"BIOS problem" solved

[Linea Recta]

22-9-2014 I started a thead "Subject: how can BIOS boot priority settings
change all by itself??"

Today customer had bootproblems again. Message: diskdrive failed.
Couldn't believe that because I had fitted a new drive.
I booted into the BIOS setup and was baffled when I saw the detected drives.
Where I expected the name of the new drive to show up, I saw a string of
strange characters.
Couldn't believe the drive had died again.
Now I decided to open the case and examine the SATA cable. I took off the
cable at both ends and reconnected it.
After this the PC booted fine again.

So those of you who suggested the cause is bad cable connection were right!



--


|\ /|
| \/ |@rk
\../
\/os

 

[Rod Speed]

Hi, Mark.


Ah, yes, the old "Pink Pearl" treatment. ;<)

Shades of the TRS-80! ;^{

Way back in 1978, when we had unexplained glitches, we would unplug
cables - usually from edge-cards in those days. Then we would give the
edge card connectors the "Pink Pearl" treatment: we would "erase" the
film of fine debris from the edge-card teeth using a pencil eraser, with
Pink Pearl being the most common brand name for such erasers. Then we
would plug the connector back onto the edge card and...problem solved.
;<) Often we didn't even need the eraser; just unplug the cable and plug
it in again, letting simple friction remove the film. That simple
solution still works surprisingly often. ;<)

It shouldn't be happening with modern standardised cables.

 

[Paul]

It shouldn't be happening with modern standardised cables.

As I understand it, this is a connector retention problem
or a kinked cable problem.

First generation cables have absolutely nothing to hold them in
place. The connector body didn't provide a compression (interference)
fit. SATA cables would regularly fall off, due to vibration.

This happened, because the dudes at SATAIO were spending all their
time designing the perfect "SATA backplane connectors". That allows
disk drives to be plugged into slots in a server. Such as the 24 bay
"things" you can get for servers, to hold drives. Those have a
SATA backplane in back, and the carefully designed SATA connectors
are inside the box.

For the desktop, they just decided to "saw that connector in half
and send the plans to China". Without even checking whether this
was a good fit for desktop systems or not.

Later, some of the little people in the industry decided to
fix this stuff. So the connector body got deformed a bit,
to provide a compression fit. And later, the "metal jaw" thing
was added. The metal jaw thing has to be designed, so if the
mating member isn't available on the other end, the parts don't
conflict.

I got a new motherboard with jaw-equipped cable, and it would not
fit on my WD branded hard drive. So I had to fall back to the
simpler cable (deformed body). Thus, there are at least two retention
features available, and sometimes both features are on the same
connector. And this is all because the SATAIO committee didn't
test their original design before shipping it. (Otherwise, they
would have noticed how bad it was.)

The contacts are gold on gold, which relies on a "wiping" action
rather than the "biting" action of tin on tin. Since the gold is so thin,
erasers are no longer recommended for "reconditioning" the contacts.
The gold is so thin (10u instead of 50u), there might already be
pinholes in the finish. If you must clean such a connector, try
isopropyl alcohol and a lint-free cloth. I would not recommend
anything stronger than that.

The SATA cable should not be bent until it kinks. This has an impact
on the dielectric insulation inside the cable. And can change the
cable impedance. It's more likely to be a problem, if one conductor
is kinked and the other one is not. When the cable changes direction,
try to make a "rounded corner" in the cable, not a pinch. I'm still
waiting to see an eye diagram or a TDR picture of a kinked cable,
to see just how bad bending them is.

And use some care when disconnecting a cable from a motherboard.
There was one brand of motherboard, with vertical connectors, where
the manufacturer didn't affix the connector body to the board very
well. A user would tug on the cable, and pull the connector
right out of the motherboard. Later connectors were fastened a bit
more securely than that.

Paul

 

[Linea Recta]

As I understand it, this is a connector retention problem
or a kinked cable problem.

First generation cables have absolutely nothing to hold them in
place. The connector body didn't provide a compression (interference)
fit. SATA cables would regularly fall off, due to vibration.

