How can you reduce the cost of your power supply design and development?

Suppose you have assign a task to develop a low cost Power Supply without compromising high reliability, safety and long life/MTBF value.

What are the first three strong points comes to your mind where you can optimize your SMPS cost?

Any new design Ideas, previous experienced, failure analysis, production, testing, business related issues/analysis are most welcome......

The key to a low cost power supply is low cost materials. For power supplies of any volume anywhere from 70-90% of the sales price is material (BOM cost). Large companies have two advantages.

First, they strictly enforce preferred parts lists and only allow engineers to use the preferred parts. Introducing a new part is a big deal and generally a high hurdle.

Second, the companies buy those preferred components in huge quantities and negotiate very low prices for those components.

For some application areas, like server power, the industry benefits because the components used are common across many power suppliers, giving the component suppliers even more advantage through high volume.

Takeaway: If you are doing a low volume design (low volume meaning less than 100,000 units per year) it will be hard to get a really low cost. You will have take every opportunity to simplify the design and get rid of every component that is not absolutely needed. Use the simplest topology available. If a simple flyback will work, use it.

Once the BOM cost is minimized, you have to minimize assembly cost. Use good design for manufacturability rules (like every component is surface mount and no manually constructed or installed wire harnesses). If you need to mount semiconductors to heat sinks use clips instead of screws. Use insulated power semiconductor packages to get rid of manual assembly of insulating pads and washers.

In summary, low cost requires the simplest design with the simplest assembly process.

It would be useful to know how many units you plan on making.

Sometimes companies want ultra low cost, and only a few 1000 units. Then you have to factor in the NRE.

Bob's points above are all good.

you have to determine how low you need to drop the BOM cost. And as Ray said, at quantities in the 1000's all this may not make much sense because every hour you invest thinking about it is a paid engineering hour and must be amortized.

If it is truly a mass product it pays to think out of the box. For example, my first mass product switcher does not contain any commercial PWM chip because those were too expensive. All just logic chips and discretes. Then the client wanted to push it down some more and we even looked at each resistor. Does it really have to be 1%? We dropped many to 10% or even untrimmed 20%. Just by that they saved the equivalent of a nice car, every year, for 15 paid consulting minutes. However, on a design that only runs 50 units/month that would not have made sense at all.

For most designs my three initial points I'd look at would be:

1. Magnetics. Is there a source that can deliver for less? Maybe in another country? I have had situations where there was a catalog inductor available but we went custom anyhow because it cost less.

2. IC. Can it be done without one? Is there a cheaper one? Many times I found that the non-mainstream companies (often smaller ones) can deliver modern parts at really good prices. Richtek, for example.

3. Look for FET and diode bargains but keep the heat sink in mind (aluminim isn't free ...). Digikey is quite useful here because you can sort by price. Drive them with gusto so you can squeeze the best out of a FET. For reasons I will never understand many switcher chips only drive at lower voltages such as 5V or 7V. I try to avoid those, or only use them if I can bypass or backfeed the internal linear regulator for the gate drive. Also, think about negative drive for a zippier turn-off. Less heat -> less required cooling -> cheaper.

Thank you all for participating in this discussion and sharing good thoughts.

I think most of the participants targeted to minimizing end product cost depending on number of production unit, assembly process, centralised BOM parts management systems etc etc. Where as some of them have mentioned about using simpler design, minimizing BOM components, magnetic, avoiding custom made parts etc . Nice ....

Aside from the NRE, what I see being done:

Single sided PCBs (I hate these, but they happen)
Discrete controllers rather than PI and ST integrated FET/Contoller
Old style controllers like 3842 with no fancy features
Magnetics from third tier suppliers (and the risks that go with that)
In some cases discrete controllers, self-oscillating (risky again, but once it is tweaked, it can work)

Along with the push to ultra low cost comes a price in terms of reliability in many cases. As manufacturers of power supplies jump around from one supplier to another to push the price down, they risk changes in parts that are significant and which don't show up until a few hundred thousand are in the field.

Classic case is that of the motherboards that used an inferior MPP core. After two years, the cores had degraded and the computers died for lack of power.