Author Topic: 5 port gigabit switch powered over POE via one of the cables  (Read 5705 times)

0 Members and 1 Guest are viewing this topic.

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #25 on: November 29, 2022, 09:13:10 pm »
Passive injectors are not standards compliant
Mind pointing out the IEEE 802.3af chapter that says so?

(As far as I know, the "negotiation" is simply detecting the DC resistance of the twisted pairs used for power delivery.)

Mind providing access to the very expensive document?
I do not have access anymore myself, that's why I asked.  (I had to scour the internet to check if my recollection is correct; it sometimes fails me. Not 100% sure here, though.)

A PSE checks for a signature resistance before enabling power. This is to avoid blowing things up. Just dumping 48V (or less) on the spare pairs is a recipe for flames.
Yes.  The former is what a proper passive PoE/PoE+ adapter does.  The latter is what a $5 cable does.

To be clear, I'm not advocating using directly spliced cables that cost $5-$10 except when you clearly mark the cable as being used for such.  I am advocating passive injectors and splicers that do specify they're IEEE 802.3at compatible.  As I mentioned, I'm a decade out of date on the exact devices, so I cannot suggest any, but I do believe there are standards-compliant devices for sale in the $20-$30 range.

(Edited: the Mikrotik POE injectors actually seem like they might do the resistance check, even though they cost < 10€.  Couldn't verify it from the manual; might be worth it to contact MikroTik about it, though.  Also, TP-Link TL-POE10R splitter (17€ here) does specifically say it is IEEE 802.3af compatible, but it does not provide an isolated DC output, which would be useful.  Essentially, its negative is tied to the ground on the POE injector or switch.)

(For example, Conrad claims TP-Link TL-POE4824G (24€) is IEEE 802.3af compatible injector, but the datasheet or manual doesn't mention 802.3af at all.  I would need to open the device and inspect it, to see if it has the proper detection before injecting voltage.  I do believe they use something like 120 Ohms or so for the minimum DC resistance – below that the target being assumed to not be PoE compatible), and that doesn't need much circuitry at all, as I'm sure you know, and at max. 2A or so, tiny MOSFETs suffice for the switching.  Definitely no microcontroller or such needed, although I'm pretty sure dedicated IC's do exist for this.)
« Last Edit: November 29, 2022, 09:22:48 pm by Nominal Animal »
 

Online Monkeh

  • Super Contributor
  • ***
  • Posts: 7990
  • Country: gb
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #26 on: November 29, 2022, 09:23:35 pm »
A PSE checks for a signature resistance before enabling power. This is to avoid blowing things up. Just dumping 48V (or less) on the spare pairs is a recipe for flames.
Yes.  The former is what a proper passive PoE/PoE+ adapter does.  The latter is what a $5 cable does.

The former is what a midspan PSE does, whether it uses mode A or B. The latter is what 'passive' is.

Quote
I do believe they use something like 120 Ohms or so for the minimum DC resistance – below that the target being assumed to not be PoE compatible)

Apply 48V to 120 ohms. Not a good outcome. The signature is a 25kohm resistor.
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #27 on: November 29, 2022, 09:47:59 pm »
A PSE checks for a signature resistance before enabling power. This is to avoid blowing things up. Just dumping 48V (or less) on the spare pairs is a recipe for flames.
Yes.  The former is what a proper passive PoE/PoE+ adapter does.  The latter is what a $5 cable does.

The former is what a midspan PSE does, whether it uses mode A or B. The latter is what 'passive' is.
That's not what I recall the IEEE 802.3af and .3at saying.  Do you have any reference for that claim?
I guess the P802.3 DTE Power via MDI Study Group mailing list might have useful info, but while they do discuss 'midspan PSE's, they don't seem to mention 'passive' at all.

Apply 48V to 120 ohms. Not a good outcome. The signature is a 25kohm resistor.
25kOhm makes more sense, yeah.  :-[

If I recall correctly, before voltage is supplied, a 1V to 2.5V test voltage is used on the PoE mode B pins.  This test voltage is well within the signaling voltage range, so won't cause any damage.  If there is no return voltage, the other end is not PoE capable, and the pins are disconnected.  The return goes to a voltage divider, and if the voltage divider output is high enough (i.e. series DC resistance of the two pairs low enough), the pairs are connected to non-PoE capable device, and no power is supplied.  If the voltage divider output is within the acceptable range (large enough resistance on the other end), the test voltage is switched off, and the 48V supply is switched on.

