Help please: Simple Ethernet signal booster/repeater

[FATechMan]

I need to design and build several low-cost "Ethernet" signal boosters, but I am unsure how to proceed because my application seems (to me) to differ in some key ways compared to advice I've seen in books and the Internet so far. I am also personally not very experienced or so knowledgable at the moment about the design considerations for high speed network systems. These devices will need to be built, tested and deployed in the field within approx. 3 weeks.

The problem and background: I must link together several seismic dataloggers - 'Geodes' - using existing CAT5 cable. The loggers use a 10BaseT 10Mbps Ethernet-like physical protocol, in that the signal levels are, I am told, +/- 2.5V differential signals. However, the loggers do not have MAC addresses and do not use IP addressing. The mating cable connectors are not the 8p8n/so-called "RJ45" ones because these logger units are designed to be used in the field, in mud and dampness; the connectors are, therefore, water-resistant circular multipole types, which also carry other signal levels and functions unconnected with the Ethernet data signalling. So it appears the manufacturers are using the Ethernet medium but not the full data protocols in a proprietary way. Additionally, while the Geodes will communicate happily when the cable lengths are 125m, they will not do so when the cable length is doubled to 250m - almost certainly because of signal degradation due to cable capacitance etc. - but certain seismic surveys require this spacing between Geodes. Multiple Geodes may be daisy-chained to 2500m or more. We have multiple spools of cable which are 250m lengths. Precise information about the workings of the Geodes is proprietary and not easy to access, so I am working rather "around the edge" of the problem and I want the solution to be non-invasive to any existing hardware.

What I think I want to do and what I've done about it so far: The 'simple' solution seems to me to guillotine the 250m cables into 2 x 125m segments and put a battery-powered signal booster between them (there is no option for power over Ethernet: it is not supplied by the Geodes nor do I want to risk injecting it in case of damage to the logger). This would enable the data to get from one Geode to its neighbour, whereupon the Geode's internal circuits will boost and relay the signal onto the next cable segment. Since the protocol is seemingly not 'official' Ethernet, essentially all that is needed is to take whatever signal is present, clean up its edges perhaps with a schmitt trigger or somesuch, increase its energy, and send it on its way otherwise unchanged, at these cable segment mid-points. It's a repeater function.

However, despite spending literally hours searching the Internet (including searching this site), trying to locate books etc. locally, I have not in this time found a definitive schematic or information about the end-to-end signal levels and interfacing needs to perform what I would have thought is a basic function. I have found several schematics in tech notes and sites (e.g. Micrel KSZ8841/42, **broken link removed**) which show magnetics used to interface to fancy 3.3V Ethernet transceiver chips and so on but I have not deciphered what the signal levels are once they are on the OEM side of the magnetics, i.e. the "digital logic" side of the circuits. By this I mean, the 10BaseT Ethernet signal appears at the connector, passes through some (often unspecified apart from being 1:1) transformer, and network of resistors and capacitors, then they hit the "TX+/-" and "RX+/-" pins of some Ethernet PHY chip or other - but I do not know electrically what the characteristics of these chip inputs are, because none of the datasheets I've seen show what the internal circuitry is. And without knowing what those inputs really are, I can't confidently envisage how I could replace the fancy chip with something simpler, just to condition and boost the signal. Questions I've asked are What will the signal levels be when they enter the "digital domain" at the chip's TX+, TX- from the magnetics, and the levels of RX+, RX- back out to the magnetics, for example? Are they simply a logic '1' voltage level (e.g. 3.3V or 5V), and 0V etc. with respect to IC supply, or some other values - maybe even tri-level? - floating between these chip supply extremes, due to the action of the inductors or the Ethernet signalling design, which these magic "TX+/-" and "RX+/-" input/output pairs deal with invisibly? There seems the same vagueness or un-expressed assumptions in all the cases I've seen, or, I'm missing something obvious.

What I would ideally like to do is cut out the fancy Ethernet chip in these circuits and replace it with a Schmitt trigger or similar to threshold and clean the signal: something that requires no programming and will not "intelligently" intervene in the data path (due to absence of MAC or IP or other information) because I don't know what the effect of that will be on the sample data communicated between Geodes. Data changes are totally unacceptable; just boost-and-go is all I want. A line-driver chip, maybe, with explicit detail on what to do to manage voltage levels.

The use of all-in-one filtered "RJ45" sockets is not an option for the physical connection as they are not rugged enough. Or, could I wire the multipole connector of each cable segment to its own short cable stub, terminating in an "RJ45" and a socket inside the box, without those stubs affecting cable impedance and exposing the signals to EMI/RFI?

Could I get away with a comparitor or op-amp interfaced to the magnetics, for example? If so, what would I need to do to ensure correct termination, signal levels and interfacing before passing the signals on to the second cable segment?

As further background - I'm not asking for help in this aspect - the booster will have to be ruggedised, as small as possible (say, 2 or 3 x cigarette packet sized), low current drain, probably 6V lantern battery powered and cost maybe $100-200 all in including that ruggedisation. As for the other signals in the multipole connectors, I can manage these much more easily.

