Heathkit W5m delay timer

[trobbins]

Retrovert, the only heater turn-on stress related statistical assessment I've come across is by Gano and Sandy from MIT:
http://dalmura.com.au/projects/Thermistors for the gradual application of heater voltage.pdf

MIT tried to minimise valve stress and failures in early valve computers such as the 'Whirlwind 1' by introducing thermistors in to the power supply. Gano did his Master's working on Project Whirlwind 1. Thermistors for filament cycling was a project by Sandy starting in 1955. The take-away message was that thermal stress related failure rate was quite low. That is the general message I get from decades of valve use in equipment. Tomer doesn't support his comments with any tangible references.

One has to be careful when interpreting Tomer's comments, as it was not uncommon at that time to use series heater connections placed directly across the mains AC supply - where variations between valves meant that some heaters are certainly stressed more than others, and the source resistance is lower.

Skizo, can you take us through the mechanism by which a GFCI can stop an amp from being killed, and what power related event (blip or quickie blink) are you referring to? That linked product has a MOV in it, for transient overvoltage clipping - but so do many amps.

 

[whoaru99]

Amongst the functions of GFCI and surge protection, it functions as an automatic disconnector. If the AC mains input to the device is interrupted, it automatically opens and requires a manual reset to again pass mains voltage.

I believe SkiZo requested information from the mfg. about the nature of AC anomalies (what dropout voltage, event duration, etc.) that causes the unit to disconnect, but don't know if such was ever received.

 

[whoaru99]

Please enlighten me: what function will be delayed with the amp?
Is it delay of B+ after power on ? ( not needed )
...

Depends if the PS caps can withstand the unloaded PT voltage x 1.414 until the tubes warm and pull down the supply voltage to nominal operating.

Arguably, the caps should be speced to stand that (a different discussion) but it's the main precaution I see for having a delay with SS rectifiers.

 

[Retrovert]

Retrovert, the only heater turn-on stress related statistical assessment I've come across is by Gano and Sandy from MIT:
http://dalmura.com.au/projects/Thermistors for the gradual application of heater voltage.pdf

Yes, I'm aware of their work. I've been trying to track down the military studies from the 1950s, which is not easy as all of those papers aged out and were discarded during the Johnson administration.

MIT tried to minimise valve stress and failures in early valve computers such as the 'Whirlwind 1' by introducing thermistors in to the power supply. Gano did his Master's working on Project Whirlwind 1. Thermistors for filament cycling was a project by Sandy starting in 1955. The take-away message was that thermal stress related failure rate was quite low. That is the general message I get from decades of valve use in equipment. Tomer doesn't support his comments with any tangible references.

It wasn't just MIT. The ENIAC had the same issue. All of the tube computers did. I'm well aware that Tomer does not, alas, cite the original studies. This hampered me in tracking down the original work.

I have read the original Whirlwind reports. Here's the relevant bit from one of the reports:

In this category may be included tube failures caused by heater burn-out. Heater burn-out is thought to be hastened by repeated application of full voltage to the filament, since the uneven heating that results sets up stresses which tend to weaken the structure. In the computer an attempt has been made to minimize failures from this cause by applying or removing heater voltage gradually over an automatically-controlled 5-minute cycle. With this precaution the number of failures from heater burn-out has been very low. Of about 300 tubes in the computer which have been turned on and off at least once each working day for the past 15 months, only 6 tubes have failed from burn out.​

The heater failure rate is low, but that does not mean that the damage to the tube is also low. It is, in fact, to the contrary.

In 1952, RCA wrote:

[T]he temperature is not uniform along the length of the heater, but may be concentrated in small sectors such as the uncoated sections of wire near the stem lead welds.​

Which leads us to heater flash. The Philips-derived tubes have heater flash which everyone handwaves it away. Really? We can ignore a heater driven into incandescence like a lightbulb through thermal runaway? Uh, no, that's just incredible stress on the tube.

That temporary heat spike degrades the surface, causing material to spall off and plate onto other tube structures. It also causes outgassing from metal, which is a gradual problem. You can't see this but it happens and the studies of the tube microstructures shows it.