The receiving end has that signature resistance between the two pairs (in mode B at least), and taps the voltage across the signature resistor.  (48V over 25k is below 2mA, or less than 100mW dissipation in the 25k signature resistor.)
The most useful devices add an isolated DC-DC converter (class II, no protective ground) and regulator, but this isn't in the IEEE 802.3af/at, just a practical nicety for us users.

If you disagree, do say so.  I'm not claiming I'm exactly right here, because I am going by recall and occasional net search.
« Last Edit: November 29, 2022, 09:50:34 pm by Nominal Animal »
 

Online Monkeh

  • Super Contributor
  • ***
  • Posts: 7990
  • Country: gb
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #28 on: November 29, 2022, 10:00:14 pm »
A PSE checks for a signature resistance before enabling power. This is to avoid blowing things up. Just dumping 48V (or less) on the spare pairs is a recipe for flames.
Yes.  The former is what a proper passive PoE/PoE+ adapter does.  The latter is what a $5 cable does.

The former is what a midspan PSE does, whether it uses mode A or B. The latter is what 'passive' is.
That's not what I recall the IEEE 802.3af and .3at saying.  Do you have any reference for that claim?

Which? The terminology of 'passive'? Every single 'passive PoE' product on the market - not a one will claim to be 802.3af or otherwise, and generally nobody specifies (outside of a note in the depths of the manual) whether the PSE uses mode A or B because it doesn't matter.

Passive is the industry term for dumb DC bias supply - there's nothing passive about any true PoE system.
« Last Edit: November 29, 2022, 10:01:48 pm by Monkeh »
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #29 on: November 30, 2022, 05:01:31 am »
Fine; I'm not sure enough either way myself.

It's just that devices like 15 watt TP-LINK TL-POE150S is often sold as "passive injector", but TP-LINK definitely says HW v4 'complies with IEEE 802.3af' (PDF); they emphasize it.  It costs less than 20€ hereabouts.
 


Online ejeffrey

  • Super Contributor
  • ***
  • Posts: 3686
  • Country: us
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #31 on: December 01, 2022, 07:35:21 pm »
This was done in another thread on RJ45 internal-magnetics jacks here recently, but the dumb power injectors are well known for blowing up the RJ45 on the other end.
:palm:

Why do you insist on calling them dumb, when the proper term is passive?

The entire idea of the 802.3af/at standard is to give users tools.  Are all tools that users can shoot themselves in the foot dumb?

Not everything has to be done the complicated way, throwing lots of money at it just so one can put the prestigious 'Enterprise' label on it.

Passive 802.3af is not a thing.  802.3af can send power over either the used or unused (for 10/100) pairs (chosen by the power source), but it always requires negotiation.  It is not compliant for the power source to send power without first detecting a compliant powered device. In addition powered devices are required to support power on either the data or spare pairs.

There are active 802.3af midspan injectors but the vast majority of injectors are passive injectors that are not compatible with IEEE 802.3 and may destroy compliant devices.  That doesn't mean they aren't useful, but they are not standards compliant and they are certainly not 802.11af.
 

Offline peter-hTopic starter

  • Super Contributor
  • ***
  • Posts: 3671
  • Country: gb
  • Doing electronics since the 1960s...
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #32 on: December 02, 2022, 06:02:04 pm »
Hmmm

The Tenda POE30G-AT does not work with the TP-LINK TL-RP108GE.

The POE30G just sits there flashing its PSE LED 5 times, then a gap. This doesn't happen if I connect it to a normal switch, so it is detecting something different about the TP-Link box.

And no lights on the TP-Link (dead as a dodo) so it isn't getting power. Maybe the TP-Link wants a passive injector which doesn't negotiate? It can take POE IN on ports 1-7 which is suspicious.
« Last Edit: December 02, 2022, 06:32:47 pm by peter-h »
Z80 Z180 Z280 Z8 S8 8031 8051 H8/300 H8/500 80x86 90S1200 32F417
 

Offline MarginallyStable

  • Regular Contributor
  • *
  • Posts: 65
  • Country: us
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #33 on: December 02, 2022, 10:22:45 pm »

From TL-RP108GE spec sheet:

"7 Passive PoE-in RJ45 Ports:"
 

Offline peter-hTopic starter

  • Super Contributor
  • ***
  • Posts: 3671
  • Country: gb
  • Doing electronics since the 1960s...
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #34 on: December 02, 2022, 11:24:08 pm »
I guess so, but it was suggested further back:

Quote
Because TL-RP108GE supports PoE over GbE, it must have active negotiation logic.

