Aluminum electrolytic capacitor lifetime and mining equipment

[64dimensions]

Here is a link from EE times: http://www.eetimes.com/author.asp?doc_id=1279791

"Aluminum electrolytic capacitors remain a popular choice in power supplies due to their low cost. However, they have limited life and are sensitive to both hot and cold temperature extremes. Aluminum electrolytic capacitors are constructed with foils placed on opposite sides of paper saturated with an electrolyte. This electrolyte evaporates over the capacitor’s lifetime, altering its electrical properties. If the capacitor fails, it can be spectacular as pressure builds up in the capacitor, forcing it to vent a combustible and corrosive gas."

Bullet points from the article:

1)  "For every 10 degree Centigrade decrease in operating temperature, the capacitor life is extended by a factor of two."

2)  "Capacitor life ratings generally are specified at their maximum rated temperature. A typical rating might be 1,000 hours at 105 degree Centigrade."


From Corsair:

"Industrial Grade Components: The TX, HX and AX (power supplies) are all rated for continuous output power even at
operating temperatures as high as 50°C (122°F) and use capacitors rated at 105°C. This ensures
unwavering performance and durability."

Comments:

1) In mining equipment, it's unknown whether the aluminum electrolytic capacitors are rated at 105°C. This type of rating is a a sort of certification of component quality. It's important enough that power supply manufacturers use this as a selling point.

2) YouTube videos showing repairs on cheaper ATX power supplies are usually replacing an electrolytic capacitor.

3) 1000 hours is 42days. So the all out "damn the torpedoes" over clocking approach may not be such a great idea for the home miner.

4) Internal temperature readings from miners are some sort of average indicator and don't say much about potential hot and cold spots.

5) Used mining electronic circuitry has lost some of it useful working life because of electrolytic capacitor aging. This of course depends on how it was cared for. This is particularly relevant to manufacturers selling out equipment from their mining operations. This may explain the (short) 90 day warranty offered.

[1715595744]

1715595744

[1715595744]

1715595744

[1715595744]

1715595744

[1715595744]

1715595744

[1715595744]

1715595744

[Prelude]

Miners should never buy ATX PSUs blindly. Always refer to jonnyguru. If it's not reviewed there, it's an unknown quality and I'd avoid it.

There's only 1 company I trust to do no wrong, and that's Seasonic. Everything else is up in the air.

[atmtech]

I think using ATX power supplies for large miners is asking for trouble.
Caps can have all the rating printed on them and still can fail at lower temperatures.
I remember 2 cases of this happening to major suppliers of PC motherboards .
They are easy to change and cheap to buy.
The 90 day warranty is cheap to offer and MOST failures occur in that time frame with new electronic products
You know the companies for these miners are not fortune 500 companys and would be pink sheet quality in the
Stock market so you may or may not get what you pay for.

[VirosaGITS]

3) 1000 hours is 42days. So the all out "damn the torpedoes" over clocking approach may not be such a great idea for the home miner.

4) Internal temperature readings from miners are some sort of average indicator and don't say much about potential hot and cold spots.

While you have some good tidbits of information and warning, you are not applying them correctly in real life situation. At the temperature they run, at worse, hot, with no cooling and bad airflow, it could probably reach 50-60C, thats still several years.

At typical temperature in a good PSU with fan on, you're looking at something like rated for 5376 days of 24/7 use. With your own number. Imagine when you have high quality ones.

So basically saying ATX PSU are a bad idea can be true, if you buy crap, but you're generalizing and also using numbers that don't make sense. A cap wont be running at 100C at all.

[QuintLeo]

Keep in mind that the temperature inside even a high efficiency (Gold/Platinum/Titanium) power supply will be noticeably warmer than the ambient temp.

 Quite a few miners see temps exceeding 60 degrees C, especially when folks push the overclocking on them.
 (I'm not counting the SP20 indicating 100+ as that's the on-CHIP temp of the Rockerboxes).


 The "hot and cold spots" are a very good point, as temps in a miner generally are measured at a VERY FEW or even ONE point.

[RadekG]

You are mismatching things together. Please read following notes:

1) EVERY AC/DC PSU (so ATX also) is made of 105C capacitors.

2) During heavy use of PSU, highest temerature is around heatsinks with diodes or transistors, temperature can exceed 100C.

3) Input and output capacitors are usually far away from sources of heat (and it is designed this way), so temperature during use is close to ambient.

4) Good quality PSU has MTBF 100k-200k hours. This is reduced by fan lifetime.

5) Lifetime of capacitors highly depends on its quality and design quality, server PSU has capacitors close to heatsinks (tight space) and lasts longer than regular PSU.   

6) Do not mismatch LED PSU and ATX PSU. Both are operated at different conditions and both are failing at different points - LED has problem with input capacitor while ATX with output one. Reason is degrading ESR which causes warming of capacitor. Good design respects it and never overheats capacitor even with degraded ESR. 