Halogen bulbs have been well-studied and fail identically to vacuum tubes. Delayed startup for halogens greatly prolongs their lifespan. Ordinary lightbulbs tend to fail at startup (bulb flashes and fails) for the same reason: the filament is overloaded during the initial heatup.

Using NTC thermistors was common in the 1950s, but this practice fell into disfavor because of cost.

One has to be careful when interpreting Tomer's comments, as it was not uncommon at that time to use series heater connections placed directly across the mains AC supply - where variations between valves meant that some heaters are certainly stressed more than others, and the source resistance is lower.

Oh, I am very aware of this.

For those who are not, a series string of heaters (constant current) is problematic because, despite their specifications, the tubes actually heat at different rates. Voltage fluctuations along the string during initial heater warm-up typically causes the maximum heater-voltage limits specified in the datasheets to be exceeded. Some filaments become so hot they actually glow white.

So, yes, series heaters stresses the tubes, but these are generally not used in HiFi.

Skizo, can you take us through the mechanism by which a GFCI can stop an amp from being killed, and what power related event (blip or quickie blink) are you referring to? That linked product has a MOV in it, for transient overvoltage clipping - but so do many amps.

Some power strips have a feature that if the power stops and is then restored the strip does not automatically power back up. So if the power is unreliable and dies for a few seconds when it comes back the amp doesn't fry, either from the inductive kick rippling through the tubes or fully charged caps.

This is why toroids are so tricky to design into power supplies. The toroid core stores flux from the previous use—for a long time—and the core magnetization polarity is additive. So when power is re-applied at the wrong portion of the phase it will create a spike at double the peak. Everything cooks. Sensing circuitry is required to engage at the right point in the waveform.

 

[trobbins]

I have read the original Whirlwind reports. Here's the relevant bit from one of the reports:

In this category may be included tube failures caused by heater burn-out. Heater burn-out is thought to be hastened by repeated application of full voltage to the filament, since the uneven heating that results sets up stresses which tend to weaken the structure. In the computer an attempt has been made to minimize failures from this cause by applying or removing heater voltage gradually over an automatically-controlled 5-minute cycle. With this precaution the number of failures from heater burn-out has been very low. Of about 300 tubes in the computer which have been turned on and off at least once each working day for the past 15 months, only 6 tubes have failed from burn out.​

Would you be able to title/date that report - I recall that there were so many docs and the search facility a bit minimal. The use of a controlled voltage start up would be interesting to appreciate more on - I recall they used a separate rotating machine supply for the heaters - as would the year of that process, as it may identify the prior failure statistics due to heaters - although it seems that they were continually improving their statistics gathering as well as even tube manufacturer feedback and related improvements over the years, which can sort of mess with the raw statistics.

Which leads us to heater flash. The Philips-derived tubes have heater flash which everyone handwaves it away. Really? We can ignore a heater driven into incandescence like a lightbulb through thermal runaway? Uh, no, that's just incredible stress on the tube.

Wouldn't that be a thermal peak during turn on - runaway suggests the outcome of each turn on results in a failed filament.

That temporary heat spike degrades the surface, causing material to spall off and plate onto other tube structures. It also causes outgassing from metal, which is a gradual problem. You can't see this but it happens and the studies of the tube microstructures shows it.

I can see that thermal stress on a heater in an indirectly heated tube would cause movement inside the cathode tube - which can cause changing contact areas of the alumina coating with the cathode tube, and cause impurity issues within the alumina coating (which seem to be an improvement path focussed on over many years). Apart from the alumina coating spalling off at the open ends of the cathode tube, the filament itself may evaporate off - would there be other material involved that could lead to material coating other sensitive tube surfaces?

Some power strips have a feature that if the power stops and is then restored the strip does not automatically power back up. So if the power is unreliable and dies for a few seconds when it comes back the amp doesn't fry, either from the inductive kick rippling through the tubes or fully charged caps.

I find it interesting to go through the mechanisms by which an amp could damaged from repeated application of mains AC.

