Data exchange via 10m unshielded multiconductor cable

[karri]

Hi,
I have a project for agricultural machine controller.
Now I have controller which is connected via 10m unshielded multiconductor cable with another unit where are relays only.
I have transistors (ULNs) in the controller, which switches on and switches off these relay in the relay box.

No I have to move some electronic to this relay box. So I will have one MCU in the controller and another in the relay box. Have you any idea, which physical interface would be good for UART for example? I tested UART on TTL, but there was lot of noise, and many errors in data transmission, so I used TTL-RS232 converter on both sides, and I was better. But I am still not sure, that RS232 will be the best solution.

I am using multiconductor (14 conductors 1.5mm2 for each) cable, where some of them carry up to 10Amps. I can not use another cable. There is only this one.

Thank you for your responds

[pickit2]

You could try opto-rs232 + re422 (rs485) converters, and use the slowest baudrate you can.

[karri]

Yes, I was thinking about some optocouplers usage on the receiver side and transistor on the transmitter side. But I am affraid of speed degradation. It would be fine if I had about 19200bad. I have some current sources for LEDs, so I am thinking about usage this current sources at the optocouplers. So we will see.

[pickit2]

You should use optos at both ends, I have seen a standing voltage in project that an engineer said "we can run power and control cables in the same duct"
we found 20 volts on one wire and 50 volts on another induced by the 230 & 480 volt AC power cables.

[Signal]

I'd think about some carrier modulation approach. http://en.wikipedia.org/wiki/Power-line_communication

[FTL]

You should use optos at both ends, I have seen a standing voltage in project that an engineer said "we can run power and control cables in the same duct"
we found 20 volts on one wire and 50 volts on another induced by the 230 & 480 volt AC power cables.

Given that this is agricultural equipment, does that mean gas powered? If so, you can be pretty sure that the installer or later on an operator literally in the field, will be sure to tie-wrap your data cable to a bundle of spark plug wires and an induced 50 volts will be least of your problems.

So ++ for optos.

Rs485 is very similar to rs232, but uses balanced lines, so they should be more noise immune.

Edit:

I was just out for a walk and started to think about this. Since high impedance circuits are much more susceptible to noise, would it work to move to a very low impedance circuit? Could you use a couple of MOSFET drivers and terminate the line with maybe a 20-50ohm resistor? the MOSFET drivers can generally drive a line with up to 1 amp of current (both high and low), and can deal with any capacitance on the line since that is what they are designed to do.

With a 20 ohm termination and 5V TTL, you would have 250ma of current, so there should be little noise that can get into that. 19,200 baud only needs 20KHz signals, so that should not be a big deal.

ADSL drivers get several megabytes of bandwidth down twisted pair (not cat-5)  for miles without a problem, so this must be possible.

Am I out to lunch or missing something?

[pickit2]

lol

Given that this is agricultural equipment, does that mean gas powered? If so, you can be pretty sure that the installer or later on an operator literally in the field, will be sure to tie-wrap your data cable to a bundle of spark plug wires and an induced 50 volts will be least of your problems.

he is already doing that in using a single cable

I am using multiconductor (14 conductors 1.5mm2 for each) cable, where some of them carry up to 10Amps. I can not use another cable. There is only this one.

Posted on: February 19, 2015, 09:54:35 21:54 - Automerged
maybe after the first law suit, they may change this

[solutions]

Current loop is the way to go, IMO

[karri]

Current loop is the way to go, IMO

I know, what current loop fundamentaly is, I used preudo-current-loop for controling machines by logical signal trhu long cables (hundred of metres). But, could you be more specific for this case (UART). Have you any verified circuit?

Thank you

[Ichan]

Search for "digital current loop" not analog or 4-20ma current loop.

But for the case, i think a plain non isolated RS485 would be enough, and if the problem is still then avoid the communication events while switching ON/FF the relay.