[64dimensions]

Comments:

1) "EVERY AC/DC PSU (so ATX also) is made of 105C capacitors."

As an example from Corsair.

Direct quote from Corsair:

"Industrial Grade Components: The TX, HX and AX (power supplies) are all rated for continuous output power even at
operating temperatures as high as 50°C (122°F) and use capacitors rated at 105°C. This ensures
unwavering performance and durability."

Corsair has other types of ATX power supplies: http://www.corsair.com/en/power-supply-units THAT ARE NOT in the TX, HX and AX series, therefore I can only conclude that they don't necessarily use electrolytics rated at  105°C. The non TX, HX and AX series types are also a lot cheaper. Otherwise why just single out the TX, HX and AX series for notice?

The only point of introducing power supplies at all was to support the notion that high quality eletrolytics are important for reliability, not to get into a discussion about power supplies.

Getting back to mining hardware (because build quality is an issue):


2) As an A6 owner my concern was the pictures shown at: http://asicminer-shop.de/Avalon-6-33-TH-s_1. Take a look at the hashing board. Electrolytic bypass capacitors are on the other side of the hashing chips. The bypass capacitors are there in order to damp down any transients on the chip power bus. They have to be located as close to the chip as possible and not in some air conditioned paradise. The use of bypass capacitors with hashing chips is no doubt standard practice. As a lucky SP20 owner and as Qunitleo notes, the "temperature indicator" that is used by the SP20 hardware can go up to 120°C. As noted in the SP20 threads, the SP20 starts to downclock at 125°C. As Quintleo points out, where this temperature is being monitored, who knows?

[RadekG]

Comments:

1) "EVERY AC/DC PSU (so ATX also) is made of 105C capacitors."

As an example from Corsair.

Direct quote from Corsair:

"Industrial Grade Components: The TX, HX and AX (power supplies) are all rated for continuous output power even at
operating temperatures as high as 50°C (122°F) and use capacitors rated at 105°C. This ensures
unwavering performance and durability."

Corsair has other types of ATX power supplies: http://www.corsair.com/en/power-supply-units THAT ARE NOT in the TX, HX and AX series, therefore I can only conclude that they don't necessarily use electrolytics rated at  105°C. The non TX, HX and AX series types are also a lot cheaper. Otherwise why just single out the TX, HX and AX series for notice?

The only point of introducing power supplies at all was to support the notion that high quality eletrolytics are important for reliability, not to get into a discussion about power supplies.

Getting back to mining hardware (because build quality is an issue):


2) As an A6 owner my concern was the pictures shown at: http://asicminer-shop.de/Avalon-6-33-TH-s_1. Take a look at the hashing board. Electrolytic bypass capacitors are on the other side of the hashing chips. The bypass capacitors are there in order to damp down any transients on the chip power bus. They have to be located as close to the chip as possible and not in some air conditioned paradise. The use of bypass capacitors with hashing chips is no doubt standard practice. As a lucky SP20 owner and as Qunitleo notes, the "temperature indicator" that is used by the SP20 hardware can go up to 120°C. As noted in the SP20 threads, the SP20 starts to downclock at 125°C. As Quintleo points out, where this temperature is being monitored, who knows?

Please consider my previous post as a fact

1) EVERY PSU has 105C rating capacitor. It is very rare to find 85C rating caps even in cheap chinese (non ATX) PSUs.

Please check following link to prove that even cheapest VX series uses 105C caps. By the way, VS series uses high quality components including fan.

http://www.legitreviews.com/corsair-vx450w-power-supply-review_544/3

2) You are mismatching regular electrolyte (wet) capacitors with solid capacitors (polymer). Avalon 6 has surely these high endurance capacitors. By the way, it is a good idea to keep your chips as cold as you can, far less than 80C. High temperature means higher current draw leading to higher temperature. Lower efficiency. Wasted electricity and overheating components.

3) if the design of PSU respects high temperature and degradation of ESR, it can work without problems very long, but if the design do not respect worser parameters of capacitors, they will be overheated and cause PSU failure.

[sidehack]

I used to do electronics refurb as a full-time job. Failed and underspec electrolytics pretty much paid my bills for about three years.

Also, regarding the Avalon6:



If that image is still representative of its design, looks like there's a ceramic on the back of every ASIC and a can right between 'em. Also, if anyone remembers the Prisma, one of the more common failures on them was the electrolytics boiling over and the cans popping off, sometimes melting or slightly setting fire to whatever they landed on. The V2 Prisma was fully shrouded more-or-less specifically to catch bursting components; it certainly wasn't to increase airflow on the board surface.

[philipma1957]

so 18  rows of 4 chips below and the ceramic cap I whited out  does look to be above the asic chip.. the 105C wet caps are set as far as possible from the chips as i see it.

