Sata cabling

[JS]

Well you know the width of a SATA cable.
Serial data transfer right.

Take 4 SATA cables in parallel to a Hard Drive
that has four SATA connectors. What do you
have ... parallel data transfer (4 wide SATA).

Now the data xfer rate should be 4 times faster
than a single SATA cable and the cable width
would be no wider than a PATA cable.

Now I know this is a design stretch, but if you
have a hard drive with Integrated SATA Electronics,
how hard could it be to repeat the circuit design 3 more
times to get what I'll call a SATA IIx4 interface.

Problem is I doubt the current level of a Hard Drive's
mechanical rotation, read/write data rates and buffer
size would be able to feed a 4 wide SATA connection.

 

[Bill in Co.]

If we're talking about a true parallel interface, and a 32 bit data bus, 32
separate conductors are required JUST for that part of the interface alone.
And for a 16 bit data bus, it would be 16 conductors, minimum. The problem
with the required size of the connectors to accommodate all that is pretty
obvious. Heck, even the normal 40 or 80 conductor PATA connectors, as small
as they were, presents a physical problem. The Ultra ATA 80 conductor ones
simply interleaved ground wires between all the data wires (to decrease the
crosstalk problem) to raise the max transfer rates.

Well you know the width of a SATA cable.
Serial data transfer right.

And it's pretty small. Especially in comparison to any PATA ones!

Take 4 SATA cables in parallel to a Hard Drive
that has four SATA connectors. What do you
have ... parallel data transfer (4 wide SATA).

But that's not true and complete parallel data transfer, which requires
separate conductors for EACH data bit being transferred. (16 for 16 bit data
bus, 32 for 32 bit data bus), to gain the theoretical potential advantage of
parallel transfer.

Now the data xfer rate should be 4 times faster
than a single SATA cable and the cable width
would be no wider than a PATA cable.

I doubt if just doing that is really worth all the increased cost and
complexity.

Now I know this is a design stretch, but if you
have a hard drive with Integrated SATA Electronics,
how hard could it be to repeat the circuit design 3 more
times to get what I'll call a SATA IIx4 interface.

Problem is I doubt the current level of a Hard Drive's
mechanical rotation, read/write data rates and buffer
size would be able to feed a 4 wide SATA connection.

And also - it would still only be 4 wide.
That's nothing like the improvement gained when going from PATA to SATA, so
in addition to what you wrote above, it's probably not worth it.

 

[JS]

Does a SATA cable look anything like a PATA cable,
no it doesn't. So who is to say a high speed Parallel
interface cable has to look anything like today's PATA cables.

 

[JohnO]

It's was deemed a less expensive solution at the time.

I have PATA 80 wire for ATA 100 and ATA 133 drives
(connected using either round PATA cables or the traditional flat ribbon)
plus SATA I and SATA II drives.

Don't get me wrong, I'm not against SATA
but they just don't deliver on the advertized speed.
Plus the connector design of SATA does not give
me confidence in constantly connecting Hard Drives
on my "Test" computers.

The HDD speed bottleneck is the platter/head transfer, not the electrical
interface. No IDE drive ever transferred data at the limits of the
electrical interface...platter/head was always slower than the ATAPI
standard of the day.

As to serial being faster, serial has many advantages over parallel when
speeds become very high. For instance, jitter and skew are major factors. As
speeds and cable length increase, reliable data transfer becomes less
likely. Longer high-speed parallel cables are impossible unless a
differential cabling system is used, such as with later versions of SCSI,
and even then the length is limited. External IDE was never practical, where
eSATA is just a new physical implementation of the existing SATA standard,
more or less.

In our experience the SATA connectors last many times longer than any IDE
connector ever could. They're far from perfect, but an order of magnitude
better than 40-pin connectors.

-John O

 

[JohnO]

Does a SATA cable look anything like a PATA cable,
no it doesn't. So who is to say a high speed Parallel
interface cable has to look anything like today's PATA cables.

The horse has left the barn...to late for that now. Serial is the new
parallel.

Best example...Ethernet. May be duplex, but serial nonetheless.

-John O

 

[Bill in Co.]

The HDD speed bottleneck is the platter/head transfer, not the electrical
interface. No IDE drive ever transferred data at the limits of the
electrical interface...platter/head was always slower than the ATAPI
standard of the day.

As to serial being faster, serial has many advantages over parallel when
speeds become very high. For instance, jitter and skew are major factors.
As
speeds and cable length increase, reliable data transfer becomes less
likely. Longer high-speed parallel cables are impossible unless a
differential cabling system is used, such as with later versions of SCSI,
and even then the length is limited.

Yup.
And the net size of the cables and connectors needed would be impractical.

External IDE was never practical,

Nope. For the reasons we've both covered, now.

where eSATA is just a new physical implementation of the existing SATA,
standard, more or less.

