Why are 6V6G(T)'s always black inside?

[fdegrove]

Hi,

I don't see how that would happen.

Neither did it to me but I'll try to find time to reread the article tomorrow and will post the translated content here.

Certainly can be, often was, there is a US JAN spec-book for tube MTBF.

It would be great if you have a reference somewhere for this, as I've been told over and over that there's just no way to predict tube failure, by people who should know.

If they're wrong than so was I.

Cheers,

[PRR]

> there's just no way to predict tube failure, by people who should know.

I'm holding a tube in my hand. There is no way to know when it will fail. Could be the next time I power-up, could be 10,000 hours from now.

I run a computer that uses 10,000 tubes. I do not know which tubes will fail, but I can bet the computer will need about 2 spare tubes every hour that it runs.(*) If the computer runs 24/7, and the tube-delivery truck comes once a month, I need to order about 1,500 tubes for each delivery.

OK, tube computers are all retired. However the USA B-52H bomber still uses electronic systems with vacuum tubes, and there are quite a few in service and expected to stay in service for years. How many spares should they have? If they can't find enough old-stock spares in warehouses, and have to contract a special-order at a tube factory, how many should they make? MTBF is for questions like that.

MTBF is not meant to say when "this" item will fail. It is a guide to how many items of a large population will fail in a given time.

You also have to know WHY things fail.

The tires on vehicles running lightly-loaded on smooth clean pavement have a very predictable wear-out time. You can replace by odometer reading slightly before they wear-out. Large well lubricated heavily loaded bearings also have very predictable lives. (However you also have the be sure each unit was made the same: a bubble in the rubber or a ding in the bearing could cause early failure.)

Incandescent light bulbs are a classic "random failure". If they could be made perfect, the filament would evaporate at a constant rate. Failure comes either when the filament erodes so thin that light production falls, or evaporated filament darkens the glass so much that light output drops. But in reality, filaments are pressed out of powdered metal. If the powder and pressing were perfect, the filament would be perfect and evaporate uniformly. But the powder and the pressing are imperfect. There are places where the powder is not making good contact one grain to the next, there is microscopic overheating, rapid local evaporation, a weak-spot, and breakage long before the rest of the filament is thinned. The imperfections are very random: some lamps fail on factory test, some in a few hours, some not for 2,000 hours. The overall trend is a bunch of early deaths due to manufacturing flaws, a long period of fairly random failures leading to 50% dead in 1,000 hours and 99% dead after 2,000 hours. One in a million will live 10,000 hours and get dim before it breaks, so there is a rising rate of failure after 2,000 hours, but 99% of lamps never get that far so who cares? The period from about 1 hour to 2,000 hours is constant failure of "all" lamps, so the average life is rated 1,000 hours. There is no way to know if "this" lamp will be a 2-hour or a 1,900-hour life.

Another random-failure system would be tires running on broken glass and sharp metal. A given tire might meet a nail as soon as it was put on, or it might run for months before hitting a glass shard at just the right angle to get punctured. It would wear-out if it ran 40,000 miles, but if on-average it gets a puncture in 1,000 miles, then wear-out is moot and it just randomly fails in 0-2000 miles, average 1,000 miles.

Tubes are a mix of random failures and wear-out. You don't study them and predict the MTBF, you have to put a bunch on life-test and supplement that with field feedback. If you identify one design flaw and fix that, you can estimate a new longer MTBF, though you better not trust it too much until you try a whole lot of them over a long time.

(*) I have assumed 5,000 hour MTBF. This is reasonable for computer tubes which are pre-tested to weed-out weaklings, and run well below their melt-down point. Tube failure in large tube computers was such a problem that they did all they could to extend tube life. OTOH, Power Tubes working at rating, especially cheap ones, may have much shorter average lives. As low as 500 hours for 6V6 in a 450V guitar-amp; but at 5 hours and $25 a night, the guitarist gets an average of 100-night service from a $50 set of tubes while earning $2,500. He can afford short-life in return for a lighter cheaper amplifier.

I have not seen my tube-era JAN Reliability books in many years and several moves. I just did a quick search and nobody has posted them (that I see).

[fdegrove]

Hi,

I run a computer that uses 10,000 tubes. I do not know which tubes will fail, but I can bet the computer will need about 2 spare tubes every hour that it runs.(*) If the computer runs 24/7, and the tube-delivery truck comes once a month, I need to order about 1,500 tubes for each delivery.

Yes, I'm pretty well aware of the Eniac and the fact that someone had to supervise the condition and working order of some 10.000
ECC84 (in Europe) to make sure it worked. And that was not the only problem either.

But...in this case your only taking into account the likelyhood that for some reason out of the 10K identical tubes (by type) operated at roughly the same params one is likely to fail every n-minutes.

Fine, so far. But when I say that there's no useful MTBF figure other than life expectancy it was meant for a single tube.

As far as I'm concerned no one can predict when a single tube is going to fail. Hence does a tube have an MTBF figure attached?

With all due respect, this is what I meant.

In other words for any user asking his dealer when his new bought tube is ging to die it's just a wild guess within the published life expectance figures.

As for the Tor figures and the vacuum issue, I am still reading as the booklet does not have a TOC, it's going to take me some time to arrive at the appropriate passage.

Cheers,

[Sch3mat1c]

Originally posted by fdegrove
As far as I'm concerned no one can predict when a single tube is going to fail. Hence does a tube have an MTBF figure attached?

Frank. It's been said for the last 10 posts, MTBF relates only to averages, not a single tube. Note it *IS* still correct to say that a particular type has an MTBF, because the very definition of MTBF assumes an average over a large quantity. And yes, it has little meaning on a single tube.

Tim

[fdegrove]

Hi,

Note it *IS* still correct to say that a particular type has an MTBF, because the very definition of MTBF assumes an average over a large quantity. And yes, it has little meaning on a single tube.

O.K. give me the MTBF for a 12AX7A.

IOW, not what's the life expectancy which does not predict when my tube is going to fail but when it would be fair to expect my tube would fail.

In short, there's no sure way to predict when a tube is likely to fail so NO MTBF for a single tube type applies.

In other words, to you and me, when we buy a tube type it can fail within hours of service or it can last way beyond life expectancy too.

In yet other words, MTBF is pretty meaningless to us, the consumer and no tube manufacturer even made such calcs with respect to consumer goods anyway. Any such claim would backfire big time anyway, which is my point exactly.

Cheers,

[Sch3mat1c]

Originally posted by fdegrove
O.K. give me the MTBF for a 12AX7A.

Originally posted by Sch3mat1c
Note it *IS* still correct to say that a particular type has an MTBF, because the very definition of MTBF assumes an average over a large quantity.

All I can do is repeat this statement. I don't know how else you can't get it.

In short, there's no sure way to predict when a tube is likely to fail so NO MTBF for a single tube type applies.

Nearly true. But since 6L6s almost never last as long as 12AX7s, clearly, there must be some mathematically-expressible difference in their lifetimes.

In yet other words, MTBF is pretty meaningless to us, the consumer and no tube manufacturer even made such calcs with respect to consumer goods anyway.

Consumers don't care about MTBF, they just know that they can go down to the corner store and grab a new one. Well, they could in the 50s and 60s. It's been mentioned before that it is most useful in large quantities such as military applications.

Oh - think half life. Nuclear material disintegrates at random, but some materials disintegrate almost instantaneously (take chromium 51 for example.. HL = 27.7 days), whereas some take a very long time (uranium 238 for example is 4.5 billion years).

Tim