This happened, because the dudes at SATAIO were spending all their
time designing the perfect "SATA backplane connectors". That allows
disk drives to be plugged into slots in a server. Such as the 24 bay
"things" you can get for servers, to hold drives. Those have a
SATA backplane in back, and the carefully designed SATA connectors
are inside the box.

For the desktop, they just decided to "saw that connector in half
and send the plans to China". Without even checking whether this
was a good fit for desktop systems or not.

Later, some of the little people in the industry decided to
fix this stuff. So the connector body got deformed a bit,
to provide a compression fit. And later, the "metal jaw" thing
was added. The metal jaw thing has to be designed, so if the
mating member isn't available on the other end, the parts don't
conflict.

I got a new motherboard with jaw-equipped cable, and it would not
fit on my WD branded hard drive. So I had to fall back to the
simpler cable (deformed body). Thus, there are at least two retention
features available, and sometimes both features are on the same
connector. And this is all because the SATAIO committee didn't
test their original design before shipping it. (Otherwise, they
would have noticed how bad it was.)

The contacts are gold on gold, which relies on a "wiping" action
rather than the "biting" action of tin on tin. Since the gold is so thin,
erasers are no longer recommended for "reconditioning" the contacts.
The gold is so thin (10u instead of 50u), there might already be
pinholes in the finish. If you must clean such a connector, try
isopropyl alcohol and a lint-free cloth. I would not recommend
anything stronger than that.

Is that what's sold as 'contact cleaner'?

The SATA cable should not be bent until it kinks. This has an impact
on the dielectric insulation inside the cable. And can change the
cable impedance. It's more likely to be a problem, if one conductor
is kinked and the other one is not. When the cable changes direction,
try to make a "rounded corner" in the cable, not a pinch. I'm still
waiting to see an eye diagram or a TDR picture of a kinked cable,
to see just how bad bending them is.

And use some care when disconnecting a cable from a motherboard.
There was one brand of motherboard, with vertical connectors, where
the manufacturer didn't affix the connector body to the board very
well. A user would tug on the cable, and pull the connector
right out of the motherboard. Later connectors were fastened a bit
more securely than that.

thanks for your historical overview.
As I wrote, the MB and cable aren't exactly new. Do you have links to
pictures of the connector types you mentioned above? I understand they are
not exactly compatible?
For the moment this computer boots OK, but I'm still thinking about getting
a new cable and I don't want to buy the wrong type.
The MB is SATA II, and I believe nowadays the standard is SATA III, if that
makes any phisical difference.
BTW I never pull at the cables, I always pull the connector itself.
There are no sharp bends in the cable, but I suppose that's no guarantee
that there isn't any invisible micro crack somwhere in the cable.



--


|\ /|
| \/ |@rk
\../
\/os

 

[Rod Speed]

As I understand it, this is a connector retention problem

Yes, normally, particularly early on.

or a kinked cable problem.

Hardly ever.

First generation cables have absolutely nothing to hold them in place. The
connector body didn't provide a compression (interference) fit. SATA
cables would regularly fall off, due to vibration.

Or due to someone poking around in the case.

This happened, because the dudes at SATAIO were spending all their time
designing the perfect "SATA backplane connectors".

Not really. Clearly that problem was fixable and was fixed.

That allows disk drives to be plugged into slots in a server. Such as the
24 bay "things" you can get for servers, to hold drives. Those have a SATA
backplane in back, and the carefully designed SATA connectors are inside
the box.

For the desktop, they just decided to "saw that connector in half and send
the plans to China". Without even checking whether this was a good fit for
desktop systems or not.

That isnt what happened.

Later, some of the little people in the industry decided to
fix this stuff. So the connector body got deformed a bit,
to provide a compression fit. And later, the "metal jaw" thing was added.
The metal jaw thing has to be designed, so if the mating member isn't
available on the other end, the parts don't conflict.

I got a new motherboard with jaw-equipped cable, and it would not fit on
my WD branded hard drive.

I've never had a problem with my samsungs.