And kudos for some further elucidation on it.

 

[M.I.5¾]

Not crazy over the SATA connectors either.
SATA or Serial ATA has another design flaw
in that by nature serial data transfers can never
be as fast as parallel.

Totally incorrect.

You may be thinking from the good old days where RS232 would never operate
anywhere near as fast a Parallel (Centronic) connection (this was a
limitation of the UART devices of the period, not the inherent technology).

But the reality is: that if you want the fastest communication possible, it
has to be serial - parallel just doesn't cut the mustard. This is because
as the data rate gets faster, the signal pulses get shorter and shorter.
With a parallel connection, for various technical reasons, the signal pulses
travel down the parallel conductors at different speeds (known as the
propagation coefficient). With long pulses, it is relatively easy to strobe
the receiver when all signals are valid. But as the pulses get shorter and
shorter, there comes a point where at no point in time do all the pulses
have a point where the data is valid because they arrive at different times.
It is worth remembering that in a transmission line that a pulse of just .01
nanoseconds duration (a long pulse by modern standards) is approximately two
millimetres long.

Serial communication, on the other hand, completely solves the problem
because, as there is only one transmission channel, the propagation
coefficient is fixed for all pulses. The principal limitation to the
transmission data rate of a serial connection these days is solely the
properties of the copper transmission line itself, but this can be resolved
by turning to fibre optic systems. Even these have limitations dependent on
the fibre material with glass being able to carry a far greater data rate
than plastic.

1.33 Gb/s is the limitation of parallel technology (even with a short cable
around a foot long), but serial ATA III* is already operating at 6 Gb/s.

* Not to be confused with ATA 3 (an unofficial designation) which is really
ATA II, .

 

[M.I.5¾]

Well, in a crude sense think of SATA
as a two lane highway compared to
PATA as an 8 lane highway. If the PATA
cable design was updated to handle higher
transfer rates I would think that PATA could
be at least 4x faster than SATA.

Not possible. 1.33 Gb/s is the practical limit for a copper parallel
transmission line.

And an updated PATA cable need not be a
giant size ribbon cable either. Just imagine if
your ram memory was serial access instead
of DDR2 or DDR3.

Actually, serial RAM memory already exists, but it is only used in very
specialised applications due to the cost. It is, of course, much much
faster than the type of memory that is encountered in a PC, but PC memory is
largely hamstrung because it is based on dynamic architecture which is far
slower (and cheaper) than static architecture. This is why your PC has a
apir of processor caches to speed RAM access up. The L2 cache is a chunk of
static RAM which is about 4 times faster than the main RAM and the L1 cache
is static RAM which is vastly faster than the L2 (and very expensive - which
is why it isn't very big).

FLASH memory is exclusively serial access.

 

[M.I.5¾]

It's was deemed a less expensive solution at the time.

I have PATA 80 wire for ATA 100 and ATA 133 drives
(connected using either round PATA cables or the traditional flat ribbon)
plus SATA I and SATA II drives.

Don't get me wrong, I'm not against SATA
but they just don't deliver on the advertized speed.
Plus the connector design of SATA does not give
me confidence in constantly connecting Hard Drives
on my "Test" computers.

It was true that when introduced, SATA I did not deliver any speed gain over
(P)ATA 133, but this was due to the choice of data encoding that they
employed which gave a real world speed of around 1.2 Gb/s, which was fairly
close to the ATA it replaced, if slightly slower.

SATA I quickly adopted something called 'Native Command Queuing' (NCQ) as an
optional feature (to preserve backward compatibility) which gave it the
intended data rate of 1.5 Gb/s. This technology was nothing new as SCSI
disc drives had always featured it.

SATA II featured NCQ as standard and thus the advertised 3 Gb/s speed was
delivered.

It should be remembered that the data rates represent the maximum burst
speed for the data, the actual real data rate is, as ever, hampered by the
drive itself having to mechanically move the read/write head and to wait for
the data block to come around. But with large on disc caches that are now
common, things are getting better all the time.

 

[M.I.5¾]

On the surface, that makes sense - but only on the surface. Let me
explain:
In practice, it's a false assumption due to the inherent crosstalk
problems between adjacent signal carrying conductors in the Parallel ATA
cable. And THAT limits the max transfer rate. OTOH, Serial cables do
NOT have that problem, since only a single line is carrying the data.
Hence, serial cables (like in SATA) can be, and are, much faster.

That's only part of the problem. Crosstalk can be minimised by good cable
design. That was the object of the 80 conductor (P)ATA cable which placed a
grounded conductor either side of each signal carrying conductor. This
reduces crosstalk, but more importantly provides the signal with a
transmission path that more closely resembles an 'ideal' transmission line.

