From working with tubes, commercially since 1995, I have learned that it is generally not very popular with manufacturers to make a worst case design for a tube amplifier. To refresh our knowledge: It means you calculate what happens if all worst case situations happen at the same time, and by Murphy's law they do, if you just try often enough or wait long enough.

When designing an elevator, you would not accept the manufacturer to determine the required thickness of the steel cable by trying it out with 11 test persons inside, and then write on the table: "Elevator for maximum 10 persons". Instead of that, you would expect a calculation by a qualified person, document the calculation, have it verified by at least a second engineer, then verify the calculations with practical tests, then have a security margin on that, and have the whole thing approved by an external company. The result is the cable thickness, that you are hanging on, in this elevator. Security margin means if the cable must be minimum 1 Inch thick, you would expect three cables of two inches in parallel, so even if one breaks, no accident happens. Well in real life that is not even enough. All safety margins must be minimum 10x for elevators, on each large or small element.

Of course we only speak of amplifiers here. So we must see what a worst case design means here.

For instance, it can happen that the amplifier runs at full output power, with low impedance speakers, in a maximum temperature room, at 5% too high mains voltage, with tubes that draw a bit more current than average, and then do so for 24 hours a day, and a user that knows only how good it sounds or not. This is a the worst case, but it can happen. So a good designer, keeps this in mind, and calculates this situation. So calculations means numbers on a piece of paper, it means not trial and error. The try part comes later, it is called: beta testing, and in case of an error, what comes is not another try, but another calculation.

You document your calculations in a so called "R&D log book", and also you document what you have done in case of a failure at first testing it, and document where you made corrective action, and how that calculates. THIS is good engineering work, and professionals call it a worst case design. (so other then the word suggests, it is not the worst design case, but a very safe design).

Every electronic engineer working in an R&D environment of a serious company has to do this, or his boss gets trouble of some kind.

So far for the theory. The practical situation, unfortunately is another one. When investigating tube defects,  the conclusion is, capital mistakes are found with manufacturers, who think they know all the classics, and attempt to break technical limits in oder to be better than others. Believe it or not, but the latest invention with amplifiers was made by Julius Futterman, in the mid 1930's when he invented the OTL amplifier. If you want to be better than others, all you need to do is your homework.

It is most suspicious when someone's claims to get "so and so many" Watt out of a tube, and this number is higher than normal. For instance some say they can get 28 Watt undistorted Single Ended power out of an 845 tube. Folks, this is fully impossible. Still you see this often written in official specifications, and for sure not based on honesty and knowledge.

Of course most manufacturers make very good products, no doubt about that.  What this paper is about,  is about the  most common made mistakes. So the idea is to explain not how to do it right,  this is just a list of most occurring errors.   How to do it right, you can learn in engineering school, and you have to go there for several years before you leave trough the front door as a good engineer.  We can not teach that quickly here. 

List of often found errors and mistakes by the department of mistakes and errors department: 

1) A design is good,  when it worked in the past,  and no problems were found before. If you really think so, you can stop reading this.

2) Filament voltage out of range. This is the most often made mistake. Examples of this are:

• Tubes filament voltage maximum and minimum specs are not given by the tube company. So amplifier makers help themselves, and conclude +10% above typical is a good value. Don't ask where they have the 10% from, I was never able to find that out. It's like asking a homeless in the park what he wants to drink, he will say "beer". Ask some people about filament voltage tolerance and he will say "10%". Also they say a lower filament voltage is no problem because this will increase the tube lifetime. You know, if tubes would live longer at 10% less filament voltage, that would be the holy grail for manufacturers, but I don't hear them say this.
• Filament voltage tolerance of the transformer is higher than the tubes can take: Like just use a 2A3 filament winding for a 45 tube. Folks, that isn't possible!
• Mains voltage tolerance is ignored: Well of course ignoring is the easy part. But you see, from tubes they expect 10% tolerance, where NO tube can do that. But.. from the MAINS VOLTAGE they refuse to accept 5% tolerance where ALL ELECTRICITY companies in the world will not guarantee anything better. Can you smell the mistake?
• So as you already expect, adding up all those nasty tolerances will make life not so easy for those who don't care about safe design. Those people will end up with broken tubes. Then they proceed very  simple:  When some tubes are known to take the exceeding numbers,  they say these  are tubes that work good. Other tubes that break down in their amplifiers, are tubes that don't work good (they say)