Transformer primary related in-rush from core characteristics would I think only impact on a mains side fuse - which if it is appropriately speced should cope with a small frequency of high in-rush levels before eventually blowing from repeated high I2t levels.

If there was a mains side thermister, then it could be considered hot, as would cathodes in tubes - and probably with a longer time constant than say bias supplies or coupling caps. But I can't see a failure mechanism there, and stress levels would be interesting to tease out from the time-constants involved.

Ciao, Tim

 

[Retrovert]

Wouldn't that be a thermal peak during turn on - runaway suggests the outcome of each turn on results in a failed filament.

Almost, but not quite.

This is where it gets really interesting. Here's a summary from my (in progress) monograph. Took me forever and a day to work all of this out and how it happens.

A tube heater (aka filament) is a positive temperature coeficient (PTC) device. So the hotter it gets, the higher its resistance. That's why the heater doesn't eventually pass enough current to become hot enough to melt or burn out. As it heats the temperature rise slows current flow and this eventually stabilizes.

A cold heater is a dead short. Portions, usually the ends because of welds to dissimilar metals and a lack of conduction to the cathode, have lower initial resistance and thus pass more current. As they heat their resistance increases, obviously faster than the heater’s remainder which is not yet hot enough to choke off all current flow. This theoretically would choke off the current and slow the heating. But that does not happen, because the remainder of the heater is cool, so its resistance remains low.

As a result the total heater resistance end-to-end remains sufficiently low to permit (actually force) current flow through the high-resistance bottleneck even though the higher-resistance ends are sufficiently higher resistance that if the entire heater had this resistance current would stabilize at the steady-state value. The low total resistance forces more and more current through the higher resistance portions. Remember, current flow depends upon total resistance end to end.

So as long as the total resistance remains low, more current flows through the hotter regions and continues to unevenly raise the temperature of the hot portions. Eventually they burst into incandescence.

This is how heater flash originates. The Philips tubes are particularly vulnerable because of the way the filament is welded in place.

So you can see this is a thermal runaway process, even though it is PTC. Ok, you asked, maybe so, but why does it stop? Magic!

Well, if by "magic" one means the Stefan-Boltzmann Law. It's pretty close to magic even though it doesn't cut ladies in half. The heater cools through radiative and conductive losses as per the 4th power of temperature, so these high thermal losses end the metastable state (i.e. runaway) and the heater eventually stabilizes at its operating temperature instead of melting into slag.

The heater temperature is roughly proportional to the 4th root of the heater power (I × E) and the cube root of voltage. So if the heater temperature increases by 5% the heat losses increase by 625%. That is the limiting factor. Basically as the point of incandescence is glowing the rest of the heater is rapidly heating to the point that the overall resistance is high enough to choke off current flow and prevent a meltdown.

Tungsten is a very tough material with a high melting point which is why it glows but doesn't melt, until one day it burns through and we hear the magic "ka-ching!" sound of the cash register ringing up a sale.

But this process is far from gentle on the tubes. See below.

I can see that thermal stress on a heater in an indirectly heated tube would cause movement inside the cathode tube - which can cause changing contact areas of the alumina coating with the cathode tube, and cause impurity issues within the alumina coating (which seem to be an improvement path focussed on over many years). Apart from the alumina coating spalling off at the open ends of the cathode tube, the filament itself may evaporate off - would there be other material involved that could lead to material coating other sensitive tube surfaces?

Remember, this isn't going on in a vacuum. (Ok, it is going on in a vacuum, but I couldn't resist saying that!)

Tungsten gets boiled off and migrates onto other structures. This impairs emission and absorption. All sorts of impurities exist in metallic tube structures (other metals, gasses absorbed in metals, etc.) and those get baked out with heat. (Which is why tube heat is very bad and tube coolers are a good idea.) Same thing for the mica insulators. With heat the cathode material (Ba and Sr oxides) itself loses oxygen which is why the cathode eventually dies. The heater itself expands and when this happens unevenly it rubs and abrades against its sleeve.

