Hi all,

we did a test with RB750UP with ubiquiti and MT equiment. Power was 28V.

Result:

1. power detection with ubiquiti works well (4 *NSM5)
2. output voltage (28V input to simulate battery backup) is low enough to stay the NSM alive.
3. powering of one RB411U without problems
3. powering with Rb411U and one NSM5 will power down all ports
It seems that the starting current of the RB411U is too high and RB750UP thinks that is a shortage


@janisk : Is the short detection of the RB750 a hardware feature or is it software controlled? If it is software controlled please switch off only the port with the "detected" short.

short detection has to work very fast or device will get damaged. Fastest possible way was to cut power for all ports. If not mistaken time was 2 to 4 ms. If cut-off time was slower from time to time device got burned.

That behaviour i weird. What card you have in the RB411U? And where there only 2 devices connected with one of them being RB411U. Is it possible to measure output voltage and power usage of the connected boards?

How can ns5m work at 28v? 24v is maximum admited and will burn the protection diode.

There is no card in the RB411.
If we have a look at the power consumption we will see:

1. Without any device: 30mA
2. With NSM5: 180mA
3. With NSM+RB411: 320 mA max

There was no viewable max current above 320mA. The RB750 pumps the devices up serveral times. After a while an with luck everything stays on. Maybe a capacitor issue. 4 NSM5 without RB411 is working well.

We are not happy about the complete power cutoff on all ports because one defective unit will take whole Micro-Pop down. Cheap netgear GS110TP will detect overload an takes the power off from this port. Maybe we can put MTOS on it


There must be a better way to solve the problem. If I have a look on the cost, I would accept if the port will be dead an replace the whole unit.

Is it a hardware power-off or can you switch off the overloaded port by firmware?

We waited a long time for the RB750UP. I want to give it a chance but with the behaviour right now I see more problems coming through us than benefit.

How can ns5m work at 28v? 24v is maximum admited and will burn the protection diode.

The voltage regulator on the NSM5 is specified to 27V. If there is any Z-Diode for protection this one will not burn out because you lost about 3V with the RB750 build in esd protection. 28V is the worst case test to simulate 24V (2 Blocks) backup voltage. If 28V input to RB750 will run several hours we could be save that the estimated max voltage of 26.5V (fully loaded new batteries) will not blow the NSM.

First we wanted to run the system with 12V but with 500mA per RB750-port you cant run the 7-8W max power of the NSM.

nice to know, I too must use 28v because of the batteries.

nice to know, I too must use 28v because of the batteries.

This was our intention to see. This works.

The lab setup is running fine with 4 NSM since 2 days. No burned out NSM5. RB750UP is not getting hot.

If we could solve the power-of-on-all-ports-if-one-fails issue this is the solution for micro pops.

well. if port is ON, and short happens, then all connected devices are powered off. After that, port scan is done and resistance measured. Resistance has to be quite high (around those 19k to 26.5k Ohms) for port to be turned on again. If port is shorted, then resistance is close to or 0 Ohms, so it will not be turned on again due to that.

It would be unacceptable if one shorted device could burn then RB750UP and all of the devices would stay OFF. Because now, there is additional reboot of the connected devices once, afterwards, they should come back on and operate normally with shorted device being off.

And once again - short detection only works on port that is turned on (AUTO or ON setting). If port is off, its resistance is continuously measured and turned on only within set parameters of the setting used.

edit you should check the RB, i have RB411U working properly with high power card (powered by RB750UP)

If it would be as you described we would see only one single reboot of all devices. What I see here is not as you described. I see (with an second RB too) that the devices are powered off several times up to that point where all devices are running. Maybe it will work as you said when there is a real short on the the cable or in the device.

Please explain how we should work with this in the field:

1. One device can take out all others including the uplink-feed for the station
2. There is no log message that shows there is a "shortage"-event
3. If luck will leave us there is water in the cable or condensing water at a plug... Happy random booting of all any time?

Why is power is switched on at RB boot and then switched off again. Is this the init of the power control device?

