The AM Forum

My feeling is that the typical "plate cap" is not very effective at radiating heat.
I didn't think much about it at all until I had to look closely at clearances in one of my old Henry 2K amplifiers, when I was trying to pop a 4-400 into a hole designed for a 3-400 (originally), finding not enough vertical clearance.

Anyhow, the shiny 1/4" (or so) diameter plate terminal as shown in the picture is actually an add-on to the actual plate wire that comes through the glass. A ton of heat is transmitted up through that plate wire from the plate. Imho, the stock terminal, that is usually held on with just a set-screw is inadequate to transfer heat from the plate wire, and also sits over the top of the plate wire to glass seal.

Now, that last bit might not be a bad thing, since you don't want really want fastchanges in the glass to metal temperature. Otoh, given that there is constant air going past, perhaps it would be best to remove heat from that juncture as effectively as possible.

In the even you want slower change, then putting silicone rubber or the standard glass bonding "lamp cement" under the cap would make sense... does anyone know which way they want it to work - cool slowly, or remove heat as fast as possible??

But, my thinking is that a clamped (drilled hole, slit along vertical axis of the hole, and screw to pull it together) plate cap would be much more effective at heat transfer than the present system that has two separate discontinuities (plate terminal to plate wire, and plate cap to plate terminal).

Also, since the air in modern air system cooling is going vertically, not horizontally, the fins on most every plate cap I've seen are in the wrong axis!! I've found some heatsink extrusions that are about 1.5" square and are "star shaped" (radial) fins (more or less) with a solid center, designed for mounting stud rectifiers (usually) that if drilled and slit (machined) as discussed would result in a much higher efficiency heat transfer system for air that is traveling vertically up the sides of the tube!

Almost all the the literature I have seen indicates that the chimney is intended to direct air across the top metal/glass seal. I think that this sort of plate cap would likely improve the ability of the air to reach the seal, and greatly increase the effective heat radiation from the plate connection on the tube...

 ;D

             _-_-bear

If anyone is interested, I might be able to machine some up, also could post a jpeg of the extrusion...

Hi Bear,

 I'd like to see that, and I think it's a very good topic for consideration, in lue of 100% Duty Cycle.

 It's a Subject that's discussed very little.

Well Done 73. jack.

Image of one extrusion that might be suitable...
best acquired without the center hole drilled, unless it happens to be undersize for your tube's plate wire... obviously for most applications you'd have to slice a section of this particularly sized piece and then do the rest of the machining to fit and clamp.

Bear

Although not necessarily optimal... in terms of the orientation of the fins.... I think that a standard finned plate cap has so much more surface area, than many of the plate caps that are in common use, that it would probably be sufficient.

The problem, I believe, is not the flow of heat through the metal parts (which are good conductors of heat). The problem is that there is not enough surface area to couple the plate to the surrounding air. So too much heat must flow through the interface between the plate wire and the glass envelope. Glass is an okay conductor of heat... but the surface area of that interface is very small... and the wire is probably getting a lot hotter than the glass envelope. Also, the glass that is within a short distance of the wire is probably getting a lot hotter than the glass that is further away from the wire. I.e., a lot of heat has to flow radially away from the wire through a very small cross sectional area (thickness of the glass envelope x circumference of the wire).

FYI, I made a plate cap with a large surface area as follows:

I purchased a standard threaded hollow rod (called a nipple) at my hardware store. These are used to make lamps and to hang glass covers from ceiling lamp fixtures etc. I was able to purchase one whose inside diameter was very close to that of the terminal on the top of a 3-500Z. Using a couple of round brass nuts, that fit over this threaded hollow rod, I was able to sandwich a fender washer between them. I drilled and tapped the threaded brass rod, at one end, so that I could slip it over the plate connector, and secure it with a small thumb screw.

To attach the plate lead, I drilled a hole in the fender washer.

