Heathkit W5m delay timer

[dcgillespie]

Retro -- Earlier you emphatically stated not to use a CC approach to power up a heater, as that to would over power the heater. But how is this significantly different from using a resistor that limits current to 300 ma, which is then electrically shorted out after an appropriate period of time? I suppose it is all relative to the ultimate current drawn by the circuit. After all, a resistor that limits a single 12AX7 operating in parallel heater configuration to 300 ma is much different than limiting a KT88 to 300 ma before the resistor is shorted out.........

I think that for "receiving" tubes, posts #52 and #53 start to really get into the meat of the matter. The study that Retro has done is clearly extensive and supported scientifically. But as with the engine whose oil and filter is changed every 5000 miles, versus one that is changed every 3000 miles -- with all else being exactly equal (oil and all) -- what is the practical return for the effort? In terms of absolutes, it is a safe bet to say that the engine whose oil is changed every 3000 miles will last longer than the one whose oil was changed every 5000 miles. The argument is, if you an make an engine last longer, then why not do it? But how much longer will the engine actually last for the extra effort extended (achieve the same amount of wear)? If the answer is 150,000 miles then a solid case can be made. But if it is only 5,000 miles, then that would be a very hard sell. And that's the rub. In our sphere of the vacuum tube world, our use cycles are likely much different than those in which the studies were conducted, begging the question "how much better is better?"

Some years ago now I published information on extending the life of power tubes, with the take away point for this discussion being that often, delayed B+ is much more damaging than if it were never delayed at all -- the point being that delaying the application of B+ in the traditional manner caused huge spikes driving the grids of the power tubes to Eg1=0 momentarily when the cathodes may not yet have reached full temp, which is incredibly damaging to the cathodes of the tubes. During such time, the electrons were being drawn directly from the cathode itself rather than from the space charge that develops around a fully heated cathode during normal operation. I was barely getting 500 hours of operation out of tubes switched in this manner before power output dropped to unacceptable levels (64% of average new power output, or "BAD" on an English scale). By revising the delay circuit so that such pulses were eliminated (among other things) yet the B+ was still delayed, I now enjoy (on average) almost 6000 hours before power output drops to this level, with most tubes maintaining well over 90% of average new power output to well over 2000 hours of use. These results are for GE 6550A tubes done over a period of the last 20 years. None of this is to cast doubt on the topic of delaying the applied voltages to a vacuum tube, or the work that Retro has done in his research. But it is to say that how that delay is applied to a design as a whole is a matter of some importance as well.

Dave

 

[Retrovert]

Retro -- Earlier you emphatically stated not to use a CC approach to power up a heater, as that to would over power the heater. But how is this significantly different from using a resistor that limits current to 300 ma, which is then electrically shorted out after an appropriate period of time? I suppose it is all relative to the ultimate current drawn by the circuit. After all, a resistor that limits a single 12AX7 operating in parallel heater configuration to 300 ma is much different than limiting a KT88 to 300 ma before the resistor is shorted out.........

I am happy you brought up this very important point, because I should have addressed this issue.

The difference, and it is a crucial difference, is limited current vs. constant current.

It took me a lot of reading about large-scale failure before I had that thunderbolt of enlightenment.

The entire point of the constant voltage supply (tubes in parallel across the supply) is that each tube regulates its own current. Now that regulation can be quite high during startup, but it stabilizes. No tube is forced to consume more current than it wants. This is key. By virtue of manufacturing differences not all tubes may need the same amount of current.

With constant current the supply forces current through the device by varying the voltage even if the device does not want that much current through it. Constant current is the string approach, where all heaters are in series., the current of each tube is fixed and the voltage varies. The difference is obvious if you think about a tube drawing 300 mA and one drawing 600 mA. In parallel they both work. In series the higher current tube destroys the lower current one by pulling too much current through it. The explanation I gave for heater flash is a constant current approach: the cooler regions force current through the hotter ones to maintain a constant current through the tube. The hotter regions have enough current but it doesn't matter.

