Hi,
I am beginning this thread to share my project to upgrade the CRS226-24G-2S+RM to a semi-datacenter class switch by adding two internal power supplies. Any questions and comments are very welcome.

I posted my first musings here:
http://forum.mikrotik.com/viewtopic.php ... 55#p488755

An interesting thread about internal power plugs (albeit on the CCR, things look pretty similar in the CRS):
http://forum.mikrotik.com/viewtopic.php?t=69422

Like the OP in the latter thread, I failed to locate the exact same power connector used in the CCR. There are no manufacturer marks; it is a 3-position connector of which the outer 2 are used. I have spent some time with Farnell's usually efficient parametric search but unfortunately that particular connector didn't show up. Of course I could choose any connector with the same pitch, or even solder a pair of leads directly to the PCB. But I'd rather go with the same connectors already used in other devices, and for the small cost of connectors I think I'd make it neater.

I'll use a pair of Mean Well RS-25-24 power supplies which look well built and are relatively cheap and easy to obtain, even from eBay. They can be solidly but easily fixed using just two M3 screws from below.

Finally, two IEC 320 panel connectors (one of which will need some cutting to the case) will provide the AC input.

I haven't measured the heat output of the power supplies, but to be safe I'll install a fan. Which leads me to the air flow chapter.
There are cuttings in the case for the fan, a little to the side where there is space for the power supplies, i.e. ideally located, so no problem physically. Our cold air comes on the side of the network connectors, so some cuttings will be needed to that side of the case in order to be able to obtain some cold air. Above the 1G connectors the metal case has openings, which are covered with the plastic sheet that has the names of the connectors, logo, etc printed on it. It will be simple to open them up with a small cutter or similar tool. Perhaps it would be advisable to open a few holes on the right side of the front panel so as to guarantee some flow over the power supplies; I will come back to this when I have things in place.

The board has what seems two connectors for fans. One has a label that reads "5V fan" and the other "Fan Uin". I haven't tested them yet, and wonder if both are indeed intended to connect fans and if there is any reason to prefer one over the other. I guess some probing with a multimeter (unfortunately I don't have any at hand now) ought to clear this.

I have used the following parts:

• - Two IEC 320 C14 power input connectors, from the dust bin
  - Two Mean Well RS-25-24 power supplies from eBay, total €39.40 with tax and shipping from it-tronics.de
  - Molex connectors... Didn't fit them this time. I wanted to test the hack reasonably quickly; next iteration I will aim for this among other things.
  - Four conical head M3x3mm screws (I think SSD screws would do, haven't tried)

As for tools, as you would expect:

• - A soldering iron (actually I have a StarTec soldering station with temperature control, nothing fancy but reliable)
  - Vertical drill (but you can easily do this job with a hand drill as well)
  - Phillips #1 and a #2 screwdriver
  - Mini rotary tool (Proxxon in this case) with a cutting disc

Initial testing is nominal. A single or both working power supplies, and the switch is alive. Any of the two fails and the switch is alive.

Probing with the multimeter on the fan leads doesn't show much interesting. The 5V fan lead shows no power, and the other fan shows 24V. Unfortunately all the fans I have around are 12V, so I was unable to leave one working.

The good news about heat is that there is not much of it. I opened the holes in the front above the rows of RJ-45 connectors and calculate that they add up to around 400 mm^2, which I think would be enough for the fan to pull, and will not harm some natural air flow without it.

@Mikrotik it would be awesome if you could make this easier, I mean making it easy for people to add dual internal power supplies without adding cost on your side if you don't want to. Some things you could do that would be great help:

• - Punching for the second IEC power connector
  - Leaving open the ventilation holes on top of the RJ-45 connectors, for which the case is already punched and need only not covering them with the faceplate.
  - Holes for fixing the power supplies
  - Some tab or something to close the 24V power input
  - Some way to use a 12V fan - is there anywhere on the board a 12V line that can be used?
  - Publish the part numbers for the internal power connectors (for neatness) - is my guess correct?

Of interest to anyone wanting to do the same with the same power supplies, the four holes for fixing them are 29 and 81mm from the side (measured from outside) and 15 and 70 mm from the front (again measured from outside).

Here are some photos:

I have verified that the CCR power connectors are Molex 10-63-4027 connectors (or copies).
I have ordered a few of these and soldered them to the CRS motherboard of the second of 3 CRS devices being fit with dual redundant power supplies. They fit perfectly. Note that relative to the nearest edge of the board, they are reversed with respect to the CCRs.

Nice work!!

Thanks @bajodel.

Some more things. Here is a photo of the Molex 22-27-2031 that I fitted to the +5V fan holes. I have added a Sunon MB40200V1-000U-A99 fan, which unfortunately doesn't have sensor output (it just happened to be available on RS when I ordered all the stuff). I don't even know if the board has the required components to read back the fan speed.