Waste of 37 quid.

Can anyone recommend something which has at least one active POE IN port?
Z80 Z180 Z280 Z8 S8 8031 8051 H8/300 H8/500 80x86 90S1200 32F417
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #35 on: December 03, 2022, 12:20:41 am »
I guess so, but it was suggested further back:
Quote
Because TL-RP108GE supports PoE over GbE, it must have active negotiation logic.
I also said that its manual or specs did not say anything about IEEE 802.3af or 802.3at, which is annoying (should have written suspect).
One of its quirks is that you need to configure the TL-RP108GE (when powered via PoE) before using it to power downstream devices, so when testing and configuring, make sure you have nothing else but the one incoming PoE connected to port 1 to 7.  The eighth one is the "PoE" output.

But don't throw it away yet, let's diagnose this thing.

Otherwise, I'll buy it off you, since you apparently bought it because of my posts.
(I'm not just some asshole who throws shit out there, and when it creates a mess, just vanishes to leave the clean-up to others.  I take responsibility for my advice, even if not understood the way I intended it.  I'm very serious about this.)

Did you check your cable is Cat5e or Cat6 rated, and not a cheapie?  And straight (MDI), not crossover (MDI-X)?  While autonegotiation handles crossover, I do not think the PoE stuff can do the same.  Also, the 100 Ohm resistance limit per pair is strict.  The Tenda could be refusing to power the downstream device, because the cable resistance is in the gray area (too high, over 100 Ohm at 48V DC per wire (4-4, 5-5, 7-7, 8-8), to be Cat5e/6), or even one of the pairs disconnected/broken.  When measured using a multimeter, mine are all in the single-digit ohm range, 2 to 3 Ohms.

I checked the Tenda POE30G-AT datasheet, and it says it is standards compliant, "802.3af/at mid-span" injector, and uses pair 4,5 for positive, pair 7,8 for negative, at max. 51V DC 0.8A.  It also has an universal supply (AC 100-240V 50/60Hz), so the injection side should work.  It doesn't mention anything about the PSE LED blinking, and implies it stays off if the downstream device is not 802.3af/at compatible.

I also checked the TL-RP108GE datasheet, and it says the first seven ports are "Passive PoE inputs" using pair 4,5 for positive, pair 7,8 for negative, at either 24V or 48V.  IEEE 802.3af/at compliant inputs have just the 25kOhm resistance between (4,5) and (7,8), so could you grab a multimeter and check that?  The device could be just broken.  In other words, that on the TL-RP108GE PoE input ports, 4-7, 5-7, 4-8, 5-8 all have about 20k to 50k of resistance, and very low between 4-5, 7-8?
 

Offline peter-hTopic starter

  • Super Contributor
  • ***
  • Posts: 3671
  • Country: gb
  • Doing electronics since the 1960s...
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #36 on: December 03, 2022, 08:32:49 am »
Nominal Animal you have helped me here so many times, I would be very happy to send you this switch for nothing :)

The relevant cables are four pair CAT5. Installed 2004. About 10m long. But FWIW I've connected it to a POE port on a real POE switch (Linksys) with a normal short cable and it looks dead as a dodo too then.

Measuring the resistance on the RJ45s shows nothing. Not 25k for sure. Just open circuit. Maybe they use diodes, or something else like MOSFETs to detect the presence of a voltage. Or may just be duff. I will send it back. It came from Amazon, so this is easy.

The Q is, what 8 port switch has a proper active POE input? This
https://www.amazon.co.uk/Ubiquiti-US-8-Managed-Desktop-Passthrough/dp/B01N362YPG
actually says so https://dl.ui.com/datasheets/unifi/UniFi_Switch_8_DS.pdf but is rather pricey.
Z80 Z180 Z280 Z8 S8 8031 8051 H8/300 H8/500 80x86 90S1200 32F417
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #37 on: December 03, 2022, 08:55:57 am »
Nominal Animal you have helped me here so many times
Nope, that's not it at all.  I am just very, very serious about standing behind my words, and put my money where my mouth is.
It is a character question for me; about who I want to be as a human.  So, I'm not doing this for you, I am just being who I want to be.