The reason I favor a limited current approach because the current draw of a tube while heating is enormous. It can be six, eight, even ten times a normal current draw. That high current stabilizes, of course, at the steady state draw. But for a brief period of time it can drive portions of the heater into incandescence. As the heater temperature increases the heater can expend several times its normal length, and abrade the insulation and the cathode sleeve.

So my proposed approach is to limit the maximum current to the current usage set forth in the datasheet. This forces the heater to gradually heat, reducing the peak temperature for any spot. Otherwise, the heater will be heated too fast. As the heater is overheated in spots it boils away material which makes it thinner, which makes it easier to overheat, which makes it overheat faster, which accelerates the cycle until burnout.

Think of a lightbulb. Why do lightbulbs fail during turn-on? The reason is that on-current is so great that it stresses any weak spot, forcing more current through in a runaway situation as I previously described. This eventually kills the lightbulb and similarly kills the tube. Heater flash is not normal and is not desirable; one is literally watching the erosion of the heater and the deposition of tungsten, and whatever else is in the heater, into other areas of the tube. I keep thinking about Dushman's statement that the lifespan of a tube at 2,000 hours is defined as having 10% of the diameter of the filament removed. Yet the manufacturers of tubes prone to heater flash lie (yes, they present information which is false and is known to be false) and claim it isn't a problem.

The current limit should be set for each class of tube. So if they all draw the same current, it's all good. If they draw different current they need to be fed from a different current limiting source. Because that 300 mA if fed from a 600 mA source is going to be drawing twice what it should be allowed to draw.

By fixing the maximum operating current from the datasheet as the maximum possible draw, not the forced current draw, the heating time increases, perhaps for several minutes, and the tube is not stressed.

We're talking a substantial difference. If a tube heats to 1,000 degrees C in 20 seconds, that's 1,000/20 = 50 degrees per second. If we make that 120 seconds, we have 1,000 / 120 = 8 degrees per second. Wow. That's impressive. And as per Miller-Larson I believe that will alter the recrystalization of the tungsten.

To conclude, limited current is not the same as constant current.

I think that for "receiving" tubes, posts #52 and #53 start to really get into the meat of the matter. The study that Retro has done is clearly extensive and supported scientifically. But as with the engine whose oil and filter is changed every 5000 miles, versus one that is changed every 3000 miles -- with all else being exactly equal (oil and all) -- what is the practical return for the effort? In terms of absolutes, it is a safe bet to say that the engine whose oil is changed every 3000 miles will last longer than the one whose oil was changed every 5000 miles. The argument is, if you an make an engine last longer, then why not do it? But how much longer will the engine actually last for the extra effort extended (achieve the same amount of wear)? If the answer is 150,000 miles then a solid case can be made. But if it is only 5,000 miles, then that would be a very hard sell. And that's the rub. In our sphere of the vacuum tube world, our use cycles are likely much different than those in which the studies were conducted, begging the question "how much better is better?"

That is a very fair question and one that I've repeatedly asked myself throughout my research. I should point out that I was initially unsure if I should believe that cathode damage was mythology and was looking for proof one way or the other. I ended up reading monographs on tube design written by the Olde Ones and eventually worked out what was happening.

Here's the argument that convinced me.

We know that Gm changes over time, typically quite suddenly near the end of the tube's life. We also know that the bias changes over time. Which is one of the advantages of cathode bias. (EFB and EFB II are far better approaches, of course, and I will be implementing both of them in various amplifiers I own.)

All sorts of things change as the tube ages. Why?

I'm not being snarky here. Why do operating characteristics change? They don't change in transistors. (Aside from peculiar phenomenon like damage to junctions, copper poisoning, etc. which were worked out and resolved.)

These changes happen because the tube ages and its characteristics change. Again, why? Because the physical structures are altered.

At the end of the tube's life the Gm really drops. This is something the tube testers look for.

