Simulating tube amplifiers

Discussion in 'Amateur Radio Amplifiers' started by W8JI, Dec 29, 2011.

Thread Status:
Not open for further replies.
ad: L-HROutlet
ad: l-rl
ad: L-MFJ
ad: Subscribe
ad: Left-2
ad: Left-3
ad: MessiPaoloni-1
  1. W8JI

    W8JI Ham Member QRZ Page

    No, it isn't. None of the components are adequately characterized, including and especially the tube and socket. This makes a model a wild unrealistic guess at a very complex system.

    The system is well understood as a system. The best model is a working operating model.

    They are. At frequencies where the grid is effectively grounded, phase shift through the tube can make the tube behave as a tuned-cathode tuned-plate oscillator. At other frequencies, where the grid is not well grounded at the problematic frequency, the tube acts like a T-P T-G oscillator.

    There are multiple modes.

    The 811 or 572 tube, for example, is reliably unstable at upper HF because of poor shielding inside the tube. Some sweep tubes, especially when the G1 is grounded to the cathode, behave the same. For example you can take a Dentron sweep tube amp, or an Am Supply LA1000, and make them oscillate in the 20-40 MHz range by increasing anode impedance presented by the tank, especially with higher impedances at the cathode. The same is true for many unneutralized 572B and 811 amps, when they use multiple tubes or higher voltage.

    This very repeatable.

    The FL2100, when used with lower mu triodes, has enough quiescent current at standby to oscillate. This is because the load is removed from the tank and input, the tube is not fully cut off, and the gain then exceeds the loss through the feedback path. The root cause is a poor bias system in the FL2100, and a very poor feedback system or no feedback system at all to null the anode-cathode feedback. There is enough phase shift both in transient time for electrons and resonant circuits, and the reactance of the feedback path, to make the normally negative feedback from anode to cathode become positive.

    This repeatable, measurable, and can be duplicated over and over again.

    The patch is to add enough bias to make sure the tube is fully cut off on standby, the actual correct fix is to neutralize the tube.

    The second mode for tubes like the 811, with large grids and long skinny leads, is where the grid is no longer grounded. That occurs up around 60 MHz to 100 MHz, depending on socketing and lead lengths, for 811's and 572's. At that frequency the mode becomes a T-P T-G oscillator. Where the very high direct feedback capacitance to the grid from anode excites the grid enough to make the system oscillate. The cure here is to either move anode resonances below the grid, so feedback phase is wrong for oscillation, or to load the anode system with losses near the frequency where the grid is resonant.

    Of course combinations of cures work.
    There are so many transmission lines and small cells of components involved, a model would be extremely complex. An air variable capacitor alone, over an ideal ground plane, looks like a series of transmission lines of varying impedances (all very low) with shunt and series lumped reactances. The same air variable, depending on which terminal post we look at, can look much different. This even has an effect on harmonic suppression.

    When we combine impedance of chassis paths, it becomes even more complex....and that is just one component.

    I have not found a good spice model of a tube, and I am unlikely to ever see on that covers the tube from dc to several hundreds of MHz.
    When we through the socket in, lead dress and routing, and the chassis, it becomes hopeless.

    The single best thing to do, and by far the most reliable, is to run a variety of tubes through a typical unit with extra high voltage and the tube slightly into conduction, and go through all tuning and load conditions.

    If you do this with Clipperton L's, FL2100's, a Collins 30L1, and dozens of other amps you will find they are unstable. They repeatedly can be made to oscillate. If they oscillate at HF, they can arc the tank system. If they oscillate well up into VHF, they simply draw extra current and can, with time, overheat things.

    If you have a gassy tube and cause extreme anode voltages, the amp can be made to arc the tube. Of course it will also arc the tube at HF. The result is wires and chokes and other components jump or flex or shake during the fault, and some external arcs occur as the negative rail tries to lift, and we hear the "big bang". The root cause is a tube arc, the result is a catastrophic fault current and high voltage through the entire path from filter caps to tube and back to filter caps.

    All of this is repeatable, and can all be confirmed. No magic involved.

    This has all been done a long time ago.

    There is not really any debate at all about this with 99% of the people who work with this stuff. In 30-40 years, I've not seen any significant disagreement between dozens of PA designers or tube engineers. To me the most striking thing is how pathological science gets any foothold at all, because the real workings, although complex with a variety of causes, are all very logical and repeatable.

    I think, though, anyone trying to do a lumped component model simulation, no matter how many components are used, is in for the time of his life. This is because at VHF the components and lead lengths inside a high power HF PA are very complex. The anode of the 3-500Z inside the actual tube, for example, sees nothing like the impedance of the termination at the anode cap because of the large anode to grid capacitance of the tube, and the series inductances of the path from chassis, through socket, to anode, and out on anode connection to the cap. This does not even consider the distributed transmission line effects.

    I can't think of a more wasteful thing than trying to quantify all this in a SPICE simulation, when the vacuum tube itself, without socket or external connections, is not characterized outside the normal working frequency range, let alone fully characterized inside the range.

    If transistors were 5 inches in diameter and 5 inches tall, and used sockets with two-inch leads, and mounted on chassis instead of circuit boards with large ground planes, SPICE would be as bad for them.

    All of this stuff has been handled for years. Virtually all people truly experienced in design and testing agree in every way. There may be very small disagreements in abstract theoretical definitions or theorems, but the core principles, failure modes, basic operation, and testing have been well understood and agreed on for at least 50 years.

    73 Tom
  2. K9FV

    K9FV Ham Member QRZ Page

    There you go again Tom - bringing facts and logic into an emotional argument.