But I will ask you now the third time: Is the power-cut a hardware controlled feature?

Is it possible to get a log event in case of failure or is it a unidirectional communication with the power control device?

I had a look on the circuit and I think that you will not able to solve the issue.

With I2C bus the I/O and ADC of the ATtiny461 is more than enough to do a propper per port current measurement and a single POE shutdown per port. Did you made a design failure?

1. One device can take out all others including the uplink-feed for the station

yes, because in case of short time it took to turn separate port of was too link and something parts can be damaged. I am not saying that will be damage as there where both cases - device got burned and stopped working or port was powered down and nothing happened.

2. There is no log message that shows there is a "shortage"-event

we are working to solve this. If log messages will appear that will solve the power down when RouterOS boots up on the device. As there are no way to monitor if power is on the port or not. Will add feature request that events are logged. If it is going to be possible, it will be added as it is useful to know.
mount and maintain your devices properly. But short detection is turned on only when power is given on the port, so if short is plugged into the port and it is set on AUTO or ON, it will still attempt to detect if power can be turned on, and if not, it will stay in detection phase, if for some reason resistance is within the correct parameters, power will be given on the port and short and overload detection turned on.
yes it is, that is whey when port overload is detected (by measuring load on the port) single port is turned off. since the tolerances allow to turn the port off in time that board will not be damaged. And that is not the case for short-circuit.

I had a look on the circuit and I think that you will not able to solve the issue.

what issue exactly (you named several).

Janisk, I don't want to fight against you or MT. I wan't to have a good solution to stay in the market. Maybe we talked over different things.

To clarify the situation:

We will use RB750UP to build a small base station to setup a micro-cell-network. We need many base stations (about 50-70)) because customers demand for bandwidth is high (today 16.000 Download). To get a competitive price we have to include voice service.


The 4 POE-Ports (they are set to 'auto') will take ubiquiti (sorry) NSM5 or RocketM5 and one uplink. We want to use a RB411 (or similar) for uplink.

My main problem is: As I described before the RB411 tooks all ports out but there is no shortage on the board. If it is powered alone it works well. If one NSM5 is powered and you will switch on the RB411 mostly any time the other ports are switched off several time. After some cycles of repowering finally all ports are running with power.

For providing only internet service I have no problem with this cause the case will be rare. Normally nobody of the targeted private customers will call us if the internet access will be interrupped for 1 min in some rarely situations.

But: With voip on the line they will call us.

I am sure that you will implement the logging of POE events and I am sure that you will indicate overloads in the interface sections.

Finally: There must be a problem in the short detection. That is the thing.

ok talked to devs, logging is not possible in near future. Maybe some solution in the future, but that is not going to be fast if at all.

If device power part has large capacitors, then they can be mistakenly detected as short. I have stock RB411U and it is not detected as a short when board is connected along with other devices. Capacitance higher than 200uF could cause that. of course, eventually when capacitor is charged up power is given over the port.

Short detection is made as fast as possible with small delay so capacitors with capacitance of up to 200uF could charge up and would not cause false positive. Short is detected differently than overload due to that, because, that delay in case of real short will hurt the board. So there is a bllance, between how fast and what devices can be turned on.

We are working on problem that all poe-out on the ports is reset when RouterOS boots, doe to this, devices faster than RB750UP can start up and start functioning and the be power-cycled again.

About voltage measurement, there was a bug and it will be fixed. measurement will show voltage on the jack.

If you try to detect a short the short will have an other signatur like the pump-up of an capacitor. If you look into an interval (you called it delay) you will see the current decreasing by pumping up. If you got an short the current will nearly be constant. By the way: Nomally there is no need to put an power supply with more than 2A at 24V. I don't believe that a 4 times overcurrent of an single port for 100ms will destroy the board. The board must handle the 2A in normal operating state. The only thing that could be damage is the switch for powering up the port. But not in a sub second time.

If you put an power supply > 2A at the board you will be right.

they are shipping with 2.5A @24V PSUs to handle 4x500mA and board itself. Also take into account that RB750UP is in the same form factor as RB750 was previously.

