[Raveninghorde]

Sure there is. It reduces P/N count, saving inventory problems and
possibly a second (perhaps impossible) pass at the pick-n-place
machine.

Exactly. Reduced number of different parts saves inventory and reduces
machine time. As I make in the UK this is probably more important
than if I was making in China. Also given that resistors come in reels
of 5000 then if you make 500 using a special value then 4500 are held
in stock until scrapped.

 

[Eeyore]

Whenever I see app notes I have always noted the strange resistor
values used when a value from the E24 range would do just as well.

Heard of E96 ? Yes, they do make them like 604 ohms. Mix with E24 and you
have a wealth of values for precision work.

Funnily enough I'm working on a project where E96 values would come in
handy for precise analogue filters. Instead I'm using 2 E24s in series
since Farnell only sell the E96s in 50s yet E24s in 10s. Seems the wrong
way round to me.

Graham

 

[Eeyore]

Some companies purposely don't use a value from the E24 series when requiring
1% accuracy because it tips off the folks assembling the board that you really
meant 1% resistors -- when just looking at a schematic, it's not always
obvious whether a 1% or 5% part was used. (Personally I try to label things,
e.g., 10.0k rather than 10k if I mean 1%... although this is a somewhat
imperfect approach since labeling a resistor 100.0k seems wrong as it would
imply 0.1% accruacy, but it's not like you're going to label a 5% 100k
resistor at 10e1 either...)


There's been some discussion that it makes sense to just stock 1% resistors
rather than both 1% and 5% since these days the price difference is pretty
negligible -- this was the policy at one Very Big Company I once worked for.
(And one of the guys there was the biggest penny pincher you can imagine --
he'd give Joerg a run for his money -- I probably should have paid more
attention to him!)


Well, see above. Also note that while IC designers have it drummed into their
heads that as much as possible everything should be ratioed with no exact
values, in discrete design sometimes 102k really is better than 100k. If you
are just worried about ratios, you might as well use E12 or even E6 values,
regardless of the accuracy needed.

If you're completely confident your assembly guys won't inadvertently
subsitute 5% resistors on you when you mean 1%, I'd say that using 1% E24
values whenever possible makes a lot sense. (Although numerous 1% values
don't exist in the E24 line... e.g., 12k, 16k, etc... this web site has a nice
chart: http://www.logwell.com/tech/components/resistor_values.html).

Good explanation. My colleague and I are only going to stock 1% low wattage Rs
from now on too.

Graham

 

[krw]

krw wrote:

[...]

But seriously, this problem was fixed with two measures:

a. All 100K resistors became 10%.

b. The contract mfg supplies stock and they've got no problem whatsoever
with inventory because they make tons of other stuff. Otherwise they
wouldn't have asked us.

We don't use contract manufacturers. Inventory control isn't free.
OTOH, if the owner was really interested in saving money, he'd insist
that IT link the purchasing/inventory database and the
engineering/production databases. ...

In larger companies that's linked. In the one where I took over a
division it wasn't at that time but we made darn sure this was done, and
soon.

In my PPoE they certainly weren't linked and you don't get much larger
than that.
... As it is, I only have a hint at

Exactly that is one reason for using contract manufacturers. They never
pay 4-5x of what they should be paying. More like 0.8-0.9x because with
their qties they have so much more negotiating power.

Except that CMs aren't shy about dumping the small guy if a major
customer come in with an unanticipated order. The small guy *has* to
be in charge of his destiny.

Got to streamline that process. At Endosonics we "electronicized" the
ECO process. No more walking a stack of papers to the next signing
person. You could have an urgent ECO signed off in under 15 minutes.

Sure, we do that with our system now. We can even undo an ECO in
under 15 minutes (as I found out this morning). oops. ;-)

Anyhow, the first time I eased off resistor tolerances for a client it
saved them 5c/unit. Doesn't sound like much. I billed 0.25 hours for my
time for this task because that's all it took. Then they began to
produce around 50k units/year, looks like this product could make it to
the 20 year anniversary. For the cost savings to date you could buy a
nicely equipped Cadillac (the owner did ...). Unfortunately this product
has now survived the owner :-(

Sure, but we'll never get anywhere close to 50K units per year. Maybe
10K lifetime (we can only build ~10/week, now). At >$10K each, 10K
units wouldn't be bad.