    Just like my first wife used to say: "Damn it Ken - forget the facts, argue with me"..... Now late in life, I've learned the key to marriage...... KEEP MY MOUTH SHUT! {grinning}

    73 de Ken H>
  3. AF6LJ

    AF6LJ Ham Member QRZ Page

    Actually Tom's explanation on his site is accurate and understandable to someone with the capability to roll their own amplifier.

    As I implied in the other thread there are two ways to build suppressors.

    Tom's method provides a smooth transition in impedance as frequency is increased, this compared to old way a little out of the box. The old method I have seen discussed in old handbooks was to wind a coil that is "self resonant" at or near the parasitic frequency, and shunt it with a resistor. While I am not inside Rich's head I suspect this is where the thought behind his suppressors came from. Over the last couple of weeks I have read just about everything worth reading on the subject and have arrived at the conclusion that both methods work. We can argue about which works best and which is easiest to incorporate into an existing design with a minimum of hassle.

    One thing I do want to say that has not been said on the subject of amplifier behavior.

    There is only one way for unusually high currents to flow in a tube.
    I'm talking currents over the saturation current that can be delivered by the cathode, there must be a parallel conducting path within the tube. This path can be ether a metallic path from anode to grid or cathode or a gaseous path that is ionized and becomes a plasma. That much gas can only come from a leak of some type, or a machincal failure that results in the evaporation of metal within the tube.

    Okay I have said my peace.
  4. W1BR

    W1BR Ham Member QRZ Page


    I've noticed that Hallicrafters insulated on leg of the HT-32 plate tune capacitor from the chassis using cardboard washers.
    Was that in order to break up VHF resonances in the plate capacitor frame? I'd be curious if unexpected resonances in
    these structures is a consideration?

  5. W8JI

    W8JI Ham Member QRZ Page

    There was a transition in thought and practice from old tubes, components, and wiring to new things.

    Early on, everything was breadboard and mostly over non-conductive chassis. Using that system, it became critical to use single point grounding. This is because there was no chassis, no groundplane, and no large surface for grounding. Tubes had long leads, even an 811 would have been an improvement, so things were much less stable at lower frequencies.

    As things changed to chassis construction, many of the senior engineers still used techniques learned when grounds had to be in a single point for each stage.

    This is evident in transmitters like the Viking Ranger and Valiant. The plate tune capacitors float, and are grounded through a buss wire to the PA tube cathodes as a common point. This 100% because of old thinking. If the plate capacitors have the buss removed, and are grounded to the chassis, many problems in the Valiant and Ranger, like TVI and chirp, diminish.

    This link for Valiants
    This link for the TL922

    The TL922 designers were a bit slow on learning proper sheet metal design for RF paths, and it was attempted to correct that by a wire from the tune cap back to the grid!

    The breadboard layout and poor tube design was actually the root of the nichrome suppressor, because the lower frequency or operating frequencies needed suppression to lower gain. This means the low frequency path needed resistance, not just the VHF path.

    When we think of how it all works, this evolution all makes perfect sense. In the early days shielding and grounding was poor. This meant single point grounds for each stage, and additional, intentional, loss at low frequencies and higher frequencies. As we progressed, most rapidly around WWII, we learned the chassis made a great groundplane. That got rid of the need for a single common ground point for each stage. Better tubes and better wiring improved LF stability, and it because unnecessary to dampen Q at or near the operating frequency.

    People were a little slow to catch on, because they got in the habit of doing things a certain way. Habitual design is why the Valiant, Ranger, and other transmitters, and some amplifiers, carried over older techniques.

    It is all logical, and all makes perfect sense when we look at how the system really works and how systems evolved, and when the older engineers brought up with poorly designed tubes on bread board chassis stopped influencing designs. If we look at it all over the years, and the age of the designers, it becomes an "ah-ha moment".

    The battle now is simulations via software, which work well with small components and groundplane style boards, but fail miserably at emulating upper HF and beyond with chassis and power grid tube components. Imagine the complexity of a TL922 chassis and socketing RF path arrangement, one of the worse around for RF current paths.

    73 Tom
  6. AF6LJ

    AF6LJ Ham Member QRZ Page

    Now you, and others know why I say the layout of a TL-922 is scary.
  7. W8JI

    W8JI Ham Member QRZ Page

    It's a good example of how to copy something electrically (the Heathkit SB220) and make it worse. Copies should be better, not worse.

    Last edited: Dec 29, 2011
  8. AF6LJ

    AF6LJ Ham Member QRZ Page

    That's true but the FL-2100 is yet another example of an amplifier that was a bad copy.
  9. G0HZU

    G0HZU QRZ Member

    First of all thanks to Tom for starting this thread :)

    I agree with a lot of what Tom says about the potential difficulties of modelling some parts of the system.

    However, I'm a little surprised why you dismiss modelling totally.

    Tom, do you know what methods a modern RF engineer would initially attempt to use to (fairly) accurately model the entire passive PA tank circuit (i.e. the passive network that the the valve will connect to) for a basic model of an unstable amplifier that needs suppressors?
    i.e. how to quickly get a good working model onto a computer despite the complexity of the pa tank network?

    It isn't as difficult as you appear to suggest.. :)
    Last edited: Dec 30, 2011
  10. G0HZU

    G0HZU QRZ Member

    We are nearly into 2012 though Tom and there have been some tremendous advances in RF analysis since the 1960s. The engineers in those days had it tough and it must have been very difficult to fuly understand potential weaknesses in even a simple circuit like an amplifier. It doesn't mean those old engineers weren't good engineers. Far from it but compared to the analysis tools available in recent years they were at a massive disadvantage.

    A shiny new 24yr old RF engineer straight from college with virtually zero experience can model the PA tank circuit VERY quickly with the right tools and make it behave similar to the real thing even up to UHF.

    I ask again, do you know how this is quickly done?
Thread Status:
Not open for further replies.

Share This Page