I'd very much appreciate any advice from those of you who have done this kind of thing before, or can say for sure where the information can be found (because of the tight timing on this project, and having searched for so long already, I'd prefer confident or definitive answers please rather than speculative ones where possible - I do hope you will understand). Thank you in advance for your help!

 

[BradtheRad]

Three weeks timeframe? You're being asked to develop something that the manufacturer ought to be of some help with.

I was about to suggest an ethernet hub/switch/router. But you can't use RJ-45 connectors.

Non-standard protocols. No MAC addresses and no IP addressing.

In other words you can't use normal computer equipment which could quickly fill in the gap.

The idea of making a repeater should work, but you must figure out when and how the signals communicate, and then figure out how to make your reconditioned signals cooperate with existing signals.

There are ethernet repeaters over fiber optic links. You have probably researched this article or similar:

Ethernet (IEEE802.3)

Have you any reason to believe your boss previously inquired at the manufacturer, and didn't like the high price quote?...

So rather than pay for proprietary repeaters, he asked you to (in effect) reverse engineer the existing equipment, and manufacture fully-compatible repeaters in 3 weeks time?

---------- Post added at 16:26 ---------- Previous post was at 15:53 ----------

This looks like an ethernet repeater (and related products):

Ethernet Extender- Cat5 repeater for FULL RATE distance extension!

Network Extender, LAN Extenders, Cat 5 Extension, Poe, PoE Plus

It uses a small amount of juice from the cable. Perhaps you could hook it up to your lantern battery for power (since you say there is no power in the cables). You would have to prevent your power from traveling to equipment 125 meters away, of course.

I suggest the manufacturer is using something similar to a null modem rather than a networking interface. A null modem doesn't need MAC addresses or IP addresses or full ethernet protocols. It needs only 3 wires. Their equipment might include a traffic manager in the central unit to ensure that data gets received correctly from each remote logger.

In an earlier day this job would be done over a serial inteface. Probably at +12V and -12V. More reliable even if there's some signal degradation over long distances. Even though the speed might be slower.

 

[FATechMan]

Hi BradtheRad,
Thank you for your excellent reply. The link to the Veracity products is especially useful - these in-line extenders could do the job if I enclose them in a ruggedized housing and pipe in some PoE to the upstream side. I'm getting a quote for these (I'd need about 12) and cost will be a consideration.
As a parallel approach I'm looking at differential op-amps e.g. AD8132 AD8132 | Low Cost, High Speed Differential Amplifier | Differential Amplifiers | Specialty Amplifiers | Analog Devices enclosed in a ruggedized box, with short RJ45 leads soldered to the back of the multipole contacts; these would then plug in to 2 x "Ethernet interface" boards from Mikroelektronika (mikroElektronika - Communication Boards) : their technical support is brilliant. I am led to believe that by soldering a short lead with CAT5 cable and RJ45-like connectors, this probably will not affect impedance much or at all.

I'll let you know which route I take and how it goes. Thanks again.

 

[BradtheRad]

As long as you're willing to consider the repeater idea, which can work...

However there may be a simpler thing to try.

Have you confirmed how many of the 8 normal wires carry signal? This requires an oscilloscope and breakout box.

Because there's a chance that just 2 wires are carrying data from remote units. This is the twisted pair method. It might not require 3 wires as a null modem might use. It may not require full networking (handshaking lines, 'data ready to send', 'data received', etc.).

There's a chance the manufacturer dispensed with all those protocols, and used the simplest possible data transfer.

Consider that the manufacturer may have used ethernet cables because it's the current technology. Despite the tiny wires and the 100 meter limitation. It would be the easy thing to use because ethernet cables and connectors are readily available.

So you might try to extend the distance by using larger copper conductors.

I say this because I once strung small lamps along a tree-lined path. They were flashlight bulbs and drew 1/4 A. I used networking cable containing several wires. After 200 feet there was significant voltage drop. The further the bulb was from the power source, the dimmer it was. Individual wires had that much resistance.

So I incorporated 2 wires into one by soldering wires together at each end. I had to do this to each stretch of wire between bulbs. It greatly improved bulb brightness. It was simple Ohm's law. Of course I had to keep track of the color of the wires I was soldering together.

The same tactic may work for you. Ethernet wires are tiny. I would not be surprised if the 100 meter limitation is due to voltage drop as a cause of signal deterioration.

So you might join wires inside a CAT5 cable, or you might use a larger cable of your choice. You first have to determine which wires carry signal. You would need to be careful not to incorporate wires that are already in use.

It also depends on whether the wire characteristics cause signal deterioration. You don't want to soften the pulse transitions.

This method will involve work. (As will any of the other methods.) It could be the most practical if there are only 2 or 3 wires involved in data transfer.

 

[FATechMan]

Thanks for your suggestions, BradtheRad - cheers.