Dear Jeff - Please re-read the entire paragraph as I originally wrote it without the elipses [....]> that you inserted:
]If you like building your own power supplies - you can split your own design ]upstream of the regulator, after the rectifier, and place capacitors at ]both ends of the cat-5 cable (it now carries the unregulated DC) - install ]your regulator circuit near the load (even inside the case since there's ]normally room for it) and you have the same result. Because of the low ]power requirement for 802.11 hardware, even linear regulators work with ]tiny heatsinks. This is fair game for any load up to 1A - above that ]range is where DC/DC converters are really required. You'll note that I only refer to using a linear regulator for breaking a standard linear power supply into two parts - the unregulated half and the regulator - with cat-5 cable in between (this is effectively what manufacturers are doing now with cat-5 "power inserters" - providing an unregulated DC voltage which is then regulated within the network device. The unregulated voltage can be any value sufficient to overcome losses in wires and connectors. I often see 12-20V unregulated supplies that are finally regulated down to 5V internal to the device. Products are built this way so that the cheapest available "wall wart" can be used. It turns out these work equally well with light loads at the end of a cat-5 cable, provided a second capacitor is placed near the regulator to handle transients. I'm nowhere suggesting the combination of a 48VDC supply with a linear regulator - you took two paragraphs, joined them with elipses, and created a concept of your own, using my words. Nobody would use a 48V unregulated voltage to get 5V with a linear regulator - those are two concepts from two different methods of supplying power via ethernet cables. The most important part that you and I both are leaving out is the standard resistance of cat-5 pairs - per the systimax guide: "A typical 24-AWG UTP cable pair has 187.6 Ohms per kilometer or 0.1876 Ohms per meter" Both of us are leaving this out of our calculations, but at 100m, it's 18.7 ohms, or another 9.5V drop in your 1/2 A example. Most of my products only run 300mA each, and I seldom max out 100M, so I leave this out as just a few extra volts lost in transit. I believe the original poster already covered using AC - except in Loren's post he uses a pair of the cat-5 cable to join the secondaries of two identical tranformers to "step down" AC and then "step up" the low voltage back to 110 at the far end. His original post proposed the following: >For an inexpensive POE, I've used a pair of lawn sprinkler timer >transformers (24 VAC at around 1.25 Amps). One transform plugs into the wall >outlet and the 24 VAC is placed on the unused CAT5 pairs with a homemade >power injector. The second transformer is placed at the radio end and wired >as a step-up transformer. A female-female power socket is used to connect >the AP's own 120VAC powercube to the step-up transformer. Be careful!, the >exposed male prongs on the transformer have 120 Volts on them. In >California, Orchard Supply Hardware carries the transformers for around $13 >each. Ace Hardware has the same product at $16. There is a few volts drop on >the 120 Volt end at the access point with a 10 - 15 watt load. I have five >radios being powered by their individual transformer pairs and they seem to >work fine. > >Loren Zemenick (note that this quote of Loren's writing does not change the original concept) Personally - I'd rather keep AC fields off cat-5 cable but it seems to tolerate AC due to the twisted pair structure. AT&T has gone so far as to couple video with 10/100 ethernet although some crosstalk is seen. Technically, designs should not use the two spare pairs for AC signals except in certain limited exceptions - but corporations have violated this rule for years with success putting just about anything on cat-5 cable. Loren was exploring alternatives for powering low power devices at the end of a 100m ethernet cable - What I've discussed is the lowest cost method of splitting a supply - it's what manufacturers wanted in POE - What the IEEE standard finally approved is significantly more complex - What I use is a fixed 48VDC supply WITH dc/dc converters - but its not the least cost nor least parts count - splitting a linear supply into two "halves" represents the lowest cost alternative. It's not compliant with the approved POE standard. Neither is my dc/dc converter usage. I use it to power a hub and 4 APs with 60 degree sector antennas at the end of a single cat-5 run -- I need the power offered by 48V supplies. They all work though. Even the back to back sprinkler transformers or placing the transformer on one end of the cable and the rectifier on the other. Everett Jeff wrote: > > On Thu, 04 Sep 2003 17:57:24 PDT, [EMAIL PROTECTED] wrote: > > >. I've been using "my standard" for 3 years now to power outdoor > >ethernet devices - but all I do is put 48VDC on the spare pairs and use a > >standard dc/dc converter to create either 5VDC or 12VDC (or both) inside the > >device. > .... > >. Because of the low power requirement for 802.11 > >hardware, even linear regulators work with tiny heatsinks. This is > >fair game for any load up to 1A - above that range is where DC/DC > >converters are really required. > > > Don't get me wrong, this is an excellent post, but I don't know of many > 802.11b cards that are less then 300ma and most of the bridges are in the > 500ma to 700ma range. My Dlink 810+ bridge, I just measured it yesterday at > 500-650ma at 5volts. > > So with a linear regulator, you looking at it to drop 43 volts to come up > with 5 volts (48-43=5). At 500ma current your looking at a heat load of 21.5 > watts..... no tiny heatsink I know of can handle that without a meltdown. > > But your practice is a sound one (using switching regulators). I might even > asked if you took it a step further and just ran raw AC up the lines (like > from a doorbell transformer) and did the filtering and regulation (with a > switcher) at the ethernet device? > > > > -- general wireless list, a bawug thing <http://www.bawug.org/> [un]subscribe: http://lists.bawug.org/mailman/listinfo/wireless