 

[Eeyore]

Best is if the assembly guys get released part numbers and are never
allowed to switch types without at least a deviation signed by
engineering. Otherwise they might swap a surge rated 4.7ohms 5% against
a regular non-surge rated metal film type and one day ... tssss ... *phut*

Don't remind me. My ex-colleagues in R&D in one company HATED writing proper
component specs. They even tried making fun of me for specifying 360 ohms. Their
idea was "can't we use the wrong value and get away with it ?". They did that once
behind my back, putting the right value on the schematic and the wrong value on the
BOM, substituting 22pf ( which was a stock item ) for 33pF. A year later about half
the 8051 clock oscillators in production stopped working. I should have left right
then.

Graham

 

[krw]

Exactly. Reduced number of different parts saves inventory and reduces
machine time. As I make in the UK this is probably more important
than if I was making in China. Also given that resistors come in reels
of 5000 then if you make 500 using a special value then 4500 are held
in stock until scrapped.

Just for kicks, I looked up the difference between 1%, 10%, and 20%
resistors at DigiKey. We use a ton of 10K resistors for pullups and
configuration setting, where 20% would certainly work. If it saved
significant money I'd look into it further (not). The 10K 20% 0603
resistors were 100x the cost of 1% and 10% resistors don't exist. I
guess that settles it. ;-)

<snip>

 

[krw]

I bit myself on this one once. I had drawn up a schematic for a module
that we used in development - a dual-port RAM that acted like a ROM in-
circuit.

Somebody told me that the assemblers could build from a schematic. Cool,
I says, I don't have to do it! Unfortunately, I had used the convention
of showing all the bypass caps on the bus, rather than at each chip with
all those damn wires; when I got the board, there were the bypasses, in
a nice neat row, right next to the power bus. It turned out not to be a
problem; it took her about 10 minutes to add a half a dozen caps, so it
was no big deal. I guess the moral is, be more specific.

I had a tech once who decided that he knew more about decoupling than
I did. He wire-wrapped all the decoupling caps, in daisy-chains (you
don't do *anything* daisy chained on WW), to a single power and
ground. I had specifically instructed him to use these teensie
shorting clips that were made to go around the wrap pins directly to
the appropriate land right next to the decoupling cap pin. They were
a PITA to install, but not as bad as they are after the 6000 wires are
on the board. The boards had to be scrapped at a cost of a couple of
thousand for the hardware, not to mention the weeks of work lost. I
was not a happy camper. He never worked for me again and I got a kid
fresh out of school to train right (the kid ended up the same level
engineer as I was, in the research division).

 

[Joerg]

krw wrote:

[...]

In larger companies that's linked. In the one where I took over a
division it wasn't at that time but we made darn sure this was done, and
soon.

In my PPoE they certainly weren't linked and you don't get much larger
than that.

PPoE? I have also seen large companies where there was no such link and
that really surprised me. Usually that means money is being left on the
table. One of the most painful questions in a design review can be "So
what's the total BOM cost here?". Of course, as a consultant one must be
careful with that kind of question in front of so many people, better to
ask outside the meeting.

All I can say is that at EndoSonics I and the engineers had that
information at our finger tips, right on our workstations. We also had
stuff like total systems cost barometers, catheter costs, and so on,
updated in realtime. Just one of the reasons why we got the stock price
from the $4-5 range to $11 within three years (that's when the company
was sold, for that price).


... As it is, I only have a hint at

Except that CMs aren't shy about dumping the small guy if a major
customer come in with an unanticipated order. The small guy *has* to
be in charge of his destiny.

Yes, have a rather painful memory about that, from China. They don't
dump you out of the blue but begin to hint that you aren't big fish to
them anymore.

Sure, we do that with our system now. We can even undo an ECO in
under 15 minutes (as I found out this morning). oops. ;-)


Sure, but we'll never get anywhere close to 50K units per year. Maybe
10K lifetime (we can only build ~10/week, now). At >$10K each, 10K
units wouldn't be bad.

Then you'll have to look at the savings, may not be worth the effort.
But if you could change 500 pull-ups to versions that cost 1/2c less
that'll still save the equivalent of a mid-size car over the lifetime.

