Evaluating the need for prototyping

[sentry7]

How many of you guys actually prototype a new design?

What's your favorite method?

Is there ever a point in a designer's/engineer's profession (or career) where they are so good at electronics that they just don't need to prototype things anymore?

[Ice-Tea]

I see engineers around me that seem to think they don't need it. I still prototype a lot. The stuff I've done before and I'm confident with, not so much, new stuff and analog stuff: a lot.

[ajb]

Depends what you mean by prototyping.  There's lashing up something on a breadboard, there's bolting together some dev boards/breakout boards in a permanent fashion, and there's building something that looks almost exactly like what you intend to actually put into production.  Arguably, the first scenario is experimentation, the second is a proof of concept, and only the last one is a prototype--but there are some gray areas in between as well. 

Some projects simply don't lend themselves to the first two.  The minimum workable hardware may be so complex to assemble (eg, something like a PC motherboard) or may require such specific and tightly controlled manufacturing constraints to work at all (RF) that you may as well start designing for production right off the bat, because a proof of concept board would be just as expensive as the production board (and has the disadvantage that you'll DEFINITELY have to pay for new tooling when you build the production board, whereas if you design for production initially you might get lucky and not have to make revisions that will require new tooling). 

Ultimately it's a risk/benefit decision.  Personally, I might build up a test setup for a very specific aspect of the circuit ahead of time, but mostly I go straight to a producible design and expect to make a revision or two--given the type of projects I work on, that's a better use of my time and development budget than trying to cobble something together a test system and then still having to do the production design afterward.

[G0HZU]

Is there ever a point in a designer's/engineer's profession (or career) where they are so good at electronics that they just don't need to prototype things anymore?

There probably will be times where you will be asked to design something in a short time and go straight to an initial production run to meet the customer's needs. This has happened to me quite a few times. This could mean designing a high performance RF circuit that would go on a fairly large multilayer board with maybe 400 components on it. A typical board would have Rogers 4003C as the outer (RF) layers and the inner layers would be FR4.

The design process can be de risked considerably through the use of modern synthesis and simulation tools allowing lumped and distributed RF BPF, BRF, LPF and HPF filters to be designed and manufactured without any need to prototype them. The same goes for amplifier circuits, RF switches using PIN diodes, couplers splitters etc etc.
But the whole process is still very expensive. I've had to design boards like the one above where I'm not even allowed time to change the component values to optimise the final design once I have the completed board in my hands. Not even a resistor value can be changed! It has had to be design/build/test and meet all the spec requirements with no changes allowed.

[MagicSmoker]

...mostly I go straight to a producible design and expect to make a revision or two--given the type of projects I work on, that's a better use of my time and development budget than trying to cobble something together a test system and then still having to do the production design afterward.

This is my approach as well. I expect to have to make a couple of board revisions (and component tweaks) to zero in on an optimized design, but usually the first one works... mostly, anyway. 

...I've had to design boards like the one above where I'm not even allowed time to change the component values to optimise the final design once I have the completed board in my hands. Not even a resistor value can be changed! It has had to be design/build/test and meet all the spec requirements with no changes allowed.

Jeez... that sounds nightmarish. What kind of pointy-haired boss allows such stupidity to occur more than once?

[G0HZU]

Jeez... that sounds nightmarish. What kind of pointy-haired boss allows such stupidity to occur more than once?

In reality, the board will be allowed to be modded if there is a serious omission or problem but you would be surprised how successful this approach can be. A typical first run might only be 10 boards and any minor warts would be fixed on the next (big) production run. If you allow documented tweaks and changes to those first 10 boards then this adds a lot of red tape and time. So the design goal is to get it right first time and not waste time on prototype X, A, B models etc.

It might all sound crazy but this is all customer driven. If you were the customer and you URGENTLY needed widgets that prevent loss of life would you want the first 10 systems as soon as possible? There could be 10 different PCBs in a complete system. Try doing a project plan for X, A and B models of all those PCBs all going through the PCB design/test process several times. It just won't meet the deadlines demanded by the customer.