reuse the work in a new design like you could with a CAD design.

Still, there are definitely some advantages to being able to whip together a test unit like this.

Sure. It depends on how much you value time, how much else you have to do, and what the goal of the design is. I don't work or lead on one-offs much anymore, myself.

These look like great tips from someone who knows their stuff.

Yes, definitely. He showed a mix of through hole, SMD, Manhattan and Wirewrap techniques proving that he knows what is the best approach for each building need. I also liked how he used copper foil to add ground planes to perf boards. That is also used in non-cheap electric guitars to screen pickups slots from induction of mains hum. Also, I couldn't recommend more the use of leaded solder, which works so much better than leaded one.

I also use standard solder at home but we used lead-free solder at work when I had was in a hardware engineering org and it really didn't seem to cause trouble. I think people who don't have good success at first with lead-free are probably using the wrong tip size, have set their iron to the wrong temperature, or have an iron with poor tip temperature control. You see the same troubles when beginners use leaded solder, but people forget about those early experiences.

Since I'm still somewhat bad at soldering: what are the symptoms of having the wrong tip size?

If you’re new to soldering, a good “tip” is to use a copper coil rather than a sponge to clean the tip. And keep the tip tinned (keep adding a bit of solder and jabbing it off in the coil.) this prevents thermal shock and prolongs the life of the iron’s tip!

https://www.adafruit.com/product/1172

If you have a tip that is too small, then you will not have enough heat, i.e. thermal flow, getting to your solder joint. Note that heat is different from temperature! A bic lighter is hotter than your oven but it will not roast a turkey.

Symptoms of too small a tip - it will take too long to heat the solder joint. Pads and traces can lift and other things thermally connected to the joint may be damaged just like they will if the tip is too hot, but your flux may not activate and your solder will flow badly (or not at all) or make a cold joint like the tip is too cold.

Too large a tip and you can't do fine work or you rapidly heat everything around your part to soldering temperature.

Recently did my first project that wasn't utterly trivial and was surprised how quickly solderless breadboards start becoming an issue

I've moved a few projects I really liked from solderless to these: https://www.adafruit.com/product/571 because they have the same form factor as the solderless breadboards.

Those solderless breadboards seem to be too much trouble.

I still like the perf boards with the little copper rings for each hole.

This is a very good form factor. There are supposed to be enclosures having a perfect fit to these boards.

In the video, the first IC he shows a spec sheet for is a MCP9700, an analog linear temperature sensor. He doesn't use it in a circuit here.

But it's not a very complex IC, plus has the thermistor within the package, and only needs 3 connection points, so the traditional through-hole TO-92 package version looks just like a simple 3-legged silicon transistor. Not smaller than a grain of rice like the SOT version.

You power it with a steady DC supply of about 3.5 to 5V on the + and ground leads, and it basically outputs the temperature as a steady analog voltage on the third lead. As the temperature of the component changes, the analog output voltage changes accordingly. For the MCP9700, over the range of 0.1V to 1.75V output, at temperatures from -40 to +125 Celsius.

There are nice handy digital displays that contain their own analog-to-digital converter, and if you hooked up the analog output lead from the little IC, to the input of the display, you would see those 0.1 to 1.75 volts digitally on the display. To make the display show the temperature in desired units like Fahrenheit or Celsius, ideally you would need to adjust the "span" of the display while subjecting the sensor to known reference temperatures such as a low of 0C (ice point of pure water), to a high of 100C (boiling point of pure water at sea level). This span adjustment not only changes the slope of the linear function being displayed, but also calibrates that particular sensor/display assembly at that particular DC supply voltage against two physical references which could be revalidated in the future if desired.

Different components or different batches are only within a few degrees of each other, according to the spec, but calibration of any one component similar to the above technique can correct for that.

As a finishing touch once the slope is set as realistically as possible against the high & low reference temperatures (or alternatively reference voltages), a final "zero" adjustment is made which now changes the intercept of the linear function being displayed while leaving the slope the same. IOW use a third reference temperature somewhere in the middle between your low & high 0C & 100C, and finally adjust the "zero" control for the display to match this "mid-point" reference thermometer. For instance if you were going to use one of these sensors to display room temperature most of the time, after setting the span, it would be good to check it against a known good thermometer at a temperature about 25C, then finely adjust the display to match the reference using the "zero" control.

If the particular display doesn't have adequate zero & span controls, you may need to come up with the proper peripheral components like variable resistors to make the adjustments. You may need to set your own decimal point too.

OTOH you could always go all the way and make your own A/D circuit with nixie tubes for display :)

They used to have sensors like this at Radio Shack and I used one to make a research temperature controller good to within 0.001F. No display, the TO-92 sensor was submerged in silicone oil in the bottom of a narrow glass tube mimicing a thermometer, had 3 wires leading out of the heated zone into a little box containing a high-gain fixed-slope comparator, and one external knob for "zero control". A separate certified thermometer in the same heated zone had to be there anyway.

Also, notice at 11:15 the way he carefully handles the lead-free solder, making sure it's all in its proper place so that problems will be greatly minimized.

Well, this certainly looks like a great bridge into the world of surface mount parts for me. I was doing PCBs using paint programs back in the 1980s for the few boards we needed to repair 1960s vintage control systems.

I'll get my $200 cnc mill going for the next step, some day.

Reply on news.ycombinator.com