Did a search, but I came up empty-handed....so maybe someone else is thinking this and just hasn't asked.
Maybe I should have re-worded the topic title?
I am wanting to know about the life of a tube.....does sonic quality diminish over a period of time?
And is there a 'lifetime' for a tube where it should be replaced? 5,000hrs, 7,500hrs, 10,000hr, or the thing is going to out-last me! I listen to my Adcom 555's(they are bi-amping my little Polk Monitor 40's) no less then 40hr per week, and when the weather is nasty....that number can be doubled!
Just wanting to know.... Still going to build one, but I'd like to be prepared for the 'if-n-when'!!
Thanks much!
Scott (there's a vaccuum between my ears too) B
Maybe I should have re-worded the topic title?
I am wanting to know about the life of a tube.....does sonic quality diminish over a period of time?
And is there a 'lifetime' for a tube where it should be replaced? 5,000hrs, 7,500hrs, 10,000hr, or the thing is going to out-last me! I listen to my Adcom 555's(they are bi-amping my little Polk Monitor 40's) no less then 40hr per week, and when the weather is nasty....that number can be doubled!
Just wanting to know.... Still going to build one, but I'd like to be prepared for the 'if-n-when'!!
Thanks much!
Scott (there's a vaccuum between my ears too) B
My experience has been small signal tubes lest about 5-7yrs when run for about 20hrs a day (6x356x20=42Khrs) power valves under the same conditions last between 1-2yrs (10K hrs).
Of course it all depends on how you heat them (I like to underheat to extend life) and how hard you push them (light duty usually).
Shoog
Of course it all depends on how you heat them (I like to underheat to extend life) and how hard you push them (light duty usually).
Shoog
Like he said.
Some can go longer and some shorter. I have seen "Red October" tubes go far longer than the projected life. It just depends on the quality control and the operating conditions.
Some can go longer and some shorter. I have seen "Red October" tubes go far longer than the projected life. It just depends on the quality control and the operating conditions.
Sounds(no pun intended) like I'll get just enough hours out of a set of tubes to keep me happy until the next build!!
Scott (once you starting building you're never done) B
Scott (once you starting building you're never done) B
To be honest, I don't see why this info appears to be difficult to obtain. Lightbulb manufacturers usually provide an expected lifespan in hours, under nominal conditions. It should be pretty standard for valve manufacturers to provide a lifespan under nominal conditions. Considering that valves have been used in industrial as well as military applications, this data must exist for main stream valve brands (for practical reasons I will make an exception for Chinese brand valves).
It's because the information just isn't there. Just as there is no answer to the question, "How long is a piece of string?", there is no one answer to "How long will a tube last?" The variables that affect it's lifespan are various and several, some which, at least, are in play in any given application.
Datasheets sometimes give an expected lifespan, and is usually about 5000 to 10000 hours. Of course, these numbers are related to the manufacturer's idea of "when if the valve no longer fit for purpose", which was usually in computing circuits. Of course, this figure does not necessarily correspond to when the valves stops working completely, and it could be perfectly good for analogue audio purposes long after the 10000 hours are up.
Back in the 1950s, life tests on ordinary receiving tubes showed that there is a burn-in period of about 1000-5000 hours, where failure rate was usually higher than normal (but also sometimes lower than normal!), and after that the failure rate was exactly exponential, that is, the causes of failure were entirely random. The average life is therefore the time taken for the first 63% of a batch to fail (100%-100%/e), and was about 30,000 hours for common receiving tubs in the 1950s, and is probably the same or better today, thanks to better pumping and computer controlled manufacturing techniques. This agrees nicely with Shoog's observations.
Another interesting point of fact about the exponential failure rate is that if you take two identical valves, one which has been running for, say, 5000 hours, and one which has been running for 50000 hours, the chances them lasting another given period of time are actually equal!
Back in the 1950s, life tests on ordinary receiving tubes showed that there is a burn-in period of about 1000-5000 hours, where failure rate was usually higher than normal (but also sometimes lower than normal!), and after that the failure rate was exactly exponential, that is, the causes of failure were entirely random. The average life is therefore the time taken for the first 63% of a batch to fail (100%-100%/e), and was about 30,000 hours for common receiving tubs in the 1950s, and is probably the same or better today, thanks to better pumping and computer controlled manufacturing techniques. This agrees nicely with Shoog's observations.
