Consumer products guide the way in low-cost product design
This month I'd like to investigate this problem: how do you build something cheap? I
don't mean like those junky plastic gizmos your offspring get with kid's meals at a local
fast-food joint. Instead, how do you bury electronics-including a microcontroller, a
numeric display, signal conditioning and power-control circuits-all into a module that
sells for $30? A great example is my son's digital alarm clock, which includes a
timekeeping/LED driver chip, four 7-segment LED digits (plus colon), power supply, case,
power cord, speaker, three circuit boards and four switches-all for only $5. The simple
answer to these questions is, I have no idea! In fact, I'm envious of these geniuses that
pull off these everyday engineering miracles. However, I've begun looking into the topic
and am prepared to begin filling you in on some of their magic.
Price is relative
To recap my experience, the products I generally design sell to the ultimate customer
for many thousands of dollars. In addition, the volumes are fairly low-a hundred of
something is good, and a thousand/year of something is great. Over the years, I've found
that the selling price for the types of products I design typically ranges from four to
ten times the "build" cost (parts, assembly, test and calibration).
However, for high-volume products the world really changes. For example, a client once
insisted that the build cost had to be no more than $17.20. If the design came in at
$17.40, he was going to cancel the project. In other words, on an annual volume of 50,000
units, being over in price by 1.2% was a product killer. Obviously, in that marketplace,
profit margins were far from the 4X to 10X range I was used to.
So what must you do to meet those margin requirements? The first step is to forget
those fancy new ASICs and microcontrollers advertised in trade magazines. The specs look
great, and they can solve all your functional design problems, but you can't afford them.
For example, ASICs seem wonderful-a part designed just for you, no second-source problems,
and you control availability. However, to use those devices effectively, a design can't
change (and we always design things perfect the first time, right?), and the volumes must
be in the tens of thousands of units. In addition, chip manufacturers seem to have a hard
time remembering that there's more to a product than their fancy chip. Just because you
can get an ASIC for $8 doesn't mean that you can sell a product for $9!
Another great area of misrepresentation is the microcontroller. It's amazing how many
microcontrollers are available for around a dollar. They can drive liquid-crystal,
vacuum-fluorescent or LED displays; encrypt garage-door openers; and control power-factor
corrections in motors. But try and use one, and you'll discover a number of caveats. The
single biggest hurdle I face is that you must buy them 5000 pieces at a time.
To understand why this limit exists, remember that to keep costs down, these miraculous
little microcontrollers use masked ROM for program memory. This design decision, in turn,
imposes some requirements: First, over the course of a year you must purchase at least
5,000 to 10,000 pieces. Second, once the program is in the microcontroller, it can't
change. Hence, if you find a programming error in the microcontroller-or want to add a new
feature-you've got to throw away your existing stock because there's no way to modify it.
Third, in addition to the cost of the chips themselves, you must pay a masking charge of a
couple of thousand dollars or more. Fourth, you're generally stuck with one supplier. Of
course, if you can meet these requirements, then you've got some wonderful parts to choose
To solve issues related to custom masking ROMs, chip manufacturers developed the OTP
(one-time programmable) device. They replace on-chip ROM with EPROM. Therefore, you
program each microcontroller the way you want it, at whatever volume level you want,
whenever you need them. If the program's wrong, you still have to throw the part away (to
hold down cost, they have no window for erasing the EPROM), but it's only one part, not
5000. These devices offer the reduced board real estate of a masked part while still
allowing some control of the program within it.
Unfortunately, while OTP parts seem like a great compromise, they have a dark side,
too. To begin with, they're expensive. Although I don't really know the reason for their
inflated prices, if you're trying to design for the absolute lowest cost, watch out for
these panaceas. For example, I once designed a circuit two different ways. The first
approach used an 8051 derivative with the I/O and memory I needed all on one chip. I then
redesigned the same circuit but this time used an 8032 (no internal ROM), an 8255 I/O port
plus an external EPROM and latch. This circuit had more parts and took more board area,
but the final cost-including assembly and test-was still 30% less than the single-chip
To pound on this point a bit more, while investigating 8051 variants, I found the
"perfect" part for my project. It had all the I/O pins I needed, the counters
and PWM as well as the right amount of memory. But when I looked up the price each unit
cost approximately $60! I'm sure there exists somewhere one special application demanding
a $60 8051-style part-but not in my designs!
High Volume 101
As I said at the beginning of this column, I don't know how to design for high-volume
applications. What's more, I've never heard of a course in a university about this topic.
Where do people learn this stuff? Are there books, journals, magazines, apprenticeship
programs on the subject?Go ahead and e-mail me your thoughts and comments, and I'll try to
write a column based on the information.
In the meantime, let me pass on some of what I've learned about high-volume electronic
- It's often cheaper to throw something away then repair it. We've all experienced this
aspect in our personal lives. If a $5 digital clock loses a segment, most people throw it
away and buy another one. In a similar vein, I've got one client who, when designing a new
product, builds ten prototypes and then produces 50,000 units. If that first lot doesn't
work properly, he's found it less expensive (in terms of the cost of a missed opportunity)
to throw away the 50,000 items than to incrementally push production up to the 50,000
point. In his market (consumer electronics) the cost of missing a window of opportunity
was significantly more than the price of the 50,000 units.
- It's better to steal someone else's ideas than to learn by your own mistakes. I was
having a terrible time getting cost down on a device that included a microcontroller,
numeric display, keypad and ac load controls. Then I bought two broken microwave ovens and
ripped them apart to learn about their design. They had the same functional
characteristics I was looking for and yielded lots of information for a minimal cost of
$20 for the microwaves.
- The $5 digital clock mentioned earlier is a engineering wonder. Past experiences tells
me to buy 7-segment displays for the time indicator. This clock, though, uses seven
surface-mount LEDs for each digit. Each LED is in the center of each segment. To make up
the four digits and two colon points brings the total to thirty LEDs. Mounted to the board
is a plastic mask about 4 mm thick with cutouts for the LED segments. On top of it is a
red translucent filter. The result is an extremely low-cost display that's
indistinguishable from a 7-segment display but doesn't create a dependence on a single
- Another example from the clock concerns switches. In the clock, the switches are all
snap domes that provide circuit contact on the PC board while at the same time providing
tactile feedback. The low-cost trick here is that the manufacturer uses scotch tape to
secure the snap domes! Why not? The switches are in a horizontal plane, so there's no
worry about them sagging from gravity. The resulting switch cost is for a piece of spring
metal, a PC-board trace and some tape. It doesn't get much simpler.
Obviously, there's a lot more to high-volume product design. For example, I haven't
touched on labor costs because even if labor were free, the parts costs in a traditional
design are expensive. Again, the secret to good design is to meet all requirements, not
just the technological ones. So until next time, I look forward to hearing from you.