### Network

P

#### Pooh Bear

Jan 1, 1970
0
Jim said:
I know digital designers who do this as standard practice... it allows
easily forcing a test signal when trying to track a fault.

That's the first good example I've heard mentioned for including a series
resistor.

Be aware that *some* pins are called N/C by the manufacturer. Best to leave these
alone.

Graham

J

#### John Fields

Jan 1, 1970
0
Unused inputs should be tied to either VCC or ground. If they are left
open, it is possible the chip will start "oscillating", if the input is
moving
between on and off. This will probably not affect the chip, but may cause
odd problems elsewhere in the circuit.

At one time, it was the military practise to attach unused inputs through
a 1K resister to VCC. Doing this is longer thought necessary, just attach it
directly.

---
That's true except for multiple-emitter TTL, in which case the
resistor should be added if the input is pulled high. It's all
spelled out here:

http://focus.ti.com/lit/an/sdya009c/sdya009c.pdf

There is, however, the curious note on the bottom of page 7...

J

#### Jim Thompson

Jan 1, 1970
0
That's the first good example I've heard mentioned for including a series
resistor.

Be aware that *some* pins are called N/C by the manufacturer. Best to leave these
alone.

Graham

One very good designer I knew (he's now deceased) not only adamantly
required resistors on unused inputs, but edicted, "No system shall
contain a one-shot." ;-)

...Jim Thompson

R

#### Robert Monsen

Jan 1, 1970
0
Aidan said:
Unused inputs should be tied to either VCC or ground. If they are left
open, it is possible the chip will start "oscillating", if the input is
moving
between on and off. This will probably not affect the chip, but may cause
odd problems elsewhere in the circuit.

At one time, it was the military practise to attach unused inputs through
a 1K resister to VCC. Doing this is longer thought necessary, just attach it
directly.

Aidan Grey

Some of the microchip datasheets say you should connect the MCLR input
to VCC through a 10k resistor. I'm not sure, but I believe they are are
afraid of SCR latchup, which can occur with CMOS inputs if there is more
than some small amount of current into the pin above Vcc. Perhaps they
have seen this happen during startup with a big cap near the power pin,
or with glitches on the power rail.

I bet JT will know if this is really possible, or just an application
engineer's fantasy. I think the newer pics don't have this in their
datasheet anymore.

--
Regards,
Robert Monsen

"Your Highness, I have no need of this hypothesis."
- Pierre Laplace (1749-1827), to Napoleon,
on why his works on celestial mechanics make no mention of God.

J

Jan 1, 1970
0
J

#### Jim Thompson

Jan 1, 1970
0
Some of the microchip datasheets say you should connect the MCLR input
to VCC through a 10k resistor. I'm not sure, but I believe they are are
afraid of SCR latchup, which can occur with CMOS inputs if there is more
than some small amount of current into the pin above Vcc. Perhaps they
have seen this happen during startup with a big cap near the power pin,
or with glitches on the power rail.

I bet JT will know if this is really possible, or just an application
engineer's fantasy. I think the newer pics don't have this in their
datasheet anymore.

Anything's possible, but this is an ESD engineer's fantasy, not an
applications engineer's.

Since tying MCLR to VCC prevents a differential, I doubt that anything
untoward can happen.

...Jim Thompson

R

#### repatch

Jan 1, 1970
0
Unused inputs should be tied to either VCC or ground. If they are left
open, it is possible the chip will start "oscillating", if the input is
moving
between on and off. This will probably not affect the chip, but may cause
odd problems elsewhere in the circuit.

At one time, it was the military practise to attach unused inputs
through
a 1K resister to VCC. Doing this is longer thought necessary, just attach
it directly.

Right, and everything will be fine, until a new program is uploaded that
zeros the tris bit for port a, when he/she meant port b... let's just say
leaving it floating is a FAR better result...

The resistor is still necessary IMHO, if only to save a person from their
own mistakes.

R

#### repatch

Jan 1, 1970
0
Anything's possible, but this is an ESD engineer's fantasy, not an
applications engineer's.