 

[Joerg]

Just for kicks, I looked up the difference between 1%, 10%, and 20%
resistors at DigiKey. We use a ton of 10K resistors for pullups and
configuration setting, where 20% would certainly work. If it saved
significant money I'd look into it further (not). The 10K 20% 0603
resistors were 100x the cost of 1% and 10% resistors don't exist. I
guess that settles it. ;-)

North American pricing, Asian pricing, two different things.

Also, trimmable resistors are a different beast. Yes, they can be
pricey, at least here in the US. The pricing info I got was most for the
delivery location Shenzen.

 

[Joerg]

Don't remind me. My ex-colleagues in R&D in one company HATED writing proper
component specs. They even tried making fun of me for specifying 360 ohms. Their
idea was "can't we use the wrong value and get away with it ?". They did that once
behind my back, putting the right value on the schematic and the wrong value on the
BOM, substituting 22pf ( which was a stock item ) for 33pF. A year later about half
the 8051 clock oscillators in production stopped working. I should have left right
then.

Ouch, that sounds like a company that is seriously out of control. If
they wouldn't put proper procedures in place I would leave, or nowadays
terminate the consulting agreement.

 

[krw]

krw wrote:

[...]

We don't use contract manufacturers. Inventory control isn't free.
OTOH, if the owner was really interested in saving money, he'd insist
that IT link the purchasing/inventory database and the
engineering/production databases. ...

In larger companies that's linked. In the one where I took over a
division it wasn't at that time but we made darn sure this was done, and
soon.

In my PPoE they certainly weren't linked and you don't get much larger
than that.

PPoE?

Yep. In fact there was no real central engineering parts database, at
least that engineers could access. Corporate purchasing engineering
obviously had them.

I have also seen large companies where there was no such link and
that really surprised me. Usually that means money is being left on the
table. One of the most painful questions in a design review can be "So
what's the total BOM cost here?". Of course, as a consultant one must be
careful with that kind of question in front of so many people, better to
ask outside the meeting.

You have no idea. One engineer had specs (recovery was the biggie)
added to an LF356 that pushed it over $25, in 1975.

All I can say is that at EndoSonics I and the engineers had that
information at our finger tips, right on our workstations. We also had
stuff like total systems cost barometers, catheter costs, and so on,
updated in realtime. Just one of the reasons why we got the stock price
from the $4-5 range to $11 within three years (that's when the company
was sold, for that price).

Up until now the BOM cost wasn't the driver. I'll have to have this
discussion with the owner again. He just said he thinks our BOM cost
is too high for one variant. I just put the real BOM together so
purchasing could get a better idea of the real number.

There are many times where I just pull an OpAmp that's convenient,
rather than the cheapest. I don't know what the real cost is. I
asked for a dump from purchasing into the engineering database once a
week, even. So far...

... As it is, I only have a hint at

Yes, have a rather painful memory about that, from China. They don't
dump you out of the blue but begin to hint that you aren't big fish to
them anymore.

As I think I've said before, our business up until now was highly
seasonal. If we miss a month, we miss a year. We miss a year and
there is no more. The owner bought the tools and people so that
doesn't happen.

Then you'll have to look at the savings, may not be worth the effort.
But if you could change 500 pull-ups to versions that cost 1/2c less
that'll still save the equivalent of a mid-size car over the lifetime.

Sure. It would take no time to do, either. A quick sort on 10K
resistors would show if any where used for something important. ;-)
As I said, there are plenty of other places to skin this cat, too.

 

[krw]

I'm kinda fond of 4.7k and 3.3k (and multiples-of-10-thereof) for 5V and 3.3V
logic since everything is then in the ballpark of 1mA (or 100uA or whatever).

Gee, I thought you'd like 4.99K, since we're all doing 1% now. ;-)

FOr some reason, the design I'm working on now is split between 10K
and 4.99K, no matter which the supply. 10K is used for all the
configuration and version ties though. Inertia, I suppose.

At least at one point in time I believe that 10k was the most popular resistor
value around.

Popular is good. There are always lots around.

 

[Archimedes' Lever]

Ah, so the taxpayer does have to eat this. Then it's "easier" to chuck
$50k worth of parts.

You do not seem to understand.

It would cost MORE to manage it, as has already been stated.

 

[Archimedes' Lever]

If you're so worried that a few 0603 resistor values are going to compromise
national security, do all of your CM's employees have "secret" level or higher
clearances?

They are all cleared if the item being built requires it. So there are
different levels of need and compliance. Most are COMSEC cleared, if not
all, and some have secret clearance.