So I had to fall back to the simpler cable (deformed body). Thus, there
are at least two retention features available, and sometimes both features
are on the same connector. And this is all because the SATAIO committee
didn't test their original design before shipping it. (Otherwise, they
would have noticed how bad it was.)

The contacts are gold on gold, which relies on a "wiping" action rather
than the "biting" action of tin on tin. Since the gold is so thin, erasers
are no longer recommended for "reconditioning" the contacts. The gold is
so thin (10u instead of 50u), there might already be pinholes in the
finish. If you must clean such a connector,

You shouldn't need to.

try isopropyl alcohol and a lint-free cloth. I would not recommend
anything stronger than that.

Any of the small molecular weight alcohols are fine.

The SATA cable should not be bent until it kinks.

That's true of any signal cable like that.

This has an impact on the dielectric insulation inside the cable. And can
change the cable impedance.

Cable impedance just isnt a problem with digital signals.

It's more likely to be a problem, if one conductor is kinked and the other
one is not. When the cable changes direction,
try to make a "rounded corner" in the cable, not a pinch. I'm still
waiting to see an eye diagram or a TDR picture of a kinked cable, to see
just how bad bending them is.

Its not normally a problem, they just don't get kinked that much.

And use some care when disconnecting a cable from a motherboard.
There was one brand of motherboard, with vertical connectors, where
the manufacturer didn't affix the connector body to the board very
well. A user would tug on the cable, and pull the connector right out of
the motherboard. Later connectors were fastened a bit
more securely than that.

A lot more securely in fact.

 

[Rod Speed]

Is that what's sold as 'contact cleaner'?

That varys, but often is.

thanks for your historical overview.
As I wrote, the MB and cable aren't exactly new. Do you have links to
pictures of the connector types you mentioned above? I understand they are
not exactly compatible?

They are with decent hard drives.

For the moment this computer boots OK, but I'm still thinking about
getting a new cable and I don't want to buy the wrong type.

They arent that expensive, get more than one and try it.

The MB is SATA II, and I believe nowadays the standard is SATA III, if
that makes any phisical difference.

No it does not with the connector.

BTW I never pull at the cables, I always pull the connector itself.

Yeah, that's important.

There are no sharp bends in the cable, but I suppose that's no guarantee
that there isn't any invisible micro crack somwhere in the cable.

That's quite uncommon due to the physics of the cables.

 

[J. P. Gilliver (John)]

In message <[email protected]>, Rod Speed

I've never had a problem with my samsungs.

Sounds painful!
[]

Any of the small molecular weight alcohols are fine.


That's true of any signal cable like that.


Cable impedance just isnt a problem with digital signals.

If it wasn't, you wouldn't be worrying about kinks above. It's an
analogue world - especially at the sort of frequencies we work at these
days. But even back in the days of 10 megabit ethernet, impedance was
relevant (75 ohm coax, IIRR). That's why, ideally, you had terminators.
[]

 

[Paul]

thanks for your historical overview.
As I wrote, the MB and cable aren't exactly new. Do you have links to
pictures of the connector types you mentioned above? I understand they
are not exactly compatible?
For the moment this computer boots OK, but I'm still thinking about
getting a new cable and I don't want to buy the wrong type.
The MB is SATA II, and I believe nowadays the standard is SATA III, if
that makes any phisical difference.
BTW I never pull at the cables, I always pull the connector itself.
There are no sharp bends in the cable, but I suppose that's no guarantee
that there isn't any invisible micro crack somwhere in the cable.

Here is a picture summary.

http://i59.tinypic.com/2vbrjns.jpg

The Western Digital drive in the lower right hand corner, the
connector still fits. But the metal "latch" feature has nothing
to engage on, on the WD drive. This is because the WD drive has
extra clearance between the housing and the connector tabs.
The cable still grips a bit, but doesn't have quite the
interference fit of the red colored cable. The red colored
cable, only the distortion of the connector body holds the
connector in place. And some of those red cables, press a
little too tight. Using compression as a method, relies
on tight tolerances when molding the connectors. I have a
set of cables here, which pressed so tight, plastic gets
scraped off the cable connector body on each usage.

