Slashdot Mirror
Will the Serial Console Ever Die?
simpz writes "Will the serial port as a console connection ever be displaced — especially for devices such as switches, routers, SAN boxes, etc.? In one sense it's a simple connection. But it is the only current port that, in order to use, you need to know about wiring / baud rates / parity, etc. It has non-standard pinouts. And it is becoming too slow to upload firmware to dead devices, as the firmware updates get larger. Also, the serial port is rapidly disappearing from new laptops — which is where you often really need it, in data centers. Centronics, PS/2, and current loop are mostly defunct. Is there any sign on the horizon of a USB console connection?"
Also there are single chip rs232 to USB solutions available for a few dollars. They are used in lots of things these days. They are great for when the device your making only needs a simple serial port connection to a pc or other device. Arduino uses one on their boards. I cant really think of any other examples at the moment. But I dont see serial console dying anytime soon, its still really useful to have around.
I work for a small electronics manufacturing company, http://www.westmountainradio.com/
and we make a number of devices that use the serial port. In recent years, we had to start including USB-serial adapters with every device for the very reason mentioned: Many newer computers simply do not have RS232 ports anymore.
The RS232 port is a very convenient way to connect with a number of peripheral devices that don't need much bandwidth. In most cases, 9600 BPS is plenty. You also have the "handshake" lines which can be used to toggle an external device on or off. We use it to drive an LED and an opto-isolator to key a ham radio transmitter, among other things.
As long as there are low-bandwidth, human-interface devices, there will still be SOME use and purpose for the RS232 port.
Willie...
I still design lots of equipment with serial interfaces inside. It is much easier to connect to a low-end microcontroller which may barely have even a single UART. And even for a higher-end processor, it's so much easier to build the interface. Developing a USB interface requires a pretty detailed understanding of USB - selecting endpoints, which transfer protocol to use, etc - so there's a big software investment and often a significant additional hardware investment to implement a USB interface. Serial is often damn close to free, so easy that it's a no-brainer to put in. And for ethernet devices like switches I can't imagine why anyone would want to bother with a USB interface when you already have 8/16/48 copies of an ethernet interface available, just plop down yet another copy of the ethernet PHY design and make that your console interface.
Point is - serial's EASY to give you, so you're gonna keep getting it for a while.
Just my $0.55 (US inflation, 1774-2008, for $0.02)
Serial is cheap, simple, works really well, and you can hook up 15+ year old equipment to it with no problem.
Is it slow? Not really, but firmware updates should be through TFTP or HTTP by now anyways for larger files.
Complicated wiring? RX-TX TX-RX, common ground.
Also RS-232 has many brothers and sisters like:
RS-422 (a high-speed system similar to RS-232 but with differential signaling)
RS-423 (a high-speed system similar to RS-422 but with unbalanced signaling)
RS-449 (a functional and mechanical interface that used RS-422 and RS-423 signals - it never caught on like RS-232 and was withdrawn by the EIA)
RS-485 (a descendant of RS-422 that can be used as a bus in multidrop configurations)
On the USB console: yeah, you can have a USB console. Most like there will be a FTDI chip, which will make your USB into a serial connection. Want an example? Arduino.....
By the way, the post is kinda mis-worded.... USB is a serial bus, so a USB console is technically a SERIAL console :)
Simplicity really is the key.
Just a few days ago I hacked together a 9600 baud serial output in like an hour to help me debug an embedded microcontroller design using only a single IO pin and a crude spin-delay based bit-bang function. It worked great, and I found the trouble.
There's no way you could add something like USB nearly as easily. FTDI makes some great chips / cables, but at the microcontroller it's still TTL-level serial IO.
Plenty of microcontrollers have lots of extra serial IO ports. Many are adding USB ports as well, but it takes an absolutely stupid amount of firmware to make USB work.
There are several microcontrollers I can do USB for, since I've done it before. However, it takes weeks of work to implement USB the first time on any new microcontroller. It's usually really prone to bugs, too. USB is just too complex for the simple dumb pipes that most embedded developers need. On top of that... most of the time the micro vendor's USB firmware examples just barely work, and aren't designed very well so they're very hard to modularize and include in another design.
The biggest problem I've found even with the good ones is timing sensitive stuff. In some OSs (Linux included), when a synchronous write to a serial port concluded, pretty much all but the last character was already out on the line. With a USB serial port, the data is probably buffered on the device; it will go out many milliseconds later (because serial is slow). So if you've built timeouts into your code, they're all wrong now.
Then you are looking at old catalogs my friend.... no, serial is not included on every piece of hardware.
No, that little RJ45-looking jack labeled 'Console' on most newer Cisco and HP gear is actually for a serial to RJ45 cable...
There's no place like
That is defective design. Receive sensitivity for RS-232 is +/-3V so the signal is within spec. There are plenty of RS232 and RS485 devices that work off 3.3V using integrated charge pumps, and this device requires 3.6?. Double fail! I bet you were pissed off when you figured this out ;-)