So are we to believe that nothing happens and then one day our friend the tube breaks a hip and hits its head? No, there was progressive damage. That is the curious fact: the damage is continuously happening, it is just not yet significant enough or is being repaired, yes repaired, to some extent. As the oxides are baked out of the cathode new Ba and Sr oxide migrate to the surface. Eventually those deposits are mined out and the cathode dies.

So my argument, and Dave, I would be happy to collaborate on testing these if you'd like to devise some tests so we should talk about this if you're interested, is that by making small changes we can reap big rewards.

These changes are: regulating the heater current at startup, slightly dropping the heater voltage (about 5% seem to be the optimal value) and regulating it, by regulating the B+ to ramp up slowly after the tube is hot, and doing a few other tweaks like ensuring the grid and screen voltages are clamped in a reasonable range (never positive), and adding the gird and screen stopper resistors to prevent arcing, and ensuring that the chokes and transformers have flybacks, adding snubbers to diodes to eliminate HF noise, adding snubbers to transformers to prevent ringing, not switching to toroidal transformers, and probably one or two other things I've omitted from here.

The less stress on the tubes the longer they will last.

I would not be surprised, and this should be borne out by testing, that by doing all of these things the tube lifespan can be doubled or trebled.

Disclaimer and disclosure: I haven't done these longevity tests. Nobody has done all of them, only some of the easier ones. The papers I've read on large-scale lifespan tests discuss the benefits of slightly lower heaters and preventing turn-on damage. As I said, I'm writing a monograph on this and I've been down this rathole for over a year now. The more I read the more I realize it is more complicated than I realized.

Continued in Part II

 

[Retrovert]

Part II

Some years ago now I published information on extending the life of power tubes, with the take away point for this discussion being that often, delayed B+ is much more damaging than if it were never delayed at all -- the point being that delaying the application of B+ in the traditional manner caused huge spikes driving the grids of the power tubes to Eg1=0 momentarily when the cathodes may not yet have reached full temp, which is incredibly damaging to the cathodes of the tubes.

Le me say that I am well familiar with this work and consider it to be holy writ on the subject of delayed B+. I also very much appreciated the work you did on arcing in tubes. All of it considerably advancing the state of the art and ignored at one's peril. Or peril to one's tubes.

What we see in short tube lifespan and tube failures is artifacts of bad design.

Let me say that again, because it is important: tubes die because designers either weren't as competent as they should have been, were about as competent as they could have been given the technology, or were about as competent as they were allowed to be when producing products for a particular price point. Regardless of the cause, the designs are bad (sub-optimal or failure-prone if you prefer engineering argot) and we should not design tube circuits this way or leave in liabilities that are not providing a sonic benefit, just shortening lifespan.

A lot of that is the luxury of looking at these circuits fifty years on and having better test equipment which is affordable or better components. All true. We have amazing technology and amazing products that the Olde Ones did not. Just look at the 1% metal film resistors and 1% plastic film capacitors we have for a price that is a fraction of what those old components cost in constant dollars. We can also use amazing voltage regulators that they just didn't have. Could do. Wasn't possible. Ok, that's some of why we can build better circuits. But the designers (usually dead, but not always) need to own their mistakes: the datasheets are very clear on certain limits for grids, screens, voltage limits between structures (heater to cathode, cathode to grid, etc.), and the danger of arcing.

Yet designers, both back in the day and today, ignored (and continue to ignore) those warnings, generally because circuit design requires some understanding of what is happening and, besides, the amplifier worked didn't it? If those limits were so important it would not have worked.

One can run a car with bad oil. It eventually dies and that point is a lot sooner than would otherwise be necessary.

