Hi,

Recently added a secondary Power supply to the CCR (Confirmed with Mikrotik staff first it was safe in another thread) on its secondary internal power connector port.

My secondary PSU is a 19.5V 5A power supply. I have tested the voltage with a multimeter both with and without load and confirmed it remains at this voltage.

Few observations :-
1)The two sets of power supply pins don't appear to be isolated. There is voltage leakage between the devices. I.e If I plug in both power supplies at once the Mikrotik correctly uses the internal power supply at 24V. If I disconnect the mains connection to this supply the green LED remains illuminated indefinitely. Multimeter on the pins confirms presence of 19.5V - from the secondary supply...

Likewise when the internal power supply is connected 24V is output onto the same line as the 19.5V power supply (resulting in the LED on the external PSU glowing much brighter). Since this was how I observed it In the first place I'm now not risking running with the two supplies that differ in voltage by this range.

2) The fan speed changes based on the voltage of the power supply, it noticeably decreases when on the 19.5V supply. Additionally on 19.5V there appears to be issues with switching between the main and auxiliary fan - when done manually the fans appear to freeze. This seems odd as surely the voltage is regulated to a lower voltage from the higher voltage incoming supply?

3) The fail over between the two power supply plugs does work fine while the unit is on with no observable disruption.

Anyone got any ideas? I'm thinking it will need at minimum some diodes to prevent the voltage flowing back into the other supply. Presumably this raises some questions about its ability to withstand a power supply short and survive although I don't feel like sacrificing a CCR to find out the answer to that one...

Thanks

Mike

Just done some more tests.

With the 24V internal supply the other connector floats at around 3V.
If the 19.5V secondary PSU is connected, both terminals measure their correct voltage.
If you then disconnect power to the internal PSU and measure the voltage on the PSU connector - it reads 19.5V. I.e. reverse voltage/current into the internal PSU. The internal PSU LED remains lit.
If you then re-apply power to the internal PSU the secondary PSU connector receives 24V.

So the logic when both power supplies are initially connected is fine, but upon failover it allows voltage & current to reverse flow into the failed power supply. This would be especially nasty to the 19.5V psu as it will given 24V in reverse.

Would be interesting to know if this can be fixed with a firmware update or if its a failing with regards to the PCB design.

Anyone got any ideas?

Thinking this would be the only safe option now :- http://www.mini-box.com/Y-PWR-Hot-Swap- ... Controller

There is enough room for it in the case and it fully blocks each channel.

Just reading the tech info on http://www.mini-box.com/Y-PWR-Hot-Swap- ... Controller

MOSFETS are rated at 3.9mOhm. At 10amps they produce 40mwatts each. At max load, forced air ventilation is required. For fan less operation de-rate the outputs by ~40%.
Peak load should not exceed 60 seconds

Additional inline fuses maybe required as 60seconds is not a lot of time before component failure and pcb track burnup?

Hi n21roadie.

I *think* it appears safe. They are rated for 10A under normal usage (or 6A de-rated if not actively cooled) and shouldn't have more than about 1A-2A during idle going to around 3A at 100% utilisation (at least on the 16 core CCR I'm using). This is based on the CCR idling at around 25W and going to 60W under load.

Looks like that is going to be the only option for safe fully blocking redundant power to the units.

Hi n21roadie.

I *think* it appears safe. They are rated for 10A under normal usage (or 6A de-rated if not actively cooled) and shouldn't have more than about 1A-2A during idle going to around 3A at 100% utilisation (at least on the 16 core CCR I'm using). This is based on the CCR idling at around 25W and going to 60W under load.

Looks like that is going to be the only option for safe fully blocking redundant power to the units.

A word of caution, the primary of a power supply may have fuse protection but check that the output of the power supply also has fuse protection
I would use 2 X 40 AMP 600 V diodes + 2 X inline fuses, these are called steering diodes and prevent one power supply effecting the other under fault conditions, the module looks like it cannot handle large current draw in fault conditions , the small pcb tracks do burn up because they cannot handle the current draw?
The nightmare scenario is when the load side in this case CCR shorts out then both power supplies are pulled down?
Also read http://forum.mikrotik.com/viewtopic.php ... de#p261791

Great advice - Thank you I will add them in.

Can you advice a part number for the 600V diodes if you have used similar before?

Presumably there will be a small voltage drop over the diode too?

Great advice - Thank you I will add them in.

Can you advice a part number for the 600V diodes if you have used similar before?

Presumably there will be a small voltage drop over the diode too?

I would use this type of diode, you could pick any other diode provided it's rating is above max current draw in fault conditions, this one just happens to available?
http://www.ebay.ie/itm/WIND-TURBINE-SOL ... 3a60e3f482

And then use this type of 20mm fuse holder
http://www.ebay.ie/itm/20mm-PCB-Mount-F ... 53f96008ee
and solder the diode to one side of the fuse holder, then you will have a fused the dc supply from power supply + steering diode to prevent one power supply fault effecting the other?

Expect about 0.6volt drop across the diode ?

I would equal DC voltage on both power supplies then you will have shared current draw from each supply, not that it makes much difference as the normal current demand will be very low for the CCR?

Also setup and test on a old routerboard first using the 2.1mm DC input jack and when all is ok transfer to CCR.

@disca
Forgot to ask but is it only just one device (CCR) you will be powering with dc supply or do you have other devices also requiring dc supply at this location?

n21roadie - Presumably I could get away with slightly smaller blocking diodes if matched with a suitably small fuse.
I.e +-10A 60V blocking diode with a standard 12/24V automotive in-line fuse holder at 5A.

Yes just one device for now - have toyed around with having a single high Amperage supply for multiple Mikrotik devices but the thought of it going wrong scares me too much so I'll keep it seperate

EDIT: Just done some more reading - looks like its recommended to use diodes good for 3x the current usage expected to keep heat down. Also Schottky diodes appear to have a lower forward voltage drop. So based on maximum power usage of 60W at +- 20V that would be 3A. Multiplied by 3 to give a good margin for low heat would seem to indicate 10A would be ok.

n21roadie - Presumably I could get away with slightly smaller blocking diodes if matched with a suitably small fuse.
I.e +-10A 60V blocking diode with a standard 12/24V automotive in-line fuse holder at 5A.

Yes just one device for now - have toyed around with having a single high Amperage supply for multiple Mikrotik devices but the thought of it going wrong scares me too much so I'll keep it seperate

EDIT: Just done some more reading - looks like its recommended to use diodes good for 3x the current usage expected to keep heat down. Also Schottky diodes appear to have a lower forward voltage drop. So based on maximum power usage of 60W at +- 20V that would be 3A. Multiplied by 3 to give a good margin for low heat would seem to indicate 10A would be ok.

http://forum.mikrotik.com/viewtopic.php ... 51#p376051
The update on this post you may find interesting for redundancy power supply design?