We have some items that are not secret until they have been programmed
with secret code segments.

Once they have, the PWA, assembly, or final product has to be tracked,
including escort, when not in a secured location under the control of the
security custodian.

If an assembly fails test, the PWA must be destroyed if the programmed
chip cannot be physically removed.

So the only testing a CM can do is power up, or such to verify a
microcontroller is up, etc. They would rarely perform any fully
functional test programs, and any turnkey items we have built are done at
secure facilities.

Most of these places have tighter hiring criteria than a police
department (obviously).

 

[Joerg]

You do not seem to understand.

It would cost MORE to manage it, as has already been stated.

I _ran_ a business and it did not cost more. Medical devices, very
agency-regulated. We must have done something right I guess. $50k of
excess inventory would have caused our CFO to explode.

 

[Rich Grise]

I had a tech once who decided that he knew more about decoupling than
I did. He wire-wrapped all the decoupling caps, in daisy-chains (you
don't do *anything* daisy chained on WW), to a single power and
ground. I had specifically instructed him to use these teensie
shorting clips that were made to go around the wrap pins directly to
the appropriate land right next to the decoupling cap pin. They were
a PITA to install, but not as bad as they are after the 6000 wires are
on the board. The boards had to be scrapped at a cost of a couple of
thousand for the hardware, not to mention the weeks of work lost. I
was not a happy camper. He never worked for me again and I got a kid
fresh out of school to train right (the kid ended up the same level
engineer as I was, in the research division).

I guess I'm a little obsessive about bypasses - I've done prototypes
where I actually soldered the cap on the opposite side of the board
to the chip; the philosophy being that when it's laid out, I can
use wide (low inductance, well, lower than 1/2" of #30 wire!) traces
for the bypasses. Remember those sockets with the bypass cap built in?

Cheers!
Rich

 

[Joerg]

We can't export some equipment if it can make an edge that goes 0 to 5
volts in under 1 ns. There are jillions of gumdrop cmos parts that
will do that now, and you can buy them anywhere in the world, by the
reel.

You could even do that in the 80's, at least positive edges, with a
BFS17 as a follower.

 

[Joerg]

[...]

I had a tech once who decided that he knew more about decoupling than
I did. He wire-wrapped all the decoupling caps, in daisy-chains (you
don't do *anything* daisy chained on WW), to a single power and
ground. I had specifically instructed him to use these teensie
shorting clips that were made to go around the wrap pins directly to
the appropriate land right next to the decoupling cap pin. They were
a PITA to install, but not as bad as they are after the 6000 wires are
on the board. The boards had to be scrapped at a cost of a couple of
thousand for the hardware, not to mention the weeks of work lost. I
was not a happy camper. He never worked for me again and I got a kid
fresh out of school to train right (the kid ended up the same level
engineer as I was, in the research division).

I guess I'm a little obsessive about bypasses - I've done prototypes
where I actually soldered the cap on the opposite side of the board
to the chip; the philosophy being that when it's laid out, I can
use wide (low inductance, well, lower than 1/2" of #30 wire!) traces
for the bypasses. Remember those sockets with the bypass cap built in?

Those were horrible. The only proper way was to have a full ground
plane. At my first employer I insisted on prototyping my stuff on a perf
board with a full ground plane. Everybody thought I was a bit too
paranoid about it. Then my board ran and others didn't until we squished
lots of 3M copper tape between sockets. Afterwards everybody used ground
plane perf board

 

[Archimedes' Lever]

I _ran_ a business and it did not cost more. Medical devices, very
agency-regulated. We must have done something right I guess. $50k of
excess inventory would have caused our CFO to explode.

Sorry, but it costs several hundred dollars just to buy a single 100 pc
bag of screws to cover a shortage, if purchasing has to get involved. So
it costs near that amount to re-introduce parts into inventory, and
change the way program managers would kit it out. They would first have
to check inventory, and then expiry dates and lead times of the CM, etc.
Then they would have to be the exact part that was spec'd by the design
group that made it. The PMs are used to buying the parts needed for a
kit, plus attrition, and "gettin' 'er done!". It is far cheaper against
the budget for a program to do that than if one bills out the costs of
all those that would be involved with retrieving unused kit tailings.
Also, each design team works independent of others, so the 0.01µF 0603
chip cap used by one group would not be the same used in another group.
Instead of forcing hundreds of engineers to waste time learning all one
like paradigm, including newly purchased subsidiaries, is far more costly
than letting them all do what they all know the way they all know it.
That way, a new acquisition need not make huge adjustments to fit the new
parent company's work paradigms.

