Page 1 of 2

Tube life for emission labs tubes. Eml20b example

PostPosted: Mon Dec 23, 2013 10:21 pm
by MaciekP
Some observations.
I have a third pair of eml20b tubes. I was a early adopter of this tube, used as Gm 70 driver.

First set was faulty. It was replaced.
Second set died after 3 years i think. One tube filament broke. The second tube has transconductance at 3.9mA/V at typical op point 400V, 42mA.
The current set is rather strong. After 3 years tubes has transconductance at 5.3mA/V when new were marked as 6.3mA/V, so i vas given a very good and stron pair.
I estmate 1000hours of usage per year, so after 3 years, 3000hours 1mA/V was lost.

What transconductance for eml20b is an indication of needed replacement?
I think 4mA/V?????
If so, a tube should last for about 5-6 years and ca 6000hous of typical usage.

Rather far from 40000hours declared for best tubes from golden era...or maybe my 4mA/v threshold was wrong?

Re: Tube life for emission labs tubes. Eml20b example

PostPosted: Sun Dec 29, 2013 11:44 pm
by Jac19
Dear Maciej,

this is just Jac from jacmusic here.

After reading your post, let me say this.

The broken heater you had, in many cases we find a relation with start up electronics.

Transconductance of 20B was measured by the factory historically in a bit strange way. I say about it: We used it for matching only.

In 2009 we changed to the Amplitrex testers, for the data that is on the tube box. They are not ideal testers, but they are good for data that you can verify easy. That is why we use those testers.

Tubes from before 2009 can be very confusing to verify the quality, since that factory test is hard to repeat. The ultimate proof of good tubes is alway the tube curves.

A very good indication is also when the two tubes of a pair are still the same. This means chances of two good tubes is high. Very high.

Also, when they still give the original gain in the circuit at full signal, that is a very good sign.

Trying to judge before 2009 tubes, by transconductance changes, I would not do.

After 2009, yes.

Generally transconductance can be 70% and they are still perfect. Down to 60% they useually work, and we have seen 40% tubes work fine indeed sometimes, but that is rare.

At the end of life (as you want to know) the movement from 60% to a much lower value comes suddenly and very fast, so they collapse. Before that, so normal life, it is more linear.

Use hours... It's the holy grail question always. We give only true and real answers at EML, regardless if influence on sales.

No vendor ever had 40.000 hours lifetime guarantee.

Lifetime on Directly heated, small tubes like ECC81 was always 5000 hours. SQ tubes (like TFK ECC801S) has 10.000 hours.

Then, if you know that DHT power tubes by principle have LESS lifetime... you can see where realistic expectations should be.

We think our tubes do much over 5k hours. 10k hours we get reported so often. If somebody reports it above, I can be true, but it becomes rare.

Remember to measure transconductance at the DC settings that belong to the tube. Not at another voltage or current, as that gives totally different values.

Hope this helps.

Best regards,

Re: Tube life for emission labs tubes. Eml20b example

PostPosted: Mon Dec 30, 2013 12:44 am
by MaciekP
Thanks Jac,
I measured tubes bought in 2010 (ca 5-5.3mA/v) and the older from broken pair bought a few years before 2010 (below 4mA/V) at the same op point, typical for this tube.

Current pair is almost identical in op points and transconductance, although the real gain is 19,5 and 20.5 under 85h load.

If tubes works ca 5000hours it is not bad at all. I just wanted to know, at which transconductance i should treat eml20b as needed replacement. I have a comfort to have a manual tube tester to check it from time to time....

Re: Tube life for emission labs tubes. Eml20b example

PostPosted: Mon Dec 30, 2013 4:32 pm
by Jac19
Dear Maciej,

Some words about tube life time....

Fot this purpose there are many good books.

With lifetime there is no absolute number as such. There is only the probability to reach that lifetime. This probability is never 100%. The same tube with 99,9% probability on lasting 1000 hours, will last 2000 hours for 98%, 5000 hours for 90%, 10.000 hours for 50% and -say- 15.000 hours for 30%. And who knows, very rare 1% of those tubes will last 40.000 hours, nice for story tellers in the internet. These numbers are imaginary, but the pattern is true. So take this tube, and I ask you what is the lifetime of this tube? You can not answer it, without accepting a failure rate.

So here comes the thing. The manufacturer can say: 5000 hours, and if so there is always somebody saying: "Not true, I had a tube dying after 1000 hours". The better answer is 5000 hours, with 90% probability. That is still nice, you buy 10 tubes like that and after 5000 hours, you need only one replacement.

So the manufacturer must position somewhere in a meaningful way, to do justice to customer problems that occur, but not stick out his head for things outside his responsibility. Standard words perhaps, but look what some are doing, you will see they do no justice to customers at all. Like new Russian + Chinese tubes have three months guarantee. In reality that is may be 1000 hours, and the manufacturer is always safe. Any problems, will come later, and the guarantee is over. The Sofia tubes, you have to BUY the 12 months guarantee. I do not comment on that, but let me say at EML you get 18 month for free now.

I have been dealing with guarantee issues so long. The positive conclusion I have: The far greater part of the customers is honest with reported problems. So normally when we get presented a situation, it is indeed what he observed. The negative part is, conclusions are based on assumptions, and the ignored schematics issue. Most of the time there is no schematic. So when we ask, they say: Please look in the internet yourself.

So sometimes we have to do with knowledgeable users, and sometimes simply not.

Now the holy grail question.... Lifetime of 20B. I must start with ECC81 first.