During such time, the electrons were being drawn directly from the cathode itself rather than from the space charge that develops around a fully heated cathode during normal operation. I was barely getting 500 hours of operation out of tubes switched in this manner before power output dropped to unacceptable levels (64% of average new power output, or "BAD" on an English scale). By revising the delay circuit so that such pulses were eliminated (among other things) yet the B+ was still delayed, I now enjoy (on average) almost 6000 hours before power output drops to this level, with most tubes maintaining well over 90% of average new power output to well over 2000 hours of use. These results are for GE 6550A tubes done over a period of the last 20 years. None of this is to cast doubt on the topic of delaying the applied voltages to a vacuum tube, or the work that Retro has done in his research. But it is to say that how that delay is applied to a design as a whole is a matter of some importance as well.

That is very true and important: you extended lifespan SIXFOLD. Wow. That's HUGE! Six Times! Yet for years people had these failures and just considered them normal. They aren't normal. But people think of tubes as a replaceable item like a lightbulb. Oh, well, it failed.

Now, your observation about lifespan goes back to what the issue, as I previously explained, about islands of charge coalescing into a smooth emission surface. Which was known! Actually known and documented back in the day. Yet it was not honored! Why? Carelessness, ignorance, stupidity, cost savings, bad technology, take your pick. It's a combination of all of them.

I believe protection of tubes must be done in a systematic way. We must consider the entirety of the circuit, including the power supply, to understand why tubes age and how to extend their lifespan. The rules we need to follow are not complicated, but they are rules created by the physics of the devices we use. As I always say, we can have our own likes and dislikes, but we cannot have our own laws of physics.

 

[peterh]

Isn't it very rare with broken filaments ( unless we discuss directly heated large tubes ) ? I have never had
one tube fail with broken filament. Thus this problem seems rather exotic, and more resources might be
made better use on other work.
I'm not criticizing retros work, i might have unusual experiences here. Comments ?

 

[drillher]

A tube that is used within it's design limits will normally last several 1000 of hours. An amp that "consumes tubes" has some inherent problem. Playing around with delays and stabilized voltages within that broken circuit will not save tubes.
Problems might be DC related, or AC(oscillation especially in power tubes), problems might be continuous or only at
certain events ( violent power on / off that causes overcurrent ).

The B+ voltage is right for the 6922 tube.What components can I check that would influence tube life?

 

[peterh]

The B+ voltage is right for the 6922 tube.What components can I check that would influence tube life?

Current ( "bias" ) , filament voltage ( not to high, not too low ) , Vkf ( voltage between filament and kathode ) , Pa ( power exposed to plate), any possibility that the grid will get positive during power on or off etc.
Publish a schematic to get more opinions.

 

[Retrovert]

Isn't it very rare with broken filaments ( unless we discuss directly heated large tubes ) ? I have never had
one tube fail with broken filament. Thus this problem seems rather exotic, and more resources might be
made better use on other work.
I'm not criticizing retros work, i might have unusual experiences here. Comments ?

This isn't at all exotic.

As I have repeatedly explained, even when the filament hasn't failed, filament material is deposited all over the other tube structures which poisons them. The tungsten evaporated from the filament must go somewhere, it is a closed system.

The overheating of the filament also overheats the cathode, baking out oxide which poisons the tube. That's where gas comes from. It also bakes out absorbed and adsorbed gasses from other metals and the mica.

 

[Retrovert]

The B+ voltage is right for the 6922 tube.What components can I check that would influence tube life?

The heater-to-cathode voltage limits must be honored. The screen limits must be honored. Driving the grid and screen positive, and thus causing current flow, will dramatically shorten the tube lifespan.

 

[Retrovert]

I read a description of a heater testing procedure RCA conducted on batches where they put the heater voltage up to 15v and 180v above the cathode then cycled them on for 1 minute and off for in 1 minute, for not less than 2000 cycles.

Tube heaters cannot be compared to light bulbs. The insulation coating acts like a heat sink and modulates the on-off thermal stress.