As the heater becomes uneven in temperature the cathode gets hotspots which damage its emitting surface (I've elsewhere described the coalescing of emitting islands) which causes it to unevenly emit and eventually fail.

Tube failure analysis is a complex, messy business. Someday I will have my writeup finished.

The takeaway is, as I previously wrote, to slow down the heaters with a current limiter (NOT NOT NOT a constant-current source, which forces current through the tube) to limit the rise to a more gentle increase per second which does not have such unequal heating.

 

[s-petersen]

I realize that the thread has gone off topic, but I was always wondering what caused the Mullard flash.
I guess a thermistor in series would be better for the heaters.The series string AA5 radios often used a pilot light as the sacrificial element in the string, they would go super bright at turn on.

 

[Retrovert]

I realize that the thread has gone off topic, but I was always wondering what caused the Mullard flash.
I guess a thermistor in series would be better for the heaters.The series string AA5 radios often used a pilot light as the sacrificial element in the string, they would go super bright at turn on.

Nope, we are still 5x5 on topic. The OP had asked:

Am switching from tube to solid-state rectification, and am installing a delay timer to avoid making bottle rockets out of my beloved Gold Lion KT66's. Bought a pair of 555-based boards with approx. 30 seconds of delay off the 'Bay, but a trusted friend says I want a good 60 seconds or more of delay. Any suggestions?

So it's not purely about B+ delay, but also about heater delay. He still wants to avoid burning up his KT66 tubes and that's what the discussion is about.

Now there is less mystery in the world about the infamous light bulb effect in Phillips family tubes (Amperex, Miniwatt, Mullard, Siemens, Valvo, etc.).

A thermistor is one way to solve this, but it isn't the best way for a few reasons. One, the thermistor has state. One hot it tends to stay hot. So when the amp is cycled (a no-no) it is still ready to go. Two, the thermistor burns power. Three, thermistors can't handle the current. A bi-metallic switch was created for this purpose called a "Surgistor" but it has the same contact problems as a relay and it burns power. Four, the thermistor time constant can't be tweaked.

A much better way is a MOSFET (driven by either a ramp up voltage from a 555 or an RC constant) in parallel with a resistor. That way the circuit is limited to, say 50% of heater power until the MOSFET turns on at which point the resistor is switched out of circuit. Yes, you could use a relay but switching high currents with a relay is bad for the contacts and one day they will weld shut and your heater delay suddenly stops working and whatchagonnado?

It wasn't that the pilot light was the sacrificial element, per se, just that its current limits were exceeded when the tube heaters were powering up and pulling incredible current through the string even for light bulbs which were not rated for such high current. This ought to demonstrate that all is not well in a series heater string.

Behold the power of runaway. It works in counter-intuitive ways.

 

[sKiZo]

Skizo, can you take us through the mechanism by which a GFCI can stop an amp from being killed, and what power related event (blip or quickie blink) are you referring to?

By "blip", I mean a full, but momentary, interruption of power which won't allow a tube amp to discharge properly. The sort of thing that is common to storm or accident related events that trigger a sub-station breaker to trip and reset, or an automatic reroute in your area's power grid. That's what can cause damage. In this case, ignore the GFCI feature of the adapter I linked - the important part for our purposes is the non-latching feature that requires a manual reset.

We had one here last week where the power blipped several times in quick succession, with a couple major brownouts thrown into the mix before the grid said uncle and it all went dark. That was actually pretty scary - was surprised not to see any smoke curling up out of the appliances. I expected to have to spend some time in the stone age before I saw power again, but service was restored a couple hours later.

I believe SkiZo requested information from the mfg. about the nature of AC anomalies (what dropout voltage, event duration, etc.) that causes the unit to disconnect, but don't know if such was ever received.

Not sure if this is what you're looking for, but I found that spec lists the unit with a trip level of 4-6 mA and a trip response time for ground fault less than 25mS. In any case, way faster than you'll be able to jump up out of the comfy chair and pull the plug.

 

[trobbins]

Retrovert, a few comments as I see it.

