Running serial port at 230K

Posted by: xml

Running serial port at 230K - 17/09/1999 16:30

According to tech the serial port can run at 230K. And stty 230400 appears to work.
On my pc if I run pppd at 230400 I don't get a connection. Anybody had any luck
connecting at 230K?
I first tried using setserial to set 230K with spd_shi but the kernel popped
out a deprecated message. Either way, I couldn't get it to work.

Paul


Posted by: Verement

Re: Running serial port at 230K - 20/09/1999 10:05

I wasn't able to manage this either.

Posted by: altman

Re: Running serial port at 230K - 20/09/1999 10:29

230400 works for us - Mike used to run with a 230400bps serial card to minimise zmodem times :)

Bear in mind that standard PC serials do not run 230,400 - even though windows offers you the option. The standard 1.8432Mhz serial clock with a x1 divisor gives max 115,200bps: you need at least a 3.6864Mhz serial clock into your 16c[56789]50 to get 230400bps.

Hugo


Posted by: Verement

Re: Running serial port at 230K - 20/09/1999 16:22

On the subject of the serial port, is the empeg operating as DCE or DTE? Put another way, is the supplied serial cable a straight-through cable or a null modem cable?

Likewise, is the serial cable coming out of the car mount harness straight-through or null modem?

Finally, can the serial port be run at speeds lower than 115200? And, can it use RTS/CTS handshaking?

Posted by: xml

Re: Running serial port at 230K - 21/09/1999 00:58

Here's a link to a site that has patches to allow various chipsets to work
at over 115K. The link to the linux 2.2 patch doesn't work so I haven't
probed my chipset yet :-( over 115k

Posted by: altman

Re: Running serial port at 230K - 21/09/1999 03:09

It's DTE, just like a normal PC. The lead supplied is a null modem one.

The harness is identical to the internal port (but with only tx/rx/gnd/power supply on DTR) but a female as opposed to a male.

The port can run at lower speeds, yes, but RTS/CTS (though connected to lines on the SA) isn't implemented at driver level.

Hugo


Posted by: xml

Re: Running serial port at 230K - 21/09/1999 04:54

Ok, I patched the patch to work on a 2.2.10 kernel, but it didn't find a
high speed UART. Oh well. Next to try it on my laptop, but that only runs
2.3...

Posted by: Verement

Re: Running serial port at 230K - 21/09/1999 18:06

In reply to:

RTS/CTS (though connected to lines on the SA) isn't implemented at driver level.


Does this mean it's possible to add software support to enable it? It appears my current application may be rather unhappy without it.

Posted by: altman

Re: Running serial port at 230K - 22/09/1999 04:12

It's possible to add it, yes, BUT (as with the 16550, though this was fixed in the 650 and above) the control lines have no relationship to the UART, in that when the UART input fifo is full it will not drop RTS and if the transmit fifo is not empty it will not stop sending when CTS drops.

What does your app require? (bearing in mind that the in-car serial port doesn't have rts or cts even connected)

Hugo


Posted by: Verement

Re: Running serial port at 230K - 22/09/1999 20:55

What kind of UART is in the empeg?

Strictly speaking, I can live without RTS/CTS but this is only because I'm running PPP and it is easily restarted. It appears that when an overrun occurs my serial device (a CDPD modem) tends to hang and then reset. The delay while it resets and PPP restarts is on the order of 10-15 seconds, but with large file transfers this tends to happen rather often, dramatically diminishing my overall throughput.

I'd like to see RTS/CTS enabled, and I'm sorry to learn the harness lead doesn't have these lines connected. Is it feasible to modify the harness to connect them?

Posted by: altman

Re: Running serial port at 230K - 23/09/1999 03:27

The UART is an internal one inside the SA1100 - it's not really like anything else, they just claim 16550-ishness as it has FIFOs (though they're not as deep as the '550 - I think it's 12 on RX, 8 on TX from memory).

Your CDPD modem - can you enable Xon/Xoff flow control and set your PPP asyncmap to make sure it doesn't use these characters? This would solve the problem.

The reason only data is wired to the docking connector is simply the docking connector: it only has 16 pins and these have to carry 4 audio outs, 2 audio ins, antenna, power, *and* serial.

Hugo


Posted by: Verement

RTS/CTS flow control - 24/09/1999 18:50

I've confirmed with the modem maker that it does not support XON/XOFF flow control -- just RTS/CTS.

If I've counted correctly, the docking connector is currently using 14 of the 16 pins. So, are there two left available to use for RTS/CTS?

Posted by: dmz

Re: RTS/CTS flow control - 25/09/1999 18:04

You haven't counted correctly. We completely dissected our docking connector's wiring in order to successfully install the Empeg into my VW Jetta, and it's using all 16 pins, as follows (note that I don't know which signals are carried on which wires in the serial cable):

1: +12V (and +12V line on serial cable)
2: Serial Yellow Wire
3: Rear Output Right
4: Rear Output Ground
5: Rear Output Left
6: Aux Input Right
7: Accessory Power Out / Serial Blue Wire
8: Antenna
9: GND
10: Serial Red Wire
11: Front Output Right
12: Front Output Ground
13: Front Output Left
14: Aux Input Left
15: Aux Input Ground
16: Antenna Ground

Lots of different grounds - but they're all necessary, as I'm sure someone from Empeg will tell you as well... :)


-----
Daniel M. Zimmerman
Caltech Computer Science
[email protected]
Posted by: NasalGoat

Re: RTS/CTS flow control - 27/09/1999 08:50

What I don't understand is - why is the connector only so many pins? It's not like they had to follow some set standard. They could have easily picked a connector of any size needed - this is *custom hardware*.

I doubt a connector with a few more pins would have cost more than a few pennies more.

Seems silly to me the way they've purposely restricted themselves for no reason.

Posted by: altman

Re: RTS/CTS flow control - 27/09/1999 09:08

The connectors cost a fortune in fact - even when buying large volumes, they aren't cheap - over $20 as it is for the pair in quantity. There are larger versions of the same connector, but the price increase is almost exponential. The amount of space it takes up is also an issue.

We really didn't think anyone would have a problem with the lack of flow control - not supporting Xon/Xoff is very short sighted from the modem manufacturer: worse still is rebooting when the outbound queue gets full!

Hugo


Posted by: Mark Petersen

Re: RTS/CTS flow control - 27/09/1999 09:21

what about fitting to small conector instead on large (USB/serial)
not incounting the space and difficulty
Ups I youst anseret my own question !

Mark
emaks rouls
Posted by: NasalGoat

Re: RTS/CTS flow control - 27/09/1999 09:35

I find it hard to believe a 20-pin connector would be more than $1 or $2 more than a 14-pin one.

Well, so be it, that's the way it is and we'll have to live with it.


Posted by: andy

Re: RTS/CTS flow control - 27/09/1999 14:21

Where did you get the connectors from ?

I'd quite like to get hold of one of them so that I can neatly rig up an FM aerial for indoor use (because although just touching the antenna pin with my fingure gives me a great reception it's not very convenient...)

__
Unit serial number 47 (was 330 in the queue)...
Posted by: altman

Re: RTS/CTS flow control - 27/09/1999 14:53

Farnell sell them: www.farnell.com, I think the part numbers are RP16 and RS16, though I don't have the farnell order codes to hand. You want an RP16 as this is the plug - a search should turn this up.

Hugo