But it isn't the cross talk that really limits the parallel system. It's
something called 'skew'. This is where the individual signals in the
individual transmission lines arive at the distant end of the cable at
different times. Low voltage diferential signalling was introduced on some
systems to reduce crosstalk even further but there was no solution to the
problem of skew. Serial systems do not suffer from skew, hence the
transmission rates are higher.

 

[M.I.5¾]

A cable redesign could eliminate
the crosstalk.

It would be possible to design a cable that eliminated it entirely. But the
ever present problem of 'skew' renders the design moot.

 

[M.I.5¾]

Well you know the width of a SATA cable.
Serial data transfer right.

Take 4 SATA cables in parallel to a Hard Drive
that has four SATA connectors. What do you
have ... parallel data transfer (4 wide SATA).

Excellent idea. Call the individual paths 'express lanes'. You could have
(say) 16 of them paralleled up to give unprecidented data transfer speed.

Hang on it's already been done.

Now the data xfer rate should be 4 times faster
than a single SATA cable and the cable width
would be no wider than a PATA cable.

Now I know this is a design stretch, but if you
have a hard drive with Integrated SATA Electronics,
how hard could it be to repeat the circuit design 3 more
times to get what I'll call a SATA IIx4 interface.

Not hard at all. Alternatively design the existing serial data link to
operate 4 times faster. Result: a cheaper drive as the connectors are
cheaper.

Commodore Computers were the first to realise this when they converted the
parallel IEE-488 interface which used a very expensive connector into a
serial interface which could use a standard 'DIN' connector which cost just
a few cents. It took them several incarnations to get anywhere near the
data rate though.

Problem is I doubt the current level of a Hard Drive's
mechanical rotation, read/write data rates and buffer
size would be able to feed a 4 wide SATA connection.

Well they are going for the 4 times faster option. SATA III is in the
pipeline (which means that SATA IV is in development). But you are right,
in that, the 6 Gb/s only represent a burst data rate which is essentially
the speed they can shift the on drive cache

 

[M.I.5¾]

Does a SATA cable look anything like a PATA cable,
no it doesn't. So who is to say a high speed Parallel
interface cable has to look anything like today's PATA cables.

It wouldn't matter how you designed the cable, 1.33 Gb/s is the practical
upper limit for parallel communication (for reasons previously discussed).

 

[M.I.5¾]

Gerry wrote:

All of this means, people will have seen a variety of user
experiences. All the way from "no problems here", to "my cable
keeps falling off, so I glued it on"

Rather than use glue, a far better material is silicone rubber. Applied to
the ends of the connector, it is resilient enough to retain the connector
without breaking as the connector moves, but soft enough that it is easily
removed when you really do want to remove the connector.

 

[JS]

What about a Fiber cable?

 

[JS]

IEEE-488 can be daisy chained or Star configuration.
In addition an IEEE-488 cable could be removed from
the hard drive in the middle of transferring a file and then
attached again and the file xfer would complete with no
data lost, this was possible over 25 years ago, try that
with SATA or PATA!

 

[JS]

No, I'm thinking of the 8 bit wide parallel
fiber optic transmission cable I worked
with years ago. Back then fiber was too
brittle to make any practical bends in the
cable but now it may be possible.

Today a single fiber cable can carry
a lot of data across multiple spectrums.

 

[M.I.5¾]

--
JS
http://www.pagestart.com




What about a Fiber cable?

[Top posting corrected]

If you mean a fibre optic cable, these are exclusively serial devices. It
would not be possible to operate a number of fibres as a parallel data link
because they suffer skew in the transmission times as well. Although the
effect is less marked than it is with copper, the ability to transmit much
shorter pulses of light brings the problem firmly back to the fore again.

 

[M.I.5¾]

--
JS
http://www.pagestart.com




IEEE-488 can be daisy chained or Star configuration.
In addition an IEEE-488 cable could be removed from
the hard drive in the middle of transferring a file and then
attached again and the file xfer would complete with no
data lost, this was possible over 25 years ago, try that
with SATA or PATA!

[Top posting corrected - AGAIN]

Although PATA is not hot puggable, SATA is. I haven't tried pulling the
plug in the middle of a transfer, but I can't think of anything that would
prevent the completion of the transfer.

The fact that I can't think of anything, don't make it so though.

 

[M.I.5¾]

No, I'm thinking of the 8 bit wide parallel
fiber optic transmission cable I worked
with years ago. Back then fiber was too
brittle to make any practical bends in the
cable but now it may be possible.

[Top posting corrected]

That was then. Data rates are now such that signal skew would prevent a
parallel fibre system operating at anywhere near the speed serial ones are
capable of.

Today a single fiber cable can carry
a lot of data across multiple spectrums.

True, but they are all serial data streams. If you wish to divide up the
bandwidth to handle multiple parallel data streams, then the total bandwidth
is still limited by the bandwidth of the fibre - or in other words, you gain
nothing.

Please don't top post - it makes it harder for others to follow the
discussion.