this is getting crazy

[Jeff L]

"Winfield Hill" <Winfield_member@newsguy.com> wrote in message
news:ddl43b0q9m@drn.newsguy.com...

Jeff wrote...

Spehro Pefhany wrote ...

I'd have been tempted to add one relatively pointless unstuffed
footprint to the 1,999 one to bring it up to an even 2,000. ;-)

I had the same thought - Find a reason for one more part. Maybe
a 0u1 cp or something.

There are rarely too many test points or ground lugs.

True. I have seen boards with a test pint on every node for "bed of nails"
or "flying probe" testing.

--
Thanks,
- Win

 

[John Larkin]

On 13 Aug 2005 08:37:15 -0700, Winfield Hill
<Winfield_member@newsguy.com> wrote:

Jeff wrote...

Spehro Pefhany wrote ...

I'd have been tempted to add one relatively pointless unstuffed
footprint to the 1,999 one to bring it up to an even 2,000. ;-)

I had the same thought - Find a reason for one more part. Maybe
a 0u1 cp or something.

There are rarely too many test points or ground lugs.

I like to plop a bunch of grounded thru-holes around my boards. We
insert a few 2-56 screws from below, with nuts on top, as scope ground
clip targets.

And right, the test point you really need is seldom there!

*Always* put test points on all chip selects!

John

 

[John Larkin]

On Sat, 13 Aug 2005 12:08:33 +0100, Paul Burke <paul@scazon.com>
wrote:

Joerg wrote:

Man, you guys do large boards. A couple of minutes ago I checked off on
the Gerbers of my last one. Under 200 parts :-(

The one I did this morning is 6.5x5mm and has 4 components on it. Just
can't compete. (I bet I'm less competitive than all of you.)

Paul Burke

I did one board that has a comparator, two diodes, and two resistors.
The whole board solders onto an SO-8 footprint.

John

 

[John Larkin]

On Fri, 12 Aug 2005 22:55:29 -0400, keith <krw@att.bizzzz> wrote:

PS if you guys ever make it down this way, look me up.


Ditto. Lots of planes pass through San Francisco.

No more than any other point.

No more than Hattiesburg, Mississippi?

John

 

[Joerg]

Hello Terry,

Kia Ora Joerg,

Is that what Maoris would say? Nice. Sounds much better than hello.

I saw the costed BOM for that PCB yesterday. total parts cost (excl. 121
LEDs) is $8 less than the micro cost in the (non-functioning) design it
replaced. Its 120 copies of the same thing.

Same here. I found that uCs are still not competitive in many designs
where they would fit nicely. I don't know, possibly the semi mfgs try to
maintain higher margins but in the end it hurts sales. Whenever I
contemplated migrating one of my high volume designs to uC it turned out
costing more than the huge pile of SMT on the discrete+logic solution.

ya gotta love SMT

Yes, except when I have to solder one of those MSOP or TSSOP packages.

Regards, Joerg

http://www.analogconsultants.com

 

[John Larkin]

On Sat, 13 Aug 2005 20:04:36 GMT, Joerg
<notthisjoergsch@removethispacbell.net> wrote:

Hello Terry,

Kia Ora Joerg,

Is that what Maoris would say? Nice. Sounds much better than hello.

I saw the costed BOM for that PCB yesterday. total parts cost (excl. 121
LEDs) is $8 less than the micro cost in the (non-functioning) design it
replaced. Its 120 copies of the same thing.

Same here. I found that uCs are still not competitive in many designs
where they would fit nicely. I don't know, possibly the semi mfgs try to
maintain higher margins but in the end it hurts sales. Whenever I
contemplated migrating one of my high volume designs to uC it turned out
costing more than the huge pile of SMT on the discrete+logic solution.

It seems to me that the cheapest way to configure a ram-based FPGA is
probably to use a uP and an eprom... cheaper that a single serial
eeprom, and you get to use the uP for other tricks.

ya gotta love SMT

Yes, except when I have to solder one of those MSOP or TSSOP packages.

Or probe one.

John

 

[Joerg]

Hello John,

It seems to me that the cheapest way to configure a ram-based FPGA is
probably to use a uP and an eprom... cheaper that a single serial
eeprom, and you get to use the uP for other tricks.

