Slowing down Ethernet

An unusual requirement, but I can see how it might be useful in some situations.

Would it perhaps be practical in the situation to use multidrop RS422 (serial) instead of ethernet?

I think someone else suggested multi-drop - or, you could just run a bunch of plain pt-pt PPP links between the systems, and run an IP routing protocol to deal with occasional link failures. You could use heavily shielded serial cables and run them just about as slow as you please.

Are you worried about RF from the wire, or from the NIC itself? If not from the NIC itself, why not just run ethernet over fibre and stick with COTS kit?

Not entirely sure if that nixes the idea of fibre or not.

rick jones

It's not the clock speed that causes the interference, it's the rise-time of the signals.

IMHO, this is going to be really difficult to do, as all of your backoff, retry, and other timing algorythms are going to have to be redone. You are going to need to engineer, build, and maintain your own NICs, hubs/switches, and client devices.

Why not use Rick's suggestion of fiber?

[...] [...]

Why not just use fibre?

best regards Patrick

What kind of RF environment are you dealing with?

How do you know your network problems are RF-related?

What king of hub/switch do you have? What distances?

Hi all,

thanks for the comments so far.

I have to admit, I hadn't considered fibre, but this would probably be too expensive. The function that we have in mind is to collect data from a Radio Telescope. Hence it has to very RF quiet.

We have considered RS423, but the attraction of ethernet is that we can use established protocols (like SNMP) etc for data acquisition and communication for control.

We are looking at a number of local modules, each with its own IP address.

But as this an amateur venture (hopefully will hit schools etc) we really need to keep the costs down. Hence the remote controller will be PiC based and as data rates will be low we could use a hardwired hub. Ideally the only thing we would need to change would be the NIC.

Distances - I guess less that 100m.

Regards

Martyn

Shielded twisted pair? I believe that you can get shielded cat 5 cable and shielded modular plugs and jacks.

Hi Thomas,

I think that would be an option and OK for longer runs, but it still can be RF leaky and would not help with noise suppression from hubs or switches or from short patch leads

KM

Try the residental IP-over-powerline stuff. No URL. I see it for sale in the big computer stores.

Have you actually measured/heard noise comming from ethernet cable?

ISTM that CAT5e or CAT6, done right, is so carefully balanced as to not radiate much. Nail the interfaces at 10Mb just to make the harmonics that much weaker.

If you are getting noise try wrapping the hub/switch in aluminum foil and grounding it.

What band are you listening in ?

communication

As others have mentioned, you should consider RS422 and PPP (or SLIP). You can use all the established TCP/IP protocols (SNMP is not tied to ethernet).

You can run it as slow as you want (as some of us are painfully aware from the good old days), and it is easy and cheap to implement (SLIP/PPP is well supported on many different microcontrollers).

You could also consider10Base-FL if you don't mind buying used equipment. People are practically throwing the stuff away on eBay (looking now, someone is selling 20 10Base-FL to 10Base-T media convertors for $82).

It would be a PITA but if you get a phone line and a DSL account for each machine you'll have a network.

If you've got two serial ports in each machine you can set up a PPP (point-to-point protocol) connection on each interface, daisychain machines together and and get 100Kbits/sec, maybe. It's been ages since I did PPP.

Then there's this;

Maybe you want to go with 10base2 coax. Then you really only have to make sure you shield the data collection points well.

Well something else is finding 10base2 hubs might be hard unless you find some in someones storage closet. :)

Might be worth checking eBay for second-hand 100 Mbit/s fibre kit. That stuff has been around long enough that it may not be all that expensive.

rick jones

But he's going to get a LOT of noise with this setup. If I recall, he's listening for signals that are weaker than the noise on almost any unshielded wire. He hasn't said what he's using for collection devices but he might want to look at some micro controllers that have tiny basic and such for programming. They are very small, very low power, and can be sealed in a box with coax coming out back to a remote data collection computer.

Won't one need to put the hubs/switches into Farraday (terminology?) cages anyway?

rick jones

Yeah, but he's looking in a band that's about the 100th harmonic removed from the frequencies involved in serial comms, and the risetimes of RS232 signals are slow that there aren't any harmonics, anyway.

Al Dykes wrote in part:

Yes. Light snow on NTSC TV ch2 from pushing 100baseT over Cat3 about 10ft away from antenna. Snow droped from objectionable to barely noticeable when cable replaced by Cat5e. But totally absent when link depowered.

10 made it go away. IMHO, Cat3 is reasonably balanced but the slower twists give less room for slewing twists so there is more crosstalk. Cat5/6 is certainly better, but more in XT than EMI.

I would worry about the PC or whatever is at the radio-telescope generating the signal. Digital equipment is noisy.

-- Robert

Agreed. With these issues about CAT5, I'm suprised he's got a computer or a TV set anywhere in the area.

Based on what he's just said, above, why isn't 10MB on CAT6 wired to the highest spec quiet enough.

In another post I suggested ethernet over house power wiring (bought at retail computer stores) or getting each PC a POTS phone line and a DSL subscription and making an IP network out of that.

Neither of these solutions have any ethernet signaling.

Al Dykes wrote in part:

It possibly is, depending on his frequencies of interest. Digitial signalling has sharp edges which generate a wide frequency range.

And by being analog signalling (digital data) they probably produce much less interference, or at least it's confined to well-known frequencies rather than scattered.

-- Robert