MTBF Links
Hey all,
I've been following this thread mostly through SPCR but saw MColeg's post here via a link and wanted to contribute. I've been toying with the idea of building or modding a small mATX based case and doing some research on PSUs. No overclocking, I'm afraid, just a high end CPU and medium graphics in a small, stable, and quiet package. Funny thing is that even low-power systems will run into the same issues of rails, ripple, reliability, etc. when you start talking about smaller SFX or mini-1u supplies. Anyway, here's some data and links in response to some of MColegs questions:
1. Lifespan numbers for a PSU?
Most PC PSU's are horribly labeled, but the better models at least provide MTBF numbers. I think Seasonic and Enhance are usually rated for 100,000 hours, for example. However, I'm not sure that's what you mean. If you're looking for ratings for specific components that go into generating those MTBF estimates, you'll either need to talk to the manufacturers or perhaps check out the relevant guides. Most industrial/telecom power supplies are rated against either MIL-HDBK-217F or Bellcore, TR-332, Issue 6. You can find a copy of the MIL handbook with the latest change notices at SQC Online (http://www.sqconline.com/download/). They also have an on-line MTBF calculator using a MIL-HDBK-217F model
(http://www.sqconline.com/reliability/index.html)
2. What's ambient?
Computer PSU data is really hard to come by. If you look at the telecom PSU's, even the short-form data sheets provide substantially more information. For example, N2 Power's XL-275 is rated at 200,000 hourse, continuous operation, maximum output loading, worst case input voltage According to Bellcore, TR-332, Issue 6 (http://www.n2power.com/705123 XL275.pdf), and they provide specific data about temperature and deratings. XP Power rates their EMA212 at 400,000 hours MTBF by the MIL-HDBK-217F (http://www.xppower.com/pdfs/SF_EMA212.pdf). They also make particular mention of temperature requirements.
Assuming your typical computer PSU is calculating MTBF using the one of these same two models (probably a safe bet, but who knows?), my guess is that the temperature requirements would refer to air temps inside the PSU, which are going to be a function of the air temperatures drawn into the PSU to cool it. I'm betting they calculated their MTBF estimates assuming that the PSU was functioning in a 25 degree environment, so if your PSU is using 30C air from inside the computer case to cool itself, kiss the MTBF estimate goodbye.
3. MTTF or MTBF?
If you're looking for some general takes on the utility of MTBF or MTTF and what they mean, you might check out the following:
http://www.evaluationengineering.com/archive/articles/1100reliability.htm
http://www.evaluationengineering.com/archive/articles/0500analyz.htm
http://www.reliasoft.com/newsletter/2Q2000/mttf.htm
Six Sigma SPC also provides a nice definition, which I'm posting here
(from http://www.sixsigmaspc.com/dictionary/MTTF-meantimetofailure.html)
"A estimate of the average, or mean time until a design's or component's first failure, (you may not want to include external failures), or disruption in the operation of the product, process, procedure, or design occurs. Mean time until a failure assumes that the product CAN NOT be repaired and the product CAN NOT resume any of it’s normal operations. MTTF is related to items such as expected and/or operating life or other items that in general are not fixed or replacement even though it sometimes may be.
In a lot of design and components, MTTF it is especially near to the MTBF, (Mean Time Between Failures), which typically is a small amount longer than MTTF. This is due to the fact that MTBF includes the repair time of the designs or components. If a design or component works for an extended time, then it fails, is repaired in a reasonable amount of time, and then once again works for an extended time, the MTTF is the average, or mean, of the amount of time that it is in operational condition. The MTBF is the average time between failures to include the average repair time, or MTTR.
MTTR is not the correct tool for all applications, but is a useful measurement of reliability."
As to why MTBF, I'm not sure. The two may not be all that different, and MTBF may simply be the norm for multi-component equipment even though no one really repairs them as you point out.
Sorry I've got no clue as to the current thing.
Anyway, hope this helps and keep up the good work. I think the lengths you've gone to collect data and experiment are absolutely fantastic, and good on you for demanding that others back up their statements with data as well.
