HD trouble?

Bill in Co. wrote:


If you buy "generic" DIMMs by the barrel full, one good test
to run, is measure the rail to rail resistance. Some failed
cheap bypass caps are a dead short, and can cause the motherboard
socket to get burned. (I've seen reports of this on Newegg.)

Again, that is a pretty limited test. Extremely limited!!


The purpose of that test, is to prevent a motherboard from
burning a contact, due to a dead short across the DIMM. I
can't say I've done this test myself, but it would have
saved a few people from a mess, after they purchased
extra memory. I would reserve such a test, for memory
products with a dodgy history (memory that comes in a
baggie, from Ebay).

Paul

Bill in Co. wrote:
Paul wrote:
Pegasus [MVP] wrote:

The multimeter test might destroy the RAM chip, like the tap on the head
with the hammer. I find it difficult to think of a more inappropriate
test for a RAM chip. Remember - it contains a hundred million or more
transistor gates operating at extremely low currents, less than the most
sensitive multimeter can detect!
You can safely test silicon devices, if you use a multimeter with
"low power ohms" setting. It applies a voltage not intended to
cause forward conduction in the silicon chips.

Paul, this only applies to checking *discrete* devices, like diodes and
transistors, it does not apply to integrated circuits.


So what exactly does an integrated circuit have inside it, fairy dust ?

Golly, I see two transistors here. It is just a big package full of
transistors, diode, resistors etc.

http://en.wikipedia.org/wiki/File:CMOS_Inverter.svg

Neither the level of voltage excursion, nor the level of
current used for testing, is an issue.

The picture in Wikipedia, is the circuit inside the chip. It would be
considered a logic primitive.

On the pad ring, the pads have clamp protection like this. There
are various kinds of networks of components, intended to prevent
input excursions above the top rail voltage, or below the bottom
rail voltage. (I had another example network I wanted to show up,
but I can't find a picture right now.)

http://www.circuitstoday.com/wp-cont...on-circuit.jpg

So when you're using "low power ohms", you would be avoiding forward biasing
one of those things. It is to make the input of the IC "transparent", so you can
measure something which is driving that input. Nobody particularly wants
to detect the clamp diode inside the IC. (And as we've already agreed, there
isn't much point to ohming across a couple data pins for example. That is
pointless.)

And even if the low power ohms did happen to trigger forward conduction,
because the test current doesn't go above 1 milliamp, there is still no
danger to the device. The input diodes are rated for 10mA continuous.

Paul

Paul wrote:
Bill in Co. wrote:
Paul wrote:
Pegasus [MVP] wrote:

The multimeter test might destroy the RAM chip, like the tap on the
head
with the hammer. I find it difficult to think of a more inappropriate
test for a RAM chip. Remember - it contains a hundred million or more
transistor gates operating at extremely low currents, less than the
most
sensitive multimeter can detect!
You can safely test silicon devices, if you use a multimeter with
"low power ohms" setting. It applies a voltage not intended to
cause forward conduction in the silicon chips.

Paul, this only applies to checking *discrete* devices, like diodes and
transistors, it does not apply to integrated circuits.


So what exactly does an integrated circuit have inside it, fairy dust ?

Golly, I see two transistors here. It is just a big package full of
transistors, diode, resistors etc.

http://en.wikipedia.org/wiki/File:CMOS_Inverter.svg

Neither the level of voltage excursion, nor the level of
current used for testing, is an issue.

The picture in Wikipedia, is the circuit inside the chip. It would be
considered a logic primitive.

On the pad ring, the pads have clamp protection like this. There
are various kinds of networks of components, intended to prevent
input excursions above the top rail voltage, or below the bottom
rail voltage. (I had another example network I wanted to show up,
but I can't find a picture right now.)

http://www.circuitstoday.com/wp-cont...on-circuit.jpg

So when you're using "low power ohms", you would be avoiding forward
biasing
one of those things. It is to make the input of the IC "transparent", so
you
can measure something which is driving that input. Nobody particularly
wants
to detect the clamp diode inside the IC. (And as we've already agreed,
there
isn't much point to ohming across a couple data pins for example. That is
pointless.)

And even if the low power ohms did happen to trigger forward conduction,
because the test current doesn't go above 1 milliamp, there is still no
danger to the device. The input diodes are rated for 10mA continuous.

Paul

The point is, a typical integrated circuit like a memory chip has thousands
of transistors inside, and there is no way you can test them, as they are
all interconnected within the IC. This is NOT the case for testing a
single discrete device, like a diode (has 2 leads) or a transistor (has 3
leads), or a resistor, capacitor, etc.

And even with that simple CMOS inverter, you couldn't check it with a
multimeter (except to see if there were shorted pins, a pretty limited
check). The only way to check it out would be by applying a logic LO and a
logic HI input voltage to the input, and monitoring the output voltage in a
test circuit.