-ichan

[Vineyards]

RS-232 Interface already uses high voltages (plus minus 15V or so) and higher than TTL current to eliminate noise and it can go up to 15 meters depending on the environmental, conductor and baud rate related factors. RS-485 is surely more immune to noise because it is differential to start with so all the common mode noise is discarded. It requires a twisted pair cable. If your cable is not a twisted pair. You can't take advantage of this technology in which case you should choose the single-ended solution which is RS-232.

If you have a twisted pair and want to choose RS-485 (you could find TTL-RS-485 or RS-232/RS-485 converters easily) then you should ground only on one side to prevent ground loops. In your case this might be difficult because not only do you require a twisted pair but also a coaxial ground wire.

If you go for RS-232, also ground on one end only.

As far as I can understand you already need a serial receiver circuit which will either translate or directly control the relay commands why don't you add a proper cable between the RX-TX points if it is just 10 meters?

[karri]

As far as I can understand you already need a serial receiver circuit which will either translate or directly control the relay commands why don't you add a proper cable between the RX-TX points if it is just 10 meters?

This is cable between tractor and accessory vehicle. There is possible only one cable. I was looking for some mixed cable (several 1.5mm2 and some twisted pairs). I found some, but these cables were very expensive. I have to produce up to 100pcs of this devices per year, so the material costs should be reduced

[Vineyards]

Well, since we don't exactly know the noise profiles of the circuits on the equipment, this becomes a step by step trial. Skip TTL and start with UART. Keep the communication one way hence discard RX on the controller side. Isolate the remaining line with an opto powering them seperately. Try the final solution thoroughly to see how it behaves in the long run.

There is always a safety concern if your provisions fall short of actual requirements. If you think saving costs won't break anything you can economize like that. However there are tangible safety and reliabity gains if you treat that like a bullet-proof industrial solution which comes with a status check and error handling provisions. Otherwise the control unit will just blindly drive the apparatus and close human supervision will be mandatory. We humans are notoriously prone to omissions and errors.

[towlerg]

Does it have to be a wired connection?

[bigtoy]

Have you considered CAN? We use it at my work; it may have some real advantages for your application.

1. It's specifically designed for noisy environments. We all think of CAN being used in cars (which is certainly a very noisy environment), but it's widely used in industrial control as well (which is where we use it).
2. It'll do long cable runs. We've run it through 1000 feet of cable. Our standard cable is unshielded.
3. It can be run isolated. This is a big deal for cable runs when you're talking between 2 pieces of independent equipment. They may not be at the same ground potential, so having a common ground becomes a huge problem (all kinds of data errors, or even wires in the cable burning out). Look at the Texas Instruments ISO1050 chip as an example if you want to build your own board.
4. It's bidirectional over a single pair. Plus you need a ground wire (which might be isolated as per the above comment). I would assume that for safety or whatever reasons, you want to be certain the thing is operating properly - bidirectional communications would make that much easier.
5. It's multidrop, if you want more than just one module at each end of the cable. You can make yourself a little network, distribute the relays / sensors / whatever.

Downside is, well, it's CAN. You need a CAN port on your micro, or, use a Microchip CAN to SPI/serial converter chip which will talk to any micro. A full CAN software stack is not necessary if you're just using CAN for its physical layer properties.

If you don't like that, consider a completely different idea:
http://www.usriot.com/Product/20.html
as an example. Cost is $12. With this you're getting a system that again provides galvanic isolation, and supports long & noisy cable runs. And is bidirectional. It requires 2 pairs (4 wires). Ethernet to serial modules are very common - this one is just an example.

[Gallymimu]

The easiest solution is hands down RS485/422 with twisted pair (ethernet cable works fine).  You can do 10mbit at 30m no problem in VERY noisy industrial environments.

[Signal]

The easiest solution is hands down RS485/422 with twisted pair (ethernet cable works fine).  You can do 10mbit at 30m no problem in VERY noisy industrial environments.