Errr, no, it doesn't. It absolutely cannot. Not possible. That isn't its job. The coating is for voltage insulation from the cathode. The alumina coating is very thin, about 1 mil for every 75 volts. ONE MIL! That's what determines its thickness: voltage insulation. This is why the heater-to-cathode limits must be honored. Most tubes have a very thin coating and it is rapidly degraded from excessive voltage. Again, why I urge heater elevation to protect the coating. Thermal stress will fracture it and cause degradation. That's where hum comes from: leakage.

Beyond that, the thermal conductivity of alumina is terrible. If it modulated thermal stress the filaments wouldn't go into thermal runaway and they wouldn't glow. That's the proof.

You are claiming a 1 mil coating is going to equilabrate the temperature and eliminate thermal stress? On a filament routinely (every single time it is powered up) in thermal runaway? No way. Just not happening.

Here's more proof. Tungsten alloys are used to minimize grain growth from the Larson-Miller effect as the filament structure changes from heating. This was a well-known and well-documented problem in the 1950s and 1960s. Reducing the rate of change reduces the re-crystallization. That was well documented. Tungsten was selected because of its high melting point.

Tube filaments are light bulbs and fail in the same way. That is exactly what they are: a wire heated to emit electrons. Both are tungsten wires. What else would the filament be?

The thickness of the wire determines the lifespan, just as it does in a tube. This is why long-life bulbs have heavier filaments. (Reducing thickness to create failure was, in fact, the main contribution of the Phoebus Cartel to destroying the light bulb lifespan. In summary: en.wikipedia.org/wiki/Phoebus_cartel )

Spot heating in the cathode damages it over time. That is known. The primary source of failure is baking out the oxygen from the alkali-metal oxides, and this liberates Ba and Sr into the tube. This heating rapidly consumes the reserves of Ba and Sr oxides in the cathode and eventually causes the cathode emissions to drop to the point that Gm suffers. That's why the tube rapidly goes downhill: one day there's just not enough oxide and the work function (ability to emit electrons from the surface into the space cloud) is impaired. The tube is dead.

Heater current climbs in the tube over its lifetime. Why is that?

Because a thinner heater better conducts (remember, this is PTC) and draws more current.

The lack of stories on these boards about heater failure says much about the rarity of a heater popping under normal conditions.

No, it does not. It proves NOTHING and is not evidence of anything.

I'm not accepting the lack of anecdotal evidence as proof of anything theory, when we have widespread evidence of heater abuse (heater flash) that isn't continually reported as a problem (people even say it is normal), we have widespread misunderstanding of the danger of applying B+ to a cold tube as being a problem only in RF tubes (a baseless meme which endlessly circulates), when we have widespread evidence of tube arcing that isn't being continually reported or even seen as a problem, and we have widepread evidence of tube failure (just look above) with no analysis or understanding of why.

The fact that someone does or does not complain on a one-off basis is not evidentiary that a problem does not exist.

All of the large-scale studies on tube lifespan, in computers and avionics, support the theory that heater stress kills tubes. Moving from on-off cycling to continually on, as in the early computers, dramatically reduced tube failures, from hundreds of hours to thousands. That was heater shock, and possibly B+ shock, whether from heater failure or from depositing the filament over other tube structures, or from stressing the emissions surface on the cathode before it was fully formed.

That is published data. And, no, I'm not dumping all of my sources here until I finish my monograph. I spent literally hundreds of hours researching this and reading through hundreds of papers, patents, books, etc. This was academic level work and it was tedious and time-consuming. I can tell you this problem was well understood and then the tube dark ages came along and the knowledge was lost.

I'd say that any little wire is subject to the vagaries of the manufacturing process and a heater that fails under normal conditions was one with some tiny flaws that eventually gave in. But most heaters are not flawed and will last way past the designed usefull life of the tube.

That theory does not account for the drop in failures from eliminating temperature cycling. It also does not explain why Gm suddenly drops in the tubes. What is the mechanism for cathode depletion?

My theory, which comes from the published work on tube lifespan, explains all of this as a unified and simpler theory. So I think Occam's Razor applies here.

