Hacker News Reader

Comment by setquk | Hacker News Reader

Trying to express a circuit design onto any fixed layout has non trivial side effects that you don’t want. Parasitics and poor grounding are a nightmare. I don’t know why people use prototype boards at all.

I am using this construction technique which is good down to DC and good up to 1GHz or so. I’ve built very non trivial things with it:

https://pdfs.semanticscholar.org/2403/db840e6874e29f55ad9a45...

View on HN

Manhattan is great for discrete components but doesn't work that well for digital with small smd ICs.

So it's best suited to analog but for digital, prototype boards -such as the one from Mike's Electric Stuff linked in the article- are pretty good for prototypes or one-off products. There were popular methods in the past, like wire-wrapping, but unfortunately, they are no longer favoured. That used to be a good way to build fairly complex digital boards.

Nowadays, going direct to PCB is also a viable solution since they are very cheap and reasonably fast to get.

Totally agree with direct to PCB if the problem domain does not fit manhattan or dead bug. I usually do that if anything is non trivial anyway as it's less of a PITA generally

When such things become a concern and when not? Should I be concerned about specific layout of traces and their parasitic properties if I'm going to just interface low-speed digital components with microcontroller? I heard about people building whole 80's era computers on breadboards with lots of long jumper wires, and even that seem to work.

The very general rule of thumb is anything greater than 10 MHz switching frequency. To get an accurate answer requires physical stimulation, or "build it and see if it works or not".

I could generalise it as anything with fast rise times. A 74lvc14 can kick out a 400ps edge even if you clock it at 1Hz for example. That's got quite some harmonic energy.

Also it's usually the really slow stuff I have that oscillates madly :(

I wired up a breadboard this week with a few sensors I'm evaluating for work, and got a nice reminder of why we never use breadboards any more - 0.1" headers on 6" wires were producing undershoots up to a volt on a 3.3V I2C bus.

It proved the concept, but possibly at the expense of some input clamp diodes.

One thing that confused me in this document is how are the traces insulated from the underlying copper sheet ? I definitely don’t have the EE foundations to see how that isn’t a giant short circuit. Could someone eductate me?

The pieces glued down are small pieces of circuit board, not just pieces of copper, so the circuit board substrate is always between the copper layers.

Are you sure? In the parent comment that links to the Manhattan style , the figures and photos seems to indicate metal on metal, with glue between the two... what detail or context am I missing?

Yes, I am sure. "Copper clad" - board with copper cladding on it. Here's a higher-quality picture taken at an angle from a different article where you can clearly see that it's not massive metal, but fibrous board:

https://i0.wp.com/www.pbase.com/daverichards/image/136603883... (from https://aa7ee.wordpress.com/2011/07/24/the-wbr-a-simple-high...)

EDIT: in the PDF, page 8 also has an image where you can sort-of see it.

But this has a METHOD only available to CUSTOMERS.

Yeah this looks terrible and way worse than even basic prototyping skills with copper clad, but hey if you can sell it to people who don't know any better, why wouldn't you?

Relevant XKCD: https://xkcd.com/670/

Uh, what? The author doesn't appear to sell them, and the files to make the boards are open-source and linked from the article.

Indeed. That comment was a likely a reference to my prototyping platform I sell now and then. It is a plastic array of holes on a 0.5" pitch, and a set of custom PCBs with common functions, adaptors for stuff like Arduinos and PIs. The idea is to deliver something hackable and easy to understand to my innovation customers who ask for this. I often get asked to build a system with off-the-shelf bits, on a fast turnaround, so it can be handed to a junior industrial designer who can fiddle with the concept until they have enough information to write me a proper requirements spec for a custom PCB.

Then I make a pre-production final-form prototype that will work, with cost optimisation.

Prototyping means different things to different people, but the customer is always right, and mine are always delighted with the results.

Crazy-proto is not a product, is open source and was just for fun, after I went off on a tangent when thinking about making a little SOT-23 PCB I needed.

And yeah, I know the parasitics are dreadful - so what?

Don't take it too seriously guys!

Reply on news.ycombinator.com