How do you prototype large projects?

[MarkS]

I posted this here because no other forum seemed appropriate, although I don't necessarily consider myself a beginner. Anyway...

Let's say, just as an example, that I want to create my own motherboard for one of Intel's new Sandybridge-E processor. This is rhetorical; I DON'T want to do this. Anyway, the CPU is a 2011 LGA and then there are the various other components like 268 "pin" DDR3 (usually x 4) and everything else. Dozens of components and thousands of connections. Just how, exactly, does anyone go about prototyping this? Clearly, breadboards are out!

I'm assuming that projects of this scale are prototyped. If not, at what scale are projects no longer prototyped and how does one go about creating and testing revisions?

[EEVblog]

They are prototyped on either a custom test pcb, or a shot at the final production PCB, depending upon your time-to-market pressures etc.
Stuff like that needs serious signal integrity you can only get with a well laid out PCB.
Depending upon the level of engineering effort put in and checking, it's easily possible to get a working system on the first spin of the PCB that you can then use for testing. Perhaps another spin or two after that to tidy things up and fixed and changes required.
For systems like this it's not uncommon to spend many months laying out the board, even for the first "prototype".

Even for my own projects, I often won't bother "prototyping", I'll simply have a shot at the final production PCB straight up.

Dave.

[mobbarley]

Even for my own projects, I often won't bother "prototyping", I'll simply have a shot at the final production PCB straight up.

Dave.

Absolutely - and with all these budget proto places now there is no reason not to.

[Teknotronix]

Even for my own projects, I often won't bother "prototyping", I'll simply have a shot at the final production PCB straight up.

Dave.

With regards to the power supply project, would you normally have breadboarded the design at say the Part 3 stage? Or was that just for teaching purposes.

[EEVblog]

With regards to the power supply project, would you normally have breadboarded the design at say the Part 3 stage? Or was that just for teaching purposes.

I usually would have been too lazy too I suspect 
But it was pretty easy, as all the components were through-hole, and it did make for another video.
Most of my designs are not through-hole like this, so they are not that easy to prototype without a proper PCB.

Something like my uWatch or uCalc for example was pointless to prototype on breadboard. I knew some switches hooked up to a microcontroller with an LCD would work, guaranteed. It was all about the form factor enabled by a suitable PCB, so you just jump straight to the PCB.

To do this 20 or maybe even 10 years ago was almost unheard of at the hobby level, as custom PCB's cost a fortune.

Dave.

[Short Circuit]

I posted this here because no other forum seemed appropriate, although I don't necessarily consider myself a beginner. Anyway...

Let's say, just as an example, that I want to create my own motherboard for one of Intel's new Sandybridge-E processor. This is rhetorical; I DON'T want to do this. Anyway, the CPU is a 2011 LGA and then there are the various other components like 268 "pin" DDR3 (usually x 4) and everything else. Dozens of components and thousands of connections. Just how, exactly, does anyone go about prototyping this? Clearly, breadboards are out!

I'm assuming that projects of this scale are prototyped. If not, at what scale are projects no longer prototyped and how does one go about creating and testing revisions?

At commercial level breadboarding is never done, except incidentally for some small sub-circuitry with high uncertaintity level.
All else goes directly to a PCB design. Dave calls this a production PCB optimisticly, but in real life it never is actually
But working woth a first real hardware reveals problems much quicker and easier than endless review sessions. And more fun too !
For really big projects I often make small tryout pcbs with various circuits on them. Pretty much the equivalent of breadboarding,
except the result is more reliable and repeatable, and there is no problem using all these ittybitty SMD packages.

[joelby]

A computer motherboard is also probably a bit of a special case. They are heavily based on reference designs, which are tested by Intel or whomever, and chances are you've produced a similar motherboard before so you can directly re-use tested sections such as power supplies.

For my own largish projects, I'll dead bug or otherwise prototype individual components that I'm not completely sure about, but for large chips this is difficult so I'll just go straight to a PCB, with the chip-specific parts generally based heavily on a reference design or evaluation board.

You can also ask the component manufacturer's field application engineers to review your designs if you're not sure you're on the right track - I've been stunned by how helpful they can be - LTC in particular.

[EEVblog]

At commercial level breadboarding is never done, except incidentally for some small sub-circuitry with high uncertaintity level.
All else goes directly to a PCB design. Dave calls this a production PCB optimisticly, but in real life it never is actually

Depends what your idea of production is though.
For probably the majority of stuff I've worked on, "production" is like 10 units 
So sometimes it's possible to get it right on the first spin. But yeah, usually that's a tad optimistic 

But working woth a first real hardware reveals problems much quicker and easier than endless review sessions. And more fun too !
For really big projects I often make small tryout pcbs with various circuits on them. Pretty much the equivalent of breadboarding,
except the result is more reliable and repeatable, and there is no problem using all these ittybitty SMD packages.

