Corporate propaganda to justify cheap Chinese caps

[Phaedrus]

Corsair (PC PSU manufacturer) got some flak recently for putting cheap capacitors (Ltec and CapXon) in some of their newer high-end power supplies (RM and CSM series). This questionable choice in components was compounded by the designs having very cramped PCB layouts (750W and 850W models basically have the filter capacitors in an airless box), and having low-RPM fans that don't turn on until the PSU exceeds 60% load, or the secondary heatsink exceeds ~100*C.

http://www.corsair.com/blog/capacitor/

They're using the life expectancy formula and the specs provided in Ltec's datasheet to claim a 15 year lifetime for their secondary capacitors.


 

[Fsck]

it's marketing. enough truth so that most people will accept it as the complete truth since they probably don't know better.

addendum: someone should challenge them to back up their 15-year life with a 15-year warranty.

[bridgerectifier]

My friend has run a PC shop for years, and he told me a few years ago that quite a number of a certain Corsair "high-end" PSUs have gone bad - I have one of his cast-offs here with me - kept for heatsinks and inductors

JUNK! Living off a name, earned by gamers who are marketed into thinking Corsair are "the best". Lol

I always like to know what COMPONENTS are inside the parts that computer engineers call "components" - IE, a PSU is referred to as a "component", as is RAM etc, but these "components" are comprised of many circuit-level components, which can be crap, but these PC people often (not always) know no better.

I know the above sounds obvious... just had to get it out

[Phaedrus]

it's marketing. enough truth so that most people will accept it as the complete truth since they probably don't know better.

addendum: someone should challenge them to back up their 15-year life with a 15-year warranty.

They claim a 5 year warranty. I'd give those caps a half life of 3-4 years. Hopefully it comes back to bite them.

[NiHaoMike]

Yet another case of the not-so-smart fan. It degrades reliability and adds another marketing feature. A double win for them, a loss for us.

I suggest buying server grade PSUs if you really need the reliability, but those don't come cheap.

[Phaedrus]

You can get reliability in the consumer market. You just have to pay for it, and do your research.

[all_repair]

At the back of their mind is: effective warranty is 12month (after that people that claim are far in between even stated as 5 year), life of product is 24month, and next product launch is 18 month that has a 2x speed, 2x memory, 200x more features.

So the capacitors are over spec.

[echen1024]

I use a Corsair 650W. I'll open it up sometime, maybe take some pics.

[Vgkid]

JG would not be happy about this.

[Phaedrus]

JG would not be happy about this.

John Gerow (the founder) works for Corsair as a consultant now and is desperately arguing with anyone who points out the absurdity of believing CapXon and Ltec datasheets.

Jeremy Schrag (the current main reviewer) deducted a point for it, but otherwise didn't call them out on it much.

[AndyC_772]

Many here may understand why this is so hilariously funny, but I'll reveal my ignorance and ask why? What's the accepted standard way of calculating a capacitors life expectancy?

Start with the rated basic lifetime of the capacitor, and apply a formula to adjust this figure based on the actual operating temperature in service. The theory presented in the blog is sound.

It's also important to take into account what constitutes 'failure', because it's not nearly as well defined a some people think. What happens is that over time, the properties of the capacitor simply degrade; the capacitance goes down, and at some point it's no longer within the spec in the data sheet. At this point it's deemed to have 'failed', even though the circuit may continue to operate perfectly well for years to come.

This gradual degradation (as opposed to 'sudden death') is one reason why most products that use electrolytic caps do actually last way longer than the rated lifetime of the parts from which it's built. If every cap that had, say, a 3000 hour lifetime rating actually did fail completely and become useless after 3000 hours, we wouldn't be complaining about the odd PSU that fails under warranty - we'd be replacing them routinely like batteries.

Of course, some caps do fail catastrophically, but they're the statistical outliers... the ones that most likely had latent defects from day one, and were never going to meet the data sheet specs for reliability. Some manufacturers may indeed have better QC than others, and gain a reputation for 'good' or 'bad' quality accordingly, but that doesn't mean the way to calculate a capacitor's lifetime is flawed.