An indirectly heated cathode typically has a short section of filament exposed between incoming terminal and cathode tube - most of that short section is usually coated in alumina, with only a little tungsten wire exposed. There may be top-of-valve loops where the heater comes out from the cathode tube. Some heaters are multiple bent sections inserted in to the cathode tube - some like the EF86 are a single heater stick (with an out and back close helix filament). The output stage valves have much beefier heaters.

Any transfer of tungsten from the filament is a slow process, even when incandescent, and would be in to the alumina rather than free to vacuum, so I can't see a significant amount of tungsten ending up as gas molecules from heater flash.

Alumina impurities are known to migrate through the coating over time, but at normal heater temps that is a slow process, and appears to show up as heater-cathode resistance change.

The only heater resistance change plot I've seen is a curve of heater current versus time in the Gano/Sandy paper, and indicates a nominal 1:4 resistance increase.

For the small % length of heater outside the cathode structure, the thermal transfer from the filament will be only from radiation - the thermal transfer rate increasing quickly as filament temp increases - the ambient temp to transfer to won't change in a short time until the glass inner surface temperature rises (but even then the temperature differential is still very large). Heater inside a cathode tube will mainly radiate to the cathode, with only small conductive spots along the length of the heater - the thermal transfer rate not increasing as quickly, as the cathode inner wall temp will also increase at a fast rate (with equilibrium temperature differential equating to equilibrium heater power dissipation). Given the same current, and initial conditions, the filament in the cathode structure would rise to a higher temp quicker than the filament extremities, and hence by far the bulk of the heater resistance would rise quickly (as per the Gano/Sandy plot. I don't think this is how you interpret the situation?

 

[trobbins]

sKiZo, in my region the electric utilities use an 'auto-recloser' switch to disconnect users for a few seconds when a fault current is experienced, and then reconnect them, with a few retries before finally disconnecting users if the fault current is persistent. This protection mechanism is great for avoiding long outages from transient lightning strikes, or wind blown aerial cables touching, back in the days when only fuses were used to protect power lines from over-current. It sounds like you have a similar protection scheme.

If it was a nearby lightning strike then AC voltage could spike high, otherwise the result is like turning off the power switch for a few seconds, and then turning back on.

As indicated in #25, I don't see that power supply toggling would normally be a fault causing event if the occurrence of such toggling was minor.

The 4-6mA and 25ms refer to the earth leakage disconnect current and time of the GFCI - that form of disconnect would normally only relate to a breakdown of insulation between AC mains active and earth within your equipment (or perhaps an errant finger poking around inside an amp).

 

[Retrovert]

An indirectly heated cathode typically has a short section of filament exposed between incoming terminal and cathode tube - most of that short section is usually coated in alumina, with only a little tungsten wire exposed. There may be top-of-valve loops where the heater comes out from the cathode tube. Some heaters are multiple bent sections inserted in to the cathode tube - some like the EF86 are a single heater stick (with an out and back close helix filament). The output stage valves have much beefier heaters.

This is true. But what you don't know is where are the hotspots inside that sleeve. You can't see them. They exist. And they overheat the cathode in those regions while the heater is equilibrating.

That damages the emission surface by baking oxygen out of the alkali oxides, and other oxides must migrate to the surface to replenish it. Overtime those spare oxides vanish, and the oxygen starts to poison the grid, and this kills the tube.

Any transfer of tungsten from the filament is a slow process, even when incandescent, and would be in to the alumina rather than free to vacuum, so I can't see a significant amount of tungsten ending up as gas molecules from heater flash.

Deposition is not restricted to heater flash, it happens with general use. It just gets a lot worse with overheating. The useful life of a heater has been defined as the number of hours before the filament is reduced by 10%, at which point the tube starts to fail. It all comes down to the filament thickness. That magical value was set over a hundred years ago. It may have been Langmuir, not Dushman, who actually came up with it; I haven't tracked down and read all of his papers on this subject. It's unfortunately not very interesting.

Every time that tube turns on it gets overheated and that boils tungsten off. Which is where the ions come from to poison the cathode over time. They are in there in the tube, part of useful things. The support structures, the mica, the cathode, all over the place.