If you used a piece (round or square) of 1/8" (or thicker) copper or aluminum instead of the fender washer... that would be even better.

Stu

Stu,

Imho, the method you proposed is not effective.

While you may have increased the thermal mass on top of the tube, the problems inherent in the standard one piece aluminum plate cap are not addressed. It is possible that because the fender washer is wider that there is an improvement in airflow below the fender washer, but perhaps not.

The problem inherent in the stock configuration remains that the plate terminal that you connected to is not connected to the plate wire itself in a manner that is of low thermal resistance in the first place.

It is far better to remove the stock plate terminal and go with a better thermal contacting method to the plate wire itself.

Cascading relatively poor thermal interfaces keeps the efficiency low, no matter how large the heatsink becomes physically.

Even a standard design plate cap with the fins in the "wrong" direction, but clamped tight to the plate wire would perform substantially better than it does now merely because of a lower thermal resistance between it and the plate wire. It now being on the outside of the plate terminal (affixed by a setscrew) and itself being affixed by a setscrew.

The solution I propose will make a very good large surface area contact with the plate wire itself (about a 0.125" diameter, typically), and have no other interfaces until it hits the air, which is then flowing in the direction appropriate for maximal efficiency.

            _-_-WBear2GCR

Bear

I agree... if the thermal resistance between the plate wire and the plate terminal is too high, then that will be the bottleneck in conducting heat away from the plate wire. Therefore, more heat will be conducted into the glass that is close to the seal.

Are you sure that the material used between the wire and the terminal has a worse thermal conductivity than glass (taking into account that it is much thicker than the glass in the vertical direction)?

Stu

Stu,

Yes.

Consider the means by which the plate terminal is connected to the wire going through the glass.

The hole in the terminal is drilled larger than the wire it slips over. Then a set screw pushes away from the wire causing a thin stripe of the terminal body to contact the cylindrical wire coming up through the glass. By definition there is a very limited contact area. That alone is sufficient to reduce the thermal transfer.

I proposed a clamp arrangement that would create a contact area of higher than 90% of the available surface area where contact can be made. The actual surface area will depend on the surface of the wire, it's flatness and the width of the vertical slit in the plate cap (that will permit clamping) once clamped.

By maximizing the surface area one also maximizes the probability that the greatest amount of heat can be conducted away... at least that is my idea.

              _-_-bear

I wonder why 833 plate caps have vertical fins but 4-XXX tubes went the other way. Surface area doesn't help unless air blows through the fins.
The plate cap job is to protect the tube seal not cool the plate.

A commercial solution for the plate cap thing is addressed here:
http://www.mgs4u.com/RF-Microwave/tube-sockets.htm

I don't know how good it is. They are sold out of the used chimneys..

Bear, I expect to remove the whatzit from the other item and send it out shortly.

It depends on the manufacturer.  I have some 833A plate/grid caps with vertical fins and some others with horizontal fins.  They will also work on 810's, 866A's, 872A's, etc.  Some are brass and some are aluminium.  But for rectifiers I use the spring clip types since they don't need to dissipate a lot of heat.

3 modes to move heat: conduction, convection, and radiation....most of the heat is radiated in a glass tube... the cooling mode of air flowing over the convoluted surfaces of a heat sink is a combination of mostly conduction and convection.... direct conduction would tend to benefit from more air flowing smoothly along channels, like thru the fins of the heat sink ... convection is not so easy to get a hold of but is an effective heat transfer mode ... my conclusion: just move air, the mo' the betta....beefus 

I would try some of the silver thermal compound used between CPU's and their heatsinks. Not sure how RF and HV would effect it, but it would transfer heat better than any air gap would.

Hey, here's where being a Mechanical guy pays off...

If you can measure the temperature at the junction(s) that you want to protect (glass to metal of the plate lead I imagine) and given the materials used to manufacture the junction you can calculate the amount of heat transfered per unit area. You can also determine mathematically what the temperature differential is (within a reasonalbe degree of error). 
For purposes of this discussion, you could consider glass to be in insulator, as it has a much lower thermal conductivity that metals. (glass having a k of 1.1 and copper having a k of ~400).