Well, yes, if you have big iron like that on the board the cost of any
uC will seem like a mere pittance.

ya gotta love SMT

Yes, except when I have to solder one of those MSOP or TSSOP packages.

Or probe one.

Oh yes. Bzzzzt .... POP. After that it's unsoldering one of those 100+
pin devices without messing up the pads.

The other issue is that it is so easy to miss a bridge. You wonder why
the board doesn't work like expected, detect a whiff of "ampere smell"
and suddenly find out that one of the chips is as hot as a halogen bulb.

Regards, Joerg

http://www.analogconsultants.com

 

[John Larkin]

On Sat, 13 Aug 2005 22:17:39 GMT, Joerg
<notthisjoergsch@removethispacbell.net> wrote:

Hello John,

It seems to me that the cheapest way to configure a ram-based FPGA is
probably to use a uP and an eprom... cheaper that a single serial
eeprom, and you get to use the uP for other tricks.

Well, yes, if you have big iron like that on the board the cost of any
uC will seem like a mere pittance.

ya gotta love SMT

Yes, except when I have to solder one of those MSOP or TSSOP packages.

Or probe one.

Oh yes. Bzzzzt .... POP. After that it's unsoldering one of those 100+
pin devices without messing up the pads.

The other issue is that it is so easy to miss a bridge. You wonder why
the board doesn't work like expected, detect a whiff of "ampere smell"
and suddenly find out that one of the chips is as hot as a halogen bulb.

Fluke has a new handheld thermal imager, only $10K, about 1/3 the
price of such gadgets up to now. Tempting.

John

 

[Terry Given]

Joerg wrote:

Hello Terry,

Kia Ora Joerg,


Is that what Maoris would say? Nice. Sounds much better than hello.

Indeed it is. the first bit is pronouned "key", the ora rolls the R

I saw the costed BOM for that PCB yesterday. total parts cost (excl.
121 LEDs) is $8 less than the micro cost in the (non-functioning)
design it replaced. Its 120 copies of the same thing.


Same here. I found that uCs are still not competitive in many designs
where they would fit nicely. I don't know, possibly the semi mfgs try to
maintain higher margins but in the end it hurts sales. Whenever I
contemplated migrating one of my high volume designs to uC it turned out
costing more than the huge pile of SMT on the discrete+logic solution.

especially the CD4000 series.

In this cxase it was a Cygnal cpu, a lightning fast 8051 derivative.
Thats the 2nd such (bad) design in a row I have seen. the first was the
LED video screen. That pissy little 8kb 8051 gets beaten to death doing
24-bit arithmetic, and costs us $0.50 *more* than a 32-bit TI DSP with
far more peripherals.....

ya gotta love SMT


Yes, except when I have to solder one of those MSOP or TSSOP packages.

indeed. thank Zeiss for binocular microscopes.

Regards, Joerg

Cheers,
Terry

 

[Terry Given]

John Larkin wrote:

On Sat, 13 Aug 2005 20:04:36 GMT, Joerg
notthisjoergsch@removethispacbell.net> wrote:


Hello Terry,


Kia Ora Joerg,

Is that what Maoris would say? Nice. Sounds much better than hello.


I saw the costed BOM for that PCB yesterday. total parts cost (excl. 121
LEDs) is $8 less than the micro cost in the (non-functioning) design it
replaced. Its 120 copies of the same thing.

Same here. I found that uCs are still not competitive in many designs
where they would fit nicely. I don't know, possibly the semi mfgs try to
maintain higher margins but in the end it hurts sales. Whenever I
contemplated migrating one of my high volume designs to uC it turned out
costing more than the huge pile of SMT on the discrete+logic solution.



It seems to me that the cheapest way to configure a ram-based FPGA is
probably to use a uP and an eprom... cheaper that a single serial
eeprom, and you get to use the uP for other tricks.


ya gotta love SMT

Yes, except when I have to solder one of those MSOP or TSSOP packages.


Or probe one.

even SOT23s are bad. No thermal mass = instant death. tack-soldering
dangly wires aint so easy either.

John

Cheers
Terry

 

[Terry Given]

John Larkin wrote:

On Sat, 13 Aug 2005 12:08:33 +0100, Paul Burke <paul@scazon.com
wrote:


Joerg wrote:


Man, you guys do large boards. A couple of minutes ago I checked off on
the Gerbers of my last one. Under 200 parts :-(

The one I did this morning is 6.5x5mm and has 4 components on it. Just
can't compete. (I bet I'm less competitive than all of you.)