The relevant cables are four pair CAT5. Installed 2004. About 10m long. But FWIW I've connected it to a POE port on a real POE switch (Linksys) with a normal short cable and it looks dead as a dodo too then.
They sound like they only have two pairs, i.e. 1,2 and 3,6 connected.  You can only use this with mode A PoE ("endspan"), and those seem much rarer than mode B PoE ("midspan", using pairs 4,5 and 7,8) or mode C (which uses all four pairs).

If you measure the cable, you should have 1-10 Ohms on each wire, on the same pin at the end connectors (as you want to use a straight cable for PoE).
If pins 4, 5, 7, 8, or any of them are not connected, the cable is at fault.  If only pins 1, 2, 3, 6 are connected, you need a mode A PoE that only uses those same pins –– but even then the cable will be limited to 100Base-T anyway.

If you could test the TP-Link and Tenda with a short, known-good cable, it would help me and my peace of mind.  Because then the problem is a question of exactly what you need to solve the problem of PoE over a cable with only two pairs, 1,2 and 3,6, connected, and as the answer to that is not "IEEE 802.3af/at", there is nothing wrong in the TP-Link or Tenda per se.  It's just that you need a specific, rarer variant of IEEE 802.3af/at to pass power over your two-pair cable.

Alternatively, you can just test the TP-Link PoE input ports.  If between pin 4or5 and pin 7or8 you see 20-50 kOhm resistance, preferably 25kOhm, its PoE inputs are IEEE 802.3af/at compatible –– but mode B, and not the mode A you need.  They're probably easier to test with a connected cable, on the connector at the other end of the cable.

I'm pretty sure that there is a cheap solution here, one that is robust and works; I'll go check, and report back soonish.
« Last Edit: December 03, 2022, 09:10:51 am by Nominal Animal »
 

Offline peter-hTopic starter

  • Super Contributor
  • ***
  • Posts: 3671
  • Country: gb
  • Doing electronics since the 1960s...
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #38 on: December 03, 2022, 09:57:13 am »
Quote
They sound like they only have two pairs, i.e. 1,2 and 3,6 connected

That issue exists only in a section of this system, and is why I can't feed POE through there (from the left in the diagram below). So I am trying to feed it in from the other end (from the right) but that involves powering a POE-powered switch.

To clarify, I am attaching a diagram.

I can see 3 approaches:

- find a POE powered switch which is not too expensive
- assume the "dead" switch is actually OK and try powering it from a 10 quid dumb POE injector (would be more money lost if it doesn't work)
- get a POE "extractor" (called "splitter") for the location of the "?" switch below, which can be powered by the Tenda box (so needs to be "standards compliant POE") and use that to power a normal DC powered switch in location "?" - this looks like it might work: https://www.amazon.co.uk/dp/B08HS4NT13 and this is another https://www.amazon.co.uk/TP-LINK-TL-PoE10R-TL-POE10R-PoE-Splitter/dp/B001PS4NWW

The current issue is when mains power is lost, the PC loses internet connectivity. Other solutions may include ETH over mains (have played with that, speed is really poor) or WIFI (speed is really poor, although the main WIFI AP is already powered via POE from the UPS).

This issue exists with remote ETH cameras and is very common, but not many people need to run a switch at the "remote" location.

« Last Edit: December 03, 2022, 02:22:23 pm by peter-h »
Z80 Z180 Z280 Z8 S8 8031 8051 H8/300 H8/500 80x86 90S1200 32F417
 

Online Monkeh

  • Super Contributor
  • ***
  • Posts: 7990
  • Country: gb
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #39 on: December 03, 2022, 05:02:47 pm »
- find a POE powered switch which is not too expensive

Some HPE units on eBay mentioned earlier, Halcyon's suggestion of an Ubiquiti switch is reasonable, you just have to play hunt the retailer.