 

[Retrovert]

One more thing.

The large-scale installations (ENIAC and AT&T) ran their tubes at a slightly lower filament voltage for much greater lifespan. Lower voltage equates to lower heater stress.

 

[ggizzy]

Retro,

This is a very good thread. I appreciate all your research. When can we expect the monograph?

 

[whoaru99]

...

You are claiming a 1 mil coating is going to equilabrate the temperature and eliminate thermal stress? On a filament routinely (every single time it is powered up) in thermal runaway? No way. Just not happening.

It must to some material degree. The heater flash that some tubes exhibit is from an uncoated portion of the filament.

 

[tcdriver]

Using a MOSFET for a heater delay was much simpler than the relay circuit since there's no relay and no flyback. Here's the rough outline.

The circuit I like is a MOSFET (as a switch) in parallel with a resistor. An RC constant charges to the point it is sufficient to turn the gate on and allow current to pass through the MOSFET. During the time the MOSFET is off, the heater is powered by a parallel resistor sized, as per Ohm's Law, to limit current to 300 mA. When the MOSFET turns on (again, as a switch) the resistor is effectively out of circuit. Otherwise it runs through the current-limiter resistor. A resistor bleeds the capacitor to ensure that if the unit was switched it wouldn't be instant on, and would go through the same delay cycle. A diode ensures the gate voltage is always within range...

I have a couple of questions about the circuit you are describing.

1) Is the circuit applicable for both DC and AC filament power?

2) Where does the circuit (555, etc.) get its power?

3) Have you built and measured the circuit?

Thank you.

 

[tcdriver]

There has been a lot of discussion, in this thread, about slow filament turn-on and applying B+ after the filaments are at temperature. What about turn-off? Is it critical to reverse the process? i.e. Must one first ramp down the B+ and then slowly ramp down the filament voltage?

 

[trobbins]

The thread starts off with the OP's aim of using a simple relay delay of B+ application as a means to save precious KT66's in a W5M. Many posters seem to accept that aim as valid, but a few recommend that such a scheme will be far worse off for the KT66's (me included). Bob, are you going to continue with your aim, or has there been sufficient discussion to possibly sway you?

The thread also has a lot of discussion on what if any delay is needed to cover heaters, and if heater warm up needs to be smoothed. The anecdotal comments I've come across from many forums over the years, from many who have lengthy experience/exposure to commercial grade valve equipment in either service or professional or hobby environments, indicates that valve heater failure rates are insignificant in the context of equipment with failed valves. This also includes from heater flash, which appears to be related to manufacturing process and subsequent tolerances, rather than a design intent. Anecdotal comments are prone to misinterpretation, perhaps as many failed valves were not really checked for cause of failure (ie. swap it, and discard if 'bad'), and very few people had exposure to a statistically large sample of the same equipment from cradle to grave.

Only one referenced paper known of publicly identifies a heater related failure statistic, and that is in a non-commercial equipment environment. Interpretation of special equipment environments such as the Whirlwind Project may be prone to misinterpretation when relating statistics to commercial equipment, perhaps due to the use of special valve types and batches, and different powering. Interpretation of those projects is also prone to poor statistical data availability - even the known paper presents a failure rate of 3.6% which appears to me to be ambiguous (it could be the rate for no special heater warm up, or with special warm up).

Discussion by Retrovert is obviously well intentioned, however a continual use of emotive words to assert a view, or reason a proof, could easily make some believe that all valve equipment needs to be modified to some better modern standard to achieve tangible improvement in reliability of their vintage equipment.

Imho, there are lots of subtle changes that can be made to vintage equipment to improve reliability from a whole range of failure situations. Even the inclusion of an AC side thermistor to benefit other issues can alleviate the peak current experienced by tube heaters.

 

[Retrovert]

It must to some material degree. The heater flash that some tubes exhibit is from an uncoated portion of the filament.