Since tying MCLR to VCC prevents a differential, I doubt that anything
untoward can happen.

Except of course when you try and in-circuit program the PIC...

R

#### Rich The Newsgropup Wacko

Jan 1, 1970
0
I read in alt.binaries.schematics.electronic that Richard Crowley

Oh, go on, intend away! No-one was ever punished for a really good pun.

You know what is paved with good intentions? (;-)

Do you know where the apathetic Buddhists worship?

So Wat.

;-)

J

#### John Fields

Jan 1, 1970
0
Right, and everything will be fine, until a new program is uploaded that
zeros the tris bit for port a, when he/she meant port b... let's just say
leaving it floating is a FAR better result...

The resistor is still necessary IMHO, if only to save a person from their
own mistakes.

I disagree. Resistors not needed for a specific reason represent
money wasted, (a penny wasted on each of a million units is $10,000 of lost revenue!) and any designer worth his salt needs to learn how to work without that expensive a net. J #### John Fields Jan 1, 1970 0 Do you know where the apathetic Buddhists worship? So Wat. --- Funny!-) Do you know where transsexuals worship? At the alter. C #### CJT Jan 1, 1970 0 J #### John Fields Jan 1, 1970 0 J #### Joel Kolstad Jan 1, 1970 0 John Fields said: I disagree. Resistors not needed for a specific reason represent money wasted, (a penny wasted on each of a million units is$10,000 of
lost revenue!) and any designer worth his salt needs to learn how to
work without that expensive a net.

Sure John, but there are plenty of us here who are lucky to see 1000 units
of what we design go into production, much less a million. I think a better
approach is to spend the extra pennies on the first 1000 units to get the
'time to market' advantage as well as having effectively bought 'insurance'
that the design will work, and then -- if time permits -- go back and start
cost minimizing.

I also find there are usually much bigger fish to fry than saving the price
of an extra resistor or capacitor here of there... things like someone using
a very high end DSP to perform a function that a dedicated FPGA and a
microcontroller could do for 1/4 the price, using a much faster processor
than needed because they don't have a good algorithm for what they want to
do (which can also quickly lead to, e.g., bigger batteries!), etc.

I've worked places where we'd spend something like $10,000 on chrome-plated stainless steel 'skins' for the machines we were building; it was really hard to get that excited about saving$10 on some $300 PCBs that went into it... (On the other hand, that machine also had a$40,000 air-bearing
stage, and we DID sit around spending time trying to cost reduce _it_!)

C

#### CJT

Jan 1, 1970
0
John said:

Presumably they don't think a lot of unloaded outputs ringing away is a
good idea.

T

#### Terry Given

Jan 1, 1970
0
Joel said:
Sure John, but there are plenty of us here who are lucky to see 1000 units
of what we design go into production, much less a million. I think a better
approach is to spend the extra pennies on the first 1000 units to get the
'time to market' advantage as well as having effectively bought 'insurance'
that the design will work, and then -- if time permits -- go back and start
cost minimizing.

I also find there are usually much bigger fish to fry than saving the price
of an extra resistor or capacitor here of there... things like someone using
a very high end DSP to perform a function that a dedicated FPGA and a
microcontroller could do for 1/4 the price, using a much faster processor
than needed because they don't have a good algorithm for what they want to
do (which can also quickly lead to, e.g., bigger batteries!), etc.

I've worked places where we'd spend something like $10,000 on chrome-plated stainless steel 'skins' for the machines we were building; it was really hard to get that excited about saving$10 on some $300 PCBs that went into it... (On the other hand, that machine also had a$40,000 air-bearing
stage, and we DID sit around spending time trying to cost reduce _it_!)

I can recount dozens of similar stories. My favourite is this one:

A little drive (<= 2.2kW) we made was designed to be very cheap. After
the basic design was up and running, a buddy of mine was tasked with the
cost-redecution exercise. He spent many weeks at the task, ripping out
parts wholesale - why use 1% when 10% will do, etc. One particular
cost-reduction involved replacing some important 100nF film caps with
Z5U, and all 10 prototypes *exploded* within an hour, but thats a whole
'nother story. The outcome was a reduction of around $4 to$35 worth of
electronics - a sizeable chunk, and well worth implementing, which we
duly did.

