[time-nuts] wtd: WWVB info
kb8tq at n1k.org
Sun Aug 9 19:33:07 EDT 2015
If you never have tried to keep an IC in production, there are some basic things
that may not be very obvious:
1) Chip geometries shrink fast. A 4 year old production geometry is essentialy obsolete.
2) Manufacturing lines either are retired or re-tooled to the new rules on a regular basis
a line that is still running the same process and gear it had 4-6 years ago is a rarity.
3) Digital stuff shrinks with some fairly simple rules. You still pay (big) bucks to
re-tool the masks for the new process. That phone call comes in about every 4 years.
4) Analog stuff does not shrink with simple rules. You re-model and (effectively)
redesign the part each time you change process. Add that on top of the charges
for the digital process.
You would *think* that keeping an old like open would make sense. Basic math
unfortunately is not on your side. Make 1/10 as many chips on that line (and 1/10 the
batch sizes) and the cost goes through the roof. That’s fine for those who can afford
to pay $100 for a chip that used to cost $2. For the rest of us, not so good a solution.
No matter how cool all the design rules get, you still have to go dice the wafer. That’s a
process that does not shrink much with time. There is a minimum size they can dice a wafer
down to. The same ~1 mm x 1 mm rule works today that worked back in 1970. If you have a
function that uses one gate, it’s still going to be on a minimum sized die.
The new process costs less per transistor than the old process. It probably costs more per square
yard of silicon. A chip that made sense at some number of devices on the old process makes
more sense at 4X that number of devices on the new process. Effectively you get a bunch
more capability for free on a minimum sized die.
A consumer IC will sell in the “> thousands per day” range at it’s peak. A successful IC will sell
at a 10X multiple past that. That’s what gets the prices down to the level that we get used
to seeing. Drop back to <50K a year and you get a phone call from the foundry about “last
Can you move your gizmo to a new foundary when you get that call? Sure you can. Expect
to pay for new masks and all the modeling runs that go with them. There’s a couple of bucks
onto the price of each IC you make.
If you are selling into a cost sensitive application, (and who isn’t), the expectation is that the price
of components comes down at a fairly steep rate every quarter (or at least every year). Start bumping
the price up and your customer is going to look for an alternative. Down goes volume some more.
None of this even begins to get into the test and quality assurance part of keeping an IC going. It
also does not look at things like field support, inventory, and marketing. None of those things are
That all seems a bit much. It’s not. I’ve been in the middle of a *lot* of these phone calls
over the years. There’s been a lot of money spent on new masks and redesigns. The amazing
thing is not that IC’s go out of production on a regular basis. The amazing thing is that any of them
stay IN production for more than 8 or 10 years.
The net result is pretty much always that the simple old devices get replaced with ever more
complex new alternatives. The new ones may be harder to find for a basement hobbyist . The guys
who use them in volume know right were to get them. The volume of basic IC’s has been dropping
for years and will continue to do so. (That’s in pieces, in dollars … wow!)
> On Aug 9, 2015, at 4:20 PM, Tom Van Baak <tvb at LeapSecond.com> wrote:
>> This data sheet is one of a few receivers from a few vendors:
>> Looking at the data sheet link shows the internals of the receiver chip.
>> This chip outputs the serial stream of the WWVB pwm data.
>> From there any MCU can decode that stream via bit-bang.
> Hi Don,
> About the U4221B -- that TEMIC series was widely used in WWVB receivers 15 to 20 years ago. The problem for many hobbyists today is that these very nice chips have long since been out of production. Note the May '96 date on the datasheet.
> I figure there minimal low-volume demand -- just not enough WWVB hobbyists in the world. And for high-volume -- companies like Sony, Seiko, Casio, and Junghans typically roll their own chips, one where they can fully integrate the receiver into the single IC that does everything else: clock, calendar, display, subcode decode, motor control, power management (solar charge).
> WWVB RCC (radio controlled clocks) and watches have evolved -- many companies now offer "multi-band" clocks that automatically handle all the world-wide LF broadcasts (WWVB,JJY40/60,MSF,DCF77). This further lowers the market demand for a WWVB-only plain receiver IC.
> Under my http://leapsecond.com/pages/sony-wwvb/ page is a full set of vintage Temic datasheets:
> And lastly, this old Temic time code document is a must-read for anyone playing with RC clocks:
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
More information about the time-nuts