The issue here I believe isn't so much the heat, but the different coefficients of expansion of the glass and the metal.  Ideally you what the two materials to grow/shrink the same amount in the same unit time, minimizing the mechanical stresses on the joint.   So in order to really be effective in this heat transfer problem, you should figure the thermal expansion for the glass and the metal of the lead, and then tailor the heat dissipation so that those two numbers are as close as possible.  This means that the two items will be at very different temperatures, but the same relative size relationship will be maintained minimizing the stress on the joints glass to metal seal.  Here glass has about half the volumetric expansion of most metals, so with out doing the math, I'd guess you could go with quite a large plate cap.

So, the glass will heat/cool slower than the metals in this system, and change size slower than the metals in this system per unit change, by a lot.  The biggest stress will occur when the temperature is changing, but that is also going to be the hardest place to control the rate of change. (how do you warm up a tube slowly?) Once things get to a steady state, you can pretty much tailor the heat source/sink relationship to get the desired results, but there will still be some degree of compromise required (as in most designs).

Good points Ed. I was thinking earlier that if you put a really good heatsink on the plate cap, you might make the problem worse because of the uneven rates of expansion and contraction you noted.

Same here Very Good points Ed, and at this moment it becomes a MFG issue on Sealage, But,  By Keeping the metal at a lower slower Heat/Growth Factor wouldn't it cause less stress on the Seal..?..Glass is only going to grow so much where as the Metal Does most of the moving from internally out.

I went back and reread Ed's thoughts, and looked at Steve's tube again, comes to mind about how many Station pulls there were years ago and Why.

The fly in the ointment on the glass analysis is that glass while being an "insulator" also looks fairly "transparent" to heat in that it permits heat (infra-red) to pass through it relatively freely. Sure, it is imperfect and so there is heating...

Seems to me that the rate of change or differential change between the glass and the metal terminal rod/wire is an issue. The question is not going to be resolved through speculation, some sort of measurements would be required in order to determine the temperature differentials and the rates of heating/cooling.

There isn't much doubt in my mind that a tube that is run hard will make the plate rod/wire become hotter than the glass very quickly, and this is the reason for the cooling requirement in the main. Nickel wire is a much better conductor than glass, so it will heat rather quickly.

Doing a quick thought experiment seems to show that if you have a wire with a fire on one side and "dry ice" on the other, the temperature along the wire will be a gradient, with some point along the way being at the 1/2 way (in degrees) point. So, the goal then is to make the wire at the seal not exceed some differential temperature, one variable being the operational glass temperature. Since we know that if you do not use a "tube cooler" on top, the glass seal will be very much more likely to break, it stands to reason that keeping the wire from overheating is then the goal.

The ambient temperature of the outside of the glass and the outside of the tube cooler, under air flow conditions are likely to be relatively close. On shut down, the stored heat of the tube would tend to rise upward anyhow, keeping the temperatures up top again relatively close.

So, I vote for more efficient cooling!  ;D

MaxGain's fins are in the "standard" horizontal orientation. But they say:
"This post is directly attached to the internal plate structure of the tube, and heat transfer / heat dissipation is GREATLY increased! You will notice that with this plate cap it is much more difficult (if not impossible) to make the tube plates glow red under high duty cycles."

Which supports my original conjecture.  ;D

          _-_-WBear2GCR

PS. Most of the heatsink compounds likely would not stand the temps that occur at that tube junction when they get very hot, and would likely "run out" - although maybe one might not reach that critical temp... my clamp idea more or less makes that unecessary I think.

You may be correct. However, I posit that this is mostly a solution without a problem. If crapped out 3-500s were widespread, we would have heard about it by now since they and it's nearly identical kin, the 3-400, have been used in amps since the early 60's.