Not long after, the marketing powers-that-be decided the original brown
cardboard box with black writing wasnt good enough, and that a fancy,
shiny cardboard box with 6 colours was required instead. They also
decided head office would sell the product thru a wholesale outlet it
owned, rather than us doing it. This box was $7 *more* than the original box, and they had a minimum run size of 30,000 boxes. So they bought them, marked with the wholesalers details. Several years later, after only selling a couple of thousand drives thru head office, we took the product off them, and sold it ourselves. We of course had to place big stickers over the (now incorrect) company details on the pretty boxes, at a cost of about$1 per box. After a few more years we obsoleted the
design, at which point we *still* had about 20,000 of these useless
bloody cardboard boxes.

When we designed the replacement product, we didnt even bother trying to
do a similar cost-reduction by shaving off $0.01 parts. Instead we concentrated on time-to-assemble, and reduced it from an hour to 3 minutes. This allowed us to spend$35 on a user interface and $20 on a micro, yet have a build cost slightly lower than the product we replaced. The cardboard box was cheap and plain brown with black writing.... although we did discover a problem with the box. In the first 3 months of the new products life, we got a dozen or more (from several thousand) back with smashed plastic - shipping damage. Our mechanical engineer had to drop the box onto a corner from > 2m to make them break (spec was surviving 1m drop test onto concrete). It turned out that for the smaller drives the package was about the right size to grab in one hand and throw into a delivery van.....rather than implement my suggestion (1kg of concrete in each box) Mike beefed up the cardboard so it would pass a 3m drop test. Cheers Terry J #### John Woodgate Jan 1, 1970 0 I read in alt.binaries.schematics.electronic that John Fields --- Funny!-) Do you know where transsexuals worship? At the alter. Keep it up, guys. Where do atheists worship? J #### John Fields Jan 1, 1970 0 Presumably they don't think a lot of unloaded outputs ringing away is a good idea. --- I think it's just an error, where: "Unused outputs of a device should not be left unconnected (open)." should read: "Unused outputs of a device should be left unconnected (open)." in order to agree with the figure. J #### John Woodgate Jan 1, 1970 0 I read in sci.electronics.design that John Fields I think it's just an error, where: "Unused outputs of a device should not be left unconnected (open)." should read: "Unused outputs of a device should be left unconnected (open)." in order to agree with the figure. I find I could occasionally do with a reminder that says: 'Used outputs of a device should not be left unconnected.' All the voltages are correct, but there's no trace on the scope! [Newsgroups reduced to three.] R #### repatch Jan 1, 1970 0 I disagree. Resistors not needed for a specific reason represent money wasted, (a penny wasted on each of a million units is$10,000 of lost
revenue!) and any designer worth his salt needs to learn how to work
without that expensive a net.

Obviously there are cases where the extra cent won't make it worthwhile.

But consider this: what if YOU are not the programmer working on the
device? Things change, people change. It's very likely that firmware will
be changed by someone else, someone who may not have the experience,
someone who test benches new firmware, notes everything is fine, releases
the firmware to the field, and all of a sudden hundreds of dead devices
are being returned. Why? The specific conditions that set the port to zero
(when tied to one, or vice versa) were never seen during this
inexperienced programmers testing.

Consider the costs then??

Or consider this: MCUs aren't infallible, do ugly things to the power
rails or expose them to ESD and it's very possible for the port direction
control bits to flip. Even strong RF can do it.

It's a pointless debate anyways, some people are comfortable with the risk
and will go for it, and there are reasons to choose that route. But unless
you have a VERY specific reason NOT to include the resistors (cost, space,
etc.) I would include it even on a product in high volume. A fraction of a
cent per product is much cheaper then a rash of RMAs, plus the damage to

Replies
10
Views
907
J
Replies
17
Views
1K
John @ Trebutech
J
L
Replies
9
Views
929
electronic-eng.com
E
P
Replies
2
Views
893
Benjamin Todd
B
Replies
6
Views
272