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: Left-3
ad: Subscribe
ad: QSOToday-1
ad: abrind-2
ad: Left-2
  1. G0HZU

    G0HZU QRZ Member

    Tom I think you are getting closer to my line of thinking.

    I am trying to look at modelling one basic SYSTEM issue of the amplifier. I'm not trying to analyse ALL potential modes of instability. I'm looking at the main reason the amplifier often needs the suppressor.

    This is INDEED because of the physics of the signal path outside the tube. You only need a simple model of a 'tube' with a basic internal feedback path to demonstate the instability mode that the suppressor targets.

    There are lots of other potential instability modes (same applies to transistors) but in this case there is a mode that is extremely easy to understand through basic modelling.

    The main contributor of this mode is the physics of the tank circuit. The tube is not the bad guy in this case.
  2. W8JI

    W8JI Ham Member QRZ Page

    I'm sorry, but this is where we totally disagree. Very clearly, with no doubt at all, the primary problem and issue is INSIDE the tube and in the tube socketing and wiring around the grid.

    The grid or grids and the lead lengths inside the tube to the chassis are where 100% of the main issues are at.

    What i understand you to be proposing, is to take a random unstable tube design and find what prevents that particular unstable tube and layout from oscillating, and use that as a general target for other systems.

    That won't work.

    Let's look at an unstable tube, like an 811 tube or 572B tube.


    The primary problem with this tube style is the single, very long, very thin, grid lead from the grid structure to the socket pin. The capacitance of the grid to the anode, and the grid to the filament, and to the outside world results in that long thin grid lead being tuned by the stray capacitance. It is just like the grid is floated on a parallel-tuned tank circuit.

    Worse, it is a distributed tank circuit. The grid has varying impedance with frequency, going though cycles of impedance.

    As for phase shift, it isn't just capacitance and inductance. The electrons themselves move at finite speed much slower than light speed, and so we have a transit time for electrons causing phase shift. This transit time is dependent on electric field potential between elements inside the tube, as well as physical distance. This means as voltages in the tube change, phase shift also changes.

    If I were allowed to change the grid connections inside the envelope of that very same tube, with no other changes, I could probably make a 572B or 811 unconditionally stable with any anode impedance. The 8877 and 3CX800A7, as well as the YC156 and 3CX3000A7, are unconditionally stable if the grid flange is directly grounded to the chassis.

    The best example is the 3CX1200D7 and 3CX1200Z7. The 3CX1200D7 requires neutralizing on upper HF, and extraordinary care in anode impedance up around 150 MHz or so. With only a change in grid connection structure, the 3CX1200Z7 is unconditionally stanble and requires no upper HF neutralization.

    What works on the 1200A7 will NOT work on the 3-1000Z, even though both use identical grid and filament structures.

    Just by altering the grid path impedance outside the socket, or changing HV, I can move stuff around far more than the anode.

    With all that in mind, I'm trying to grasp how only knowing and considering termination impedance at the anode connection point, which is nothing like the impedance at the anode itself in a tube with long anode leads, predicts anything meaningful.

    73 Tom
  3. G0HZU

    G0HZU QRZ Member

    I'm aware of the 'basic' physics inside the tube wrt long skinny wires and the fact the grid to ground impedance will vary with frequency etc.

    This affects the feedback.

    But for this particular mode of oscillation (involving the poor physics of the tank circuit) those impedances inside the tube are effectively only a bit part player when it comes to basic analysis of why a classic RL suppressor is often required.

    What is IMPORTANT to establish for BASIC modelling is the net capacitance to ground at the true anode at VHF(inside the tube) and the net inductive reactance shown down the line to the dummy load at VHF.
    Last edited: Dec 31, 2011
  4. G0HZU

    G0HZU QRZ Member

    For sure if you were able to magically reduce the grid to ground impedance at VHF the amplifier would appear more tolerant of the poor physics of the tank circuit.

    I think that is your stance on this. (and I don't disagree)

    But you don't NEED to produce a very complicated model of the tube to produce a working model that demonstrates the instability. Once you do this then the analysis of why the suppressor is needed is simple.

    I earlier produced a basic SPICE model of the tube (using simulated grid inductance) and posted up the circuit demonstrating the instability mode. Where it oscillates is mostly dominated by a net capacitive shunt reactance model at the anode and the (equal and opposite) net inductive reactance model of everything else down the line to the dummy load.

    That's why a dip meter can prove a useful shack tool for analysing this (and ONLY this) mode of instability

    It really is that simple.
    Last edited: Dec 31, 2011
  5. W8JI

    W8JI Ham Member QRZ Page

    Correct me if I misunderstand. What I understand you are saying is a certain tank impedance at some frequency arbitrarily determines stability.

    If so, I disagree. Experience with actual systems flies squarely in the face of such reasoning.

    There is no consistent or repeatable correlation to a dip on a dip meter at one part of the tank system, or even to the impedance looking into the tank, and likelihood of an oscillation or instability. This can easily and factually be demonstrated and repeated, over and over.
  6. W1BR

    W1BR Premium Subscriber QRZ Page

    I'm probably out to lunch, but unless the self-neutralization frequency of the tube is know, and unexpected feedback paths,
    I don't see how grid dipping for resonances proves anything. There is a huge, reasonably high Q resonant
    plate tank that should produce some nice dips; but we don't expect the amp to self oscillate at it's design frequency. Maybe
    we need nichrome wires on the cathode and grids.:rolleyes:
    I'd bet you could find UHF or VHF resonances in the frames of the larger plate tune and load caps, if one looked hard enough.

  7. G0HZU

    G0HZU QRZ Member

    What I am saying is that there is an easily demonstratable SYSTEM limitation due to the physics of the tank circuit that makes the tube very prone to instability at a predictable frequency based on modelling of the tank circuit and a very simple model of the tube at the problem frequency. I believe that the physics of this SYSTEM limitation are easy to understand and model in order to gain better understanding of this mode of oscillation. Also once you do this you can understand (and model) how the suppressor works.
    Last edited: Dec 31, 2011
  8. W1BR

    W1BR Premium Subscriber QRZ Page

    "Simple" models of the tube would be difficult, the Chinese tubes have higher gain. We have 2SC2290 transistors that wouldn't work
    in earlier designs.

    Almost everyone understands how the suppressors work. The plate tank is a small part of equation, what about the grid?
  9. KL7SG

    KL7SG Platinum Subscriber Platinum Subscriber QRZ Page

    For a first order approximation, one could assume the speed of the electrons will be around 1/100 the speed of light.

    The following text I believe discusses the speed of electrons in a tube: "Thermionic Valves", A H W Beck, Cambridge University Press, 1953. Tom is correct in his position regarding the speed of the electrons; but, I think this can be factored in to a more advanced model and be usefull with respect to predicting phase shifts.
  10. G0HZU

    G0HZU QRZ Member

    In order that we all have the same idea about suppressor action can you tell me the anticipated (ballpark) BEFORE and AFTER impedance at the anode with and without the suppressor at a frequency close to the problem area?

    You can answer this with text or draw dots on a crude smith chart :)
    Last edited: Dec 31, 2011
Thread Status:
Not open for further replies.

Share This Page