Steve said:
Interesting. I have a set of Sylvania 3-400Z's where the plate pin comes out the top. A heatsink, (some kind of alloy, not aluminum, its very heavy) is screwed fast to the pin, Then the parasitic suppressor flat stock is screw to the top of the heatsink. Very different from the standard Eimac HC-6 plate cap which is the pagoda style aluminum cap.

Very interesting. Here are two Eimacs, side by side with the 3-400 on the left and the 3-500 on the right.

A picture (or two or three) is worth a thousand words......

On a side note, there are two versions of the 3-500Z. One which is very similar to the 3-400Z envelope and the one you show Steve. Yours uses a bigger chimney whereas the older one uses the same chimney as the 3-400Z.

TNX for posting. Very instructive.

I kind of get this impression from reading the Eimac Care and Feeding manual. They mention a few times about the proper amount of airflow and designed static pressure over the surfaces of the tube. I figured some engineer had a pretty good reason and left it at that.

I personally don't want a giant cap sitting on top of the tube anyway. It just increases the chance of bumping it and breaking the vacuum, or even worse of someone bumps into it while the juice is on God forbid.

I agree with those (starting wirh Bear) who have pointed out that it would be interesting and informative to have a computer simulation that we could use to model the effect of adding a heat sink... and also effects of various types of airflows... on the strains that occur in the glass, and around the seal, during the transition from standby to transmit, and transmit to standby.

I would like to make a conjecture which I think is true, because it follows from the linearity of the heat flow equations:

My conjecture is that the strains that occur in the glass itself (due to uneven heating), and between the glass and the seal between the glass and the plate lead, are similar in magnitude during the transition from standby to transmit (heating up) compared to the transition from transmit to standby (cooling down).

This is crudely analogous to the response of a circuit consisting of a bunch of resistors and capacitors in various series and parallel combinations when an input current source is turned on and turned off. When the current source is turned on, currents will flow through the various resistors (current is analogous to heat flows), and the various capacitors will charge up (storing electrical energy in a capacitor is analogous to storing heat energy in an object... which is the process that causes the object to warm up)

If you look at how fast the capacitors in such a circuit will charge up... and the rate of growth of the voltage differences that will develop between them (analogous to temperature differences)... you will note that the sizes of those voltage differences (magnitude and how quickly they change) are essentially the same during the time after turn on and after turn off.

Thus, I believe that the strains (caused by temperature gradients) on the glass and the seal will be the same during turn on (standby - transmit) and turn off (transmit - standby)... even though one might intuitively (but incorrectly) think that the strains during turn-off would be larger.

The above would be true regardless of the air flows and the size of the heat sink. But, what would be different if you used a larger heat sink, and/or more air flow would be the peak strains that occur in the glass and the seal during the transition from standby to transmit (which will be the same as the peak strains that will occur during the transition from transmit to standby).

It seems to me that we might make some progress in resolving this, by drawing a few circuits to represent the heat flows (less resistance = more heat flow) and the ability of the pieces to store heat (more capacitance means that the part heats up more slowly as it stores heat energy up... because it has a large heat capacity)... and then modeling how the voltages (which represent temperatures) change as we turn on a current source which represents the heat per second (watts) being produced in the tube when we go from standby to transmit.

Stu

Attached is a simple example of what I have in mind.

One would set:

R1 to a low value (low resistance to the flow of heat in the plate wire)
R2 to a low value (to represent a big heat sink with lots of air flowing past it) or a higher value (to represent a smaller heat sink with less air flowing past it)
R3 to a high value (because there is very little surface area for the heat to flow through at that point; and, also, glass is not as good a conductor as the plate wire or the heat sink)
R4 to a lower value than R3 because there is much more surface area for heat to flow through at the interface between the glass that is 1 cm away from the wire (radially) and the rest of the glass envelope
R5 to a lower value (because the glass envelope has a large surface area through which heat can flow to the surrounding air by convection, conduction, and radiation)

etc.

Picking the (relative) values of R1-R5 and C1-C3 to model the associated parts of the tube and the heat sink (etc.) will, of course, be the tough part... but one could develop a lot of intuition about the answers to the questions that have been raised by "playing around" with different values.

Stu

Stu,

The validity of these sorts of analysis will depend on any assumption that are made or included.

Imho, these sorts of failures are due to tube overheating more than simple temperature differentials during normal operation without exceeding normal operating temperatures. I think if you poll folks who use the 4-400 or 3-500Z (etc) glass tubes and that never "run them hard" that they get years and years of service without ever seeing a glass to metal seal failure. My hypothesis is that the goal of a more efficient plate cooler is to limit the maximum over temperature rise that can happen on the plate wire/rod, and that the rest of the time the temperature differentials are well within the range that the glass/metal seal is capable of handling. I don't see the failure time as being turn-on/turn-off.

W1ATR, I'm not so much talking about a "larger" plate cap as I am talking about a different profile with more efficiency. If ur tubes can get bumped or hit when installed or on, I think something is amiss that is not related to a given plate cap design at all. The tubes should not be stored with the plate cooler installed, imho.

                    _-_-bear

Bear

The heat flow schematic was helpful to me in thinking through the question of whether a bigger/more effective plate cap heat sink would cause more or less strain around the wire-to-glass seal during turn-on and turn-off.

Looking at the simplified heat flow schematic, and thinking about its implications I now believe that the following is provably true, regardless of the details of the "component" values in the schematic:

1. The sizes of the temperature differentials that form between and across the components during turn-on and turn-off transients (i.e. until steady state is reached) are the same... so the strains that occur on turn-on and the strains that occur on turn-off (whether or not they cause any damage) are the same.

2. If the heat sink is making a good thermal contact to the plate wire (as the simplified schematic implies) then a larger heat sink and/or more air always help to protect the area around the seal by reducing the maximum temperature differentials (and strains) that occur during the turn-on and turn-off transients, and by reducing the maximum temperature rise that occurs in the steady-state when the transmitter is switched from standby to on.

3. If the thermal contact between the plate wire and the plate terminal (or between the plate wire and the heat sink) is not good enough... then one would have to modify the heat flow schematic... as attached below.

4. In the "modified" schematic, even though the larger plate heat sink might reach its steady state temperature more rapidly ... depending upon the "RC" time constant" associated with the larger plate cap's heat capacity: C1 (Joules of energy per degree of temperature rise) and the thermal resistance from the larger plate cap to the air: R2 (degrees of temperature differential per watt of heat flow between the plate cap to the air)... I believe that the schematic implies that a larger plate cap would always improve the situation with respect to:

a. The maximum temperature differentials (strains) that occur around the seal during the turn-on and turn-off transients

b. The steady state temperature rise of each of the components (e.g. glass envelope in the vicinity of the seal, the plate inside the tube, etc.) that occurs then the transmitter is switched from standby to "on".

Best regards (I've learned a lot by reading the posts in this thread and thinking about this problem)

Stu

Stu,

WRT item 4:

It seems to me that a larger heat cap/cooler (ignoring the transfer to air) will have a higher thermal mass, and as you noted it will take more Joules in order to bring it to a given temperature. Similarly, due to the higher thermal mass it will hold its temperature longer (unless it also has an increased ability to transfer heat energy to the air).

Merely having enough thermal mass seems sufficient to sink extra temperature rise from the plate rod/wire, up until the point where that mass reaches a temperature that we could deem "unsafe" (too high). But lacking a means to transfer that added heat energy, it would then try to sink it back down the plate rod/wire once the tube starts to cool...

So, it seems like the optimal solution would take into account both the thermal transfer to the air WRT time and the rate that the thermal mass of the heatsink can change temperature.

How to do that is a good question...

             _-_-bear