Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
Hardware

Low Power Ethernet Hubs? 24

mike.851 asks: "I'm an engineer working in the embedded systems field. For the past few years, I've been using RS-232 to interconnect various modules into the embedded systems that I build. Lately, I've found that RS-232 simply isn't flexible enough for my growing needs. I need to switch to something else, and I don't think that USB is really an option for me. I've been looking into embedded ethernet as an alternative. I've found several great low-power platforms for hardware development including several of the uClinux kits as well as products from Rabbitsemiconductor. However, I'm having trouble finding ethernet hubs that meet my power requirements (my goal is 2W or better for the hub). So far I've found Ctrlink and W-linx, and I've considered replacing the power supplies in commodity hubs with switching regulators. Does anyone have experience with these products? Does anyone know of other low-power hubs? Thanks."
This discussion has been archived. No new comments can be posted.

Low Power Ethernet Hubs?

Comments Filter:
  • you could always invent wireless power... just a thought
    • you could always invent wireless power... just a thought
      Tesla so beat him to that...
      • Re:well... (Score:1, Offtopic)

        you could always invent wireless power... just a thought
        Tesla so beat him to that...
        Really? No wires at all? I hope you're not talking about a Tesla coil! Otherwise, what is it a coil of?

        At any rate, wireless power exists everywhere, even before Tesla. It's in T-storms with their lightning. It's in the water. It's in the wind. It's in the earth, the sun, and all atoms. And that's just the stuff we can get at right now! Wait until someone makes a Dyson sphere!

        • Wireless power = converting RF energy to electrical energy.

          =)

          And yes, it's already done. Sorry to spell it out for you. Depending on the technology in question, it might even be an appropiate system, such as, say, in smartchips for passcodes.
          • I'm sorry to respond, but it is just a pet peeve of mine when people correct me and I am not wrong. (You said "Sorry to spell it out for you") Perhaps you meant to respond to another post, but I don't think so. Anybody who has made one of those old style transistor radios knows about turning RF engergy into electrical energy.

            But, if that's the case, then every transformer on the planet is using "wireless power". You see, you're making the definition broader than most EEs would agree with, but my whole point was that it is shortsighted to say that the only power is that which we have already gotten into our wires. Everybody already knows about what you're saying, and you add nothing to the discussion.

            I should have just left it at that, but you corrected me, and your comment was based on a poor reading of my comment. It's just annoying when people do that. Have a nice day.

        • Re:well... (Score:1, Offtopic)

          by autocracy ( 192714 )
          Yeah... let's just hit stuff with lighting... THAT works well.

          Tesla coils (haha, funny joke) were groundbreaking because they managed to place electricity in a usable form that didn't require a wire. Blowing wind across a room doesn't count either - you'd need a generator to convert it then. BTW Joe... exactly what is it that you accelerated into? It appears to have had an impact on your logic. (yeah, a pun, no kidding)

          • Yeah... let's just hit stuff with lighting... THAT works well.
            I guess you've never seen "Back to the Future"! ;-)

            And when we play Quake 3, I'll hang on to the lightning gun, and you can just have the gauntlet.

            Tesla coils (haha, funny joke) were groundbreaking because they managed to place electricity in a usable form that didn't require a wire. Blowing wind across a room doesn't count either - you'd need a generator to convert it then.
            I think you're pissed that your version of wireless power has wires, while mine doesn't. You seem to believe that electrical power is the only kind of power, but I still bet you'd run to the storm cellar during a tornado!
            BTW Joe... exactly what is it that you accelerated into? It appears to have had an impact on your logic. (yeah, a pun, no kidding)
            I guess you could say that I accelerated into slashdot with my Mega-Cool low User ID, and yes, hanging around here that long might have a lasting impact on my brain functionality. Duh! Oh well, at least you got in within the first 200000 users!
            • I guess you could say that I accelerated into slashdot with my Mega-Cool low User ID, and yes, hanging around here that long might have a lasting impact on my brain functionality. Duh! Oh well, at least you got in within the first 200000 users!
              Despite which you only recently started posting comments - as indicated by the fact that you only have 84 down, and 24 of those were in 3 months. To maintain that rate, you've been on less than a year. Based on those 24 comments I could go back and read, the damage must have gotten worse over time - you at least had the sense to keep quiet before.

              Sure, my UID is not the lowest, but at the least I've made an effort to contribute in a positive manner along with my smart-ass remarks.

  • What about FireWire faster than fast ethernet and is peerless. Apple just released a reference driver.
  • by Pauly ( 382 ) on Friday August 23, 2002 @08:01PM (#4130839)

    You could always take a step back in time and use 10base-2. Yeah, coaxial cable sucks and you're limited to 10 Mbps.
    Also, I have no idea if any embedded ethernet controllers support BNC connectors, but this would allow to you take one device out of the power consumption equation.

    Starting references: here [wown.com] and here [tldp.org].

    • Then you asume that a computer using 10base2 uses the same amount of power or less than a 10baseT network client. I didn't read the ethernet specs, but from a page like this [cewindows.net] which I found on google shows that most PCMCIA networkcards need about 100mA more when used on a 10base2 network.
  • Ethernet is designed for relatively long distances, and has high power requirements to support those distances. I've worked building embedded systems that used lower power trancievers on the backplane and on each board to save power. Obviously if you do something like this you won't be standards compliant anymore, but if you're only talking to your own devices it could work for you. If not, there are high speed serial interfaces availble on telecom chips that you could use. Search google for HDLC and you'll find exactly what you're looking for.
  • by Ranc0r ( 12107 ) on Friday August 23, 2002 @08:36PM (#4130983)

    I ran into a similar perdicament last year where I needed a low power ethernet switch/hub to connect two wireless units on a tower where the cable lengths to the ground were too long.