Speaking of connectors, this time I went for the luxury in power input and ordered two Schurter 6100.3300 IEC power input connectors which fit nicely (with M3 inox screws and bolts) and even have an option to clip the cable to prevent accidental disconnection, which I didn't buy. They cost 75 cents apiece, a princely sum due to the low volume purchased (just 4 units). The first connector in the case fits in the provided holes; for the second one, for which the case is not prepared, I re-used two of the honeycomb holes which happen to be almost exactly the required 40 mm apart between centers and opened a hole with a rotary tool as before.

I am only awaiting the crimp terminal tool to finish this one. For the first half of my life I used ordinary pliers for crimping and I have grown tired of that game; a white haired lady once told me that the older we grow, the more comfort creatures we need.

Meanwhile, the first modified unit has been running for two days straight (had to reboot to upgrade software) without any issues so far, current temperature 42 degrees Celsius.

Cheers,

Thanks @bajodel.

Some more things. Here is a photo of the Molex 22-27-2031 that I fitted to the +5V fan holes. I have added a Sunon MB40200V1-000U-A99 fan, which unfortunately doesn't have sensor output (it just happened to be available on RS when I ordered all the stuff). I don't even know if the board has the required components to read back the fan speed.

Speaking of connectors, this time I went for the luxury in power input and ordered two Schurter 6100.3300 IEC power input connectors which fit nicely (with M3 inox screws and bolts) and even have an option to clip the cable to prevent accidental disconnection, which I didn't buy. They cost 75 cents apiece, a princely sum due to the low volume purchased (just 4 units). The first connector in the case fits in the provided holes; for the second one, for which the case is not prepared, I re-used two of the honeycomb holes which happen to be almost exactly the required 40 mm apart between centers and opened a hole with a rotary tool as before.

I am only awaiting the crimp terminal tool to finish this one. For the first half of my life I used ordinary pliers for crimping and I have grown tired of that game; a white haired lady once told me that the older we grow, the more comfort creatures we need.

Meanwhile, the first modified unit has been running for two days straight (had to reboot to upgrade software) without any issues so far, current temperature 42 degrees Celsius.

Cheers,

Good work I also thought about PSU failover however I would never use two of the same make/model of SMPS as one issue which effect one is sure to effect the other (not talking about PSU failure)
For example at remote sites AC voltage can contain voltage spikes and noise etc. SMPS do not really filter this out completely which can result in having to reboot routers
So I asked myself what is the cleanest DC supply available - answer battery supply, so I setup a large battery bank at the sites and used a float charger to keep the batteries charged, simple but very effective and has given me several years of trouble free PSU operation at remote sites,
Last year at one site, lighting in the area blew out the AC side of the float charger but the equipment just kept working as it was on batteries and continued until I swapped out the float charger
Just my opinion on the most stable power supply to routers?

Good work I also thought about PSU failover however I would never use two of the same make/model of SMPS as one issue which effect one is sure to effect the other (not talking about PSU failure)
For example at remote sites AC voltage can contain voltage spikes and noise etc. SMPS do not really filter this out completely which can result in having to reboot routers
So I asked myself what is the cleanest DC supply available - answer battery supply, so I setup a large battery bank at the sites and used a float charger to keep the batteries charged, simple but very effective and has given me several years of trouble free PSU operation at remote sites,
Last year at one site, lighting in the area blew out the AC side of the float charger but the equipment just kept working as it was on batteries and continued until I swapped out the float charger
Just my opinion on the most stable power supply to routers?

Good advice. My modified switches are intended to be used in datacenters where the electric supply is exquisite, hence the fact that I was only looking for reasonable robustness and efficiency.

For remote locations, it is perhaps worth the time looking at UPS vendors catalogues, as many sell line conditioning units that filter out those problems but without the UPS, thus with less utility but also much less weight, volume and cost. They also sell simple and cheap spike discharge units intended to protect equipment from the effects of the most severe spikes, i.e. lightning, by discharging spikes to earth.

Looking for as different failure modes as possible, you may want to use one SMPS and one linear PSU, for example.

The only thing I haven't yet managed is to start the fan. I am using a nice Sunon 2-wire 5V fan, Maglev kind. But for all I got from it, I could have used a piece of wood to the same effect.
I found in the wiki that If you have fan with only 2 wires (V+,GND) then you have to set fan-mode to manual, and I did so but still no power to the fan, even though temperature > fan-on-threshold (I set the latter to 30C for testing).
Probing with a multimeter, I read 0.00 volts coming out of pin 2 relative to pin 1 (these are supposed to be V+ and GND), no matter what I set the fan values to.
Any ideas?
BTW, here is a photo of my latest modified CRS. I used the connectors that took me so much to find and tied down the long 24V power cables to the back honeycomb. There is also a still useless fan

The only thing I haven't yet managed is to start the fan. I am using a nice Sunon 2-wire 5V fan, Maglev kind. But for all I got from it, I could have used a piece of wood to the same effect.
I found in the wiki that If you have fan with only 2 wires (V+,GND) then you have to set fan-mode to manual, and I did so but still no power to the fan, even though temperature > fan-on-threshold (I set the latter to 30C for testing).
Probing with a multimeter, I read 0.00 volts coming out of pin 2 relative to pin 1 (these are supposed to be V+ and GND), no matter what I set the fan values to.
Any ideas?
BTW, here is a photo of my latest modified CRS. I used the connectors that took me so much to find and tied down the long 24V power cables to the back honeycomb. There is also a still useless fan

I remember asking a question some time back about using two internal PSU's, what would happen in the scenario where
one PSU output was short-circuit would this short on say connector 1 pull down the output voltage of the second PSU connected to connector 2 on the router PCB, the reply from Mikrotik was it would not effect the second PSU voltage but sometimes the only to know is to check what really does happen ?

