Not really, I've never seen a good comprehensive book about it. Two that come to mind:
a. ARRL Handbook. I have only old ones but that's what I learned my stuff from when I was young. Plus from mentors. Ok, that book is mostly about RF stuff but with modern analog parts just about anything is RF. An oscillating amplifier won't care about the fact that it was just meant to amplify a 30kHz signal :-)
b. Howard W.Johnson "High-Speed Digital Design, A Handbook of Black Magic". Not for prototypes and contains stuff I won't quite agree with but it's a very good resource for rookies. That is the reason I have it, to be able to point client engineer to chapter five point something. The book shows where pitfalls are and what to do about them.
But most of all, line up one or more consulting mentors. People whom the rookie engineer can send a schematic, sketch, photo, scope plot, whatever, and ask them "Now why does this go berserk on me?". Yeah, it may cost an hour or a few in fees but that's better than having a guy agonize over a recalcitrant circuit for days while your schedule is floating down the Klondike. Sometimes 15 or 30 mins of billed time are enough to cause a reaction like "Oh dang, that was it!". But it must be someone who is able to relay the "story behind the scenery" so it won't happen again next time. Just like a good fiberoptics guy thinks that everything in life is an etalon, an analog guy must learn that everything in life is a loop and comes with undesirable inductance.
Get lots of copper clad and never, ever, let the guy build anything on a proto-board that has no ground plane. Perf board without a plane has been the source of many outbursts of anger and hissy fits for engineers. If you must use it for a somewhat neat looking prototypes get the stuff with at least one full plane. This can be helpful:
The ARRL Handbook has quite a number of techniques in their building chapters, from dead-bug to circuit boards.
These days I do the "make a production-quality board" technique that John L. is advocating; if I'm worried about all or part of it I intentionally make the board really big, and make sure that there's plenty of space for green wires, and even dead-bug parts, around every part (or at least every part that I'm worried about).
For circuit elements where you're particularly concerned about unknowns you can often get eval boards from the chip manufacturer, distributors (well, at least Avnet), and places like SparkFun, and rope them together as necessary, with cables. I currently have one circuit for a client with a Jr. engineer that consists of three eval boards (ADC, processor, DAC) screwed down to a piece of plywood, with a terminal strip (remember those?) serving as a power entry point. Usually once you get things working at the "eval boards + cabling" level, you're pretty close to just making a final schematic and running with it.
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Most of the stuff we - all of us - design these days is fairly complex. We are using parts like fine-pitch, leadless, BGA, that are hard to breadboard. We have to lay out a real board sooner or later, so I figure go "sooner" and skip the prototype stage entirely. That has advantages:
You get to check the board-related issues too: placement, footprints, bypassing, impedances, thermals.
You can probably get the real thing built as soon as you could cobble up a prototype by hand. It may wind up being cheaper, too.
You may be able to sell the first version. Ideal time-to-market.
You can make several first articles easily. Give one to the FPGA guy, one to the programmer, one to the test department, keep one in case somebody blows theirs up. Take a picture of it and press release it ASAP.
It teaches good habits, namely *think* about the issues and bugs, and don't expect to test for them. Develop a culture of expecting to get it right the first time.
Basically I agree 100%. But ... there are times especially when it come to RF stuff where you want to try an idea right now. Maybe there is a bunch of very worried folks in the board room who want to know whether a proposed solution will do the trick before they head home on their afternoon flights.
Even if you have a fat budget and can throw all the rush charges they want at it, the Fedex truck with the boards in there won't pull up to your delivery ramp until some time the next day. Or the day after that. That is too long for some of my stuff.
I do breadboard simple circuits that I don't have confidence in, like to characterize a part or an unusual circuit. I don't consider that to be a "prototype" because I'm not trying to validate an entire product design, just one little piece of it. Call it Proof of Principle maybe.
Sure, but you could spend a long time prototyping something complex. We just go for rev A, as close to final and sellable as we can make it.
Nice. I use Vector 8007 for that sort of stuff, so I can use through-hole parts when possible. It's all 0.1 inch pitch, which makes small stuff harder. A bit of that busboard stuff might be just the ticket.
I remember a time at an ultrasound company where stuff wasn't working and a major group of big kahunas was announced. And that the new machine better be running by then. We worked like crazy, building some major new stuff on copperclad. We had the system running and buttoned up by about
10:30am and the group waltzed in the door less than an hour later. Because the first thing they wanted to do was to check out that new system. Whew.
Most of the time I use Wainwright Mini-Mount strips. You can snip them to length, peel off the sticky tape and press them onto copperclad. Not sure if the manufacturers still exists, at least I can't find them anymore.
I've used the Bellin Systems ones, but they have such long traces on them that RF stuff oscillates easily. Also the holes are plated, which means that you have to worry about their shorting to the ground plane.
A new favourite part of mine is the Skyworks SKY65050 pHEMT. Low gate current (for a pHEMT at least), runs well at IDSS, about 0.5 pF input capacitance when cascoded with a BFP560, doesn't need a bead on the gate. The BFP650, on the other hand, does need a bead. The bad news is that the Skyworks part has a 1/f corner of about 5-10 MHz. :(
I measured the input capacitance by using a 1 pF input coupling cap, measuring the gain, putting 3 pF in parallel, and measuring it again. The voltage divider ratios give Cin. The nice thing about that method is that it works at normal bias, and doesn't cause gain funnies at high frequency that might make the measurement inaccurate.
Sometimes referred to as the "get it right second time" approach. There's quite a lot of merit in it since you're bound to uncover something in the final production version anyway if it differs much from the prototype.
Neat, I 'reinvented*' this trick of wrapping a bit of coated wire around the leg of some input.. (through hole). hitting it with a step, to couple in a known bit of charge. (Much more repeatable than a pmt pulse.)
*the inspiration was the variable 'gimmick' capacitor.
I assume you have an old stash of vectorboard. At least I do and I hoard it. But if you had to price new vectorboard over a made in China PCB, I think vectorboard costs more. That 4x6 board is just shy of $30, and you still have the surface mount issue.
Surf boards are rather pricey too. I saw the post about Sparkfun, and have bought components on their boards. Sparkfun doesn't have that much of a markup, and a Sparkfun board is usually way cheaper than what a manufacturer charges for a demo board.