As far as the severity of the problem, well, it's real. I've explained about tungsten filament lightbulbs and halogen bulb lifespan. So we know that the tungsten gets plated onto the inner surface of the bulb. Regulating temperature with a lower voltage greatly prolongs lifespan in bulbs and a tube can be treated as a bulb for purpose of the filament.

Alumina impurities are known to migrate through the coating over time, but at normal heater temps that is a slow process, and appears to show up as heater-cathode resistance change.

Yes, that's one source of cathode poisoning. But it's also the oxygen that gets baked out of the alumina which sticks to the grid. There's all sorts of stuff in the tube that gets boiled out of metals. A lot of work went into removing absorbed and adsorbed gasses prior to assembly. For example, I vaguely remember a paper about adsorbed material during solvent cleaning killing klystrons. Turned out the clean tubes failed faster.

For the small % length of heater outside the cathode structure, the thermal transfer from the filament will be only from radiation - the thermal transfer rate increasing quickly as filament temp increases - the ambient temp to transfer to won't change in a short time until the glass inner surface temperature rises (but even then the temperature differential is still very large). Heater inside a cathode tube will mainly radiate to the cathode, with only small conductive spots along the length of the heater - the thermal transfer rate not increasing as quickly, as the cathode inner wall temp will also increase at a fast rate (with equilibrium temperature differential equating to equilibrium heater power dissipation). Given the same current, and initial conditions, the filament in the cathode structure would rise to a higher temp quicker than the filament extremities, and hence by far the bulk of the heater resistance would rise quickly (as per the Gano/Sandy plot. I don't think this is how you interpret the situation?

Mmmmm, I believe that cannot be the case. Here's why. If it were true, the heater middle would be heating faster than the ends. So the resistance of the heater as a whole would be so great as to limit the current at the endpoints, and they couldn't glow. We know the only way to force a current through those spots is if the rest of the heater resistance is sufficiently low as to pass a high current. Which couldn't occur if the rest of the wire was hot, as it would be high resistance. So the middle must be cooler than the ends to permit that current flow and thereby heat the ends. QED.

Again, we also know that the heater draw is not equal across the length and it heats unequally. But what we don't know is where all of the internal hotspots are. They exist, but we can't see them since they are covered by the cathode sleeve. (Thermal imaging would turn this up. I don't have one of those thermal cameras.) We can see the external ones, which is how I can demonstrate that the phenomenon of unequal heating exists. Those are particularly problematic spots. It is known that the heater does not evenly heat and the cathode has hotter spots, typically towards the middle and cooler at the ends.

The heating in the wire is from current flow. Losses from radiated thermal transfer AND from direct conduction into the cathode sleeve keep it from melting. So we have two ways to lose the heat. Remember, the wire touches the sleeve. It isn't all radiated.

 

[sKiZo]

If it was a nearby lightning strike then AC voltage could spike high, otherwise the result is like turning off the power switch for a few seconds, and then turning back on.

No few seconds about it ...

- We're talking instantaneous cuts, which I'm sure are followed by a spike.
- We're also talking several "blips" followed by a prolonged outage in many cases. I've also seen one good blip attributable to an auto accident down the road, with a pole hit and the power rerouted automatically.

The 4-6mA and 25ms refer to the earth leakage disconnect current and time of the GFCI - that form of disconnect would normally only relate to a breakdown of insulation between AC mains active and earth within your equipment (or perhaps an errant finger poking around inside an amp).

Not sure what feature of the unit those numbers apply to, but I imagine it's pretty much a max response time to any event that will trip it and kill the wall power. And yes, whatever is happening on the incoming line, it DOES trip any time something questionable or mysterious occurs in the power system, including brown outs, so I'm quite happy with what it DOES do, and not all that concerned with how it does it.

PS ... I also have a "whole house" protector installed between the meter head and service box. That seems to do a pretty decent job - I had a lightning strike on the pole outside the house that knocked out the transformer and no damage inside other than a stain where I was sitting ...