Paul Burke


I did one board that has a comparator, two diodes, and two resistors.
The whole board solders onto an SO-8 footprint.

John

John wins the competitive PCB layout stakes

in the early 90's we played an april fools trick on the PCB assy staff.
We got the prod. and Q. mgrs to come up with the appropriate
documentation, then gave the head of PCB assy a tube of SOIC14 74HC14s
to use in lieu of DIP14, saying they were the only ones available. An
hour later, Shannon and her ladies had soldered the entire tube onto
PCBs, beautifully done. They'd placed a small spacer under the chip, and
wired them in with stripped W/W wire then glooped over it. The prod mgr
saw this, and exeplained the "joke". She wasnt amused, and hit him in
the stomach, hard enough to knock him to the ground (small guy, big
lady). IIRC we used the PCBs that way

Cheers
Terry

 

[Terry Given]

John Larkin wrote:

On Fri, 12 Aug 2005 22:55:29 -0400, keith <krw@att.bizzzz> wrote:



PS if you guys ever make it down this way, look me up.


Ditto. Lots of planes pass through San Francisco.

No more than any other point.


No more than Hattiesburg, Mississippi?

John

Or the maldives? Or Jupiter? I like this game

Cheers
Terry

 

[Terry Given]

Winfield Hill wrote:

Jeff wrote...

Spehro Pefhany wrote ...

I'd have been tempted to add one relatively pointless unstuffed
footprint to the 1,999 one to bring it up to an even 2,000. ;-)

I had the same thought - Find a reason for one more part. Maybe
a 0u1 cp or something.


There are rarely too many test points or ground lugs.

one TP for every meaningful node on the PCB. After the hassles we went
thru laying out the PCB twice, this was the last thing on our minds.

Cheers
Terry

 

[John Larkin]

On Sun, 14 Aug 2005 12:09:03 +1200, Terry Given <my_name@ieee.org>
wrote:

John Larkin wrote:
On Fri, 12 Aug 2005 22:55:29 -0400, keith <krw@att.bizzzz> wrote:



PS if you guys ever make it down this way, look me up.


Ditto. Lots of planes pass through San Francisco.

No more than any other point.


No more than Hattiesburg, Mississippi?

John


Or the maldives? Or Jupiter? I like this game

Cheers
Terry

Boring, Oregon? Waterproof, Louisiana?

John

 

[Joerg]

Hello John,

The other issue is that it is so easy to miss a bridge. You wonder why
the board doesn't work like expected, detect a whiff of "ampere smell"
and suddenly find out that one of the chips is as hot as a halogen bulb.

Fluke has a new handheld thermal imager, only $10K, about 1/3 the
price of such gadgets up to now. Tempting.

You might even be able to modify a cheap digital camera for that. I
built a CCD camera from scratch as my master's project in the 80's. I
had to try hard to reduce the IR sensitivity, meaning expensive optics.
Else you could hold it out of the window and see whether someone in
another building in the distance was smoking. You could also "see" a
person in a pitch dark room. It was amazing.

Regards, Joerg

http://www.analogconsultants.com

 

[Jeff L]

"Joerg" <notthisjoergsch@removethispacbell.net> wrote in message
news:7kuLe.10995$O07.9475@newssvr23.news.prodigy.net...

Hello John,

It seems to me that the cheapest way to configure a ram-based FPGA is
probably to use a uP and an eprom... cheaper that a single serial
eeprom, and you get to use the uP for other tricks.

Well, yes, if you have big iron like that on the board the cost of any
uC will seem like a mere pittance.

ya gotta love SMT

Yes, except when I have to solder one of those MSOP or TSSOP packages.

Or probe one.

Oh yes. Bzzzzt .... POP. After that it's unsoldering one of those 100+
pin devices without messing up the pads.

Not a problem, I've "changed" such IC's in under 5 minutes! Probing, make
sure the probe is nice and sharp. Most newer Tek scope probes are sharp
enough. The biggest problem I find is keeping track of which pin to probe -
it's easy to miscount a pin or two when all the traces look identical and
there are 20 to 50 pins on each side spaced 0.5 mm apart.