One might find 100 different 0.01µF 0603 caps in our system, with a
whole array of differing specs all within that one value. Each having
their own P/N. So ANY that one finds upon a search must be verified for
use by the PM that bought them. Maybe it will sink in a bit now.

Certain programs flat forbid any such action as re-use of any part of a
BOM from their design.

It is far cheaper for a company that is large and growing still to use a
system that allows each engineer to fly through life with the (his) stick
in *his* hand. Of course any huge screw up will cost him and the
company, but you would be surprised at how low that number is.

 

[Joerg]

Sorry, but it costs several hundred dollars just to buy a single 100 pc
bag of screws to cover a shortage, if purchasing has to get involved. ...

That sounds like some organizational change is in order. Our purchasers
would issue that kind of order in 2-3 minutes on their computation
machines and then move on

Ok, then the incoming inspection guy would spend another 2 minutes on
the arriving bag of screws and stash it in inventory. So 5mins total.
Those guys didn't make several thousand Dollars an hour. With my dentist
I am not so sure, $800 for a root-canal, ouch.

... So
it costs near that amount to re-introduce parts into inventory, and
change the way program managers would kit it out. ...

That's the next problem. When I took over we abandoned this whole "idea"
of kitting (which I never understood). Migrated production to a Kanban
system. _Huge_ cost savings resulted, mostly by inventory reduction and
WIP reduction plus all the people that used to do the kitting all day
long could be assigned to other more productive jobs.

Seriously, every time I see kitting and other more archaic processes
happen in companies I politely suggest a visit to a Toyota plant or a
similar enterprise. Ok, requires a trip to Japan but for a company with
large scale production it is worth it.

... They would first have
to check inventory, and then expiry dates and lead times of the CM, etc.
Then they would have to be the exact part that was spec'd by the design
group that made it. The PMs are used to buying the parts needed for a
kit, plus attrition, and "gettin' 'er done!". It is far cheaper against
the budget for a program to do that than if one bills out the costs of
all those that would be involved with retrieving unused kit tailings. ...

We just didn't have any unused kit tailings because there were no more kits.

Also, each design team works independent of others, so the 0.01µF 0603
chip cap used by one group would not be the same used in another group.

Big problem, should be fixed asap (we did). All it takes is some minor
education and most of all a common database that is accessible to all.
Heck, we even had that in the early 80's and that was a business that
must be very similar to yours (defense). I had my pick of a very small
array of 0.01uF caps and if I needed a specialty one it would have to be
formally released and I better had a good explanation handy because that
had to be written onto the ECO.

Back then there were VT-whatever terminals for that next to the data
book shelves. Which wasn't bad because you had to get off your kiester
and exercise the leg muscles so it wasn't all that bad.

Instead of forcing hundreds of engineers to waste time learning all one
like paradigm, including newly purchased subsidiaries, is far more costly
than letting them all do what they all know the way they all know it.
That way, a new acquisition need not make huge adjustments to fit the new
parent company's work paradigms.

True, aquisitions need to be left alone initially so folks don't get
spooked. But not forever, there needs to be a gradual migration towards
more parts commonality. We've acquired two companies so I know how that
goes. Or can go.

One might find 100 different 0.01µF 0603 caps in our system, with a
whole array of differing specs all within that one value. Each having
their own P/N. So ANY that one finds upon a search must be verified for
use by the PM that bought them. Maybe it will sink in a bit now.

Not really. I do not think that is a terribly efficient system. At least
there is potential for improvement.

Certain programs flat forbid any such action as re-use of any part of a
BOM from their design.

Sure, when it's secret stuff. But it certainly won't forbid using the
same 0.01uF cap in other designs.

It is far cheaper for a company that is large and growing still to use a
system that allows each engineer to fly through life with the (his) stick
in *his* hand. Of course any huge screw up will cost him and the
company, but you would be surprised at how low that number is.

Screw ups may be low but the scenario you describe can and usually does
lead to much higher cost. In gvt or defense land that often doesn't seem
to matter to the stakeholders because it's often cost plus accounting.
And the taxpayer isn't asked anyhow, he just pays up ...