Exceptions??: The ground’s of each the send device can’t always be tied together, as they may be on separate electrical systems entirely. But in this system, each recieve device is usually tied to the send device so this isn’t as much of an issue, they are treated like 2 separate systems. However, When they are connected as in a network A and network B, it is best to have a autmoatic switch to toggle which device they are connected to at any given time, and the switch should be grounded how? TO both systems?. To one system? Still confused…

I am trying to remote a single serial sensor (232) about 100 meters from a PC. For this purpose, I purchased two 232/485 converters. The PC’s serial port is connected to one, which is connected via 3 wires (A, B and Ground) to the other (and out comes the 232 to the sensor).

I am also passing power to the remote converter and sensor on the same cable that caries the signal (4 wires in total, 3 one 5v supply).

The problem is, that this setup “basically” works. Meaning, it can be up for an entire day, but at some time it will fail. Sometimes a simple power cycling (off/on) will restore communications, and sometimes it will not. Sometimes I have no idea why, it just comes back on! Then, often, it stays off for hours. I really don’t get it.

What might be wrong? This is the first time I am using 485.

Might it be a problem to pass power along with the signal? Maybe the fact that the ground of the power is common to the ground of the signal? (but anyway they are shorted in the converter box so I don’t see how it might mater).

Help!

Thanks guys.

PS - Ah yes, the in-built terminators on the converters are enabled.
PPS - I have not connected the shielding of the 485 cable to ground
PPPS - I am using regular 4 core shielded wire, not twisted…

I wish I could get more people to understand that carrying a third wire between RS485 devices was necessary. I wish everyone who implemented an RS485 device had put the third wire on the connector so you could connect to it.

Apropos PROFIBUS: Profibus may be RS485, but the reason it works is because the purple cable and connector is EXACTLY DEFINED and full network performance is only achievable using EXACTLY THAT CABLE (and connector). If you grab any old twisted pair cable use it to hook up your Profibus network, you deserve the problems you have.

I agree with Douglas, a third wire is not always necessary.
As long as the difference in ground potential between all the devices, including noise, is within the common mode range for TIA485 specification then it should work without a ground.
In this situation is side A more positive than B or vice versa, this is what is meant by differential signals.
This common mode range of TIA485 was mentioned early in the article, -7 volts to 12 volts.
If this is exceeded in the system than one or more isolation devices will be needed.
Though personally I would always use a ground.

Biasing resistors, not to be confused with terminating resistors, only need to be small enough so that noise on the signal line does not get through the receiver, at least 200mV.
This has become confusing over time as the original specification required the receiver chips to have a defined input resistance which means that there is always going to be current flowing into the input pins. This limited a bus to 16 devices owing to the drive available from the driver chips.
Since then there has been quarter power, eighth power and even sixteenth power devices, so resistors for biasing with early devices are now too low in value to work properly in large systems, the driver will become overloaded. This means that later systems will have higher value resistors, and can lead to providing insufficient bias for older chips allowing noise through the receiver and affecting the integrity of the communications.

Trevor

Your point about Profibus is a good one but I think it re-inforces the the fact that rs485 requires 3 conductors.

The reason I say this is that Profibus is a standard that encapuslates RS485 and imposes other electrical requirments on it (like the isolation). Thus you get a bus with devices with a uniform electrical interface.

This is not the case with other RS485 applications. For example BACnet MSTP merely references the RS485 standard and does not impose other physical layer requirements.

I have to agree with Wilfried, viasing resistor calcs should not be required for field installations.

I’m sorry but basically you are wrong.
Voltages are not defined by themselves, they are only defined as the difference in the force of the electrical field. To measure a voltage you always need two points to put your probes on.

In RS485 one point is the signal wire A or B, and the second is the local ground of the receiver circuit. When you have an isolated device, the voltage of the isolated ground compared to the local ground of your device is undefined. This is especially true for the voltage difference between the signals and the isolated ground, because the signals are generated in respect to the local (isolated?) ground of the *transmitter* circuit. When these two grounds are completely isolated, there is no guarantee the receiver understands the data. It often happens the difference is too big. One proof of that is the real-world experience, transmissions become successful when you just add the ground line.

Biasing resistors do not really help, because for the purpose you describe you have to attach them to *every* device connected to the bus. This requires a recalculation of their values every time you add or remove a device from the bus.
(fact: calculation of resistors is not done by average customers.)
And in fact in this way you use the data lines to clear all voltage differences between the isolated grounds of the devices. At least you connect them via the resistors and the Data- line, which is pull-downed.
Even without bias resistors there may be some kind of parasitic connection between the data lines and the receivers ground, which also clears the differences between isolated components. That may also be the reason why in many applications the communication is fine with only two wires.

I agree, when using bias resistors as you describe it, there is only a reasonable current when a new isolated device is attached to the network. Later the isolated ground planes share the same electric potential, and there are only very small currents, if any.

By the way, even maxim-ic says the ground wire is required.