Quote
- assume the "dead" switch is actually OK and try powering it from a 10 quid dumb POE injector (would be more money lost if it doesn't work)
- get a POE "extractor" (called "splitter") for the location of the "?" switch below, which can be powered by the Tenda box (so needs to be "standards compliant POE") and use that to power a normal DC powered switch in location "?" - this looks like it might work: https://www.amazon.co.uk/dp/B08HS4NT13 and this is another https://www.amazon.co.uk/TP-LINK-TL-PoE10R-TL-POE10R-PoE-Splitter/dp/B001PS4NWW

Either of these I'm sure would work fine. A proper PoE switch or the splitter is preferred over passive. I use a TP-Link injector and splitter combination to supply my ONT, so far so good.
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #40 on: December 03, 2022, 05:10:04 pm »
Okay, so to recap:  You need 802.3af (or 802.3at) mode-A PoE injector/PSE, and a mode-A PoE-in switch, or a mode-A PoE splitter/PD plus compatible switch, across that two-pair section.

TP-Link says all its PoE switch PSE's are all mode A.  However, TL-RP108GE uses mode B, and only accepts power via PoE mode B.  That is, it always uses pins 4,5 and 7,8 for power.

If we are to believe TP-Link community pages and several other online sources, then TP-Link TL-PoE150S and TL-POE10R pair will do IEEE 802.3af over mode A, i.e. using pins 1,2 and 3,6, only, for both data and power.  Yet, the datasheet does not say for sure.  (If someone has these, testing them with a cable with 4,5,7,8 disconnected would be excellent.)

On the downstream/PD side, The Pi Hut in the UK has some 802.3af mode-A compatible splitters:
 - 105026: 5V 2.4A USB-C for £9.50
 - Adafruit 3238: 12V 1A for £10.10
 - Adafruit 3239: 12V 1A, 9V 1A, or 5V 2A, for £25.40
In addition to the splitter, you will need a switch that can be powered from the splitter output.  The voltage is the key. I do know D-Link DGS-105GL and Zyxel ZyXEL GS-105S v2 (gigabit 5-port switches under 20€ here) take 5V barrel jack at less than 1A (2.5W nominal, so 0.5A).

You can find similar ones at e.g. Amazon.co.uk, like DSLRKIT 802.3af PoE splitter, 5.5x2.1 barrel jack £9.50, (same with USB £8), which claims to support both modes A and B.  (DSLRKIT does make other PoE mode A stuff, too, but I haven't physically verified the devices, of course.)

You can get the above from anywhere else, of course, I just wanted to provide specific links so you can check.

Next, the supply/PSE side.

Unfortunately, the Tenda POE30G-AT is midspan/mode B only, and only provides power via pairs 4,5+ 7,8- so that won't work with your two-pair cable.

Devices that say they can supply IEEE 802.3af or at in mode A (1,2+ 3,6-):
 - TrendNet TPE-115GIv2.1 injector
 - TrendNet TPE-TG44g, 8-port switch
 - DSLRKIT PSE4P1UGK, a Chinese gigabit switch with four downstream PoE ports
 - If we believe TP-Link customer service, then e.g. TL-SF1005LP, a five-port 10/100 802.3af/at PoE switch should work

Since GbE uses all four pairs, shouldn't all GbE switches providing IEEE 802.3af/at support mode A?  There is even mode C, which uses all four pairs for both data and power. Well, UniFi ones use mode B only.



And finally, you can fix your current situation using not too many quid, if you don't mind hacking it together.

Pick up a fake PoE injector cable pair, like this £3 pair off the Pi Hut, and two Abracon ARJMxx-104 MagJacks.  Mouser sells the latter for around £4 apiece.

The idea is that we connect your Tenda POE30G-AT to the splitter's RJ45 with a standard Ethernet cable, all four pairs connected.  The splitter RJ45 should have 24.9k 1% resistance between 4/5 and 7/8, so that the Tenda will supply it with power.  Cut the wires around midway on the splitter, soldering the wires to the ARJMxx-104 MagJacks, power side to VC2+ and VC2-.  VC1+ and VC1- won't be connected, because your wall cable has no wires there.  You connect the MagJack to your wall socket with a short patch cable.

On the other side, another short patch cable from your wall socket to the second MagJack.

Here is a choice to be made.  You can reuse the injector to make this side a nonstandard injector (by connecting the MagJack data and VC2+ and VC2- to the injectors RJ45, by basically cutting the injector cable and soldering the its RJ45 side to the Magjack).  The PSE always sees the signature resistor, so this side would be supplying the 48V regardless of whether the connected device can handle that or not.  If you clearly mark the cable for the TL-RP108GE only, then there should be no problem, though.  Inside the wall, the two-pair cable will have 48V bias on its pair, so that needs to be marked so too.  Whether the direct voltage is okay for the TL-RP108GE, would be easy to check if you have one end of an Ethernet cable, and a lab voltage source capable of 48V at 100mA.  If you connect 48V to RJ45 pins 4,5 and 0V to RJ45 pins 7,8, the TL-RP108GE should just work.  If not, it may blow up, but see my previous message, that still holds.