Another interesting point of fact about the exponential failure rate is that if you take two identical valves, one which has been running for, say, 5000 hours, and one which has been running for 50000 hours, the chances them lasting another given period of time are actually equal!
Last edited:
JRN: Not true, the analogy to "a piece of string" does not apply to mean time between failures of industrial parts.
I admit I may have made my statement incorrectly. I should not have said "To be honest, I don't see why this info appears to be difficult to obtain.", because maybe it IS difficult to obtain. However, I am convinced the data does exist. Simply because it must.
Manufacturers of any industrial equipment always specify longevity under NOMINAL conditions, because it is an industrial requirement to do so (after all, which sane company would ever buy equipment for industrial applications that cannot be planned for maintenance?).
Please note "under NOMINAL conditions", which means under specified testing conditions used by the manufacturer, which means, specific loading, specific input signals, specific voltages etc (usually what is stated on the spec sheet). How do you think the industry can maintain their equipment if manufacturers just decided to not provide mean time between failure data?
Vacuum tubes have been industrially used for a long time. Therefore, this equipment needed to be maintained. Therefore maintenance schedules have been generated. To do so, information on Mean time between failures MUST be provided by the manufacturer. Therefore the data exists.
I'm not saying the data is easy to obtain, I am saying the data exists.
(Again, with possible exception of chinese made valves.)
I admit I may have made my statement incorrectly. I should not have said "To be honest, I don't see why this info appears to be difficult to obtain.", because maybe it IS difficult to obtain. However, I am convinced the data does exist. Simply because it must.
Manufacturers of any industrial equipment always specify longevity under NOMINAL conditions, because it is an industrial requirement to do so (after all, which sane company would ever buy equipment for industrial applications that cannot be planned for maintenance?).
Please note "under NOMINAL conditions", which means under specified testing conditions used by the manufacturer, which means, specific loading, specific input signals, specific voltages etc (usually what is stated on the spec sheet). How do you think the industry can maintain their equipment if manufacturers just decided to not provide mean time between failure data?
Vacuum tubes have been industrially used for a long time. Therefore, this equipment needed to be maintained. Therefore maintenance schedules have been generated. To do so, information on Mean time between failures MUST be provided by the manufacturer. Therefore the data exists.
I'm not saying the data is easy to obtain, I am saying the data exists.
(Again, with possible exception of chinese made valves.)
Merlin is correct in his rationale. What's more, in the 50's process control was not as developed as it is now. Process control nowadays have a huge effect on controling failure rates. Automation is part of process improvement (as you take out the variable called "human"), and therefore caters for a significant change in process improvement.
Back to the main topic: SDB, have you got a specific tube type in mind? this may make it easier for us to help find the data for you...
Back to the main topic: SDB, have you got a specific tube type in mind? this may make it easier for us to help find the data for you...
I don't dispute your explanation of MTBF being a more-or-less knowable value in controlled conditions. You are quite right, but not in the context of the OP, wherein the term "nominal conditions" is not defined. George (Tubelab) regularly demonstrates (much to our entertainment) that any such figures are hopelessly inaccurate when a tube is subjected to real world (or in George's case, unreal) conditions. This being a forum for tinkerers of similar, if less extreme, "ingenuity", the "piece of string" question is most appropriate.
I would strongly question the contention on this thread that quality control and manufacturing processes in tube manufacturing have actually improved since the 1950/60s - I believe the reverse is true and that generally MTBF for current production tubes is much lower than it was 40 - 60yrs ago. It is very unlikely given the nature of the industries currently manufacturing tubes and where they are generally located that modern statistical process control techniques are in use (cp, cpk, mean, standard deviation, etc) or that any significant portion of the manufacturing process has been automated.
Tube manufacturing has always been a highly labor intensive operation that does not lend itself very well to high levels of automation.
Production Inspection and QC isn't what it was either and rejection criteria are pretty low based on some of the tubes I've purchased.