Normally when I design a product using electrolytic caps - say, a PSU output filter - I'll start with a reasonable estimate of how hot I think it's going to get in service, and then design in enough capacitance such that even if the parts do degrade over time, there will still be enough for the product to work correctly once the design lifetime of the product has expired. Then, once a prototype has been made, I'll hook up some thermocouples and check the actual temperature of the capacitor in service, to make sure that MTBF estimate was accurate.

[Phaedrus]

They're using the life expectancy formula and the specs provided in Ltec's datasheet to claim a 15 year lifetime for their secondary capacitors.

Many here may understand why this is so hilariously funny, but I'll reveal my ignorance and ask why? What's the accepted standard way of calculating a capacitors life expectancy?

Several problems,

1. This is an idealized formula that doesn't take into account impurities, build defects, and other issues with the capacitor.
2. Environmental factors such as changes in temperature, humidity, and pressure, length of storage, and treatment during soldering will effect the lifetime of the cap.
3. It uses an average ripple current figure and doesn't take into account transient events that will inevitably occur due to rapid load changes or input surges.
4. Chinese capacitor manufacturers regularly lie or inflate the ratings of their components, so plugging their numbers into the formula is a rookie trap.
5. Even if everything else is accounted for, this is at best a rough mean. It doesn't give any useful statistical information, and when you make tens of thousands of power supplies with dozens of caps each, you have to look at the worst case in the second or third standard deviation, as a 0.1% capacitor failure rate can still equate to hundreds of unit failures in the field.


The lame excuse that John Gerow and George Makris (Corsair reps) give is that every other power supply in the last 20 years that has failed due to using Ltec, CapXon, or other off-brand caps must have been designed by an idiot with too high of a ripple current. Additionally they try to smear Japanese caps by pointing out that some are now made in Malaysia, while admitting quietly that the reason Japanese caps are better is extra margin, better materials, and QC. None of which is affected much by the location of manufacture. And while Chinese manufacturers may be getting access to higher quality aluminum (in a year or two, not now...) and electrolytes, the majority are still way behind on QC, and they do not provide the safety margin that Japanese manufacturers do, which is a large part of what gives them their legendary reliability.

[XOIIO]

I think I'll stick with cooler master or.. hrmm, there's another brand but I can't place my finger on the name right this second.

Oh well. I've always thought corsair had a bit too much hype around it.

[peter.mitchell]

Yet another case of the not-so-smart fan. It degrades reliability and adds another marketing feature. A double win for them, a loss for us.

I suggest buying server grade PSUs if you really need the reliability, but those don't come cheap.

If you don't mind modifying some connectors and the load noise of fans, they aren't THAT expensive...

They're using the life expectancy formula and the specs provided in Ltec's datasheet to claim a 15 year lifetime for their secondary capacitors.

Many here may understand why this is so hilariously funny, but I'll reveal my ignorance and ask why? What's the accepted standard way of calculating a capacitors life expectancy?

I'd say there is a pyramid of reasons, starting at the unreliability of these manufacturers. These companies fudge the numbers on the datasheet a little bit at the best of times, let alone the quality control.
Next up would be them using near best case numbers for the basis of their calculations, eg;

...and are operating at temperatures around half of what they’re rated at (44° to 53°C versus 105°C)..

Not a snowflakes chance in hell they would be running at those temps in a standard computer case, i mean, Wilfred, its 43 degrees ambient here today for crying out loud!
Then you have to take into consideration things like was the wave soldering profile to spec, how long were they in storage, how well is the ripple current balanced between the caps in the design, how do they age, how do they handle thermal cycling?

[tszaboo]

Chinese manufacturer said it in the datasheet.

[Phaedrus]

I'd say there is a pyramid of reasons, starting at the unreliability of these manufacturers. These companies fudge the numbers on the datasheet a little bit at the best of times, let alone the quality control.
Next up would be them using near best case numbers for the basis of their calculations, eg;

...and are operating at temperatures around half of what they’re rated at (44° to 53°C versus 105°C)..

