I am making a control board for a 10 pound battery powered UAV. Every oz and every watt counts, and the ethernet signals only need to go a few inches from the two radios to the autopilot and the payload. Is there a way to split the ethernet signals with an unpowered, ultralight circuit so that all 4 devices can speak to one another? They will all be on the same circuit board.

 

Not really. You need something to serve as the "traffic cop".

A switch is better than a hub.

With a hub, it's the equivalent of a pizza delivery guy walking into a crowded room and yelling "Who ordered a pepperoni?"

With a switch, the pizza guy knows which person ordered the pizza, and delivers it right to them without asking. You get far fewer collisions that way.

I have a Netgear FS105 5-port switch, but it uses 7.5W (1A @ 7.5v); kind of steep for your requirements.

You might look to see if Microchip has some Ethernet-enabled uC's that you could use to build a low-power hub/switch.

 

Total network data traffic is about .1mbit/sec peak. That's point-one megabit... And yes I already said what you are thinking: why does it need to be ethernet. The answer is because managers like complicated things.

I literally just need to get the packets there. There will not be any collisions. No switching is needed. The space and weight requirement I have is very small, it needs to be about the size of a quarter.

 

Does this help?: http://www.google.com.ar/search?q="...s=org.mozilla:es-AR:official&client=firefox-a

 

You can do that pretty darn small if you use SMD components and if you do not need the rj45 jacks.

A quarter-size would be do-able.

 

Very interesting; hadn't seen that before.

Here's a 6-pin two-pair series diode SMD package you might consider: http://www.diodes.com/datasheets/ds30386.pdf

You'd need 8 of them for 4 Ethernet connections.

 

Depends in on the Ethernet devices you're using, a four port passive hub by simply properly wiring the Ethernet lines together may work, depends on the protocols you're using as well as the bandwidth you're using, you'll get collisions but again that's where the protocols and how they handle it matter.

Why are you using ethernet for inter device communication on that scale? It's not particularly practical.

 

And yes I already said what you are thinking: why does it need to be ethernet. The answer is because managers like complicated things.

Bosses. Who needs 'em!?

 

How about connecting them all in a ring, and then have each device forward (retransmit) any packet that it didn't send?

Richard

 

What about using token ring instead of RJ45?
All you need then are some resistors to act as terminators

 

He said the devices are only inches apart, I'm not sure why such a complex transmission medium is being considered, a simple SPI interface would work. But if the devices being used communicate using Ethernet, then that's what he's got to work with. A simple passive hub will work, but collision detection will be required.
If TCP/IP communication is being used then everything will sort itself out automatically, so long as the bandwidth of all the devices isn't too high.

 

Funny, I've been fooling around with Ethernet for years, but I really don't recall what the specs are for the signal levels on CAT5/RJ-45 connectors. Anybody? Bueller? Bueller?

Wiki's entry: http://en.wikipedia.org/wiki/Fast_Ethernet
IEEE Standards (free download): http://standards.ieee.org/getieee802/802.3.html
I just don't feel like digging through a lot of documentation at the moment. However, I hope our OP has looked at which particular flavor of Ethernet their interfaces will have.

I'm also thinking that the way that diode hub was wired, one needs an odd number of connections (either three or five); since the original used diodes with a Vf of ~1v, a 5-connection passive hub would need to use diodes with a Vf of 3/5 or 0.6v at the same current level. Perhaps 1N914 or 1N4148 diodes would be worth considering. I simply don't know offhand.

 

Whilst unsuitable for this job, rj45 splitters have been around a while.

I often use these to add printers when there is only one cable run to a location.

http://cpc.farnell.com/jsp/search/browse.jsp?N=411+1003006+521&isRedirect=true

 

I didn't dig into the specs too far but ethernet is only a couple of volts, somewhere between 2 and 4, and the actual far end detection threshold is only a few hundred mv's. So a diode drop or two or even three shouldn't hurt anything. I don't remember ANY of the details but I do know when I first used ethernet that passive four port hubs were queit common and for close quarters 4-8 ports were available.

As far as collisions go which is the only real problem the bandwidth used by each device and the protocol they use to communicate is hyper critical.

Traffic collisions and voltage drop make them impractical for large networks or any decent distance.

 

Here is an option. It should be pretty easy to build an ultra small PCB to fit 24 diodes "surface mount?".
http://www.eeweb.com/blog/circuit_projects/building-a-passive-ethernet-hub

 

Here is an option. It should be pretty easy to build an ultra small PCB to fit 24 diodes "surface mount?".
http://www.eeweb.com/blog/circuit_projects/building-a-passive-ethernet-hub

This thread is 5 years old.

 

Yes the thread is old, I was looking for the same info Nickw1881 was looking for. Found something that worked and thought it would be appropriate to attempt an answer so others are not spinning wheels, if they happen to look to this thread 5 years from now.

 

Yes the thread is old, I was looking for the same info Nickw1881 was looking for. Found something that worked and thought it would be appropriate to attempt an answer so others are not spinning wheels, if they happen to look to this thread 5 years from now.

Agreed. But on this site, no good deed goes unpunished. Thread will likely be closed and your post may be copied to a new thread and you will, for some unknown reason, be accused of hijacking.