Time is money and therefore a lot of tubes are manufactured and pumped down as rapidly as possible, quality of vacuum IME in modern power tubes in particular is quite poor compared to vintage practices. I've run into plenty of "soft" (marginally gassy) power tubes. When MPD took over from GE sometime in the 1980s manufacturing GE6550 the level of gas left in those tubes increased rather significantly, so much so that they started to fail in ARC and other amps I was working on at the time. Most current production power tubes IMHO are not as hard as they could/should be.
I'm also surprised at the high hour claims made by some here, not borne out by my experience. I find current production EL34, 6550, 6BQ5 to degrade measurably and audibly typically within a few thousand hours of use, and sometimes less - some may do better. So far the only exception I have found to this has been a couple of pairs of JJ 300B I bought nearly 10yrs ago, at approximately 5K hours per pair these still sound and measure good. (I'm surprised) Small signal tubes I generally replace at 5K hours or so, or when I can hear or measure a reason to replace them. The only exceptions I make are small NOS DHTs, D3A, 6J5G, and 6SN7 which seem to last (much) longer.
YMMV
Tube manufacturing has always been a highly labor intensive operation that does not lend itself very well to high levels of automation.
Production Inspection and QC isn't what it was either and rejection criteria are pretty low based on some of the tubes I've purchased.
Time is money and therefore a lot of tubes are manufactured and pumped down as rapidly as possible, quality of vacuum IME in modern power tubes in particular is quite poor compared to vintage practices. I've run into plenty of "soft" (marginally gassy) power tubes. When MPD took over from GE sometime in the 1980s manufacturing GE6550 the level of gas left in those tubes increased rather significantly, so much so that they started to fail in ARC and other amps I was working on at the time. Most current production power tubes IMHO are not as hard as they could/should be.
I'm also surprised at the high hour claims made by some here, not borne out by my experience. I find current production EL34, 6550, 6BQ5 to degrade measurably and audibly typically within a few thousand hours of use, and sometimes less - some may do better. So far the only exception I have found to this has been a couple of pairs of JJ 300B I bought nearly 10yrs ago, at approximately 5K hours per pair these still sound and measure good. (I'm surprised) Small signal tubes I generally replace at 5K hours or so, or when I can hear or measure a reason to replace them. The only exceptions I make are small NOS DHTs, D3A, 6J5G, and 6SN7 which seem to last (much) longer.
YMMV
Jrn, I totally agree with you there. Any mtbf is pretty useless if the teSt conditions are not known.
As for quality control on tubes. If that has gone down hill, I expect that to be due to a shift from industrial use to More luxurious use (audio applications). However, I was previously refering to the general industry, rather than tube specific.
As for quality control on tubes. If that has gone down hill, I expect that to be due to a shift from industrial use to More luxurious use (audio applications). However, I was previously refering to the general industry, rather than tube specific.
Everyone says that, but I think too many are unfairly comparing modern tubes with the likes of Mullard / RCA etc. But there were plenty more 'cheap' tubes made in the 50s, in India, Korea, China and so on, which were probably no better than today's offerings.
Again, they always have been. The acceptable gm tolerances were -40% + 20% even in the 1960s. It just so happened that Mullard/RCA etc were better at producing large quantities of spot-on tubes than the rest. (And those that weren't quite so up to scratch were sold off and re-labelled under cheaper brands- sneaky!)
But that's still within the normal burn-in time. You have to expect valves to change measurably during that time, now as well as in the 1960s.
Last edited:
Some newer Output valves, like the Golden Lion KT88 reissue (made in Saratov, Russia) certainly do improve the quality of pumping and quality control generally, and these new valves give high confidence in their lifetime and reliability.
But otherwise, the quality picture is mostly degraded compared to the 1960s, as Kevin has said. This is because the big customers (ie guitar amp makers) pressure them about cost. Saratov can make good valves, as the Lions show, but if your cost is under pressure, you must reject only the really bad production samples, AND pump them only for the minimum time you can get away with.
If you don't believe this, just try operating a new valve at high voltage & normal current, and check the grid current - your indicator of gas.
As for power valve lifetime, just check out the Soviet era Saratov data sheet for the 6L6Gx family equivalent, the 6П3С-Е.
This is a 6L6Gx/5881 equivalent with a design-lifetime of 5000 hours. But underneath that line, it tells you that there's a 90% chance it will do 10000 hours. Cool.
