This is quite an interesting area. We do a lot of smaller volume projects for Australian Electronics Manufacturers making niche products and it is important to take an approach that will deliver real profits. You can't use the same manufacturing approach for 5,000 units that you would for 1,000,000 units. Some things to consider are:
- the engineering and design costs which must be amortised over the production batch
- the component costs
- the assembly, test, calibration and packaging cost
- the elimination of rework
- the elimination of warrantee returns
When you are making 5000 units the production test strategy become critical. You can't afford to spend $50,000 on a LabView based ATE with standard fixtures, racks and ancilliary equipment. It has to be a lot cheaper than that. And if you don't select a suitable strategy BEFORE you do the electronics design then they can't design that it from the beginning. Which leads to redesign late in the project. This is often treated as a design overrun when it really is poor project management.
Component cost is often given a lot of attention but you also need to consider the cost of assembling the product, testing the product, calibrating it if required, and preparing it for shipping. Sometime you can save $0.25 on component cost and add $1.25 of these production processes because of it. Looking at the total picure helps a lot.
And if they all just work and you don't have any rework to do, that is also a huge benefit in reducing the real cost. So this means things like a fully toleranced design where there are no part or parameter issues, preventing inappropriate substitutions at the purchasing stage, PCB layouts that take the PCB loaders capabilities into account so you don't get thermal capacity issue leading to poor joint quality. This is a big one. Sometimes we have to do redesign of other peoples work and the only real problem was they didn't take the manufacturing capability into account when they laid out the PCB. All the hand loaded or low volume produced samples were OK but the production batches had too high a percentage failure rate.
The final area is warrantee returns. No-one wants this. So good design and good manufacture helps but it is also a feature issue. I can give a good example of this.
Pump controllers are the first thing swapped out when a new pump installation is commissioned and it doesn't pump. Why? Because it is the easiest thing to swap out. But 80% of the time the problem is the check valve being installed the wrong way around. So most of these pump controllers go back to the distributor or factory and present with 'no fault found'. This is very annoying for everyone. So one of our clients recognised this issue and added a fault indicator to the pump controller. This would check that the pump was running and there was no pressure and if the priming period expired would give a fault indication. On both the case of the pump controller and in the manual this particular fault list causes and the top one was 'check valve reversed'. So now the pump controller didn't get swapped out automatically. And most of the time the problem was sorted on site and quite quickly. There were other causes also listed. The short verion was:
- check valve reversed
- flow valve still closed
- outlet valve still closed
- pressure switch stuck
If the pump was not turning then a different fault would show.
Their return rate dropped from 11% to under 1% and the cost of adding the display and the software logic that drove it was trivial by comparison with that outcome.
Low Cost Electroncis Manufacture requires attention is paid to all these areas, and particularly if it doesn't have high volumes to amortise the captial expenses against.
I've also done the two extremes: spent four years on a spectrometer before the overall power switch was turned on (debug took only three weeks thereafter), and OTOH designed multimedia powered loudspeakers for OEMs that shipped in multimillion quantities, and where, as r3son says, every penny counted and schedules were invariably tight as well. I found both activities exhilarating.
I have worked on both sides. While I agree that the design of low-volume or one-off products is liberating from a technical sense (bigger budgets, less restriction on the types of components used, higher technology level), designing for mass-market is much more challenging. Any engineer can design one high end "thing" to work well, given the time and budget, but try making hundreds of thousands or millions where every fraction of a penny counts.
Blog Doing Math in FPGAs Tom Burke 22 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...