reliability bathtub curve
I teach some reliability courses at the university and also have worked extensively in the field of microelectronics at IMEC (Belgium) dealing with reliability of microelectronic devices. Just for your information.
The speed at which components fail is described with the bathtub curve (google for graph) . Initially, there are some products failing for some difficult to understand reason. We call that infant mortality or extrinsic failure (= your new car breaks down after a few miles) and are process, production related. After a given time there is a time window where the speed at which components fail is rather constant. We call that the operational life region. Next to this period comes the wearout period where the speed of failure is increasing. This is the region that you need to avoid! Don't invest money in a car that reached that region altough it runs fine because you will be faced with predictable failures in a short time span. In general, the trick is to weed out the early failures. Manufacturers achieve that by doing a burn in on your components at accelerated environmental conditions. As such, failures take place at a higher speed and bad components can be removed within a relative short time span. The customer can start at the beginning of the operational life as it should be. These failures are called intrinsic failures and the speed of failure is of course very low and you will be a happy user.
Kind regards, //WDC
www.monolithmagnetics.com
I teach some reliability courses at the university and also have worked extensively in the field of microelectronics at IMEC (Belgium) dealing with reliability of microelectronic devices. Just for your information.
The speed at which components fail is described with the bathtub curve (google for graph) . Initially, there are some products failing for some difficult to understand reason. We call that infant mortality or extrinsic failure (= your new car breaks down after a few miles) and are process, production related. After a given time there is a time window where the speed at which components fail is rather constant. We call that the operational life region. Next to this period comes the wearout period where the speed of failure is increasing. This is the region that you need to avoid! Don't invest money in a car that reached that region altough it runs fine because you will be faced with predictable failures in a short time span. In general, the trick is to weed out the early failures. Manufacturers achieve that by doing a burn in on your components at accelerated environmental conditions. As such, failures take place at a higher speed and bad components can be removed within a relative short time span. The customer can start at the beginning of the operational life as it should be. These failures are called intrinsic failures and the speed of failure is of course very low and you will be a happy user.
Kind regards, //WDC
www.monolithmagnetics.com
capacitor sound check
Think of the following experiment that you could conduct. Start with 3 identical pairs of capacitors
Pair 1 : as manufactured
Pair 2: a surviver of the burn-in test
Pair 3: as manufactured
Compare 2/1 and also 3/1 top to be sure that identical caps sound identical to your ears.
I have conducted reliability tests on film capacitors very very long time ago. We monitored the leakage current as a function of time at max voltage at high tempearture. The survivers of this test had their leakage current decreasing. I suppose this could mean that they sound different form the original caps. Perhaps I am able to dig up these results if you were interested.
Kind rergards, //WDC
www.monolithmagnetics.com
Think of the following experiment that you could conduct. Start with 3 identical pairs of capacitors
Pair 1 : as manufactured
Pair 2: a surviver of the burn-in test
Pair 3: as manufactured
Compare 2/1 and also 3/1 top to be sure that identical caps sound identical to your ears.
I have conducted reliability tests on film capacitors very very long time ago. We monitored the leakage current as a function of time at max voltage at high tempearture. The survivers of this test had their leakage current decreasing. I suppose this could mean that they sound different form the original caps. Perhaps I am able to dig up these results if you were interested.
Kind rergards, //WDC
www.monolithmagnetics.com
Those who want to believe in burn-in and understand circuit theory should consider how caps are used in circuits and estimate what sort of change might make a difference. For example, a coupling cap could affect LF but not HF (assuming that, like most caps, it functions well up to RF frequencies). A small shunt cap (e.g. for equalisation or loop stability) can affect HF but not LF. If you do this you will find that most claims are simply not possible.
You can then extend the process by considering stray capacitance. Here it may be found that the physically large caps favoured by 'audiophiles' can sometimes create problems due to strays: hum pickup, noise pickup, HF instability, even HF rolloff. They may sound different, but in most cases it will be a degradation not improvement.
What if caps really do need burn-in? What is happening? Well, it won't be the metal film itself, as major easily measurable changes in that won't disturb most audio circuits but would show up first in RF. It could be the connections from the film to the wires (cold-weld?) if the manufacturer has weak quality control and routinely makes poor joints, but again this would show up in RF circuits as extra losses or intermodulation. It could be something (plasticiser?) migrating within the dielectric - but that would only affect LF and subsonics.