I take ECC81 because it's a standard. There is guaranteed lifetime, it used to be three months. So, 2500 hours worst case. Or 1000 hours in many cases. So you see, they do this for a 5000 hours tube. Already after 1000...2500 hours the failure rate becomes too high to finance it for free. Then we have perhaps more than 5% or 8%, I don't know where their pain limit was, but if it was only a 1% percent after three months, the would rather have have given 6 months guarantee.

So we do have a really usefull benchmark here. On a 5000 hours tube, they give 1000...2500 hours guarantee, and an acceptable, few percent dies within that time, which they replace.

Now for the 20B.

Heater breakage is not the normal way for the 20B to die. They are quite immune to that, so if it happens we must look at the heater electronics. We replace them for free, as we always do in case of doubt.

The normal end of life, is by old age, so by loss of emission. Not a bad getter apperance, so no gas problems, or broken heater.

We have a guarantee now of 2500 hours. We guarantee that for all tubes. So this indicates, the expected lifetime for "most" of the tubes you buy, is much higher. How much higher, depends on the failure rate you accept.

Our estimation is 5000 hours you will get for the far greatest part of the tubes you buy. 10k hours is no exception.


When you have the possibility to measure gain, this is very nice, as gain will not go down as long as the tube is in it's useful life. So when the transconductance goes down a little bit, the normal way, there comes a moment where plate impedance goes up. As gain = plate impedance * Transconductance this will even out for a great part. Until the moment comes, where this is not so any more. Then, the tube collapses quickly, and stops doing the first thing you want from it: Gain = 20. You need to do this at full signal, because at the end of life, the tube curve ends loose steepness, and this reduces gain. So you must sweep through the entire curves. Is the gain fine, the tube is probably good. Not so good is small signal gain testing, as tube testers do. Full signal at an amplifier is harder for the tube, so it tells you more.

Measuring gain in the circuit is fine, but then you should monitor the new tube gain, against used tube gain, in that same circuit. In a circuit, the tube is loaded with something, and this reduces gain a little.

Please make very good note, transconductance depends VERY much on the operating point. Just because the curves are no straight lines, and there are many curves. So at which Grid voltage curve are you measuring, and what is the voltage and the current. That will give totally different transconductance. So when you have a friend with a tube tester, you must use the EML recommended bias settings, to measure the transconductance.

Don't forget to read this: ... itions.htm


Re: Tube life for emission labs tubes. Eml20b example

PostPosted: Mon Dec 30, 2013 7:12 pm
by MaciekP
Wow, nice post Jac!
Well..I started looking closer to measurements, when i couldn't get more than 10W out of one GM70 tube where normally i get about 36W. Gm70 operated normally at normal bias, but when i connected generator and scope i noticed that at 10kHz signal was not stable and distortions appears at 33v rms on gm70 grid. It was a defective tube, but i wouldn't be possible to find it out without osciloscope. Trying to find a problem i startedt estign all the singnal path elements. To measure gain to makes thing simpler and not calulate vpp from osciloscope, i used frequency for which my hadnhald multimeter give proper AC voltage values, i measured rms voltage on eml20 grid and voltage on gm70 grid after the the eml20 tube. I devided them and got the gain. The same could be done from scope (xDIVx DIV value) for vpp or converted to rms *0.3535.
Of course, when measuring transconductance you have to take tube current into consideration.
Between 30mA and 42mA change is not huge, probably around 0.5mA/v, but if you go below 20mA Rp rises because transconductance goes down.

Talking about broken heater. From the very begining i use voltage stabilization on 1084 adj chip. It happens. They didn't have gold plated grid at that time....

Re: Tube life for emission labs tubes. Eml20b example

PostPosted: Thu Dec 10, 2015 11:04 pm
by MaciekP
another report for 4th set of eml20b tubes, now v4 octal socket version.
measured at 420v/42mA.
when new gm=6.2mA/v
after 14months i have in both tubes 5.0mA/v, so around 1mA/v per year down with around 1000h/year max usage.
tube operates at around 450v/22mA/-15v, so far from max anode dissipation.
it looks like 2500h useful life.
it it 2500hours because modern tubes doesn't use dangerous and rare materials like old once or mercedes case from 80's when too reliable cars amost made a producer a bancrupt? :)
I have a very positive example with avvt av8b tubes, which survived 10 years heavy usage! After 5 years where still measuring like new! it is possible!

Re: Tube life for emission labs tubes. Eml20b example

PostPosted: Fri Dec 11, 2015 10:02 am
by J. Leven
Hi MaciekP,

which operating point and output transformer do you use at the GM70 ? ,

the voltage stabilization on 1084 adj chip with standard wiring is not a good way ,


Re: Tube life for emission labs tubes. Eml20b example

PostPosted: Fri Dec 11, 2015 10:51 am
by MaciekP
I use tentlabs now for EML20B and CLC for GM70 with Lundahl filament choke.
GM70 now fixed bias as 930V/100mA/-80V with 7k OT or autobias at 870V/110mA/-66V

Re: Tube life for emission labs tubes. Eml20b example

PostPosted: Fri Dec 11, 2015 12:19 pm
by MaciekP
BTW, as output power is not as important (for me) as damping factor and bass control, probably 10k OT would be more appropriate here, but good 7k is still fine.

Re: Tube life for emission labs tubes. Eml20b example

PostPosted: Fri Dec 11, 2015 1:12 pm
by J. Leven
The OT with 7k, is that a custom transformer and
have you measrued the damping factor,?
I estimate which is approximately at 3