The other issue is that it is so easy to miss a bridge. You wonder why
the board doesn't work like expected, detect a whiff of "ampere smell"
and suddenly find out that one of the chips is as hot as a halogen bulb.

Regards, Joerg

http://www.analogconsultants.com

 

[Terry Given]

Jeff L wrote:

"Joerg" <notthisjoergsch@removethispacbell.net> wrote in message
news:7kuLe.10995$O07.9475@newssvr23.news.prodigy.net...

Hello John,


It seems to me that the cheapest way to configure a ram-based FPGA is
probably to use a uP and an eprom... cheaper that a single serial
eeprom, and you get to use the uP for other tricks.

Well, yes, if you have big iron like that on the board the cost of any
uC will seem like a mere pittance.


ya gotta love SMT

Yes, except when I have to solder one of those MSOP or TSSOP packages.

Or probe one.

Oh yes. Bzzzzt .... POP. After that it's unsoldering one of those 100+
pin devices without messing up the pads.



Not a problem, I've "changed" such IC's in under 5 minutes! Probing, make
sure the probe is nice and sharp. Most newer Tek scope probes are sharp
enough. The biggest problem I find is keeping track of which pin to probe -
it's easy to miscount a pin or two when all the traces look identical and
there are 20 to 50 pins on each side spaced 0.5 mm apart.

I put little markers by every tenth pin, and write down one of the
numbers at each corner. As well as a rectangular pad for 1, rounded
rectangles for the rest.

I've seen some consumer stuff with silk-screen on the underside. great
for techs, esp. when they write down test voltages, waveforms etc.

The other issue is that it is so easy to miss a bridge. You wonder why
the board doesn't work like expected, detect a whiff of "ampere smell"
and suddenly find out that one of the chips is as hot as a halogen bulb.

Regards, Joerg

Cheers
Terry

 

[Michael A. Terrell]

Jeff L wrote:

"Joerg" <notthisjoergsch@removethispacbell.net> wrote in message
news:7kuLe.10995$O07.9475@newssvr23.news.prodigy.net...
Hello John,

It seems to me that the cheapest way to configure a ram-based FPGA is
probably to use a uP and an eprom... cheaper that a single serial
eeprom, and you get to use the uP for other tricks.

Well, yes, if you have big iron like that on the board the cost of any
uC will seem like a mere pittance.

ya gotta love SMT

Yes, except when I have to solder one of those MSOP or TSSOP packages.

Or probe one.

Oh yes. Bzzzzt .... POP. After that it's unsoldering one of those 100+
pin devices without messing up the pads.


Not a problem, I've "changed" such IC's in under 5 minutes! Probing, make
sure the probe is nice and sharp. Most newer Tek scope probes are sharp
enough. The biggest problem I find is keeping track of which pin to probe -
it's easy to miscount a pin or two when all the traces look identical and
there are 20 to 50 pins on each side spaced 0.5 mm apart.

The other issue is that it is so easy to miss a bridge. You wonder why
the board doesn't work like expected, detect a whiff of "ampere smell"
and suddenly find out that one of the chips is as hot as a halogen bulb.

Regards, Joerg

http://www.analogconsultants.com

I used to use ultrafine "Sharpie" markers in red and black to mark
every fifth and tenth pin in alternating colors down the side of a chip
to help keep track of pin numbers. A Q-tip and IPA removed the ink when
I was done but it made troubleshooting a lot easier.

--
Link to my "Computers for disabled Veterans" project website deleted
after threats were telephoned to my church.

Michael A. Terrell
Central Florida

 

[PeteS]

Win wrote:

"There are rarely too many test points or ground lugs."

The corollary is
"There is never any space left for sufficient test points and ground
lugs"

I always try to add sufficient *easily accessible* test points, but I
am sometimes thwarted by the size contraints imposed by the product.

I just finished a board (small really - about 700 components, 100mm x
60mm) and there are over 300 test points - but they are the 0.035 inch
variety so they can fit. Some of the parts on there are quite large (we
still have a long way to go in miniaturisation).

Cheers

PeteS

 

[Joerg]

Hello Michael,

... A Q-tip and IPA removed the ink when
I was done but it made troubleshooting a lot easier.

IPA? I didn't know that India Pale Ale removes Sharpie marks

Regards, Joerg

http://www.analogconsultants.com