 

[trobbins]

The link to John Harper's webpage http://www.john-a-harper.com/FilamentHeating/ was interesting for the experimental results of the indirectly heated cathode 6SN7, which show a knee in the inferred filament temperature rise versus time plot for constant V application.

In this situation, the temperature of the filament mass only stabilises when the input power to the filament is equal to the heat dissipation away from the filament, and during turn on of power all those variables are changing from initial condition to final steady-state condition. In a microscopic scale, the heater has only a very low percentage of its surface touching the cathode inner wall - The 1962 RCA Tube Design book (https://archive.org/details/RCA_1962_Electron_Tube_Design), page 220, fig 4 has a good diagram - and so radiation transfer of heat dissipation away from the heater is dominant and very dependant on temperature differential. By inspection of the parts, the cathode would appear to have a similar or somewhat greater relative mass than the heater.

During the first few seconds after voltage application, the filament and surrounding alumina coating temp is rising rapidly. The knee represents the changing ability of the heater to dissipate heat away to the cathode tube. In the 300B directly heated example, the filament radiates to a near constant temperature surface (mainly the large plate structure, and possibly the inner surface of the tube glass, which is basically opaque to thermal radiation when at room temp - both structures have a very large thermal mass compared to the filament). In the 6SN7 example, the cathode inner surface temperature is rising substantially (be it at a slower rate than the filament due to different thermal mass and its ability to radiate heat away on the outside of the cathode), and hence the heater heat dissipation capability is changing significantly when compared to the 300B heater situation.

 

[Retrovert]

I'm unsure what point you are trying to make.

Yes, most of the heat is radiated. But the heater touches the cathode sleeve in a number of places. Those drawings are for a COLD filament, and it does expand when hot. The mechanical stress from heating-induced stretching of the heater is considerable, and the abrasion inside the sleeve was a known problem.

But none of this is particularly relevant to the issues I raised.

What matters is that the heater does not evenly heat (flash is just one example) and reducing thermal stress prolongs life. So ramping up the heater voltage during the initial heating cycle reduces the stress on the tube. Just as delayed B+ reduces cathode stress and damage. And, finally, just as cooling keeps the other tube structures from having gas baked out of them.

Be nice to your tubes and they'll live longer.

 

[trobbins]

I think the intent of the RCA diagram was to indicate that at a microscopic scale, even if the heater appears to be continuously touching the inner wall of a cathode tube, in fact the roughness of the alumina coating means that very few points are in direct contact. Somewhat similar to using mill finished heatsinks when using mill-finished semiconductor device pads. So given that just a 'line' on the heater surface could ever be in close contact with the cathode tube, then radiation transfer likely applies to >>99% of the heater surface.

I think it is worth having a reasonable awareness of basic operation in order to rationalise the wide spread opinions that can be put - of which the need for heater start up ramping is one, and the perceived disadvantage of heater flash is another. This also applies to the perceived need for a delay relay as well, and what if any benefit can be obtained for most amplifier applications. The presumption is that all amps without relay delay cause output stage valves to fail as 'bottle-rockets'. If there are well appreciated reasons for using a relay delay, then that appreciation will direct the diy relay constructor to such things as the delay time, and how it should handle resets/power cycling to the amp, and what other changes may be in play such as adding a thermistor/resistor slow start, or mute switch contact.

 

[kward]

I've been casually following along on this thread. Help me understand really how important this issue is to address if manufactures like Fisher, Scott, Harmon Kardon, which used diode rectification in their amps, never worried about B+ delays or slow ramp up? Up until several years ago when I replaced the output tubes due to age only, my HK A500 amp was running the original 1963 output tubes without any problems. Also, I recently acquired a Fisher 500C, and three of the four output tubes sure look like originals--worn and tired, but still chugging away. No B+ delay or B+ ramp in those amps either.

I can see how a B+ delay might be important if you are trying to address the situation of quick turn off then turn back on while the output tubes are warm. (example, if you have a toddler who is within arms reach of the power switch). But even in that case, it seems to reason, that it only seems applicable if you have a fixed bias amp where, on power off, the output stage loses bias voltage quicker than the tubes cool down, so that on re-power on a second or two later, output tubes are still warm (and ready to conduct) but bias is not fully reestablished yet, so output tubes conduct like mad and blow the fuse.