No. I previously addressed this. It arises because of either a dissimilar material from the weld: "A cold heater is a dead short. Portions, usually the ends because of welds to dissimilar metals..." or because the wire is damaged in some region. Welding will do that.

When spot heating occurs inside the cathode sleeve it is invisible because it is covered. My research into filaments for light bulbs suggests it occurs at any thinner spot in the wire and these hot spots arise because of the manufacturing techniques.

 

[Retrovert]

There has been a lot of discussion, in this thread, about slow filament turn-on and applying B+ after the filaments are at temperature. What about turn-off? Is it critical to reverse the process? i.e. Must one first ramp down the B+ and then slowly ramp down the filament voltage?

Yes. I had not specifically addressed this, but the answer is yes. Gradually reducing the voltage, both filament and B+, were well-known lifespan issues for transmitting tubes. This required a motorized variable transformer (trade name Variac) as those monsters took a while to shut down. I was going to propose it for B+ in amplifiers as the logical extension, because this allows the islands of charge to gradually reduce so there is no point emission into the space cloud, but I do not believe this to be a significant risk for the tube provided no current is flowing through the tube.

That's key: no current must be flowing so that the emission surface cannot have one region emit more than another. The ideal ideal power down sequence moves the grids (and screens) very negative to put the tube fully into cutoff. That ensures no current can flow, so it matters less what the cathode emission surface is doing as nothing is leaving the space cloud. But one must still ensure that the space cloud collapses as the inverse of powering up.

 

[dcgillespie]

Retro -- OK. So a resistor that limits current flow to the rated value will in fact be the same as a CC device to maintain that same current flow -- but only at the instant of turn on. After that instant, the resistor (being fixed) causes current flow to drop as the tube heats, where as the CC device will keep it constant. Got it. The CC device is way better than nothing at all (for parallel connected heaters), but not nearly as good as a resistor that is then later shorted out. I suppose the best possible approach then would be a circuit that ramped the voltage up over time (as you have previously suggested) without any (later) shorting required. All to protect the heater and prevent evaporation of the coating. Sound reasoning supported by theory to extend tube life. Long term tests would be needed then to gain some understanding as to the life extension produced.

Understand that in my case, the six fold increase in life expectancy must be put into context: This was in the particular design I was using, and is as measured against using a traditional approach to B+ delay which produced the significant event pulses noted. It is undetermined what the increase in life expectancy would have been had the tests been measured against using no delay at all. I suspect that due to the factors I was citing (turn-on/turn-off pulses), it would not be as impressive -- but still significant enough -- for the following reason:

To tc's question, turn-off pulses are just as destructive as turn-on pulses are. Power supplies with massive reserves of capacitance in NFB designs are particularly prone to them, and they occur moments after turn-off as the cathode is cooling down. Again, this is very destructive to the tubes. My ultimate answer to the question of B+ delay is to (just as with equipment pieces themselves) "stage" the application of B+: Small signal stages are first to get the B+ applied, and the output stage the last. At turn off, the output stage is immediately disconnected from B+ (relay contacts after the PS caps) at the instant of shut down. With this approach, all turn-on and turn-off pulses are eliminated in the output stage. This, coupled with the use of Screen Stability resistors and operating the heaters at precisely 6.30 vdc (slow start with regulation) produced the significant increase in output tube life expectancy that I observed.

This is excellent work you are doing!!

Dave

 

[Retrovert]

Retro, This is a very good thread. I appreciate all your research. When can we expect the monograph?

(throws up hands in classic gesture) Wish I knew. It's done when it's done. The more I dig into this problem, the more I find. I have librarians and archivists looking for material for me. I'm trying to get it all into a form that is comprehensible and then get it out for some review. I'm not being paid for this, so that's a factor, too.