Not a snowflakes chance in hell they would be running at those temps in a standard computer case, i mean, Wilfred, its 43 degrees ambient here today for crying out loud!
Then you have to take into consideration things like was the wave soldering profile to spec, how long were they in storage, how well is the ripple current balanced between the caps in the design, how do they age, how do they handle thermal cycling?

They use even cheaper capacitors in their Australian (APAC region) units than they do in their US ones. This is evidenced by the Corsair unit Dave opened when he was upgrading his PC. It used lower tier capacitors than the ones used in the same SKU in the US. And of course US spec units are the ones that go to reviewers. They forget that ozzies also have screwdrivers and soldering irons.

[peter.mitchell]

Wouldn't these factors be taken into account in the statistical sampling of the manufactures lifetime test capacitors? I don't know, but I just assume they would select a statistically significant sample size to form a test batch.

Even if they did, if you design around average rather than minimum it will bite your ass.

Aren't these beyond the control of the manufacturer? How would you select a value for each as a typical value to use? I concede they will effect the lifetime but many different products have an expected lifetime under test conditions that are not typical of all circumstances. Car air filters are suggested to last a certain time and the manual will always advise more frequent replacement in dusty conditions. Speaking of dust you could also claim that dust buildup on the fan/heatsinks and case ventilation holes should be factored in as well.

No, if you place the caps against heatsinks, away from vent holes, behind components blocking airflow ect, the average temperature may swing from ambient +5 degrees at full idle to ambient +80 degrees full load.
Since you're a massive manufacturer you can afford different units for different locales, if you're sending it to a humid locale, use more humidity resistant parts.
Treatment during soldering is easy, you're the manufacturer, you can choose exactly that.
Same with length of storage, buy minimal over-stock and buy new lots for new batches.

I don't know how the local usage conditions, again, can be calculated by the manufacturer. What percent of the time is a transient condition present? Is it so infrequent and "transient" that it can be considered statistically insignificant? Or is it so frequent that it really isn't a transient and is more properly considered part of the average ripple current. In the later case is it still an issue?

Transients are transients, they're hard to appropriately estimate, however, you can always calculate for expected worst case + margin, rather than calculating to the mean.

Having made tens of thousands of power supplies, wouldn't you then be well placed to test experience against the "rough mean" and determine if the calculation method is flawed?

Sure you would, and the company would go "hey, were making shit tonnes for money and people aren't complaining much, lets just keep doing it" - doesn't make it good for the consumer though.

[Phaedrus]

Since you're a massive manufacturer you can afford different units for different locales, if you're sending it to a humid locale, use more humidity resistant parts.

Actually you can't really. Not unless you're as massive as, say, Delta Electronics. The same batch of power supplies, divided between distributors in the US, may end up in climates as diverse as Malibu, Denver, Miami, and Minneapolis. Though I would consider it ideal for PSUs sold in the APAC region (usually includes southeast Asia and Australia, occasionally India or East Asia) to have higher-spec caps to account for the heat in this region, the fact that most of the countries in that area are poorer means that most companies cut cost by using cheaper caps.

[NiHaoMike]

Actually you can't really. Not unless you're as massive as, say, Delta Electronics. The same batch of power supplies, divided between distributors in the US, may end up in climates as diverse as Malibu, Denver, Miami, and Minneapolis. Though I would consider it ideal for PSUs sold in the APAC region (usually includes southeast Asia and Australia, occasionally India or East Asia) to have higher-spec caps to account for the heat in this region, the fact that most of the countries in that area are poorer means that most companies cut cost by using cheaper caps.

Get rid of the "smart fan" and it will run cooler. Of course, that would make the unit too reliable for future sales...

A trick you can do (if you're not willing to spend the extra for a server grade unit) is to pressurize the case with a powerful fan.

[Phaedrus]

Assuming that capacitor failures approximate a bell curve, designing for the average +/-1 standard deviation will give you a 15% failure rate. Obviously not a good plan.