Or, splitters and devices that support mode A ought to handle being supplied the 48V, so any mode-A splitter could be connected to the wall socket, and that then to a switch (that takes 48V in).

Or, you can take the splitter's unused RJ45 (the plain one), and solder that to the chip side of the MagJack.  The VC2+ and VC2- from the Magjack you solder to a male barrel jack cable.  You connect the plain RJ45 and barrel jack (soldered to the MagJack) to a proper 802.3af injector's inputs, so that it does the downstream PoE capability checking.  This active PoE 802.3af injector should be mode B, since that's what the TL-RP108GE wants, which should be common.

Or, you can take the splitter's unused RJ45 (the plain one), and solder that to the chip side of the MagJack. This is then a normal non-PoE RJ45.  Use a DC-DC converter that can take 30-55V DC in, and drop it to 5V (since that's what most small switches want), and power a small switch using that.



All this said, I would personally just replace the damn two-pair cable with a shielded one, using the existing cable to pull the new one in its stead.  Problem solved.
« Last Edit: December 03, 2022, 05:19:01 pm by Nominal Animal »
 
The following users thanked this post: peter-h

Offline peter-hTopic starter

  • Super Contributor
  • ***
  • Posts: 3671
  • Country: gb
  • Doing electronics since the 1960s...
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #41 on: December 03, 2022, 05:35:57 pm »
Fortunately, and I am sorry I did not make this clear in my diagram, only the cable section marked "only 2 pairs connected here" has only 2 pairs connected. All the rest has 4 pairs connected, so the cable fed by the Tenda injector should be OK.

I just need to get either a splitter or an 8 port switch compatible with the Tenda injector which I already have.

So I bought this: https://www.amazon.co.uk/TP-LINK-TL-PoE10R-TL-POE10R-PoE-Splitter/dp/B001PS4NWW and the 12V 2A (2.1mm DC coax power plug) should feed an 8 port Netgear switch I have (12V 1A power requirement, and I measured the socket to make sure it is 2.1mm).

I can see I could use an injector capable of going via 2 pairs only, and a compatible splitter, and power the injector from the L.H. UPS, but that introduces more variables because these are clearly less common, and the data sheets are ambiguous.

However I may have to revisit this one day because I have four cables with just 2 pairs connected elsewhere :) So then I will definitely need to do POE over just 2 pairs. Or rewire the sockets...

This has been a learning experience, not least because the product I am developing will get blown up by any POE injector which is compatible with that TP-Link switch ;) And same is true for many industrial products with an RJ45, not sold as POE compatible. It rapidly burns out the 75R resistors, but ETH may continue working... especially with short cables.

« Last Edit: December 03, 2022, 06:01:01 pm by peter-h »
Z80 Z180 Z280 Z8 S8 8031 8051 H8/300 H8/500 80x86 90S1200 32F417
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #42 on: December 03, 2022, 06:03:58 pm »
Wait.  The Tenda should be able to power the TL-RP108GE switch just fine, if you just use a normal Cat5e/6 all-four-pairs wired Ethernet cable.  Just use port 1-7 on the TL-RP108GE to connect to the Tenda, and not port 8.

If they don't play together, then one of them is definitely broken.  They absolutely should work fine with each other, given the datasheets they provide.  (Both explicitly describe using mode B pins only.)

And then, even a 2-pair only cable between the Linksys LGS116P to TL-RP108GE should work, because the Linksys doesn't power the TL-RP108GE, only the Tenda would.  That connection will be 100M max. due to just two pairs, but there is no reason for it to carry any current since you have the Tenda on an UPS on the other side.



