1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Electronics Undervolting a network switch

Discussion in 'Modding' started by Miser, 23 Mar 2007.

  1. Miser

    Miser New Member

    Joined:
    1 Oct 2005
    Posts:
    47
    Likes Received:
    0
    As some of you know, I am integrating an 8-port network switch into my fileserver. The AC/DC adapter that came with the switch outputs 7.5 VDC. Do you think there would be any negative repercussions from running it on 12-5=7VDC off my power supply? Would noise be an issue? Should I just integrate the transformer into the case?
     
  2. kbn

    kbn New Member

    Joined:
    17 Mar 2004
    Posts:
    603
    Likes Received:
    0
    using 12v would probably be safer... making 7v from 12-5 would cause problems with the final voltage being 5v higher than for every other device connected to it..

    there will be another regulator inside, so the undervolting will depend on the output of that. if its too close it wont work.
     
  3. Wolfe

    Wolfe New Member

    Joined:
    7 Sep 2003
    Posts:
    776
    Likes Received:
    1
    As long as the only connections going to it (aside from the power) are ethernet cables, you'll be fine.


    Ethernet over UTP cables uses differential signaling, and is isolated at each end with a transformer.
     
  4. whypick1

    whypick1 The über-Pick

    Joined:
    17 Oct 2002
    Posts:
    2,014
    Likes Received:
    2
    As kbn stated, there's probably a regulator in there, and dollars to donuts it's some form of 5v regulator. What I'd do is crack open your router and see if there is one (should be a square-ish IC right next to the power jack). Odds are, the thing is just a linear voltage regulator, so excess voltage is shed in the form of heat, to put it crudely.

    If you're going to put 12v across it, you need to make sure that the IC won't overheat and shutoff (bad) or overheat and die (very bad). You can find this out by finding the datasheet for the regulator, then finding the power dissipation of the regulator, measured in °C/W, as well as the maximum temperature of the regulator. The wattage the regulator will have to dissipate is (Vin - Vout) * I (use the current on the power supply, so you get a worst-case figure). If this number is less than the max temperature minus room temperature, you're set. If not, there are alternatives, but I'll bog you down with them only if the above fails.
     
  5. r4tch3t

    r4tch3t hmmmm....

    Joined:
    17 Aug 2005
    Posts:
    3,166
    Likes Received:
    48
    Use 12V with a couple of diodes to trim the voltage down a bit. Using six of these would bring the voltage down to 7.8V. Or you could use a voltage regulator, but assuming you need an AMP constant for most switches (mine do) it would be cheaper to use the aforementioned 2.5A diodes which would be cheaper.
     
  6. Miser

    Miser New Member

    Joined:
    1 Oct 2005
    Posts:
    47
    Likes Received:
    0
    First off, thanks everybody for your replies.

    I guess since nobody picked up on the noise bit I threw in it is not a problem. I guess all of that would be handled well by the regulator, the five 16v 470uF capacitors, and the other assorted caps on the board.

    kbn - I think I understand what you are saying about the 12-7=5: the onboard logic would still operate at the 7v potential difference, but any connected devices (ie switches, hubs, NICs, mobos) would get the 12v...but Wolfe clarifies that this shouldn't matter since the port voltages are transformed on each device anyway. I also get that the input voltage to a regulator has to be a certain degree higher than the output for it to function.

    That being said, Whypick1- I have definitely acquainted myself with the board, and I could only find one IC that I think could be the regulator. The others are processing, port controllers, etc. The IC in question is marked 34063 5D01. It is 8-pin, rectangle, about 8mm x 5mm, non-heatsinked, and not resting on the board(guess this means it isn't dissipating much heat).

    I have no clue where to find datasheets for this, and google yielded little. If I can't find information that says the regulator can drop the wattage required(power brick is rated 1000mA, btw) and you guys advise I don't 12v-5v it, I will see about finding a regulator on Mouser or persue r4tch3t's way.

    To show how serious I am about this project, and to show you what I am working with, I have included a picture:

    [​IMG]

    Sorry about the picture quality, it is late and the camcorder and firewire was just sitting there... thanks again!
     
  7. Macaba

    Macaba New Member

    Joined:
    4 Mar 2006
    Posts:
    107
    Likes Received:
    1
    I think I may see an inductor in there. So it may be a switch mode IC. In which case, its quite likely that it can handle inputs upto 35V. Perhaps it may not. But 12V is a far safer voltage than 35!
     
  8. cpemma

    cpemma Ecky thump

    Joined:
    27 Nov 2001
    Posts:
    12,328
    Likes Received:
    55
    I'm surprised there's no obvious regulator IC, my router-modem managed a couple for 5V and 3.3V chips. But the idiot-proof, no guesswork, safe method is to pre-regulate the 12V down to 7.5V with a 317T and a couple of resistors, no need for a trimpot, accuracy is not of the essence AFAICS. Or the string of 6 series diodes would work.
     
  9. Cptn-Inafinus

    Cptn-Inafinus Member

    Joined:
    3 Mar 2007
    Posts:
    601
    Likes Received:
    12
  10. kbn

    kbn New Member

    Joined:
    17 Mar 2004
    Posts:
    603
    Likes Received:
    0
    I forgot too add in my first post that the main thing that is badly designed in network stuff is heat output. My last switch died because it got hot enough for the solder to melt in several places - several components moved slightly from the contacts, but i think it died due to heat rather than creating a short.

    Most regulators/regulator circuits can handle inputs at least up to 20v (usually 35v+ for a regulator ic), if you don't consider the heat output (which can be huge on a linear circuit).

    IMO you should run the switch without its case from 12v, if nothing gets hot (which it probably wont due to switchmode regulation) then its good. if it does you might want to add diodes however they usually get very hot and will annoy you with placement/portability issues... (a small diode disipating more than 0.5W will also need some form of cooling... )
     
  11. Miser

    Miser New Member

    Joined:
    1 Oct 2005
    Posts:
    47
    Likes Received:
    0
    Ok, I have done a little looking around, and I have decided for safety and reliability's sake that I am going to regulate the voltage down to 7.5v using your method, cpemma. The problem I am running into - though I'm sure it's no big deal - is the way Mouser lists its LM317s. When I view the details for a unit, in the general specs to the left, it will say something like between 500mA and 1.5A; but on the the right it will list something like 10mA. I want a 1A capable, since that is what my original transformer is rated. Can I get a hand with this? Also, what wattage resistors do you recommend for R1 and R2 per your calculator? I already know I need them in 240ohm 1200ohm.
     
  12. r4tch3t

    r4tch3t hmmmm....

    Joined:
    17 Aug 2005
    Posts:
    3,166
    Likes Received:
    48
    [​IMG]
    Using this circuit will get you 7.5V.
    It is the same circuit I am using for the PSOne LCD screen, but I am using a LM350 as it is rated to 3A.
     
  13. cpemma

    cpemma Ecky thump

    Joined:
    27 Nov 2001
    Posts:
    12,328
    Likes Received:
    55
    You want the LM317T, TO-220 case, 1.5A max rating. Only 2 to pick from at Mouser, Fairchild or ST, no difference apart from price AFAIK. It will need a small heatsink, at 7.5V 1A you'd be wasting 4.5W so one rated around 12-14°C/W should be OK.

    R1, R2 anything from 1/8W up will do so common 1/4W are OK.
     

Share This Page