 

[Retrovert]

Retro -- OK. So a resistor that limits current flow to the rated value will in fact be the same as a CC device to maintain that same current flow -- but only at the instant of turn on. After that instant, the resistor (being fixed) causes current flow to drop as the tube heats, where as the CC device will keep it constant. Got it. The CC device is way better than nothing at all (for parallel connected heaters), but not nearly as good as a resistor that is then later shorted out.

Yes and no. The constant current is worse for two reasons: (a) it varies the voltage to maintain current and (b) it forces current through the device, as per (a), to maintain the current once the ostensible operating current has been reached.

A limiter does just that: limits current. It still allows the tube to regulate its own current, which may not always be ideal, of course, as was set forth.

I suppose the best possible approach then would be a circuit that ramped the voltage up over time (as you have previously suggested) without any (later) shorting required. All to protect the heater and prevent evaporation of the coating. Sound reasoning supported by theory to extend tube life. Long term tests would be needed then to gain some understanding as to the life extension produced.

Yes, a ramp is superior to switching the resistor out of circuit. That's why I wanted to use a sawtooth from a 555 or an exponential curve, and why I thought an Arduino programmed with a curve would be better yet. And yes, research is needed.

I suspect that the very early stages of heating will be highly non-linear and consequently peculiar in shape, but this needs experimental data to verify.

Understand that in my case, the six fold increase in life expectancy must be put into context: This was in the particular design I was using, and is as measured against using a traditional approach to B+ delay which produced the significant event pulses noted.

Oh, of course. The point is you stopped blowing up tubes at an alarming small number of hours. My point was not that this would happen everywhere, but that look at what a simple fix did to lifespan. Tube stress really matters a great deal.

Many amplifiers are noted for short tube lifespan. Guitar amplifiers in particular eat tubes like snacks, because they drive them so hard.

Why is "red plating" even in anyone's vocabulary except as "turn that thing off and fix it!" I was elsewhere reading a thread about metal tubes being overheated to the point the paint blistered. That's just really bad design. It shouldn't matter if one can see into a tube, the grid should not be driven positive, and the tube should never be driven this hard.

I have read the warnings from manufacturers about positive grids from the early days. Yet these warnings were routinely dishonered and continue to be dishonored.

To tc's question, turn-off pulses are just as destructive as turn-on pulses are. Power supplies with massive reserves of capacitance in NFB designs are particularly prone to them, and they occur moments after turn-off as the cathode is cooling down. Again, this is very destructive to the tubes.

Exactly so. I believe I responded to this above, but will re-iterate that the design practices for RF tubes involve increasing grids and screens bias to shut off conduction, and then ramp down the voltages. This was a well-worked out and well-settled practice before most of us were born.

It was never implemented in consumer electronics because of cost. That's key. We are accustomed to cost-saving design because these practices would be a luxury.

Now we have inexpensive MOSFETs to do the work! Go MOSFETS!

I'm sure someone will make a comment about "no sand in my amplifier" to which I say, yeah, sure, whatever, go blow up your tubes, just don't use any of the stock I'll ever want to purchase as it drives up the price.

If you want to see really bad design look at the commodity AM radios and lower price-point televisions from the 1950s and 1960s. From the wonderful people who brought you heater strings and no DC restoration in the video signals.

My ultimate answer to the question of B+ delay is to (just as with equipment pieces themselves) "stage" the application of B+: Small signal stages are first to get the B+ applied, and the output stage the last. At turn off, the output stage is immediately disconnected from B+ (relay contacts after the PS caps) at the instant of shut down. With this approach, all turn-on and turn-off pulses are eliminated in the output stage. This, coupled with the use of Screen Stability resistors and operating the heaters at precisely 6.30 vdc (slow start with regulation) produced the significant increase in output tube life expectancy that I observed.

Yes. And that is pretty much what the RF tube procedure was. Gradually reduce B+ to zero, gradually reduce filament to zero.

This is excellent work you are doing!!

Thanks. I often feel like I'm proposing an alternate to the Ptolemaic system and I see the collection of kindling and hear the angry shouting.