However, to get such a long life, you must respect the operating conditions that qualify the lifetime figure. From the 2-columns at the bottom left you can see the conditions that give 500 hours, and the conditions that give 5000. They're not very different!
In each case, you must keep Va & Vg2 to 250V. but increasing bulb temperature Температура баллона в наболее from 180 to 210 deg C, running high power in the screen: мошност, рассеиваемая, второи сеткои at 2,75W in lieu of 2, or slackening tolerance on the heater voltage Напряжение накана from 6..6,3V to 5,7..7,0V degrades life expectation from 5000 down towards 500 hours.
Given that guitar amps run these at about 500V and near to 100% of rated anode/screen dissipation, is it any wonder that they last only a few hundred hours? and why pro stage guitarists change them twice or more per year? And since ac-heated 6L6s will vary in voltage due to line voltages, how many can keep their mains to 5% accuracy?
So, if you have some nice old valves and you need them to last a long time, run the voltage and power levels LOW. 60 to 80% of rated maxima on each. Use current sources for heating, and ventilate the bulbs fully, or use fan cooling. Like Kevin, I've run many thousands of hours on old JJ 300Bs, and have followed these rules. In a guitar amp, that ignores the rules (it has to, to sound right) expect bad reliability.
When you design an amplifier, you sometimes have to decide whether it sound quality, power output, or reliability that comes first. Like many things in life, you can't always have everything.
But otherwise, the quality picture is mostly degraded compared to the 1960s, as Kevin has said. This is because the big customers (ie guitar amp makers) pressure them about cost. Saratov can make good valves, as the Lions show, but if your cost is under pressure, you must reject only the really bad production samples, AND pump them only for the minimum time you can get away with.
If you don't believe this, just try operating a new valve at high voltage & normal current, and check the grid current - your indicator of gas.
As for power valve lifetime, just check out the Soviet era Saratov data sheet for the 6L6Gx family equivalent, the 6П3С-Е.
This is a 6L6Gx/5881 equivalent with a design-lifetime of 5000 hours. But underneath that line, it tells you that there's a 90% chance it will do 10000 hours. Cool.
However, to get such a long life, you must respect the operating conditions that qualify the lifetime figure. From the 2-columns at the bottom left you can see the conditions that give 500 hours, and the conditions that give 5000. They're not very different!
In each case, you must keep Va & Vg2 to 250V. but increasing bulb temperature Температура баллона в наболее from 180 to 210 deg C, running high power in the screen: мошност, рассеиваемая, второи сеткои at 2,75W in lieu of 2, or slackening tolerance on the heater voltage Напряжение накана from 6..6,3V to 5,7..7,0V degrades life expectation from 5000 down towards 500 hours.
Given that guitar amps run these at about 500V and near to 100% of rated anode/screen dissipation, is it any wonder that they last only a few hundred hours? and why pro stage guitarists change them twice or more per year? And since ac-heated 6L6s will vary in voltage due to line voltages, how many can keep their mains to 5% accuracy?
So, if you have some nice old valves and you need them to last a long time, run the voltage and power levels LOW. 60 to 80% of rated maxima on each. Use current sources for heating, and ventilate the bulbs fully, or use fan cooling. Like Kevin, I've run many thousands of hours on old JJ 300Bs, and have followed these rules. In a guitar amp, that ignores the rules (it has to, to sound right) expect bad reliability.
When you design an amplifier, you sometimes have to decide whether it sound quality, power output, or reliability that comes first. Like many things in life, you can't always have everything.
Attachments
Last edited:
Merely automating something has NOTHING to do with employing well designed (meaning considered/measured) SPC processes. Secondly, merely employing SPC will not make the product "better" or more "reliable". It certainly may increase repeatability of end product output. In other words, simply because an end product may have been produced to meet a specification AND within 6 sigma tolerance bands has nothing to do with the questions of original product design suitability or optimization to end use. These are totally different realms. Ask Dick Sequerra, or Nelson Pass, John Curl, SY, etc.....
I can make beautifully consistent scrap that is 6 sigma stable!!!!! Perfect... " it meets the specification.....".