Cheers,
Zinj
Hey all,
I've been following this thread mostly through SPCR but saw MColeg's post here via a link and wanted to contribute. I've been toying with the idea of building or modding a small mATX based case and doing some research on PSUs. No overclocking, I'm afraid, just a high end CPU and medium graphics in a small, stable, and quiet package. Funny thing is that even low-power systems will run into the same issues of rails, ripple, reliability, etc. when you start talking about smaller SFX or mini-1u supplies. Anyway, here's some data and links in response to some of MColegs questions:
1. Lifespan numbers for a PSU?
Most PC PSU's are horribly labeled, but the better models at least provide MTBF numbers. I think Seasonic and Enhance are usually rated for 100,000 hours, for example. However, I'm not sure that's what you mean. If you're looking for ratings for specific components that go into generating those MTBF estimates, you'll either need to talk to the manufacturers or perhaps check out the relevant guides. Most industrial/telecom power supplies are rated against either MIL-HDBK-217F or Bellcore, TR-332, Issue 6. You can find a copy of the MIL handbook with the latest change notices at SQC Online (http://www.sqconline.com/download/). They also have an on-line MTBF calculator using a MIL-HDBK-217F model
(http://www.sqconline.com/reliability/index.html)
2. What's ambient?
Computer PSU data is really hard to come by. If you look at the telecom PSU's, even the short-form data sheets provide substantially more information. For example, N2 Power's XL-275 is rated at 200,000 hourse, continuous operation, maximum output loading, worst case input voltage According to Bellcore, TR-332, Issue 6 (http://www.n2power.com/705123 XL275.pdf), and they provide specific data about temperature and deratings. XP Power rates their EMA212 at 400,000 hours MTBF by the MIL-HDBK-217F (http://www.xppower.com/pdfs/SF_EMA212.pdf). They also make particular mention of temperature requirements.
Assuming your typical computer PSU is calculating MTBF using the one of these same two models (probably a safe bet, but who knows?), my guess is that the temperature requirements would refer to air temps inside the PSU, which are going to be a function of the air temperatures drawn into the PSU to cool it. I'm betting they calculated their MTBF estimates assuming that the PSU was functioning in a 25 degree environment, so if your PSU is using 30C air from inside the computer case to cool itself, kiss the MTBF estimate goodbye.
3. MTTF or MTBF?
If you're looking for some general takes on the utility of MTBF or MTTF and what they mean, you might check out the following:
http://www.evaluationengineering.com/archive/articles/1100reliability.htm
http://www.evaluationengineering.com/archive/articles/0500analyz.htm
http://www.reliasoft.com/newsletter/2Q2000/mttf.htm
Six Sigma SPC also provides a nice definition, which I'm posting here
(from http://www.sixsigmaspc.com/dictionary/MTTF-meantimetofailure.html)
"A estimate of the average, or mean time until a design's or component's first failure, (you may not want to include external failures), or disruption in the operation of the product, process, procedure, or design occurs. Mean time until a failure assumes that the product CAN NOT be repaired and the product CAN NOT resume any of it’s normal operations. MTTF is related to items such as expected and/or operating life or other items that in general are not fixed or replacement even though it sometimes may be.
In a lot of design and components, MTTF it is especially near to the MTBF, (Mean Time Between Failures), which typically is a small amount longer than MTTF. This is due to the fact that MTBF includes the repair time of the designs or components. If a design or component works for an extended time, then it fails, is repaired in a reasonable amount of time, and then once again works for an extended time, the MTTF is the average, or mean, of the amount of time that it is in operational condition. The MTBF is the average time between failures to include the average repair time, or MTTR.
MTTR is not the correct tool for all applications, but is a useful measurement of reliability."
As to why MTBF, I'm not sure. The two may not be all that different, and MTBF may simply be the norm for multi-component equipment even though no one really repairs them as you point out.
Sorry I've got no clue as to the current thing.
Anyway, hope this helps and keep up the good work. I think the lengths you've gone to collect data and experiment are absolutely fantastic, and good on you for demanding that others back up their statements with data as well.
Cheers,
Zinj
Thanks for the great information, I bet it will be mind clearing for some of us, specially for mcoleg. On Ambient, basing myself entirely on common sense, I would assume it means the hottest air supply, weather if its from the inside or the outside (although the last one is really unlikely, and if it happens then you must be doing something wrong (or Extreme cooling?)).