That is, I do believe this should work (redrawn for clarity only, and because I can't count to eight):

  ADSL                               :
   │                                 :  Powered via UPS 1
  LinkSys LFS-116P                   :
   ┊                                 :
   ┊
  two-pair Ethernet cabling
   ┊
   ┊                                 :
  TP-Link TL-RP108GE                 :
   ║ │ │ │ │ │ │                     :
   ║ ? ? ? ? ? └─ ?                  :
   ║ 3 4 5 6 7    8 (PoE out)        :
   ║                                 :  Powered via UPS 2
   ║                                 :
  Tenda PoE30GAT                     :
   │                                 :
  PC                                 :

where indicates normal Ethernet cable, the wonky cable, and PoE over Ethernet being used.
« Last Edit: December 03, 2022, 06:50:40 pm by Nominal Animal »
 

Online Monkeh

  • Super Contributor
  • ***
  • Posts: 7990
  • Country: gb
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #43 on: December 03, 2022, 06:25:11 pm »
If they don't play together, then one of them is definitely broken.  They absolutely should work fine with each other, given the datasheets they provide.  (Both explicitly describe using mode B pins only.)

The injector is not going to supply power to a non-PoE load. The TL-RP108GE is explicitly not a standard PoE device.
 

Offline peter-hTopic starter

  • Super Contributor
  • ***
  • Posts: 3671
  • Country: gb
  • Doing electronics since the 1960s...
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #44 on: December 03, 2022, 06:32:01 pm »
Quote
The Tenda should be able to power the TL-RP108GE switch just fine, if you just use a normal Cat5e/6 all-four-pairs wired Ethernet cable.  Just use port 1-7 on the TL-RP108GE to connect to the Tenda, and not port 8.
If they don't play together, then one of them is definitely broken.  They absolutely should work fine with each other, given the datasheets they provide.  (Both explicitly describe using mode B pins only.)

I tried the Tenda to the TL-RP108GE via a ~10m cable and the TL is dead. The Tenda PSE LED flashes 5 times, then a gap, then repeats.

I also tried a Linksys POE switch to the TL-RP108GE via a ~1m cable and the TL is dead.

I am pretty sure the TL-RP108GE doesn't negotiate POE, so no standard POE source will feed it. It takes some finding but here
https://www.lambda-tek.com/TP-Link-TL-RP108GE~sh/B44812266&viewSpec=y#product-view
it states
"Power over Ethernet (PoE) type supported:   Passive PoE"

So only the cheap noncompliant chinese POE injectors will feed the TL-RP108GE.

Z80 Z180 Z280 Z8 S8 8031 8051 H8/300 H8/500 80x86 90S1200 32F417
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #45 on: December 03, 2022, 06:44:52 pm »
The injector is not going to supply power to a non-PoE load. The TL-RP108GE is explicitly not a standard PoE device.
That is horrible, and completely unexpected for me.  So, this would be required:

  ADSL                               :
   │                                 :  Powered via UPS 1
  LinkSys LFS-116P                   :
   ┊                                 :
   ┊
  two-pair Ethernet cabling
   ┊
   ┊2                                :
  TP-Link TL-RP108GE                 :
   ║1│ │ │ │ │ │                     :
   ║ ? ? ? ? ? └─ ?                  :
   ║ 3 4 5 6 7    8 (48V out)        :
   ║                                 :
  Passive injector                   :  Powered via UPS 2
   │      power ↑                    :
  Active splitter                    :
   ║                                 :
   ║                                 :
  Tenda PoE30GAT                     :
   │                                 :
  PC                                 :
which means an active splitter whose output is either 24V or 48V is needed (because that's what TL-RP108GE expects).

Why the hell would anyone make the Ethernet inputs non-compliant, valid only for dumb voltage injectors?
The only reason I can think of they saved a few cents by using non-PoE-capable magnetics.  Or maybe no magnetics at all.  The price is way, way higher than the quality of the device!

So, instead, peter-h is going for

  ADSL                               :
   │                                 :  Powered via UPS 1
  LinkSys LFS-116P                   :
   ┊                                 :
   ┊
  two-pair Ethernet cabling
   ┊
   ┊                                 :
   ┊ ? ? ? ? ? ?                     :
   ┊ 3 4 5 6 7 8                     :
   ┊ │ │ │ │ │ │                     :
  Netgear switch  non-PoE            :
   │ power ↑                         :  Powered via UPS 2
  TL-PoE10R                          :
   ║                                 :
  Tenda PoE30GAT                     :
   │                                 :
  PC                                 :

which should work, assuming TL-PoE10R is not a fake too.
(And instead of Netgear, any 8+ port GbE switch with 5V/2A, 9V/1A or 12V/1A DC input should work; or even 10/100, if there isn't much internal traffic.)
« Last Edit: December 03, 2022, 07:06:44 pm by Nominal Animal »
 

Offline peter-hTopic starter

  • Super Contributor
  • ***
  • Posts: 3671
  • Country: gb
  • Doing electronics since the 1960s...
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #46 on: December 03, 2022, 07:21:36 pm »
Yes, exactly.