Short detection is made as fast as possible with small delay so capacitors with capacitance of up to 200uF could charge up and would not cause false positive. Short is detected differently than overload due to that, because, that delay in case of real short will hurt the board. So there is a bllance, between how fast and what devices can be turned on.

Overload detection also seems to be triggered by a small rise in supply voltage (when AC power returns and battery starts being charged), which should be completely harmless. Not all devices are turned off then, only some of them - so it's probably not detected as a short, just overload. But it's still bad. I can reproduce it reliably with RB750UP powered from a CyberPower CS24U12V-12 supply and powering UBNT radios. It seems a bit too fast. You may need different current thresholds and delays, to get close to a real fuse (I^2)*t characteristic (blows quickly with large overload, slowly with small overload). Not sure if the hardware is capable of that - do you use the ADC, or analog comparator (faster, but only one threshold), or both? I'll try to see if I can slow down the voltage rise by adding large capacitance on the power supply output.

use PSU that was supplied with the unit. You will have better results if you use 24V

I am not sure if they can program the atmel onboard with new ROS version. You will not able to power ubnt with 500mA with 12V without problems

if by they you mean us, and if we can - yes we can.

you check how much your device needs and check if that is within those 500mA. with high power cards and 2 chains - i seriously doubt that.

Janisk, you sounds like Obama "Yes we can"

Ubiquiti is specified with 8W max. (typ. 6-7W in production) Impossible with 500mA at 12V!

use PSU that was supplied with the unit. You will have better results if you use 24V

Unfortunately, the supplied PSU doesn't have battery backup which I need.

I use UBNT radios at low power (staying within legal limits), and three of them (1*NanoBridgeM5, 2*Bullet5 - each with ~15m of good quality copper FTP cable) have been running for over 9 hours from RB750UP powered by a 12V 12Ah battery after a long AC power failure - not bad. They were cut off when the RB750UP voltage monitor reading was about 9.4V (adding the known measurement error, real battery voltage could be a bit over 10V), the RB750UP itself continued to run though (it still has 3 weeks uptime after power finally came back). So the radios draw about 400mA each (calculated based on battery capacity), and the overload protection protects the battery from too deep discharge (current rises at low voltage), which is a good thing. I'd use 24V with longer cables though.

The overload and reboot of radios when AC power comes back earlier seems to be caused by the capacitive load, adding a few milliseconds delay to the overload protection should fix it without burning anything.

BTW, a feature request: please add configurable limits for the voltage monitor, so that PoE is turned off below VL, and only attempted to turn on again if it rises above VH (some hysteresis is needed to avoid cycling). Ideally, cut off the radios one at a time with some delay - the least important one first (say, port 2) as reduced load may allow the other radios to still run for some time (port 5 is the backhaul, turned on first and turned off last).

well, that delay is already there.

consider that there is voltage drop on protection chains and resulting voltage can be lower than you actually expect. As result overload protection can happen faster. For example, 12V to 10V is 17% drop that results in 17% current increase.

So best course would be to go for 24V batteries or connect 2 12V in series. Giving more headroom will result in more stability.

consider that there is voltage drop on protection chains and resulting voltage can be lower than you actually expect. As result overload protection can happen faster. For example, 12V to 10V is 17% drop that results in 17% current increase.

Why so much voltage drop on the protection?

OK, yes it is a bit out of spec, but 9 hours for RB750UP + 3 radios on 12V 12Ah battery is still not bad. It's not the low voltage itself, is the voltage transient (step rise from 12V to 14V when AC power comes back, even after only a few minutes of power failure) that triggers the overload protection.

what you can try - unplug just charger from mains and plug it back. Check what is happening with devices powered by 750UP. It could be that after power failure a lot of devices are turning on and as result you get noisy AC, that in turns generates peaks on the PSU output.

Also, you can put some induction coil coupled with large capacitor on RB power input to filter out possible peaks and lows from your power circuit.