Or better yet, by applying a pulse input waveform while simultaneously
monitoring the output waveform on an oscilloscope (a dynamic check)

Bill in Co. wrote:


The point is, a typical integrated circuit like a memory chip has thousands
of transistors inside, and there is no way you can test them, as they are
all interconnected within the IC. This is NOT the case for testing a
single discrete device, like a diode (has 2 leads) or a transistor (has 3
leads), or a resistor, capacitor, etc.

And even with that simple CMOS inverter, you couldn't check it with a
multimeter (except to see if there were shorted pins, a pretty limited
check). The only way to check it out would be by applying a logic LO and a
logic HI input voltage to the input, and monitoring the output voltage in a
test circuit.

Or better yet, by applying a pulse input waveform while simultaneously
monitoring the output waveform on an oscilloscope (a dynamic check)


I don't see a point to probing LSI devices *at all* with a multimeter.
But there will be situations where you're probing other elements
on the PCB, and the wiring just happens to be connected to a large
IC as well. By using the low power ohms scale, you might avoid interference
from the large IC. If it is "transparent" and not affecting your measurement,
then you may be able to get a better reading on something, say, next
to that chip.

Again, in the lab, I didn't spend a lot of time probing boards with
ohmmeters. I would typically use a multimeter for checking VCC voltage
feeding some chip, to see if the chip is powered and that the voltage
is within spec (+/-5%). Much more of my time would be spent with a
four channel storage scope, or a logic analyzer. The logic analyzer
on some occasions would even be tied to the storage scope,
so I could take an analog voltage snapshot of something happening
when a certain digital event was captured by the logic analyzer.
One of our analyzers, even had a storage scope built in, for those
kinds of mixed captures.

There isn't much point in adding a stimulus to a modern PCB, because
the LSI chips are more than capable of generating useful patterns
you can verify. I worked with a guy programming the LSI chips
(FPGAs) - he'd cook up a new design file, and later in the day,
I would verify the patterns coming out of the chip, then leave
him some notes on things that needed fixing. That would be
a typical scenario now. If the FPGA designer is good with
testbenches, in fact his design might work on the first try.
(You can simulate everything on the computer, leaving nothing
to chance.)

Years ago, we had an ancient kit in the lab, perhaps made by HP.
It consisted of a logic pulser (would sink up to 0.5 amps to ground
for a period down in the nanoseconds). The second item in the kit
was a logic probe. The intention was, to test simple logic gates,
like a 2-in NAND. The logic probe would touch the output pin,
while the pulser probe would drive an input low (and drive it
so strongly, as to overrule the previous chip in the chain).
Since the pulser 0.5 amp pulse was so short, there was no
damage to the circuit it was strong-arming. So that is a toy
we'd have used in the "jelly bean era". I think I played with
it once, just to see "history in action". But for modern LSI
components, that isn't useful any more. It is more a matter
of observing the system with instruments. Frequently a test
stimulus comes from some software program being run on
the hardware. Like writing a register in a chip somewhere,
to create a sequence on some I/O pin. If your logic analyzer
is set up to catch the write to the register, you can then trace
in time, the response coming from the LSI.

Paul

"Paul" wrote in message ...

The purpose of that test, is to prevent a motherboard from
burning a contact, due to a dead short across the DIMM. I
can't say I've done this test myself, but it would have
saved a few people from a mess, after they purchased
extra memory. I would reserve such a test, for memory
products with a dodgy history (memory that comes in a
baggie, from Ebay).

I'm kinda confused now. Would you not recommend getting any DDRAM from
Ebay then?

Bill

Bill Cunningham wrote:
"Paul" wrote in message ...

The purpose of that test, is to prevent a motherboard from
burning a contact, due to a dead short across the DIMM. I
can't say I've done this test myself, but it would have
saved a few people from a mess, after they purchased
extra memory. I would reserve such a test, for memory
products with a dodgy history (memory that comes in a
baggie, from Ebay).

I'm kinda confused now. Would you not recommend getting any DDRAM from
Ebay then?

Bill

I wouldn't. It's a gamble. You may save $20, over the price
at a retailer. But you may not have a warranty, and RAM
does seem to have a pretty significant failure rate, for
something that doesn't wear out.

At the very least, look at the reviews for various SKUs
on Newegg. Even some manufacturers with relatively good
reputations, have products with high failure rates or a
lot of DOAs. So at the very least, you'd want to identify
a product with a good track record. Then your exposure
on things like testing or warranty, won't be quite so bad.

But if a module is advertised on EBay as "generic" with
Samsung chips, you don't know what care and attention
there was to testing or packaging (with proper ESD precautions).
They use the word "Samsung" to win your trust, but
Samsung had nothing to do with the actual DIMM
manufacturing steps. Samsung just made the chips.

I've bought generic RAM locally, and had failures in both
cases. Now I go with branded stuff, and do a check to
see if the SKU in question is a good one or not.

Paul