No need for a gigabit switch there because the first part of the cable has only 2 pairs wired :)

Quote
Why the hell would anyone make the Ethernet inputs non-compliant, valid only for dumb voltage injectors?

I have never designed a POE powered product but it is obviously a lot cheaper to do it passively i.e. with diodes, or perhaps "diodes" implemented with MOSFETs so the RJ45 draws power only when >=24V is presented.

I looked at POE for my product a while ago. I already have a fully isolated power supply in there and would need to add at least an isolated DC-DC converter. And that is for passive (noncompliant) POE which is hardly the way to go. Example:
https://www.analog.com/en/product-category/poe-powered-device.html
Look up LT4295. That's a sh*it load of parts, to get POE power input with isolation. In my case since I already have an isolated PSU so could feed the input of that directly but not from 48V (it does 12/24V). And after this little exercise I know a lot more about it :)

And now implement this across seven ports of a switch. It's a lot of expensive parts. Hence "real" POE powered switches seem to accept POE on just one port.

Hence POE powered products tend to exist at the higher end of the price range. I have a load at work: Draytek wifi APs, Snom phones at 150 quid, etc.
« Last Edit: December 03, 2022, 08:08:04 pm by peter-h »
Z80 Z180 Z280 Z8 S8 8031 8051 H8/300 H8/500 80x86 90S1200 32F417
 

Offline MarginallyStable

  • Regular Contributor
  • *
  • Posts: 65
  • Country: us
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #47 on: December 03, 2022, 09:15:27 pm »
Quote
Can anyone recommend something which has at least one active POE IN port?

Already did. Netgear GS108Tv3
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #48 on: December 03, 2022, 09:20:50 pm »
LTC4295 is IEEE 802.3af/at/bt, up to 70W, so it is a bit unfair comparison for something as low-power as a switch...

A switch doesn't need much power – 8-port gigabit ones less 5W, many rated at 2.5W max; TL-RP108GE at max. 4W if not supplying power – and something like LTC4267 is about 4€ apiece at Mouser in sets of 100.  With a RJ45 female with full magnetics and PoE support (so Gigabit MagJack with 4PPoE), you can do a PD with 10/100/1000 and IEEE 802.3af mode-ABC. The one shown in the datasheet uses PA1133 for power isolation.  But, the 4PPoE MagJacks are quite expensive, with four-port ones 20€ even in larger sets at Mouser, so looks like costs as much as the rest of the parts put together.

Oooh, now I understand.  TL-RP108GE is a variant of TL-SG108E (v2 and later).  Amazon.co.uk sells TL-SG108E for the same price, £35-£39, as others sell TL-RP108GE.  The latter is obviously intended for some kind of NAS or such from multiple cameras or whatever powered from different 24V or 48V injectors, so that if any of them are powered, that switch will be powered too, and provide 5V or 12V to some kind of central monitoring or alarm device.
For normal users, it is the  TL-SG108E (depending on version, v2 and later are managed!) that people want instead of that one.  At least v5 and v6 take 5V max. 0.6A in, so something like the TL-POE10R in front of it makes a POE-powered switch.

So, yeah.  If you had gotten the TL-SG108E with TL-POE10R instead of the TL-RP108GE, you'd not lost money, and didn't have to use the Netgear switch.
Are you sure you can't swap the TL-RP108GE for TL-SG108E anymore?
 

Offline peter-hTopic starter

  • Super Contributor
  • ***
  • Posts: 3671
  • Country: gb
  • Doing electronics since the 1960s...
Re: 5 port gigabit switch powered over POE via one of the cables
« Reply #49 on: December 03, 2022, 10:31:19 pm »
Yes; all true. Can be done for less.

No; it came from Amazon and there is no "support" or "customer service". You send the box back.

Quote
Already did. Netgear GS108Tv3

Sorry; missed that. Good price too. Yes that looks like it would have worked also. A rare box. Not a lot of people stock it, and finding it in a search is almost impossible. I always look at Netgear first...
Z80 Z180 Z280 Z8 S8 8031 8051 H8/300 H8/500 80x86 90S1200 32F417
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf