No, that's not actually how the capacitor size limit arises. With a low filter input capacitance, an arc may happen, but it won't sustain. As soon as the AC input crosses zero, the arc goes out. When the capacitance is over a certain limit for each tube, the arc once started can sustain itself - and lead to tube damage and destruction.
All tubes occaisonally spark over - due to loose cathode material, microscopic particles left in during manufacture, etc. In audio it results in an audible click in your speakers, but in the tube days long before CD's, people assumed it was dust on the record or if in a radio, "static". When transistors became available and designers begain to mix transistors and tubes, the transistors were significantly less reliable than when used in non-tube applications. It was realised that this was due to the occaisonal internal spark in the tubes causing transients that could take out an associated transistor. Certain types of transistor, such as the older type of diffused alloy are much more tolerant than epitaxial types. In TV sets, specific devices were included (spark gaps) to prevent CRT internal sparks from causing destructive transients in transistor circuitry.
The victims started arcing after some use.
It's certainly not reversible. On one, the filament damage was visible.
Iload was 60mA dc at 450V, but I suspect that the dc voltage was lower at the time of the arcs, and the current scaled down too.
Rod is right, the only thing limiting the size of capacitance is the peak repetative current capability of the valve. You can increase the capacitance provided you keep the peak current under the usual limit. Some data sheets even point this out:
http://www.shinjo.info/frank/sheets/074/5/5R4GY.pdf
http://www.shinjo.info/frank/sheets/127/5/5R4GY.pdf
Of course, adding limting resistance means wasting a lot more power, which is why manufacturers usually assumed you wouldn't want to do it. Their recommended maximum capacitance is a sort of optimum between efficiency and smoothing.
The cap value is determined by the peak current, as I said.
See the RCA 5R4WGB data sheet.
Attachments
A long ago, I donated a 5R4GY to science making the thermionic emission experiment, after that, designing a regulated 450V PSU; on early tests, the only caps at hand was 220μF/450V, commonly used in TVs, then I put two in series and using a 550V-0-550V trafo the glow at start up was very funny, with no other load!
This is another face of ion bombardment, an extreme case when a lot of ions interact with electrons and the cathode, the blue-violet glow can vouch for that.
The solution was quite easy, a 30s delayed relay between rectifier and caps, and two birds with one stone, all valves are pre-heated on the amp.
Thanks, interesting. There is a graph Fig 60 Harmmann/Waganer Vol1 1950 pp106-118 which aims to illustrate how sparking vulnerability might develop/change with use, presumably for a rectifier.
60mA@450V is squarely in area 1 for a 5R4GY......so presumably the area I/II boundary might change depending on peak repetitive current, and really it's a series of curves?
Last edited:
Running simulations using Duncan Amps PSU Designer tool reveals what Rod Coleman's 5R4GY might have been asked to do in the 32uF/4uF warm/cold start test - at least for pre-warmed start. The issue is initial charge of the empty electrolytic capacitor. The first conduction requires peak Ik of c 1.57A with 32uF, versus 0.53A with 4uF. Such peaks stabilise to c 0.25A peak repetitive current within a few cycles for both 4uF and 32uF.
When the cathode is started from cold, some parts of it achieve emission before others, and cathode current demand is satisfied solely from these parts - if Ik demand is high enough this generates local heat in the cathode coating increasing emission which generates further heat until cathode coating locally evaporates (H/W Vol 1 pp106-118). Evaporation realises gas and ions which degrade vacuum resistance to arc formation when the rectifier enters blocking phase ((H/W Vol 1 pp106-118). This fits Rod Coleman's observations, I think. And simulation provides some insight into likely current demands likely to invoke such mechanisms, which ultimately relate to conditions at the cathode and current demand for a relatively short time.
In power triodes/pentodes can similar cathode conditions ever exist at cold cathode start, and as the cathode warms up when the cathode is in saturation and grid control doesn't operate? I think not - current would be limited by impedance of OP tx primary before magnetising current is established, ie big inductance and some DC resistance. But what do you all think ?
Are 5R4GY cathodes similar enough to say EL34 cathodes to equate current levels at which such effects might be invoked, based on Rod Coleman's tests ?
When the cathode is started from cold, some parts of it achieve emission before others, and cathode current demand is satisfied solely from these parts - if Ik demand is high enough this generates local heat in the cathode coating increasing emission which generates further heat until cathode coating locally evaporates (H/W Vol 1 pp106-118). Evaporation realises gas and ions which degrade vacuum resistance to arc formation when the rectifier enters blocking phase ((H/W Vol 1 pp106-118). This fits Rod Coleman's observations, I think. And simulation provides some insight into likely current demands likely to invoke such mechanisms, which ultimately relate to conditions at the cathode and current demand for a relatively short time.
In power triodes/pentodes can similar cathode conditions ever exist at cold cathode start, and as the cathode warms up when the cathode is in saturation and grid control doesn't operate? I think not - current would be limited by impedance of OP tx primary before magnetising current is established, ie big inductance and some DC resistance. But what do you all think ?
Are 5R4GY cathodes similar enough to say EL34 cathodes to equate current levels at which such effects might be invoked, based on Rod Coleman's tests ?
The grid still provides significant shielding from the stong electric field of the anode, even if the 'correct' bias takes a while to stabilise. An ordinary AC-coupled circuit will cause at worst a strong pulse of anode current, but it won't reach the levels possible in a rectifier; a power valve will just shrug it off. The problem is if the grid is pulled very positive, e.g. more than 100V.
The 5R4Gx observations do support the idea of this non-uniform warming - because if the rectifier is preheated, there is no arc.
We know from H & W that the cathode surface is very rough and uneven, and so portions of the coating will reach emission temperature before others. Modern production DHTs and Beam Power Tubes are the same in this regard.
If the load's current demand is in the Ampere region, as may be found in rectifiers, we are likely to see an arc.
With output DHTs and Beam Tubes, the demand is less likely to be in the Ampere region, even at startup. As Merlin notes, the grid would need to be strongly positive for this to happen.
However, even if the grid remains at 0V (a very optimistic case for the usual DIY ac-coupled amplifier) you still present a current demand to the power valve which is higher than the normal peak current because the (unloaded) supply voltage is higher, and at zero volts, the grid is at the maximum demand level, with an auto-bias amplifier.
In this situation, an arc is unlikely, because the current demand is not sufficient, but that is not the end of the risk.
In this case, the risk is of what is described in H & W on pp109: at the tip of some high-point in the emission surface, which has warmed-up an instant before its neighbours - we find that all of the current-demand is placed on this local hot-spot, leading to a thermal-runaway at that point, causing local evaporation of the coating, with appreciable out-gassing.
So the power valve survives the switch-ON, but at what cost? At each switch-ON the process repeats, and the valve suffers cumulative loss to the coating, and maybe more damagingly, evolution of gas. Both of these effects reduce the life of our expensive power valve.
Given the cost of power valves of the 30-40W class, I don't know why anyone would wish to run this risk.
In grid valves during warm-up, to support the conditions for runaway leading to hotspots, conduction simply has to be sufficient to support saturation from a small part of the cathode I think. ie field near the possible hotspot merely has to support transport away of enough carriers to saturate emission there locally. g-k bias holds the key then. Crudely 'normal' g-k curves should apply well enough to work with I think, if anything field will be more positive (conductive) than in normal operation perhaps.
Scenario 1 : Cold start, cathode bias with bypass. g-k initial potential is zero, and becomes more negative only as bypass C charges. Rate C charges is set by saturation current of cathode conduction. During warm-up cathode might be locally saturated I reckon - until g-k potential reaches control point. And such conditions seem similar enough to Rod Coleman's 5G4GY test scenario I think, because it is the time cathode spends in saturation whilst charging capacitors that is the issue in both cases I think - for the rectifier charging 4uF through 450V, for the power grid valve charging say 1000uF through 30V - devils advocate choice of bypass C/smoothing C decides the winner if saturation current is similar.
Scenario 2 : Cold start, normal grid bias, grounded cathode. g-k potential is immediately more negative than scenario 1. No bypass C to charge, time cathode spends in saturation far shorter in principle. Seems obviously advantageous over scenario 1.
Scenario 3: Cold start, deep cut-off grid bias, grounded cathode. No conduction, cathode never in saturation.
But I think whether scenario 1 might ever mimic Rod Coleman's 5R4GY test (from cathode hot spot development point of view, not arcing) might be mitigated by limiting impedance of the anode circuit during warm-up. ie whether saturation current can be provided by the anode circuit. It's also devil's advocate in assuming cathodes are similar. But it doesn't seem out of the question, IMO.
Scenario 1 : Cold start, cathode bias with bypass. g-k initial potential is zero, and becomes more negative only as bypass C charges. Rate C charges is set by saturation current of cathode conduction. During warm-up cathode might be locally saturated I reckon - until g-k potential reaches control point. And such conditions seem similar enough to Rod Coleman's 5G4GY test scenario I think, because it is the time cathode spends in saturation whilst charging capacitors that is the issue in both cases I think - for the rectifier charging 4uF through 450V, for the power grid valve charging say 1000uF through 30V - devils advocate choice of bypass C/smoothing C decides the winner if saturation current is similar.
Scenario 2 : Cold start, normal grid bias, grounded cathode. g-k potential is immediately more negative than scenario 1. No bypass C to charge, time cathode spends in saturation far shorter in principle. Seems obviously advantageous over scenario 1.
Scenario 3: Cold start, deep cut-off grid bias, grounded cathode. No conduction, cathode never in saturation.
But I think whether scenario 1 might ever mimic Rod Coleman's 5R4GY test (from cathode hot spot development point of view, not arcing) might be mitigated by limiting impedance of the anode circuit during warm-up. ie whether saturation current can be provided by the anode circuit. It's also devil's advocate in assuming cathodes are similar. But it doesn't seem out of the question, IMO.
If the Valve is biased such that damage may be done (however small), it will probably occur at the instant that the first location to reach emission has reached enough temperature. So I believe that the cathode bypass cap will be uncharged at this point.
Yes, this is the set-up that will protect your valves from risk of any degradation, even with solid-state rectifier diodes. I would recommend the practice, if non-thermionic rectifiers are chosen.
But, you have to be sure the bias appears instantly.
The energy required to do some mischief to the cathode is probably small. Since any such damage repeats at every switch-ON, it is undesirable.
The OT inductance will limit the rise-time of current, but may not limit the current substantially.
Of course you are correct in practice.
What we have here is one of those arguments I call "standing on a sheet of paper". It goes like this...
We all know that because of the curvature of the Earth if you are taller or if you are standing on something that is tall you can see further. We also know that the taller you are the better chance you have of being able to see over some random object. So I argue that if I stand on a sheet of normal paper borrowed from my ink jet printer I can see farther. Next I'll argue that a sheet of cardboard would work many times better allowing me to see many times further than the paper allowed. We then get into a 200+ post long thread arguing over the best brand of cardboard.
The cathode stripping thing is just like this. A 200+ post long thread about some effect that might in theory have some effect but in practice does not maters at all in normal audio gear. (different story in very high powered radio transmitters and antiques.)
Here is my CHALLENGE: Use NUMBERS. How this any of this effect things you can MEASURE. No more hand waving So when talking about differences in tube life is this 8,000 vs 8,000.00001 hours or 8,000 vs 16,000 hours?
If you do find a study that measured the effect, where they using the same kind of tube you are using? Was the B+ voltage comparable?
Without numbers this is exactly like arguing about the brand of cardboard.
Member
Joined 2009
Paid Member
I love the analogy!
And more then 'numbers', you actually need 'statistics' - anecdotes about a few amplifiers here and there doesn't really cut it either, you need proper sample sizes from the right populations of parts. These days, we don't always know the provenance of the parts we buy which adds the risk we have a 'bad batch' pawned off on ebay.
Oh, so you are certain that it is absolutely safe to apply B+ to cold cathodes?
Then I will beat you with your own stick.
Where are your NUMBERS that give you this assurance?
In your assessment, how many cold starts gaive a certain degradation, and how does this sample compare with the sample that was preheated?
Of course, you have no numbers, and are adding nothing whatever to the debate.
(I would be delighted to be wrong about this, so if you really have evidence, bring it on).
Member
Joined 2009
Paid Member
The cold start is 'innocent until proven guilty'. We've been looking at the physics, some numbers involved, and it hasn't been proven guilty to my satisfaction (not saying that I'm the only judge around here).
Personally I'm interested in getting to the bottom of all the possible mechanisms(s) in play first, because that's the only way to predict any likely benefit in the scenario where NOS valves are irreplaceable and one is interested in the difference between say 3,000 hours life and 12,000 hours life - totally arbitrarily of course. There's no opportunity to use statistics in the case where one has only a few at best samples, plus sacrificing them to experiment is self defeating. Also, on the face of it, much might depend upon the circuit at least as far as cathode hot spot development is concerned - specifically bias arrangement and cold anode impedance.
I think its at least fair to say there is a case to answer, and it would probably get to court. Evidence is still being assembled though, I think.
- Status
- Not open for further replies.