Something that audio people need to take on board is that we mainly use the same type of components as other much more critical applications of electronics, and they don't have this problem. The sort of issues which audiophiles claim to be a problem would render communications receivers unusable if they were real. There it is possible to have aN IP3 spurious-free dynamic range of over 90dB (roughly corresponds to 0.003% 3rd order distortion at max signal) and even then it is known that the limits are imposed by mixers and crystal filters, not capacitors and resistors.
You can then extend the process by considering stray capacitance. Here it may be found that the physically large caps favoured by 'audiophiles' can sometimes create problems due to strays: hum pickup, noise pickup, HF instability, even HF rolloff. They may sound different, but in most cases it will be a degradation not improvement.
What if caps really do need burn-in? What is happening? Well, it won't be the metal film itself, as major easily measurable changes in that won't disturb most audio circuits but would show up first in RF. It could be the connections from the film to the wires (cold-weld?) if the manufacturer has weak quality control and routinely makes poor joints, but again this would show up in RF circuits as extra losses or intermodulation. It could be something (plasticiser?) migrating within the dielectric - but that would only affect LF and subsonics.
Something that audio people need to take on board is that we mainly use the same type of components as other much more critical applications of electronics, and they don't have this problem. The sort of issues which audiophiles claim to be a problem would render communications receivers unusable if they were real. There it is possible to have aN IP3 spurious-free dynamic range of over 90dB (roughly corresponds to 0.003% 3rd order distortion at max signal) and even then it is known that the limits are imposed by mixers and crystal filters, not capacitors and resistors.
He's Back 
Tomchr - I looked at your site which indicates great hands-on knowledge and experience. It does prompt me to ask a rather simple question.
With all that high level equipment at your disposal - are you saying that in testing a cap, say this one, as delivered by the postman, will show no change on any of the readouts with any test of your choice, any piece of equipment - between the first application of electricity (fresh state) and subsequent readings after a sustained/prolonged application of a current ("burn-in" or "forming").
This is not a challenge or a disagreement - just a straight forward question based on awareness of your real world lab background.
Tomchr - I looked at your site which indicates great hands-on knowledge and experience. It does prompt me to ask a rather simple question.
With all that high level equipment at your disposal - are you saying that in testing a cap, say this one, as delivered by the postman, will show no change on any of the readouts with any test of your choice, any piece of equipment - between the first application of electricity (fresh state) and subsequent readings after a sustained/prolonged application of a current ("burn-in" or "forming").
This is not a challenge or a disagreement - just a straight forward question based on awareness of your real world lab background.
Maybe you're over-simplifing...
This is completely true only if you consider 'ideal' capacitors...
A real world capacitor has an impedance (frequency dependant and not linear) and Dielectric absorption that modify (sligtly) the signal.
All this reasoning is impaired by the preceding simplification, IMHO.
Who said this is a problem?
Again, who said this is a problem?
What we call burn-in simply imply that a capacitor in the first tens of hours of use slightly changes his behaviour until it stabilize.
This has nothing to do with stabilty or correct operational behaviour...
Only a really poor design woud be affected by such small variations.
Last edited:
SY:
Noted (but don't be surprised if in 6 months time I have forgotten!).
Put a non-ideal cap in a typical circuit. Estimate what degree of problem would be noticeable.
DA only affects subsonics, but it often gets dragged into the argument - I'm not sure why.
Noted (but don't be surprised if in 6 months time I have forgotten!).
No, maybe I did not explain myself properly. My point is that a non-ideal capacitor normally appears in circuit positions where only gross non-ideality would be an issue. Coupling capacitor non-linearity can only affect low LF and subsonics. Coupling capacitor series resistance could only affect HF if large enough to be easily measurable and feeding a highish capacitance too.ClaveFremen said:
Put a non-ideal cap in a typical circuit. Estimate what degree of problem would be noticeable.
DA only affects subsonics, but it often gets dragged into the argument - I'm not sure why.
I thought you, and others, did? If there is no problem, then what is break-in for? If there is no problem, what are 'audiophile' components for? If there is no problem, then just use any reasonable quality cap with an appropriate dielectric and solder it straight into the circuit.
