It is useful for the community to know. Yes, do let MikroTik know directly. They probably source these power supply units and do not make them themselves.

Good to know, thanks for sharing with us. Indeed probably an isolated batch where the PSU units were sourced from a different, probably Chinese supplier who in turn did not deliver according to required/promised specifications. Ideally MikroTik should have conducted their own tests on the batch before passing them on to customers, but who knows what happened. Not the best sign either way

Also for your company is a good idea to test the new releases because you fix one issue and broke a 4 things that working good.

It is useful for the community to know. Yes, do let MikroTik know directly. They probably source these power supply units and do not make them themselves.

The world should know, Mikrotik is NOT reliable.

My CCR1036 (Old design with only one PSU) has 4 PSU failed in a row for each year. I just need to switch to another router to keep the internet on.

It's ridiculously faulty design for a vital network device supposed to run 24X7 365 days a year, failed each year on PSU.

The world should know, Mikrotik is NOT reliable.

My CCR1036 (Old design with only one PSU) has 4 PSU failed in a row for each year. I just need to switch to another router to keep the internet on.

It's ridiculously faulty design for a vital network device supposed to run 24X7 365 days a year, failed each year on PSU.

Can happen with any brand.
But this right here "My CCR1036 (Old design with only one PSU) has 4 PSU failed in a row for each year" I call bullshit.
If 4 different PSU fail on you, I put the blame 100% on you for not protecting your equipment with a good UPS, if at all.

I manage a small number (19 units) of CCR1036 and only one PSU failure in 10 years now. The failed PSU was not on UPS.
Now, almost anything is possible.
Like having n% of PSUs fail from a batch. I get that.
But changing 4 different PSUs in one single unit and having them all fail...I don't know Rick.

So your anecdotal experience that without UPS CCR1036s fail is not bullshit, but without any knowledge under what circumstances tons of other users' CCR1036 have PSU failures are. Got it.

The world should know, Mikrotik is NOT reliable.

My CCR1036 (Old design with only one PSU) has 4 PSU failed in a row for each year. I just need to switch to another router to keep the internet on.

It's ridiculously faulty design for a vital network device supposed to run 24X7 365 days a year, failed each year on PSU.

Can happen with any brand.
But this right here "My CCR1036 (Old design with only one PSU) has 4 PSU failed in a row for each year" I call bullshit.
If 4 different PSU fail on you, I put the blame 100% on you for not protecting your equipment with a good UPS, if at all.

I manage a small number (19 units) of CCR1036 and only one PSU failure in 10 years now. The failed PSU was not on UPS.
Now, almost anything is possible.
Like having n% of PSUs fail from a batch. I get that.
But changing 4 different PSUs in one single unit and having them all fail...I don't know Rick.

So your anecdotal experience that without UPS CCR1036s fail is not bullshit, but without any knowledge under what circumstances tons of other users' CCR1036 have PSU failures are. Got it.

I personally know people with lots of CCR's and none complained so much so to make it obvious it is a trend.
Is there some statistic regarding this trend of failures?
Or if someone states they don't have a recurrent problem with specific Mikrotik gear, it is anecdotal, but somehow, if someone states "Mikrotik is NOT reliable.", then that is the rule and not the exception?

If i need to replace PSU for every 1.5 year, how about charge me the license fee annually and upgrade the PSU to a more stable one?
It's pain to remove all cables and re-plug them to another router, remove CCR1036 and carry it to the vendor for break-fix.

The world should know, Mikrotik is NOT reliable.

Do you have any other equipment from Mikrotik (ie, another CCR1036 or something else similar) ?
Can you establish a pattern regarding PSUs dropping dead from all CCRs from Mikrotik ?

Do you have replacement router, once that 1036 drops dead ? I surely would have one, considering it failing once a year. Is it another Mikrotik or something else ?

EDIT:
Considering that you run some serious network, I would suggest that you replace that aging router for something newer, or at least, other piece of the same used device.

Do you have any other equipment from Mikrotik (ie, another CCR1036 or something else similar) ?
Can you establish a pattern regarding PSUs dropping dead from all CCRs from Mikrotik ?
Do you have replacement router, once that 1036 drops dead ? I surely would have one, considering it failing once a year. Is it another Mikrotik or something else ?

The world should know, Mikrotik is NOT reliable.

That up there is a bold statement, considering it's coming from 1 piece of equipment failing.
At the moment, I have exactly 48 various Mikrotik products (routers, switches, APs,) running 24/7 all across my network for years now. Apart from one RB1100AHx4 dead from lightning surge that came through vDSL modem copper wire (I blame ISP) I yet have to see even power brick from Mikrotik failing, yet alone entire device. And I know that 48 devices is not big number for a pattern, that's why I won't say to the world that Mikrotik IS reliable. But it is reliable for ME.