I remember asking a question some time back about using two internal PSU's, what would happen in the scenario where
one PSU output was short-circuit would this short on say connector 1 pull down the output voltage of the second PSU connected to connector 2 on the router PCB, the reply from Mikrotik was it would not effect the second PSU voltage but sometimes the only to know is to check what really does happen ?

I think (talking from memory right now) that the power input goes through a fat diode. If that is the case, a short circuit will bias it in reverse, blocking it, and therefore nothing would happen.

If I get the chance, I'll do the experiment for you though.

I remember asking a question some time back about using two internal PSU's, what would happen in the scenario where
one PSU output was short-circuit would this short on say connector 1 pull down the output voltage of the second PSU connected to connector 2 on the router PCB, the reply from Mikrotik was it would not effect the second PSU voltage but sometimes the only to know is to check what really does happen ?

I think (talking from memory right now) that the power input goes through a fat diode. If that is the case, a short circuit will bias it in reverse, blocking it, and therefore nothing would happen.

If I get the chance, I'll do the experiment for you though.

Be careful not to cause any damage to your router by experimenting on this?

Also another fault condition is where one PSU it's output voltage is pulsing a high voltage output due to a component failure on load and this high voltage although short in duration could cause the router to cut out as working voltage 8-28V + is exceeded?

Hi,

I'm sorry but I was in a hurry to send these boxes to production use and I don't have a spare one to test a short to the power output.

As mentioned already, it shouldn't be a problem, as the input is protected by a diode (or so it seems) and the PSUs can be shorted without permanent harm to them.

Anyway, this failure mode doesn't worry me much. I'm more concerned that the PSUs are actually redundant and the box works fine with only one powered up. That I did test, and the result was satisfactory.

Sorry to dig up this old thread.

Did you ever manage to fix the fan not turning on problem? I was looking at adding a fan to it to provide some form of active cooling at least. I was looking at using a noctua 40x40mm 5v fan for it.

But you mentioned you couldn't measure any voltage on the 5v FAN connector? Does the other fan connector work maybe? Did you test that one too?

I was totally unable to get the fan to turn. Maybe the components required had been omitted, who knows, but my advice would be to play safe and hook directly to the power supply output.

you can modify the CCR to be silent too and the CCR1036 also needs a better PSU. WIth a bit of tape you can use silent fans with the CCR. With a better PSU you can have more silent fans, and stable usb power.

I like what you did with your CRS. Adding a fan would cool it down significantly but there still is no STP and the switch chip keeps hanging whenever i swap wires between the gateway as the SFP+ port is unreliable.

Can't agree more on the big no-no that is the lack of STP, which I think I read somewhere to be scheduled for version 7.
We didn't find reliability issues like the one you describe, but we did find a few annoying bugs.
I also find VLAN configuration unintuitive and would much prefer something simpler, perhaps à la Cisco but without the dangerous catches like the difference between "switchport trunk allowed vlan 7" and "switchport trunk allowed vlan add 7".

Is there a way to just wire the fan directly into the PSU's and not worry about On/Off functionality? Assuming the fan doesn't spin so fast as to create too much noise, why not just run it all the time?

The reason not to wire directly to the PSU is that the mainboard seemed to provide a 3-wire connector. This has the benefit that you can query the device for the fan speed (and get an alert if it stops). Besides, the cabling is neat.

However, if the mainboard does not provide power... Then direct to PSU is actually the only option, of course. And you can perhaps save a few pennies buying fans with no speed reporting wire and electronics.

I was totally unable to get the fan to turn. Maybe the components required had been omitted, who knows, but my advice would be to play safe and hook directly to the power supply output.

But did you try both ports? Or only the one labelled 5v? Next to the other one I do see a mosfet on the board.

Can't recall now and I can't access the modified devices either, but the photos should tell which one I used.

I have verified that the CCR power connectors are Molex 10-63-4027 connectors (or copies).
I have ordered a few of these and soldered them to the CRS motherboard of the second of 3 CRS devices being fit with dual redundant power supplies. They fit perfectly. Note that relative to the nearest edge of the board, they are reversed with respect to the CCRs.

Hey Alfredo, can you tell me where you purchased the Molex 10-63-4027 connectors? And do they come as a set of the solder in male pins and the matching female plastic clip in that can be soldered/wired to the PS line feed? Or is that a separate Molex item number?

Thanks!!

I think I ordered them from RS Components. Could also have been Digi-Key.

The part number quoted is for the male side. The Molex site will helpfully link these with their female counterpart and the useable contacts.

You need all three parts: Male (pins with their housing); female (hollow plastic); and contacts (that are installed inside the female).