Changes what aspect of its behaviour? Given that gross changes in behaviour (e.g. doubling/halving capacitance) would not make much difference in most circuit applications, what change is taking place and how does this affect the circuit?
Last edited:
We 'break in' both caps and even entire amplifiers, because we have to, in order to make them sound their best. It is a reality for designers like me, who hate the very high effort that it takes to do so. It is NOT a marketing tool. In fact, IF we can, we design out any serious 'break-in' by alternate component selection, if practical and possible.
To shorten 'break-in' we might cryo the more sensitive components before assembly into the final product AND NOT tell the lay public or our competitors what we are doing.
This is the reality of serious hi end design that I live with day by day.
To shorten 'break-in' we might cryo the more sensitive components before assembly into the final product AND NOT tell the lay public or our competitors what we are doing.
This is the reality of serious hi end design that I live with day by day.
Hi DF96,
IMHO a capacitor has a (maybe small) effect on signal even outisde his designed working area (i.e. in case of a filter in the 'unfiltered part of the signal).
Dielectric absoption has a role, IMHO, whenever the signal passing through the cap is not symmetrical (in such case DA voltage would tend to be 0 in average).
Maybe we misunderstood....
Really, burn-in it's not a problem, just a thing we consider when subjectively evaluating devices and/or parts, for a fair evaluation/comparison we wait some tens of hours of use.
Thats it!
I can't see no harm in this.
Obviuosly there are also some snake oil vendors that try to fool people stating that their gear sounds bad because it's still in the burn-in phase and they must wait 1000 hours more to judge the device...
IMHO a capacitor has a (maybe small) effect on signal even outisde his designed working area (i.e. in case of a filter in the 'unfiltered part of the signal).
Dielectric absoption has a role, IMHO, whenever the signal passing through the cap is not symmetrical (in such case DA voltage would tend to be 0 in average).
Maybe we misunderstood....
Really, burn-in it's not a problem, just a thing we consider when subjectively evaluating devices and/or parts, for a fair evaluation/comparison we wait some tens of hours of use.
Thats it!
I can't see no harm in this.
Obviuosly there are also some snake oil vendors that try to fool people stating that their gear sounds bad because it's still in the burn-in phase and they must wait 1000 hours more to judge the device...
you seem to have jumped on to an example about which you have taken almost no time to research.
Which of these change with use when in circuit, but do not change on the shelf?
How much do these change with use?
How much do these changes affect the performance?
Hi Andrew,
it was not directly related to burn-in.
If I understood correctly DF96 stated that filter caps were pretty transparent in the unfiltered part of the signal, so an hypothetical variation due to burn-in would have been null.
I didn't agree with him... IMHO, DA and impedance (frequency dependant) apply on the whole signaland not only to the filtered part.
it was not directly related to burn-in.
If I understood correctly DF96 stated that filter caps were pretty transparent in the unfiltered part of the signal, so an hypothetical variation due to burn-in would have been null.
I didn't agree with him... IMHO, DA and impedance (frequency dependant) apply on the whole signaland not only to the filtered part.
Any signal detectable by human ear, is detectable with microphone. The converse is not true.
I would be truly excited to be proven wrong with double blind testing.
Audible difference after dribbling electrons on and off capacitor plates?
I'm taking a deep breath now, but somebody else is has to hold it for me.
I would be truly excited to be proven wrong with double blind testing.
Audible difference after dribbling electrons on and off capacitor plates?
I'm taking a deep breath now, but somebody else is has to hold it for me.
I've seen posts referring to hi-rel components being screened via burn-in thermal cycling, dwv, absorption, etc, for reasons of infant mortality and reliability.
But I havent seen mentioned that these components dont just have to meet spec after the screening, they have to show that their parameters did not change substantially when comparing before and after screening parameters (much tighter than their absolute specifications).
Capacitors do have high internal electric fields and impurities and can / do eventually fail (or age) but its really hard to imagine why they would start to sound (perform) better.
Thanks
-Antonio
But I havent seen mentioned that these components dont just have to meet spec after the screening, they have to show that their parameters did not change substantially when comparing before and after screening parameters (much tighter than their absolute specifications).
Capacitors do have high internal electric fields and impurities and can / do eventually fail (or age) but its really hard to imagine why they would start to sound (perform) better.
Thanks
-Antonio
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.