You may find some motherboard connector, that doesn't like the
metal-work on a fancy modern cable. I don't have a large enough
collection of computer crap, to give examples.

You can use the connectors with the metal latches. And some motherboard
connectors mate nicely with the metal latch. But I've discovered
at least one hard drive here, where the metal latch "tooth" has
nothing to grab onto. The cable still stays in place, because
the connector body is still an interference fit. Just not a
super-tight one.

One of the reasons the back of the WD drive is so goofy looking,
is WD attempted to market their own cable assembly. It consisted
of a monolithic "block" with both power and signal in the same
connector. As well as a "square part" that fits in the square hole
in the WD drive housing. So that's the history of how the WD drive
housing ended up that way. It was to help WD sell their cable
assembly. The sample Seagate drive in my picture, is just designed
for the more average cabling. Seagate didn't try to sell a special
looking cable. The Secure Connect was WD's answer to the first gen
cables falling off. It's an inventive solution, but I bet WD
wishes they hadn't done that now. I don't expect they made
that much money on the cables.

"Western Digital Secure Connect cable"
http://www.legitreviews.com/images/reviews/309/westerndigital_secure3.jpg

Paul

 

[Paul]

In message <[email protected]>, Rod Speed

I've never had a problem with my samsungs.

Sounds painful!
[]

Any of the small molecular weight alcohols are fine.


That's true of any signal cable like that.


Cable impedance just isnt a problem with digital signals.

If it wasn't, you wouldn't be worrying about kinks above. It's an
analogue world - especially at the sort of frequencies we work at these
days. But even back in the days of 10 megabit ethernet, impedance was
relevant (75 ohm coax, IIRR). That's why, ideally, you had terminators.
[]

OK, here's a couple waveforms. The smooth one is what the unbent
cable will look like. The one with ringback, would be when
the cable is kinked. Try as I might, including the word "SATA"
in my Googling, does not dig up actual SATA waveforms for me to post.
So I have to make do with whatever waveforms I can find.

http://ts4.mm.bing.net/th?id=HN.608039572705642875&pid=15.1&P=0

http://www.micron.com/-/media/image...nt_image_anatomy_of_a_terminated_data_eye.png

That's the kind of measurements I'd want to see, if bending
the cable and studying what kind of a difference it makes.

The smooth waveform in those examples, shows the pulse template
on the screen at the same time as the waveform. The dark blue diamond
in the center, checks eye closure. The top and bottom blue rectangles
check for overshoot. The waveform is automatically normalized to fit
in the scope window (so you don't have to adjust the gain and offset
knobs on the scope). The pulse template is met by TX and RX designers,
and is an "agreement" as to what their end has to do, in order to work.
So you check your design (disk drive or motherboard), to see if it is
making the right quality of waveform. If the waveform touches the
template, it doesn't mean "instant death". But if two marginal pieces
of equipment get connected together, the room is hot, the +5V voltage
feeding the equipment is low, there might be transmission errors. Passing
the template means you have some margin against conditions like that.

I would want to see the kinked cable tested, to see how much
waveform distortion results from kinking one of the two conductors.
Kinking both of them the same amount, while it degrades the waveform,
might not be as effective as kinking just one conductor.

The SATA cable is a miniature dual twinax (one pair for serial TX,
one pair for RX). And when I say "kink", I'm referring to the
compression of the shield around one of the two conductors in
that pair. The twinax conductors are differentially terminated
at the receiver, in 100 ohms or so.

(SATA cable cross-section)
http://www.satacables.com/assets/images/IMG_188353.jpg

Paul

 

[Rod Speed]

Sounds painful!

Nope, completely painless.

If it wasn't, you wouldn't be worrying about kinks above.

I don't worry about kinks above.

It's an analogue world
Nope.

- especially at the sort of frequencies we work at these days.

Not in the sense that the cable impedance matters.