We target a failure rate <0.5% over the warranty lifetime of the product. A product that gets more than 1% failure rate is considered troublesome, and more than 2.5% is considered awful--this is what we get for the cheap non-80PLUS PSUs we used to sell before we realized warranty replacement costs exceed our sales margin... The worst products we've ever sold only hit about 10% failure rate and are considered abject failures.

Anyway, you want to target a few standard deviations below the average if you want your product to be at all reliable. Then you have to figure that that lifetime calculation is per cap, not per unit... It adds up. Add in the inflated specs in Chinese data sheets, unforeseeable transient events, base manufacturing error rate, storage and environmental degradation, potential partial failure of other components, and a host of other issues.

You can use cheap capacitors in your products. You can't claim that they are as reliable, or even nearly as reliable, as their Japanese and upper-end Taiwanese equivalents.

[iampoor]

Assuming that capacitor failures approximate a bell curve, designing for the average +/-1 standard deviation will give you a 15% failure rate. Obviously not a good plan.

We target a failure rate <0.5% over the warranty lifetime of the product. A product that gets more than 1% failure rate is considered troublesome, and more than 2.5% is considered awful--this is what we get for the cheap non-80PLUS PSUs we used to sell before we realized warranty replacement costs exceed our sales margin... The worst products we've ever sold only hit about 10% failure rate and are considered abject failures.

Anyway, you want to target a few standard deviations below the average if you want your product to be at all reliable. Then you have to figure that that lifetime calculation is per cap, not per unit... It adds up. Add in the inflated specs in Chinese data sheets, unforeseeable transient events, base manufacturing error rate, storage and environmental degradation, potential partial failure of other components, and a host of other issues.

You can use cheap capacitors in your products. You can't claim that they are as reliable, or even nearly as reliable, as their Japanese and upper-end Taiwanese equivalents.

Hey Phaedrus

Not to hiajk this thread, but were you one of the stuff at overclock.net who did all the PSU reviews?

[poorchava]

One word: Seasonic.

(performance+lifetime)/price approaches infinity. Look for teardowns of those: you won't find any capacitor brand other than Chemicon, Rubycon, Panasonic, Nichicon. Brand name parts, even fans are SanyoDenki.

Actually, I think all of the Corsair and Antec PSUs from "higher" end are also made by Seasonic 

[TMM]

Actually, I think all of the Corsair and Antec PSUs from "higher" end are also made by Seasonic 

Used to be, now they just seem to go with whatever OEM is the cheapest at the time.
http://whirlpool.net.au/wiki/psu_manufacturers
Delta is reasonable. FSP is 'okay', don't expect them to last 5+ years. I dunno what CWT are like these days. They made the Antec TruePower(1) series which was popular around release but after about 5 years they all failed from bad caps.

FWIW i have had a Seasonic fail and take some harddrives with it but still regard them as generally well built supplies.

[Phaedrus]

Assuming that capacitor failures approximate a bell curve, designing for the average +/-1 standard deviation will give you a 15% failure rate. Obviously not a good plan.

We target a failure rate <0.5% over the warranty lifetime of the product. A product that gets more than 1% failure rate is considered troublesome, and more than 2.5% is considered awful--this is what we get for the cheap non-80PLUS PSUs we used to sell before we realized warranty replacement costs exceed our sales margin... The worst products we've ever sold only hit about 10% failure rate and are considered abject failures.

Anyway, you want to target a few standard deviations below the average if you want your product to be at all reliable. Then you have to figure that that lifetime calculation is per cap, not per unit... It adds up. Add in the inflated specs in Chinese data sheets, unforeseeable transient events, base manufacturing error rate, storage and environmental degradation, potential partial failure of other components, and a host of other issues.

You can use cheap capacitors in your products. You can't claim that they are as reliable, or even nearly as reliable, as their Japanese and upper-end Taiwanese equivalents.

Hey Phaedrus

Not to hiajk this thread, but were you one of the stuff at overclock.net who did all the PSU reviews?

Yes

[VK3DRB]

You close your electronics factories, sack your neighbours, lose your skills and get your stuff made in a third world country like China. Don't expect good quality. Or sympathy.