We're putting together a network of 10 environmental sensor stations that connect to a central terminal (ie. a PC). Each sensor generates about 500 bytes/sec. Unfortunately RF is not an option, so we have to use a wired network. The maximum distance between sensors is 1 mile. We'd like to be able to run an IP network if possible. Fibre-optic cable is the first option that comes to mind but is fairly costly. Does anyone here have any suggestions for some other options that would be worth looking into?
How about telephone cable? Your serial protocol must be prepared for lots of data errors at this distance in addition to potentially large DC currents induced in your cable every 11 years due to sun spots.
Preventing nearby lightning strikes from blowing up your equipment may be your greatest challenge.
RS485/422 will work at that distance, environmental effects ( lightning etc.. ) will be your main problem.. wireless may be a better solution.. why do you say it is *NOT* an option ? Presumably you will implement a data compression alogorithm to reduce your bandwidth requirements ??
Is this central terminal in the middle of the area, i.e. the cables would fan out to different directions from the central terminal or are the sensors in a singe line with the maximum distance between the central unit and the last sensor could be up to 15 km ?
If they are all on different directions, RS-422 or 20 mA point to point connections would be simple alternatives and only about 5 kbit/s would be needed. If there are several sensors along a single line, it would make sense to use some RS-485 multidrop connections, but the date rate would have to be much higher to transfer the same amount of data from multiple sensors and the protocol overhead.
Fiber optics would nicely solve any ground potential differences as well as lightning problems.
RS422/485 is specified for up to 1200 m with ordinary wires, however, with thicker wires and possibly higher than normal termination resistance (which is an option at least at 5 kbit/s, since the length is still much less than the wavelength) might be used.
Galvanic isolation would be essential due to ground potential differences and voltages induced by lightnings. If the sensors are operating from floating power supplies, this might be enough and proper optoisolation at the central unit. To protect the central unit, it might even make sense to use an RS-4xx/fibre converter, use a few meters of fibre before connecting to the PC with an other converter.
Data compression over unreliable links is not a good idea, since you will need an intermediate layer that will guarantee a 100 % transfer (e.g. with repeats) in order to get _any_ data through.
In typical sensor applications some measurements are periodically transmitted and if some measurement is corrupted, it is usually not a big deal, since the next measurement will most likely get through and the missing value can be interpolated. The important thing to know if a measurement is valid or not, which can be done with CRC checks etc. With a bad quality link, it makes sense to use short frames with CRC checks or a long message (e.g. due to half duplex delays) and split it into sections with independent CRC words within the message. This way a single bit error does not corrupt all measurements for a sample period.
You can buy little RJ45 jacks with builtin microcontrollers and everything. I think they are called smart ports or something like that. Shouldn't be to hard to find. We have some that are no bigger than a typical normal plastic right angle RJ45 jack.
Thanks everyone for the info. I think the issue with any copper-based solution is the difficulties associated with running it outdoors over long distances. Fibre gets around all these issues, but then we have to have fibre-ethernet conversion in each weather station. We don't have alot of experience running these kinds of distances. Has anyone here seen this work with a copper-based solution?
Your problems are no different from what telephone companies do with their $multi-million computer. If damaged, there is no telephone service for a week. Interface electronics can cost upwards of $10,000. Copper based communication - reliable and without damage - is well proven, inexpensive, routine, and found in virtually every town. Yes fiber can be used. But you stated 'low cost'. Fiber becomes low cost only when data rates are massive.