[Ftdi-usb-sio-devel] FTDI Serial Driver Question

[Edward D. <Ed....@no...>]

Hello,

I work for NOAA (National Oceanic and Atmospheric Administration) in Oak 
Ridge, TN and we've got a multi-port serial device (ByteRunner
USB-8COMi-RM-120V) URL:

http://www.byterunner.com/byterunner/product=name:USB-8COMi/title=USB+to+8+Port+Serial+Adapters+(RS422,+RS485) 


with the FTDI chipset running with the latest version of Fedora Core 6. 
I don't know the exact kernel, but it is at least 2.6.16.  We have been 
having a few problems about which I would like to ask your opinion.

I'll start by outlining the serial loading of each serial port and 
describing the problems we are having, followed by some questions. 
Hopefully you can give some insight as to what might be going on.

Below are the devices and a description of the amount of serial data and 
the timing of the data we receive:

Device 1 - 35 bytes every 20 ms, baud rate 460800, RS-422
Device 2 - 35 bytes every 20 ms, baud rate 460800, RS-422
Device 3 - 26 bytes every 50 ms, baud rate 115200, RS-232
Device 4 - 12 bytes every 20 ms, baud rate 19200, RS-232
Device 5 - 36 bytes every 50 ms, baud rate 9600, RS-232
Device 6 - 318 bytes every 100 ms, baud rate 115200, RS-232
Device 7 - 238 bytes at 00 ms, 168 bytes every 100 ms until next even 
second, baud rate 115200, RS-232
Device 8 - 56 bytes every 100 ms, baud rate 115200, RS-232

Here is another look at the sum total of the data, as a function of time:

00 ms - 756 bytes
20 ms - 82 bytes
40 ms - 82 bytes
50 ms - 62 bytes
60 ms - 82 bytes
80 ms - 82 bytes
100 ms - 682 bytes
120 ms - 82 bytes
140 ms - 82 bytes
150 ms - 62 bytes
160 ms - 82 bytes
180 ms - 82 bytes
200 ms - 682 bytes

... and the pattern repeats until the next even second, when the 756 
byte burst comes again.

I have written c-code to gather data from each device and print it to 
the screen as a test to see if things are working.  The devices are 
tested individually with separate test programs and once everything is 
working I run a multi-threaded program to simultaneously collect data 
from each device and store it to a common data file.  I am using an 
intermediate serial library (SIO from Tim Hudson, 
http://svn.gnumonks.org/branches/scez-ng/sio/) to provide linkage 
between the serial drivers and my application code.  One other key point 
is that the traffic is all one-way, receive only.  I don't send any data 
to these devices so the computer just sits back and listens to each 
device and archives data.

Now for the problems...  The test code for the first 5 devices works 
fine when plugged into the USB serial ports.  But devices 6, 7, and 8 
fail intermittently when plugged into the USB serial ports.  In the case 
of device 7, data flow is fine for several seconds, followed by fairly 
regular checksum errors, then the program terminates with a segmentation 
fault error.  I can plug the same device into the built-in serial port 
on the computer (/dev/ttyS0) and the test code runs flawlessly, without 
the CRC errors or the segmentation fault.  I get very lethargic 
performance with device 8 when it is plugged into the USB serial ports 
(but at least no segmentation fault!), but flawless performance when 
plugged into the built-in serial port (again, same test code in both 
cases).

My questions: Are there any differences between the drivers for the 
"built-in" serial port (/dev/ttyS0) and the FTDI USB devices?  I'm 
thinking about buffer sizes in particular.  Also, are there any "tweaks" 
I can do to the buffer sizes for the USB FTDI driver that might help my 
particular situation?  I'm thinking about handling the serial burst 
loads that will appear every second and 1/10 second when the full system 
is up and running.

I also noticed some flags in the FTDI serial structure regarding 
throttling and non-throttling of serial ports.  Could there be anything 
there that might degrade (or improve?) performance?

I'm not sure whether the USB interface for this device is 1.0, 1.1, or 
2.0, or whether it would make a difference.  At this point, any 
suggestions would be appreciated!

Thank you for your time and for developing the FTDI serial drivers!

Best regards,

Ed Dumas
NOAA/ATDD