design process advice

hello i need advices from any experienced people here... if you have a microprocessor-based embedded system board (sbc like), you got design on schematics (but you dont know if it's going to work, because it's on paper only and you can't test it), how do you validate the design before you send it to pcb manufacturer for prototype?

also, can you tell me what the basic procedure involved in designing embedded systems, like from sketch, component wiring, etc, until production? and is there any way we can save money by avoiding mistakes before sending it to pcb maker ?

If you are copying from a development board, make sure you've copied accurately. You might be able to try things out with a breadboard or lash something up with matrix board. Go to someone cheap (PCB Pool, Olimex etc.) for the prototype and assume that you are going to get things wrong before they come right. That's going to happen if you are inexperienced anyway, and probably when you are experienced too.

Blag what you can from reference designs. Read the data sheets and app notes very carefully. Think carefully about layout: you need a good low- impedance power supply system, where switching in one part won't cause glitches in another part. Think of where the return paths for signals (esp. high speed ones) go. Make sure bits that don't belong together are separated, electrically as well as physically. Have a good CAD system. Keep your notes up to date. Have someone handy to explain every bit of the circuit to as you go along... there's nothing like saying it out loud to expose the fatal flaw in the best laid plans.

Check what you are doing, how it works, and how you've implemented it very carefully indeed. You might actually SAVE money by getting a prototype fairly early to make sure the concept works... nothing worse than finding the big misconception a week before you are due to demonstrate a working unit to the customer.

Paul Burke

You build a breadboard.

My old boss used to say 'The only people not making mistakes are those who are doing nothing'

Ian

--
------------------------------------------------------------------------------
|   itb@iname.com          | Windows - just another pane in the
glass        |
|   Ian Thompson-Bell     
|                                                 |
------------------------------------------------------------------------------

[...]

Experience will help you avoid mistakes .

You gain experience by making mistakes and learning from them...

Regards,

-=Dave

--
Change is inevitable, progress is not.

Saving money on a PCB is quite easy these days - we just had the pleasure using

.... 99-USD for two prototype boards - shipped with Fedex overnight.

If your schematic is made - and the most important sub parts of the circuit have been carefully evaluated (e.g. bread boarded) ..... hit and trial - the first PCB will never be perfect anyway. Make sure you surf for hints on PCB design before the layout is done.

Kristoffer

I use some manufacturers in Brazil:

Digicart - snipped-for-privacy@digicart.com.br ($66 for 5 prototype boards) Micropress -

(more expensive, but makes multi-layer boards) PCI -

Verification of the design could be achieved in 3 ways either in combination or in isolation.

1. Simulation
2. Prototype
3. Peer review.

Each method has a limited scope of coverage and each have associated pros & cons.

Typically the engineers at my work, go from schematic->PCB layout->

straight to copper. From experience they have found that breadboard prototyping is a waste of time. Particularly if you're using some inferior medium like Veroboard or similar (talking about micro designs here). One seems to be debugging the breadboarding rather than the design.

So the only verification done at my work is design review of the schematics and probably a sanity check on the PCB layout. Also (if possible) plan (in time & budget) for at least another interation of the board.

To perform a proper review, the reviewers must: a) Be in a correct frame of mind & had inspected the circuits prior to the review meeting. There's nothing worst than going into a review meeting and having the reviewers only just viewed the design. All participant must be familiar with the material in question prior to the meeting. If this is not the case, then reschedule the meeting until everyone is better prepared. b) Setting out a clear & concise list of criteria. Throwing a circuit at an engineer & saying "Here, check this" is not the right approach. Possibly each review member should be given a particular task such as checking the supply rails & current consumption or checking that all unused inputs are tied low, for example. c) Reviews should NOT be confrontation or "slinging" matches. The aim of the review is to expose errors in the design. Although difficult, "style" should be removed from the agenda or better still have style issues be part of a company guideline, so that there won't be any disputes. d) The convenor of the review should be strict & keep the discussions in scope. e) "Share the responsibility". Have reviewers sign-off a Review Sheet

-- having people put their signature on a piece of paper is always a reality check.

Experience is invaluable and the use of prior designs which have been tried & proven is a plus.

Ken

+====================================+ I hate junk email. Please direct any genuine email to: kenlee at hotpop.com

Good advice. I can't breadboard my 357-pin BGA processors, so things need to get reviewed frequently. Get the symbols reviewed for pin accuracy, get the schematic reviewed before release to the layout guy and get the layout reviewed before it goes to the PCB house.

When building the first prototype, build it in sections. Place the power supply, processor and memory down first. Then fire those up slowly, and keep an eye on the current meter. Magic smoke is precious....keep it in the part. Add another circuit, and keep going until you get the whole thing together. If you are new to the process, you will probalby have to kludge stuff together.

Elroy