    We ended up using industrial ethernet switches from Sixnet. The price wasn't _too_ bad, I seem to recall $500 for a 5 port switch. Their site is: here

    Taking a look at the specifications located here [sixnetio.com] they have units in the 2W Range.

    I hope this helps. BTW, we've had no problems with the unit, even after a lightning strike that nuked two of the connected radios.

  • Lots of choices.... (Score:3, Informative)

    by lotussuper7 ( 134496 ) on Friday August 23, 2002 @09:30PM (#4131173) Homepage
    While Ethernet is one option, it might be a poor fit for your needs.

    Some of the things you left out were the ability to survive in harsh environments, dirty power tolerance, price considerations, distance, etc.

    There are a number of other bus systems that have made a few trade offs in these areas vs. total speed. Many of them are derived from work for the aerospace and automotive markets.

    A few systems you might consider are:

    LIN - very very low cost, but probably not fast enough for you.
    CAN - up to 1 mb/Sec, lots of low cost off the shelf parts available, developed mostly by Bosch.
    Flexray - faster yet, parts are more expensive, but it is faster
    MOST - very fast, but targeted at moving multi-media data.
    TTCAN - a modification of CAN for real time.

    Most of these will support power on connects/disconnects, and are very fault tolerent.

    Places to go:
    www.flexray.com
    www.can.bosch.com
    www.ami-c.org/downloads1.htm

    These are automotive focused (I'm NOT a "Car guy"), but that is because the industry is very cost aware and very big into reliability. What they tend to use is 10 year old technology that has been beat to death, has all the bugs worked out, has bunches of development tools, and all the hard parts (chips, in this case) are commodity items.
  • is about as low-power as it gets, heh.
    • If you reall only need the equivalent of an RS232 connection, then you may only need a crossover cable. Similarly, it may be cheaper to provide two or three ethernet trancievers and crossover cables to the associated units, than it would be to pay for a third-party low-power hub.
  • Have you considered PoE-based solutions? Here's one from 3Com, the NJ100 Network Jack [3com.com] (and a review of it by Tom's Hardware [tomshardware.com]). While the NJ100 is not intended for use inside a closed box, using one of these would allow you to route power along with data between the various modules in one of your systems.
  • ArcNet? (Score:1, Informative)

    by Anonymous Coward
    Depends how much width you really need- I've heard the venerable ArcNet technology is quite popular among the embedded set, and that updates to the spec have catered to that market, as regards noise-resistance, power, etc.
  • stay away from 100Mb (Score:3, Informative)

    by Matt_Bennett ( 79107 ) on Saturday August 24, 2002 @09:48AM (#4132870) Homepage Journal
    Stay away from 100Mb ethernet- by design, if you have a 100MB link, there is a continuous stream of a pseudo-random code being transmitted by each side, which means, basically, 40mW per port, no matter the activity. 10Mb uses much less frequent link test pulses. You can also go in and remove the LEDs, which are probably unnecessary in your application. If you really want to go hard-core, you can hack your way into the part using the MDIO interface (Similar to I2C) and disable unused ports individually.

    The Altima AC205 [altimacom.com] is an "ultra low power" part that may be used in the hub you want. It is 100Mb, but to use the power hungry 100Mb mode, both sides have to be 100Mb.

  • The same I2C protocol used in PCs for environmental monitoring and whatnot is useful for embedded too. You can string several devices on a two-wire bus, you get 2Mbps of speed (I think, it's been a while), and I've seen embeeded systems boards as well as devices like lcd display units that support this protocol for generic communication. There's also the I2C-variant "SMBus", not really sure what the difference is between the two.
  • PCI-NIC+Hubs? (Score:1, Informative)

    by Anonymous Coward

    It's probably not applicable for your use, but there are at least 4 manufacturers out there producing a combined PCI NIC (rtl8139) and 5 port 100Mb switch or 10Mb hub.

    One port is used internally for the NIC and the PC presents 4 ports out the back, with one having a rollover switch for uplinking.

    My own experience with small external hubs is that there's more power dissipated in the wall-wart transformer than in the hub itself. This also applies to modems. Most of the wall warts are pretty ineficiently designed and produce about 9-10W worth of heat.

    (This is significant when you have 70 modems in a cabinet. As a small ISP, we converted the modem pool to one big-ass transformer and a bus feed, which dropped power consumption by 600W and the AC bills by a few dollars a month.)

    If every mW counts, consider clipping the leds.

    Have you actually measured the current consumption of a DC hub to see how much they're chewing?

    Most of the DC hubs(*) I've encountered are being fed from 9VDC/250mA PSUs, so can't be exceeding 3 watts in the first place.

    (*)AC-fed ones are more common, usually with 9-12V 250mA PSUs feeding them, so the power consumption stacks up to much the same thing.

  • Maybe the "picnic [members.vol.at]" would be useful to you... its an open source TCP/IP stack for a PIC microcontroller...

UNIX is hot. It's more than hot. It's steaming. It's quicksilver lightning with a laserbeam kicker. -- Michael Jay Tucker

Working...