Serial better than parallel?
August 23, 2006 11:13 AM Subscribe
Was there an important technological breakthrough leading to recent widespread use of serial communications over parallel? E.g., USB as opposed to parallel port; SATA over ATA.
I'm somewhat aware of the timing synchonicity problem with parallel communications. Is there anything more to it?
I'm somewhat aware of the timing synchonicity problem with parallel communications. Is there anything more to it?
Best answer: Pin count. Improvements in IC making technology mean you need less and less silicon to make a chip that does the same thing. This is being used to integrate many different functions on one chip, but to allow each of those functions to communicate eith the outside world, you either need hundreds of pins on the chip (very expensive), or you reduce the number of pins needed for each function.
And obviously, the more pins a chip has, the more difficult to solder it becomes, and the more complex the PCB needs to be, and the more complex the connectors and cables need to be.
So in short, mechanical complexity.
posted by cillit bang at 11:37 AM on August 23, 2006
And obviously, the more pins a chip has, the more difficult to solder it becomes, and the more complex the PCB needs to be, and the more complex the connectors and cables need to be.
So in short, mechanical complexity.
posted by cillit bang at 11:37 AM on August 23, 2006
smackfu said it.. and as everyone else.
Serial can run much faster with no line interference. Parallel was good back in the day with low speed connections (Parallel is what, 1.5mb tops?). Now with modern highspeed connections (1ghz links), parallel generates too much crosstalk.
Computer interfaces still run at highspeeds over short traces because they can limit crosstalk on the line.
posted by SirStan at 12:06 PM on August 23, 2006
Serial can run much faster with no line interference. Parallel was good back in the day with low speed connections (Parallel is what, 1.5mb tops?). Now with modern highspeed connections (1ghz links), parallel generates too much crosstalk.
Computer interfaces still run at highspeeds over short traces because they can limit crosstalk on the line.
posted by SirStan at 12:06 PM on August 23, 2006
Which should get you wondering why we haven't been serial since day one. We used parallel connections for stuff like printers and hard drives because with low clock speeds you can jam a smeg of a lot more infomation over a parallel connection.
posted by Mitheral at 12:25 PM on August 23, 2006
posted by Mitheral at 12:25 PM on August 23, 2006
Response by poster: I was aware of clock skew ("timing synchronicity") but not the crosstalk problem. It sounds like the change was an evolution of higher clock speeds and more integration. Thanks for your answers.
posted by exogenous at 1:14 PM on August 23, 2006
posted by exogenous at 1:14 PM on August 23, 2006
For external devices, serial interfaces make a lot of sense from a mechanical standpoint. The cables are physically thinner, lighter, more durable, and have smaller and more durable connectors. Compare a USB or firewire cord to an ancient parallel port cable. People like cables on external / portable devices to be as small, easy to carry, and easy to plug/unplug as possible.
posted by Rhomboid at 3:03 PM on August 23, 2006
posted by Rhomboid at 3:03 PM on August 23, 2006
Cheaper silicon is a big factor that hasn't been emphasized enough yet, I think. Basically, the data is still parallel where it is used, so to transmit serially you have to convert. It is now virtually free to pack very high end conversion, including error correction, and who knows what other low level processing, onto every chip.
posted by Chuckles at 7:08 PM on August 23, 2006
posted by Chuckles at 7:08 PM on August 23, 2006
This thread is closed to new comments.
I've never heard of a single tech advance that led to widespread serial use, though there have been a number of small advances (signalling techniques, differential lines, et cetera) that have probably contributed. I believe the primary cause is ever-increasing clock speeds.
posted by j.edwards at 11:35 AM on August 23, 2006