Kinda' sounds like a certain bunch habitaiting Capitol cities everywhere in the world!!!😱
Last edited:
My comments on statistical process control techniques probably needs to be clarified, you can use these techniques to vet the robustness of the manufacturing and quality process through burn in and testing, using statistics to determine whether or not you are building product that meets the criteria you have selected. (Which could be anything you choose to measure over a sufficient sample size) I doubt there is much effort going on in this area due to the additional time and cost involved in doing this.
Tube life
I've got a lot of working hifi and Hammond organ signal tubes that are 40 or 50 years old. Output tubes in the Hafler Ultralinear circuit, would wimp out (produce low wattage) after about 7-10 years of 5 evenings a week use. Also rectifier tubes wimp out, get high resistance and cause low power amp wattage after 7 - 10 years. This organ I just bought, manuf 1968, has high output 7591's and 6BQ5's, a new Sovtek 5AR4 didn't even pump up the volume, put the old one back in. All the electrolytic capacitors were landfill ready, however, replaced them all except the harp circuit (115 ea. expensive 15 mf @ 20 +-10%). Hoping the new Soviet and Slovak tubes are seriously manufactured.
TV's of the fifties ran their tubes harder, I changed about half of the tubes in a 1954 Philco B&W TV in the late sixties.
I've got a lot of working hifi and Hammond organ signal tubes that are 40 or 50 years old. Output tubes in the Hafler Ultralinear circuit, would wimp out (produce low wattage) after about 7-10 years of 5 evenings a week use. Also rectifier tubes wimp out, get high resistance and cause low power amp wattage after 7 - 10 years. This organ I just bought, manuf 1968, has high output 7591's and 6BQ5's, a new Sovtek 5AR4 didn't even pump up the volume, put the old one back in. All the electrolytic capacitors were landfill ready, however, replaced them all except the harp circuit (115 ea. expensive 15 mf @ 20 +-10%). Hoping the new Soviet and Slovak tubes are seriously manufactured.
TV's of the fifties ran their tubes harder, I changed about half of the tubes in a 1954 Philco B&W TV in the late sixties.
kevinkr,
forgive my snippy and pedantic comment. I well knew your tack.
Your point about additional time cost is of course exactly the crux of the problem.
No cost was too great when the end customer was either a military or high reliability cost no object client, and or the product was at the nexus of state of the electronic art.
Alas, we are now relegated to floor sweepings status by virtue of the global size of the market for our beloved tubes/vales. Now darn surely if the Blackberry or I-think-a-majiggy in ones pocket was vacuum tube driven .......... well you know!!!!!
Anyway, this is an interesting thread on a subject not often fleshed out that we may think about our ability to keep our much loved equipment running and at what cost.
Thank you all for making the topic breathe.
We only just stopped using CRTs. You could expect a colour TV to last 5 years, 8 hours a day, easy. 15,000 hours. 10 years was not that uncommon. Often it wasn't the tube that failed.
I believe it was not uncommon for small signal devices to survive 100,000 hours+, 'course that's nearly 12 years continuous.
The data you want was collated, at least in some cases. In the past, manufacturers held some information much closer to their chests than they do now. Much, I fear, has been lost in the mists of time. Some undoubtedly still survives in paper form, but who knows for how long...
The single parameter most commonly identified as having an adverse affect on service life is bulb temperature.
w
Even now it can be hard to get information. Recently I was curious about some surface mount analog switches (intended for mobile phone use) which look to have great performance, but the manufacturer is vetting who they permit to download datasheets. If you're not a mobile phone manufacturer, forget it.
I believe it was not uncommon for small signal devices to survive 100,000 hours+, 'course that's nearly 12 years continuous.
The data you want was collated, at least in some cases. In the past, manufacturers held some information much closer to their chests than they do now. Much, I fear, has been lost in the mists of time. Some undoubtedly still survives in paper form, but who knows for how long...
The single parameter most commonly identified as having an adverse affect on service life is bulb temperature.
w
Even now it can be hard to get information. Recently I was curious about some surface mount analog switches (intended for mobile phone use) which look to have great performance, but the manufacturer is vetting who they permit to download datasheets. If you're not a mobile phone manufacturer, forget it.
- Status
- Not open for further replies.