But even back in the days of 10 megabit ethernet, impedance was
relevant (75 ohm coax, IIRR). That's why, ideally, you had terminators.

But PATA didn't and SATA doesn't either.

 

[Rod Speed]

J. P. Gilliver (John) wrote

OK, here's a couple waveforms. The smooth one is what the unbent cable
will look like. The one with ringback, would be when the cable is kinked.

Not with SATA.

Try as I might, including the word "SATA" in my Googling, does not dig up
actual SATA waveforms for me to post.

Because cable kinking isnt a problem with SATA.

So I have to make do with whatever waveforms I can find.

http://ts4.mm.bing.net/th?id=HN.608039572705642875&pid=15.1&P=0
http://www.micron.com/-/media/image...nt_image_anatomy_of_a_terminated_data_eye.png

Those arent relevant to SATA.

That's the kind of measurements I'd want to see, if bending
the cable and studying what kind of a difference it makes.

It clearly doesn't if you can't turn them up with google.

The smooth waveform in those examples, shows the pulse template
on the screen at the same time as the waveform. The dark blue diamond
in the center, checks eye closure. The top and bottom blue rectangles
check for overshoot. The waveform is automatically normalized to fit
in the scope window (so you don't have to adjust the gain and offset
knobs on the scope). The pulse template is met by TX and RX designers,
and is an "agreement" as to what their end has to do, in order to work.
So you check your design (disk drive or motherboard), to see if it is
making the right quality of waveform.

No point with SATA.

If the waveform touches the template, it doesn't mean "instant death". But
if two marginal pieces of equipment get connected together, the room is
hot, the +5V voltage feeding the equipment is low, there might be
transmission errors.

Not with SATA.

Passing the template means you have some margin against conditions like
that.

And that is what SATA ensures.

I would want to see the kinked cable tested, to see how much
waveform distortion results from kinking one of the two conductors.

That isnt going to happen with a SATA cable.

Kinking both of them the same amount, while it degrades the waveform,
might not be as effective as kinking just one conductor.

And is what happens with a SATA cable.

The SATA cable is a miniature dual twinax (one pair for serial TX, one
pair for RX). And when I say "kink", I'm referring to the
compression of the shield around one of the two conductors in
that pair.

That isnt going to happen with a SATA cable.

 

[Paul]

That isnt going to happen with a SATA cable.

Thanks for providing the incentive for me to keep searching.

http://h20566.www2.hp.com/portal/si...ac.admitted=1415269744471.876444892.199480143

"Never bend a SATA data cable tighter than a 30 mm (1.18 in) radius.
Never crease a SATA data cable."

Notice that SATA power cables are not mentioned. This is because SATA
power cables are not coaxial or twinaxial in nature, and there is
no AC impedance to worry about in that respect. A SATA Power cable should
not be bent so tightly as to break the copper wire inside.

Paul

 

[Rod Speed]

Rod Speed wrote

Thanks for providing the incentive for me to keep searching.

"Never bend a SATA data cable tighter than a 30 mm (1.18 in) radius.
Never crease a SATA data cable."

That isnt for the reason you were talking about.

 

[Gene E. Bloch]

Thanks for providing the incentive for me to keep searching.

http://h20566.www2.hp.com/portal/si...ac.admitted=1415269744471.876444892.199480143

"Never bend a SATA data cable tighter than a 30 mm (1.18 in) radius.
Never crease a SATA data cable."

Notice that SATA power cables are not mentioned. This is because SATA
power cables are not coaxial or twinaxial in nature, and there is
no AC impedance to worry about in that respect. A SATA Power cable should
not be bent so tightly as to break the copper wire inside.

Paul

Apropos of nothing (although I offer thanks for being motivated to look
further above!), I still remember as a child not understanding how
electricity could still get to the light bulb when there was a knot in
the wire.

Although some here might think otherwise, by "child" I mean when I was a
kid of 4 years of age, not when I was 19 years old.

 

[Paul]

Apropos of nothing (although I offer thanks for being motivated to look
further above!), I still remember as a child not understanding how
electricity could still get to the light bulb when there was a knot in
the wire.