3) Plate dissipation chosen  close to the tube's absolute maximum. The bigger the tube, the closer you want to go to the maximum limit, but...the bigger the tube, the more risky this gets, when you don't know what you're doing. Now here is an old formula from General Electric, from 1942.  It seems this problem happened before. They say: Chosen Tube Dissipation  =  Absolute Maximum,  minus Safety margin,  minus tolerance for mains variation, minus tolerance for tube variation.  Check here.  As you already see,  there is a lot of "minus", and it brings you quickly below 80% of absolute maximum.  Several amplifier building companies want to hear this, and think that if a tube is for instance maximum 40Watt, then 36 Watt is fine. This 10% margin comes from.. eh.. yes where from? They don't know.

4) Grid resistor is  not specified by the tube manufacturer.   So the amplifier builder takes a "nice" value that seems to work fine with some tubes when he tried it. Taking values from a classic schematics is regarded an elegant solution, because somehow it seems right.

5) Building a prototype with components that perform at the limit of what is "good" is seldom  done.  The proof for that is simple, we are only very rarely asked for prototypes of such tubes. That tells it all.    I think a good amplifier must work good on all tubes,  but I am quite alone with that opinion.    For instance a new 12AT7 is good above 7mA and below 15mA.    Typical is 10.5mA.  So, tube factories do sell them from 7...15mA, and these are simply new, and fine tubes.  So a good amplifier must work good if the tube has 7....15mA.  So you see, it would be extremely useful to construct your amplifier, and try if it works good with a (selected) 7mA tube and another (selected) 15mA tube. It must work 100% good with both tubes, and then the end user can use random, unselected, normal and good tubes from the shelf. We sell tubes, but we almost never have an amplifier manufacturer asking for those "limit value" tubes. Nobody cares. We also offer calibrated tubes, so you can at least verify your tube tester. But again.. who cares. We sell only two of those tubes per year.

6) Tube testers are never send for calibration. Some companies don't have this problem, because they don't have a tube tester :)

7) Fixed bias designs  ignore that for fixed bias you must derate the maximum power with 20%. You don't believe this? Then learn that this is not based on believe, but this question is answered by the tube datasheet. (and not by the amplifier builder) If you tell them this, they ask for proof, as if we are responsible for sending people "proof". When we do send it, they throw it away and say: "We use this and that tube since 25 years, and never had a problem" That's a way to tell they stopped thinking somewhere between now and 25 years ago.

8)Switch on / Switch off surge current are not measured. Nobody knows how high it is. Typical excuse is: "We never had a problem before".

9) Here is a classical one for the end of this list. "We have this sound problem with a stereo amplifier....So we swapped the left channel and right channel tubes, and then the problem followed the tubes. So, the problem is the tube. It is really a-m-a-z-i-n-g how serious engineers mean this when they say it.

Conclusion:

1) We need more awareness that trial and error is no good way to design power electronics.

2) When blaming the used materials for problems, you may be right. And you may just as well be wrong. To say something realistic, great opinions will not be of great help. It is important to check if the amplifier design passes each of the the above points, step by step, and sure we have not even listed up everything here. When you are the beginning of this: At least a measurement with an uncalibrated instrument is better than swapping tubes and try to draw conclusions from that.

Here is what an airplane pilot must consider, for the amount of fuel in the tanks before taking off. This the official check list, he must keep records of it, and he can be jailed for ignoring this and causing an emergency landing this way.

CHECKLIST:

• Calculate with maximum fuel consumption specified for that type airplane engine
• Calculate with worst case wind speed.
• Check passengers number before taking off
• Calculate with Maximum passenger weight and maximum luggage weight
• Now add the safety margin, personally check the fuel amount by hand, and write this in the pilot's log book. Measuring is knowing.