High Speed RS232 Link

Thread Starter

dhooker

Joined Jun 9, 2009
11
We have been using the attached circuits for many years for an RS232 current loop isolated transmission link at 9600 baud. We have tested it successfully over 1200 mtrs. We now need to increase the baud rate to 19200 baud and are experiencing problems transmitting over a distance of about 600 mtrs. Our installation has to use existing cabling which, no doubt, has a higher capacitance/mtr than if we were able to install new cable at as low a capacitance/mtr as possible. Our original designers are no longer with us so I would like to ask if anyone can identify the component(s) in our circuits that needs to be changed to acheive our requirement.
 

Attachments

Papabravo

Joined Feb 24, 2006
21,159
The trouble might be in the opto isolators. They degrade with age and the current transfer ratio may be too low. If it were me, I'd get some low delay optos and some RS-485 transceivers.

On the other hand I might just start from scratch with a set of requirements and design something for the 21st century. 1488's and 1489's went out of fashion in the mid 1980's. You really have a choice here to do it right or to continue screwing around with an ancient unsupportable technology from a bygone era. 19200 c'mon -- you should be aiming at 115200 and above.
 

THE_RB

Joined Feb 11, 2008
5,438
The line drivers look a little wimpy. On LCC2 the line seems to be powered by TR5 and TR7 which are constant current sources that only supply about 12mA. It doesn't show how the line power at ST and SR is actually attached to the line drivers??

Also if it is connected in a way that ST and SR pull down when it transmits a baud then TR6 and TR8 will cause a very slow rise time. Can you confirm how the ST and SR supplies are attached to the line drivers??
 

Thread Starter

dhooker

Joined Jun 9, 2009
11
On the LCC2 "master", ST is connected to T+ and SR is connected to R+. T- is then connected to the first LCC3 "slave" R+ and R- to the first LCC3 "slave" T+. The LCC3 R- then goes to the next R+ and the LCC3 T- to the next T+ and so on until the last "slave" is connected to TRET and RRET. This circuit supports 8 slaves. We have no control of the cable capacitance as this particular application is an existing comms link (25 years old).
 

THE_RB

Joined Feb 11, 2008
5,438
Ok, I think I got it now. ;)

Looks like there is a 12mA constant current supply powering the transmitter T+ then all the receiving LEDs in the receiving optos are in series, grounding at Ret.

I would say that circuit was designed for much lower data rates maybe even as low as 1200 baud.

There is an obvious filter shown in LCC2; R2/C1, I'd speed it up by changing R2 to 10k or so. Might not help that much but it's a start.

Then you might think about shorting out the indicator leds in the signal path, these will act as resistors when driving the capacitive load of that cable, and every little bit of resistance you can remove will help.

Then with the driver PSU you could change R12 and R14 from 47ohm to 33ohm to bump the current drivers from 12mA to 18mA, again not a huge help but those few little mods might add up.

The way i read the operation of TR6 and TR8 they seem to be slow rise circuits which is effectively giving you a lot less voltage to power your line drivers once it starts sending.
Once the first start bit occurs the line drive voltage is very quickly pulled down to less than half of its normal value then all bits in that stream afterward must be sent with an underpowered line driver. I would be very tempted to reduce C10 and C11 from 220uF to a small value of just REMOVE those caps all together to test. That will probably be the single biggest change you can do with this circuit to drive that capacitive load.

Likewise you can reduce the values of C12 and C13 but the time constant of these is not too bad at 19200 baud. If you wanted to test you could reduce them to 33n.

Or otherwise as papabravo said you can swap for better optos, or redesign by adding a more powerful line driver which would only take a few transistors, or fully redesign with something that was not optimised for 1200 baud haha.
 

Thread Starter

dhooker

Joined Jun 9, 2009
11
Many Thanks. I can understand the mods you are suggesting. We have tried changing R1 on the LCC3 to 2K7 and R7.R9 on the LCC2 to 1K5. This also has made a significant improvement. I agree that the current transfer ratio on the optos needs to be consistent and as high as possible. We were finding on the application that if we changed an LCC3 we could reduce the failure rate (for that slave) which suggests a component related solution.
 

THE_RB

Joined Feb 11, 2008
5,438
Hmm, sounds like the optos might have aged. Please post if you find a solution, I would like to hear how you finally sort it out.
 

Thread Starter

dhooker

Joined Jun 9, 2009
11
Well, we have found some old test notes that indicates that the Resistors I mentioned previously should be selected depending on the current transfer ratio of the optos so looks like we are in the right ball park (for this application). Perhaps we will look at a redesign for the future. Thanks for all your help.
 
Top