As I promised, I'm back with the results. I have tested the RB260GSP input/output characteristics in a lab.
Test environment:
- Power supply is a precision laboratory power supply that is able to output 24V 20A. It also has adjustable overcurrent protection and adjustable output voltage.
- I have added a variable resistor in series with the power supply, and then it went into the DC input jack of the RB260GSP switch. This resistor was simulating a longer ethernet cable on the POE-in line.
- The test load on the POE-out side of the switch was a HAP-AC2 router.
Here are the results:
- If the resistor is very small (<0.5 Ohm), then everything works fine
- When the resistor reaches about 2Ohms, then the switch will detect "short circuit" and shut down its POE output.
- This behaviour seems to be independent of the static voltage level. Does not matter if you use 12V input or 24V input or even 30V input. I have tried all of them.
- It is clearly not a power supply problem - the lab power supply produced exactly the adjusted voltage, and it could have given 10 times the required power
- I have also tired to add a 22000uF capacitor in paralel to the power input jack of the switch. This was helping a bit - then I could increase the resistance to about 3Ohms before it started to shut down its outputs.
- When I circumvented the whole thing with POE injectors and splitters, then I could power up both devices. In fact, I could add another load resistor and I could easily power another 20W of load. I have also tried to turn off/on this load resistor periodically to see how it affects the switch and the router. Not suprisingly, the voltage was fluctuating, but the devices had no problem with that.
The conclusion is that RB260GSP measures the derivative of the power supply voltage. If it goes below a certain negative value, then it shuts down its power output. I strongly believe that the short circuit detection is faulty. It only works if the input power supply has very low internal resistance. When you are using POE-in to power the device, then the power supply will always have a higher internal resistance. (Well not always, but in all cases when POE input is needed anyway...)
It is very natural to have some voltage drop on longer POE cable. In fact, this should be expected! The short circut detection should be designed in a way that allows to use power supplies with higher internal resistance. POE works that way, by design. I understand that this would require current measurement on the POE-out ports, instead of voltage drop measurement over time. Such protection circuit would be more difficult to implement, and it would cost more.
Anyway, RB260GSP can only be cascaded only if its power supply (including the POE-in line) has a very low resistance (<0.5Ohm), e.g. connected with a very short ethernet cable. This is a fact. The use cases are very narrow. When you need to use POE to power a device, then almost always you will have a longer ethernet cable and that precludes cascading of these devices.
The rumours about using a different circuit for its DC input jack and POE-in jack are untrue. I have measured that too. There is no apparent difference between the two power inputs. The root of this myth is probably the way we are using DC input jacks - we tend to plug in a local power supply with a short cable, and that has a low internal resistance. So it would seem that the DC input jack behaves differently, but it is not.
The second sentence on the main product page (
https://mikrotik.com/product/RB260GSP ):
> Ports 2-5 can power other PoE capable devices with the same voltage as applied to the unit (passive PoE over datalines only).
The specification also says that the unit can be powered from POE-in. What it does not say is that you cannot do both. Technically you can, but only in very special, rare cases.
I think that this should be documented! In my concrete case, I have bought this device because - according to the specification - I was confident that I could cascade two of these devices. But in practice, I had to create this abnomination:
https://imgur.com/a/hxm0Xdx
Using the cables above, I can power 3-4 devices without problems. The power supply is fine. The power lines are fine. The power consumers are also fine and they are stable. The problem is clearly with the switch. In fact, I'm not using the POE part of the switch at all, simply because it does not work as it should.
I think that MikroTik should at least add a note about this in the product specification. I think that I'm not the only one who had problems with this. I really think that adding such note would be benefical for both MikroTik and for the customers.