EDIT:
Considering that you run some serious network, I would suggest that you replace that aging router for something newer, or at least, other piece of the same used device.

If i need to replace PSU for every 1.5 year, how about charge me the license fee annually and upgrade the PSU to a more stable one?
It's pain to remove all cables and re-plug them to another router, remove CCR1036 and carry it to the vendor for break-fix.

You can also install a better power supply, especially if you are replacing it every 1.5 years.. You are the only one that has reported them failing that much.

24v 4 amp power supply, from memory..

Do you have any other equipment from Mikrotik (ie, another CCR1036 or something else similar) ?
Can you establish a pattern regarding PSUs dropping dead from all CCRs from Mikrotik ?

There is another thread here that has become active again where even Mikrotik admitted early power supply failure in many of the early CCR r1 routers.. Bad caps..

The 'green' caps was the design change/fix, and they were certainly better after.

Do you have replacement router, once that 1036 drops dead ? I surely would have one, considering it failing once a year. Is it another Mikrotik or something else ?

EDIT:
Considering that you run some serious network, I would suggest that you replace that aging router for something newer, or at least, other piece of the same used device.

Our CCR1036 that seems to be dying, We have a CCR1009 that will be swapped in so I can fix the CCR1036.. We outgrew the CCR1009 but it was left in the rack as a spare/backup.. This issue does have us looking at a CCR2116 again but that Marvell 98DX3255 looks like a major handicap, which is why we haven't (yet).

Someone from Mikrotik posted that's because the overheating of the capacitor, the next generation will be better.

Someone from Mikrotik posted that's because the overheating of the capacitor, the next generation will be better.

Heat is the biggest enemy of capacitors.

And the later generations have been better.

You said you are able to, so I suggest you do, replace with the exact capacitors I gave you earlier.. When you see search results, check that you are looking at the same model, sometimes searches return other model numbers that are 'close' because your chosen vendor has them in stock and they are similar.

Are you running your CCR1036 in a warm/hot environment?

1. I checked the previous failed PSU PCB, the bulge up C10 capacitor (1000UF, 35V, 105 degree Celsius max tolerance), it's already a good spec withstand 105 Celsius, i can only replace it with the same model, i have no idea will it fail again in 1.5 year...

1. I checked the previous failed PSU PCB, the bulge up C10 capacitor (1000UF, 35V, 105 degree Celsius max tolerance), it's already a good spec withstand 105 Celsius, i can only replace it with the same model, i have no idea will it fail again in 1.5 year...

Yes, what I sent you before is a higher (over) spec cap..

Brand matters, some are better than others. Some brands specs are closer to wishful thinking. Panasonic and Rubycon are really good. There is a third brand that is also really good, but I can't think of it without looking them up.

Higher voltage cap is better than a lower one.

Expected lifetime in hours matters too.

Or look for a 24v MeanWell power supply and put it in its place.

the same 105 Celsius, 7,000 hours roughly equal to 291 days only.

Incidentally, 7000 hours @ 105ºC is an uncommonly good cap for an aluminum electrolytic, not junk. There’s better available, but not much better.

the same 105 Celsius, 7,000 hours roughly equal to 291 days only.

Only under a naive reading of the specs.

First, those two specs are inextricably paired. If you drop the temp, the lifetime rises proportionately. The accepted rule of thumb is that lifetime is expected to double for every 10ºC drop in operating temperature.

Second, while that rule of thumb is based on solid empirical science, it isn’t a physical law. Different cap designs and formulations will have different temp vs lifetime curves, even under the same front-page specs.

This is why posters above were pressing you on operating temperature inside the device. It doesn’t matter that the cabinet’s floor-level inlet temp is 25ºC if you put the router at the top of a heat chimney. If the fan inlet temp at the router is 60ºC and the internal components add another 15ºC over ambient, you’ve cut expected lifetime by 5 doublings; 2⁵=32.

At this posited 75º operating temp, you’d expect 3 doublings of the lifetime spec. 2³=8, so 56k hours, or 6.3 years.

(It doesn’t stop being a cap at 6.3 years, either. It’s merely allowed to drift out of spec at that point without invalidating the data sheet claims.)

Working it inversely, if 105ºC 7000 hour caps are dying in 1.5 years, that’s only about one doubling, so unless these caps are being run at WTF temps (~96ºC!) they either don’t meet their specs or don’t follow the rule of thumb above.

Incidentally, 7000 hours @ 105ºC is an uncommonly good cap for an aluminum electrolytic, not junk. There’s better available, but not much better.

My latest failed PSU capacitor already the newer capacitor as my vendor replaced the PSU, where you cannot understand i replace the same spec capacitor will quite possibly last the same??