I am curious as to what the temp reader is on the photo below
I get 3 readings per avalon 6

34 76 72

34 = Intake?

76 = board 1 ?

72 = board 2 ?

[sidehack]

"As far as possible" means approximately one centimeter. Not to defend Bitmain's engineering practices, but at least all the ASIC-adjacent large caps on every large miner they've sold were solid-state. That said, I have seen a few blown electrolytics on S3 boards right next to the buck drivers.

[RadekG]

"As far as possible" means approximately one centimeter. Not to defend Bitmain's engineering practices, but at least all the ASIC-adjacent large caps on every large miner they've sold were solid-state. That said, I have seen a few blown electrolytics on S3 boards right next to the buck drivers.

Black caps are wet, lilac are solid.

[QuintLeo]

4) Good quality PSU has MTBF 100k-200k hours. This is reduced by fan lifetime.

 MTBF is a oft quoted stat that often has ZERO basis in actual real-life experience.
 I've seen TONS of devices or items with 6 digit rated MTBF that failed to AVERAGE even 1 year (8760 hours appx) in ACTUAL usage.



 I have no clue why the "LED PSU" crap is even being mentioned, miners don't use "LED" PSUs on their miners as those PSU don't have the voltage/current capasity needed to power current miners and are NOT efficient.

[VirosaGITS]

4) Good quality PSU has MTBF 100k-200k hours. This is reduced by fan lifetime.

 MTBF is a oft quoted stat that often has ZERO basis in actual real-life experience.
 I've seen TONS of devices or items with 6 digit rated MTBF that failed to AVERAGE even 1 year (8760 hours appx) in ACTUAL usage.



 I have no clue why the "LED PSU" crap is even being mentioned, miners don't use "LED" PSUs on their miners as those PSU don't have the voltage/current capasity needed to power current miners and are NOT efficient.

And that fan thing is kind of untrue(or can be), fans typically dont even spin on my PSU. Its really all about build quality. When a PSU's warranty cover 10 years it give you a hint compared to the 2 years ones, too.

[RadekG]

 MTBF is a oft quoted stat that often has ZERO basis in actual real-life experience.
 I've seen TONS of devices or items with 6 digit rated MTBF that failed to AVERAGE even 1 year (8760 hours appx) in ACTUAL usage.

This has nothing to do with MTBF. If you take 1000 classical light bulbs, you will see 500 dead after 750hrs of working hours. But, you will see probably 100 of them dead within first 250 hours. This is MTBF. Nobody can guarantee how long last any product, but someone can measure, how long it usually lasts.

If you were man repairing lights, you would see much more dead light bulbs even during first 250 hours of their life than other people. If you were car serviceman, you would see much more failing engines than any other people.

If you see really so many early dying devices, you should consider use them properly.

I do not want to contiune this way of talking, this thread is about capacitors. My opinion is: design is the key, quality of components doesn't matter so much. We can see very old products still working, even with low quality components, because designer was clever.

Avalon has solid capacitors which aren't exploding, all (corsair) PSUs have 105C capacitors, most of them last years (I have 700W Fortrons working under 70-90% load 24/7 two years).
So, what are we talking about?

[QuintLeo]

The problem is not that I don't use the devices properly.

 The problem is that the "standards' for MTBF calculation are non-existant (outside of mil-spec) and the quoted MTBF on most gear I've seen (INCLUDING mil-spec) tend to be anywhere from optimistic to "marketing hype amounting to an outright lie".

 Note - I specified AVERAGE actual lifetime, not "I start seeing failures in less than a year".
 Any sort of sleeve-bearing FAN in particular tends to be wildly optimistic on MTBF, for my favorite example, especially fans commonly used in low-cost power supplies (that often tend to cause the entire PS to fail shortly after the fan dies and causes the PS to start overheating badly VERY quickly).


 I agree that length of warenttee is *usually* a much better clue. When you actually have to pay to repair your stuff, you pay a lot of attention to how long you're willing to guarentee it (or you end up in a world of financial hurt, ALA Clipper Windpower).

[RadekG]

 Any sort of sleeve-bearing FAN in particular tends to be wildly optimistic on MTBF, for my favorite example, especially fans commonly used in low-cost power supplies (that often tend to cause the entire PS to fail shortly after the fan dies and causes the PS to start overheating badly VERY quickly).

Sleeve fan bearing lifetime is extremely derated by temperature. For example, Radeon reference fans have sleeve bearing, they never failed on my R9 GPUs at max speed for more than two years, so about 20k hours at temps reaching 45C during summer. PSU should always use 2Ball bearing. Good PSUs have overtemperature protection.

I think it take sense if the PSU has good warranty and is equipped with protections. Changing fan after warranty is much easier than repairing blown mos-fets and burned SMD parts around.