Very noisy industrial environments have their standards for distance between parallel power and control cables. Half a meter as I recall. I am not able to convert 10Mbit|30m|0.5m to Xbit|10m|1mm at glance. Could you?

I think any of suggested variants will work with the help of protocol. We do know very little about actual environment and requirements (acceptable delay for example) so mainly it is a competition of words.

[bobcat1]

Hi Karri

You can use isolated RS485 (not optical) with isolated power supply to avoid ground loop ,TI & Analog manufacture special chip's who can do the work simply

All the best

Bobi

[Vineyards]

Actually there are two main types of isolators: analog and digital. Analog isolators have three subcategories: optical, capacitive and inductive. An ordinary optocoupler can be considered as a digital isolator as the transfer factor is not completely linear. There are specially designed, optically compensated isolators using two leds which provide good accuracy up to a certain threshold. Capacitive one modulate and demodulate signal through a very small value (like one pF) capacitor. A good example to those is ISO124 series. They are pretty good. I have used them many times in the past. Inductive isolators are also good but they can pick up noise if they are placed in the vicinity of other inductive sources.

Depending on your requirements your selection of isolation type will dictate radical changes in your interface design. In other words, isolation is a crucial step.

[Gallymimu]

Very noisy industrial environments have their standards for distance between parallel power and control cables. Half a meter as I recall. I am not able to convert 10Mbit|30m|0.5m to Xbit|10m|1mm at glance. Could you?

I think any of suggested variants will work with the help of protocol. We do know very little about actual environment and requirements (acceptable delay for example) so mainly it is a competition of words.

I don't really understand what you are saying there, BUT we did 10mbit, 30m, pretty close to the 50KW 5000V power cables that were driving industrial plasmas.  Even during heavy arcing the communication worked well.

[karri]

Hi,
I am working on next generation of my old project for agriculture machine controler.
I have device separated to two parts:
1) Controler - with buttons, procesor and LCD
2) Power part - with shift registers, ULNs and relays

Controler has transistor outputs which switches optocouplers on power part. This optocouplesr drives shift register, which drives relays.
I have optocouplers there because I am affraid of noisy envinronment. I have 10m long cable, where are no twisted pairs, so I can not use RS485.
But I am not sure, If optocouplers are good solutions because of power consuption (10mA for each - in my case 4x - 40mA).
Have you any idea of beter solution for this case?
Thank you.
Ondrej

[Gallymimu]

RS485 does not REQUIRE twisted pair.  It will just work better if twisted and balanced impedance.  If you are using a low enough baud rate it doesn't matter that much.

since we don't know your power constraints no one can say if 40mA is too much.  If you are wall powered then it's no big deal, if you are running off a small battery it is a lot.

10 meters isn't very long, almost anything will work as long as the data is slow and the noise isn't too terrible (you didn't suggest what the noise environment was like.

I am generally a big fan of RS485 for robust communication in most cases.

[pickit2]

At my place of work, we use an Vishay V018A at 1mA, maybe a lot lower at slow speed.

http://www.vishay.com/product?docid=83432

[bigtoy]

As Gally said, RS485 doesn't require twisted pair. It helps with noise immunity but it's not a requirement. I would start with the RS485 transceivers.

If you play with optos, just make sure you allow them lots of margin, so that as they age and their CTR drops, your circuit continues to work. I've seen too many designs containing optos which failed over time / temperature / component variation simply because that wasn't taken into account.

[solutions]

Optocouplers for noise is BS - optocouplers are used for galvanic isolation, not for noise immunity, so I have no idea where you pulled that one from (well, I do...you're sitting on it). You are confusing fiber optics with optocouplers...they have next to nothing in common in terms of noise immunity characteristics as links.

If you want a decent response from anyone apart from guesses, tell us what you are building in more detail so we can solve your problem properly. You've told us next to nothing about this secret Lex Luthor contraption you plan to conquer the world with. You fear noise, but don't tell us the source. Certain types of noise make your shift register implementation a disaster waiting to happen.