Junk grade is 1000 hours @ 85ºC. And yes they really exist even from top-tier name-brand suppliers like Panasonic.

i cannot assess how hot the nearby heatsink of the transistor can be,

i cannot assess how hot the nearby heatsink of the transistor can be,

You work in an industrial setting and nobody around has a DMM with a thermocouple you can borrow?

The internal health readings should get you into the ballpark, at least.

It’s no accident that this menu is called “health,” by the way.

which of the following "health" showing the temperature of the heatsink of the transistors??

who would open up a production firewall (router) to check heatsink temperature?

Wouldn't it a design defect to allow an overheating heatsink to reduce the capacitor life thus blown up the whole PSU?

I wrote "ballpark" for a reason. It's unreasonable to expect die temp readings on every transistor in the device, but it's equally unreasonable to suppose that the PSU is running at a wildly different temp from the rest of the device. It's likely to be somewhere in the range you show.
And if true, then 45°C ambient means we should expect 105-45/10=6 doublings, or 64×7k hours = ~52 year lifetime.
Therefore, something else is going on other than heat, if your caps are meeting their specs.
If we're starting a pool, I put dibs on voltage spikes from the collapse of the inductive windings on the M-G set next door.

Me, but then, maybe I simply nerd harder than you do.

Wouldn't it a design defect to allow an overheating heatsink to reduce the capacitor life thus blown up the whole PSU?

Engineering is all about tradeoffs. There are no absolutes.

[admin@MikroTik] > system/health/print
Columns: NAME, VALUE, TYPE
# NAME VALUE TYPE
0 power-consumption 50.8 W

CCR1036 (the CCR1036-12G-4S variant) has rated max power consumption at 60W. So the reported power consumption indicates that power supply is running at 80%+ capacity and I'd expect it to have a bit higher temperature.

So what can you do to extend power supply lifetime? Reduce load on router. Yes, this suggestion is outrageous, but it would definitely help.

[admin@MikroTik] > system/health/print
Columns: NAME, VALUE, TYPE
# NAME VALUE TYPE
0 power-consumption 50.8 W

CCR1036 (the CCR1036-12G-4S variant) has rated max power consumption at 60W. So the reported power consumption indicates that power supply is running at 80%+ capacity and I'd expect it to have a bit higher temperature.

So what can you do to extend power supply lifetime? Reduce load on router. Yes, this suggestion is outrageous, but it would definitely help.

1. My CCR1036 is not in high demand, only a few people will connect through it, therefore, i already adjust down the CPU frequency to lower the operating temperature. However, consider the capacitor overheating theory, the heat comes from the nearby power transistors to regulate the current, it seems reduce load (lower frequency) not working here.

However, it's even the other way around: ROS regulates fan speed (and thus air flow) according to temperature of CPU (and port area). But indirectly this also affects cooling of CPU (part of air does flow through PSU) and keeping fan speed low means higher temperature inside PSU.

The load on router doesn't come from number of users, it comes from throughput and concurrent number of connections. Number and actual speed of used ports also is a factor in power consumption. CPU frequency only partially affects power consumption. And PSU temperature is proportional to power delivered by PSU.

"However, it's even the other way around: ROS regulates fan speed (and thus air flow) according to temperature of CPU (and port area). But indirectly this also affects cooling of CPU (part of air does flow through PSU) and keeping fan speed low means higher temperature inside PSU. The degree to which this actually happens largely depends on thermal design of chasis and airflow inside chasis, so increasing airflow may not affect PSU temperature much if airflow avoids PSU."

"BTW, telecom grade equipment tend to run fans quite high ... so using telecom grade equipment in office/residental environments, where noise is generally not welcome, may actually not be the best idea, it's better to use equipment, designed to emit low noise and thus cooling has to be designed in a different way."

Generally you shouldn't replace capacitor with a part that has different capacity (neither smaller nor larger). It can be rated for higher voltage (that doesn't hurt) and definitely can be rated for higher temperature.

For power regulation, capacitor with larger voltage rating and larger capacitance usually work better, for example, to better smooth ripple.

For power regulation, capacitor with larger voltage rating and larger capacitance usually work better, for example, to better smooth ripple.

Sure, all else being equal, but all else is not equal, because you've got a fixed area of PCB space to install it in. Greater capacity and higher voltage tolerance in the same space must trade off something. TANSTAAFL.

My bet? ESR.

For power regulation, capacitor with larger voltage rating and larger capacitance usually work better, for example, to better smooth ripple.

Sure, all else being equal, but all else is not equal, because you've got a fixed area of PCB space to install it in. Greater capacity and higher voltage tolerance in the same space must trade off something. TANSTAAFL.

My bet? ESR.