I mean, I don't know, that's my reasoning on this. But help me understand if I'm missing something important.

 

[Retrovert]

Sigh. This is going in circles. I've repeatedly explained the issues and am basically done repeating myself. Believe it or not, do your own research, take your own counsel.

The fact that manufactures of consumer electronics:

(a) did NOT understand this issue in the 1950s or 1960s
(b) could not add circuitry that competitors did not and still remain cost-competitive
(c) repeatedly defied specified limits for tubes and other components
(d) could not simulate designs
(e) used design topologies we would not​

does not constitute "proof" that a problem does not exist.

Just look at the number of amplifiers which blatantly ignored or openly defied the tube manufacturers' recommendations for screen supply limits, grid and screen stopper resistors, grid leak resistors, etc. Then there's the lack of snubbers on transformers and diodes. And a lack of inrush limiters, which were often added after-market with the same excuses of, oh, if the manufacturer needed that it would have been included. Same for DC heaters.

The fact is that we can fix those problems because we better understand them and are not limited by cost multipliers in the supply chain.

Here is a simple example. Consider DC restoration to eliminate blocking distortion. Very few amplifiers (i.e. none), to my knowledge, did this. It's an easy fix. When DC restoration is omitted from a television it looks terrible. Yet many TV sets omitted it for cost-savings reasons. Doesn't mean it isn't necessary.

Dynaco used a single rectifier when similar ampifiers used two.

As I have previously explained, the islands of charge on the cathode must (a requirement) fully coalesce to form a proper emitting surface. Not allowing that to happen before pulling current damages the cathode and shortens the tube's lifespan. The large-scale studies done by the military and computer builders in the 1950s demonstrated the value of heater regulation, delayed heaters, delayed B+, and tube coolers. If you read the original papers on cathode emission it is clear that electron emission occurs from cathode hot spots and higher spots, and this improper emission gradually degrades the coating.

A transmitting tube, which pumps far more electrons from cathode to plate, can be destroyed in days, even hours, by pulling B+ before the cathode surface is properly configured. That's why people repeat the mythology and hearsay that the phenomenon is limited to RF tubes. It isn't, of course, that's just ignorance and internet "lore". Like many physical behaviors, it just happens much faster with higher voltage. Temperature and voltage accelerate nearly all reactions. The cathode work function is such that it readily emits more electrons when hot, far more than a plain tungsten filament.

The same goes for heaters. I've read the reports for the MIT Whirlwind and what I could find about the ENIAC. These designs investigated and then honored limits for both voltage and current through the tube to prolong lifespan.

The issue isn't hot cycling an amplifier; that's a different issue and largely a red herring. While that happens with power glitches, it is uncommon. Destructive, but uncommon. The issue here is not stressing the heater with high inrush current. High inrush current stresses capacitors, rectifiers, and transformers, too. It's always a bad thing.

A relay is not the proper way to create a slower heater voltage, as I previously outlined, because the heavy current damages the contacts. A MOSFET fed heater with a parallel resistor to lower inrush current is the correct circuit.

As I said, you can believe this or not. I've extensively researched this and someday I will finish my monograph on the subject. I'm still tracking down the original papers by the military, which is not easy as they've all aged out and were discarded decades ago.

 

[kward]

Ouch, it appears I may have unintentionally poked a nerve. Looks like you've done a lot of research and given a lot of thought to this.

 

[Thunderbud]

Retrovert, I Like Music, and all who posted here in response to my query about changing the amount of delay on a timer/contact set ;

Thank you all for your time and attention. Much of the discussion was far beyond my training and experience, but that just proves to me that I brought my question to the right folks. LOTS of talent and helpful attitudes here. In the end, I emailed the manufacturer of the little timer boards I'd purchased. I got a same day response (!!) identifying which resistor needed replacing to change the total delay. Without your combined ideas and suggestions I would have been swapping it by " rote " without understanding what I was doing. Again, thanks to ALL who replied. !

Bob Shively.