And if you are prototyping to get some performance measurements for example, then the exact layout can be vital.
Not uncommon to development entire prototype test boards and systems just to test the performance of one chip in a proposed design.
Other times you might need the proper layout to test thermal performance package issues or some such.
There are almost countless permutation and combinations when it comes to electronics design, and commercial and industrial and military designs etc can differ greatly in the "prototype" requirements.

Dave.

[MarkS]

Thanks everyone!

That really is the only logical answer, but I didn't know for sure.

[bfritz]

When I worked in the Portables group at Compaq, about 15 years ago, we would produce a PCB for the proto.  It often would exclude circuitry that might be needed in the final design, such as power supplies, battery charging circuitry, etc.  It was often meant to prove out some of the toughest layout challenges, and toughest signal integrity problems.

This board was also typically designed with a couple extra inner layers, that would extend past the normal layout size.  This was used for connections at various points for scope probes, logic analysers, and for connectors to power the board from bench supplies.  We would make the area within the normal size of the board as production ready as possible, and pull out the couple layers that would have connections with test points and other items for the final product.  It worked pretty well.

A new portable design, with a new processor family, might take 3 such layouts before the final form factor units, and that might have 1 more layout for EMI/RFI.

[SoftwareSamurai]

All prototype PCB that I've seen are always very large (in comparison to the final PCB) and always contain lots of connectors and parts that are solely for prototyping/debugging. In fact, one EE that I worked with told me specifically that when he creates a prototype PCB, he always exposes all signals on the top or bottom layers specifically so he can cut them and deadbug fixes/changes directly on the board.

If the circuit you're working with is already "proven" (i.e. it's already been PCB'ed and shown to work), then you'd be pretty safe to go with what you would consider a final PCB layout. But if you're working with a circuit that's new and untested, plan for a prototype PCB that's physically 3-10 times larger than you think the final PCB will be. (You should see the pictures of the first "prototype" of the first Sony Playstation. It was as big as a refrigerator!) Do things that make it easier to R&D on the board. (e.g. Keep important signal traces on the top and bottom layers, put in plenty of test points, even consider putting in rows of header through holes for things like JTAG, etc.)

Also, for "prototype" boards, I would suggest that you take a long look at your circuit and ask yourself "what if I have to change this part to X". Try to plan the PCB to make things like that easy. (i.e. Keep SMD parts farther apart to make it easy to get to them with soldering tweezers/iron; Plan to use IC socket packages instead of SMD packages so you can easily replace chips who's performance you're unsure of, etc.) One thing that I've found that helps a lot is to pepper your prototype PCB with LEDs (and resistors) for certain sub-circuits. (e.g. You can wire up a GPIO pin on a uController to turn a LED on and off. Then you can use that LED as a debugging tool in the code that runs on the uController.)

Good luck!

[Neilm]

When prototyping a circuit, I try to keep the layout as I would like it in a final layout. This is because I do it for work and have to meet various EMC standards. I like putting these into the EMC chamber to see how they perform from an EMC point of view - it allows me to spot any problems before it gets into the unit.

Neil

[Zad]

I think a lot of it depends on the specific application. In the case of a PC motherboard, there is already a lot of IP with regards to design rules, constraints, connector / component libraries, modules (CPU PSU, PCI-E array, additional USB + SATA etc). Intel and AMD also have their own reference design, which manufacturers adhere pretty closely to.  In other words, it is just an evolution of existing designs.

In my case, I do a lot of RF and "mixed signal" work. Inherently, these lend themselves to a modular design. So, in the early stages, you can hack together your own module, or buy an off the shelf "whatever"; for example a DDS oscillator, mixer module, 10.7MHz IF strip etc. Just how far you want to go with prototyping is down to the engineer's judgement of their own ability and experience. It also depends how quickly other people need real hardware to start developing software with, or for hooking up to other hardware.

[gregariz]

I would second the idea of modules or tryout boards above. I have a bin full of them. Most analog/rf chips I find often have overly optimistic spec sheets so theres often no real way to get at what performance specs you may actually get on a complete layout. If I do a complete layout I'll typically place test ports all around the signal paths so I can get in to test the various sections of the circuit. If the circuit is heavily based on a previous design that is stable, I will go for the final layout first time. If it uses new chips and a new layout I'll proto with modules first. On the rare occasion I don't do either of those I'll actually use a breadboard - but it'll likely be for something discreet.