Although some here might think otherwise, by "child" I mean when I was a
kid of 4 years of age, not when I was 19 years old.

Transmission line cables are a little different, in that
abusing the dielectric, changes the characteristics of
the cable when passing GHz signals.

Transmitter ---- R0 ---- Coax -------+---- Receiver
(z0) |
R0
|
Ground

What that diagram is meant to convey, is both the transmitter
end and the receiver end, should match the AC impedance (Z0)
of the cable. If you do that successfully, no signal
bounces off the end at the receiver and goes back towards
the transmitter. The signal then has "perfect integrity"
and has the same shape (potentially with high frequency loss)
as the source. If the cable has extremely long length,
you use line build-out, things like pre-emphasis, to
make a good looking signal.

http://en.wikipedia.org/wiki/Pre-emphasis

The SATA standard uses short cables, and no fancy signal processing.
So rather than repair any high frequency loss issues, they limit
the length of the cable. To allow the ESATA cable to be
slightly longer, they increase the launch budget (still
with no correction for high frequency loss).

SAS supports much longer transmission, and has some line buildout
capability. Which allows the thing to adjust to whether a
long or a short cable is in usage. Some config bits, allow
the SAS interface to change the emphasis settings.

On SATA, the seven pin interface would look like this.
And R3 (differential) is meant to match the R1 and R2 thing
at the other end. The receiver has enough sensitivity,
to compensate for the line loss. SATA supports full duplex,
and packets can be moving in both directions between
disk drive and motherboard, at the same time.

(drain wire)
TX+ --- R1 --- Twinaxial -----------+--- RX+
TX- --- R2 --- Conductors -------+ |
| R3
| |
+---+--- RX-

(drain wire)

RX+ ---+-------- Twinaxial ------- R1 --- TX+
| +---- Conductors ------- R2 --- TX-
R3 |
| | (drain wire)
RX- ---+---+

Compare that to the "peeled" diagram for the cable.
The center drain pin, goes to two physical drain wires,
in the construction of the cable. Construction would
have been marginally easier, with an eight pin connector.
I'm sure somewhere, the person who had to solve that
issue, must be cursing the SATAIO people.

http://www.satacables.com/assets/images/IMG_188353.jpg

That wiring diagram is just meant to give a transmission
line view of the interconnect. There are other circuit
details not include - AC coupling, 8B10B coding...

HTH,
Paul

 

[Gene E. Bloch]

Transmission line cables are a little different, in that
abusing the dielectric, changes the characteristics of
the cable when passing GHz signals.

Transmitter ---- R0 ---- Coax -------+---- Receiver
(z0) |
R0
|
Ground

What that diagram is meant to convey, is both the transmitter
end and the receiver end, should match the AC impedance (Z0)
of the cable. If you do that successfully, no signal
bounces off the end at the receiver and goes back towards
the transmitter. The signal then has "perfect integrity"
and has the same shape (potentially with high frequency loss)
as the source. If the cable has extremely long length,
you use line build-out, things like pre-emphasis, to
make a good looking signal.

http://en.wikipedia.org/wiki/Pre-emphasis

The SATA standard uses short cables, and no fancy signal processing.
So rather than repair any high frequency loss issues, they limit
the length of the cable. To allow the ESATA cable to be
slightly longer, they increase the launch budget (still
with no correction for high frequency loss).

SAS supports much longer transmission, and has some line buildout
capability. Which allows the thing to adjust to whether a
long or a short cable is in usage. Some config bits, allow
the SAS interface to change the emphasis settings.

On SATA, the seven pin interface would look like this.
And R3 (differential) is meant to match the R1 and R2 thing
at the other end. The receiver has enough sensitivity,
to compensate for the line loss. SATA supports full duplex,
and packets can be moving in both directions between
disk drive and motherboard, at the same time.

(drain wire)
TX+ --- R1 --- Twinaxial -----------+--- RX+
TX- --- R2 --- Conductors -------+ |
| R3
| |
+---+--- RX-

(drain wire)

RX+ ---+-------- Twinaxial ------- R1 --- TX+
| +---- Conductors ------- R2 --- TX-
R3 |
| | (drain wire)
RX- ---+---+

Compare that to the "peeled" diagram for the cable.
The center drain pin, goes to two physical drain wires,
in the construction of the cable. Construction would
have been marginally easier, with an eight pin connector.
I'm sure somewhere, the person who had to solve that
issue, must be cursing the SATAIO people.

http://www.satacables.com/assets/images/IMG_188353.jpg

That wiring diagram is just meant to give a transmission
line view of the interconnect. There are other circuit
details not include - AC coupling, 8B10B coding...

HTH,
Paul

What does all that have to do with my youthful "understanding" of wire?

 

[Paul]

What does all that have to do with my youthful "understanding" of wire?

Well, my young friend, this is wire with blue
chewing gum inside. Try a bite...

http://www.satacables.com/assets/images/IMG_188353.jpg

Pauk

 

[Linea Recta]

Apropos of nothing (although I offer thanks for being motivated to look
further above!), I still remember as a child not understanding how
electricity could still get to the light bulb when there was a knot in
the wire.

Although some here might think otherwise, by "child" I mean when I was a
kid of 4 years of age, not when I was 19 years old.

--

I always thought that uploading to a friend in Switserland took more time
than the other way round because it had to go up the mountains!



--


|\ /|
| \/ |@rk
\../
\/os

 

[Paul]

So are you saying reflections off a kinked cable will be too small to
matter or that the cable is so short that reflections don't matter?

It almost sounds as if you're saying, the more ludicrous thing - that
the cable doesn't have a characteristic impedance.

1) The cable does have a characteristics impedance. The SATAIO
committee has some test procedures for the cable, if you want
more details about what they test for. The Z0_diff is likely to
be 100 ohms. (There are a number of high speed differential
interconnects that work the same way. LVDS is where the fun
started. We had some proprietary interfaces at work, which
predate LVDS, but they don't count, as they were a trifle
crappy and a lot slower.)
2) The receiver has a differential termination resistor inside
the chip housing. That terminates the line. If there wasn't
a resistor in there, you could do fly-by termination.
3) Kinking or creasing the cable, compresses the blue-colored
dielectric in the twinaxial thing. You can get a reflection
off such a discontinuity.
4) The receiver is differential, and subtracts the signal from the
two conductors in the twinax. When common mode noise is present
on the cable, the receiver can eliminate the noise, by taking the
difference. That also makes it a little difficult to predict
what will happen if one or both conductors has a reflection on it.
My suspicion is, kinking just one of the two dielectric areas,
makes for a worse result.

I'd hoped that someone on the Internet, had taken their SATA test
gear, and bent a cable to see how much effect it has.

If such a test result is not available, SATA packet transmission has a CRC
error detector, and you could use counted cable errors as an indication
of cabling damage.

http://mindshare.com/files/ebooks/SATA Storage Technology.pdf

"In contrast, every packet of information sent across the SATA bus
includes a Cyclic Redundancy Check (CRC). These checks are designed
to detect every sin-gle and double-bit error that occurs. There
exists an extremely high probability that CRC will detect
virtually every error."

So there is a way of determining that conditions are not the
best on the cable. I thought one of the SMART indicators (one
which cannot be cleared or reset), counted those on the
drive end.

While that Mindshare sentence expresses some optimism about the
CRC code, I don't share that optimism. CRC can suppress errors
and make the overall error characteristic about three orders of
magnitude better than it might otherwise be (you can request
retransmission if an error is detected). The presence of
the CRC, does not imply that data corruption will never occur.
Which is why, when CRC errors are detected and counted, you
want to fix or replace the problematic cable. Before a multi-bit
error burst does slide through.

Paul

 

[Gene E. Bloch]

I always thought that uploading to a friend in Switserland took more time
than the other way round because it had to go up the mountains!

You also have to deal with a redshift due to relativistic effects. It
wild be a blueshift going downhill.

Or is it the other way around?