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-3
ad: abrind-2
ad: Left-2
  1. AF6LJ

    AF6LJ Ham Member QRZ Page

    One of the problems with modeling tubes is obtaining inductance values for the elements.
    Take for example the grid structure above for a 3-500Z family of tubes.
    Can you accurately model the inductance of that component.
    We can measure interelectrode capacitance directly.

    Transistors are built to very tight tolerances both mechanically and electrically, it is the nature of the modern processes. When transistors are made in this day and age the amount of doping applied (through ion implantation) is knows to a shocking degree of accuracy. While they are mass produced by the billions transistors are a remarkably precisely built product. You cannot say that for hand assembled vacuum tubes. While hand assembled tubes can for practical purposes be consistent in performance and characteristic, they don't exhibit the same consistent performance as we increase frequency that we see in transistors.

    I have no doubt such an accurate model for the behavior of a 3-500Z could be developed, the problem arises from the usefulness of such a model outside of academic circles. This is a fifty year old tube and it's application has long been set in stone. It's limitations are well documented and their use even in amateur circles is dying.
  2. VK6ZGO

    VK6ZGO Ham Member QRZ Page

    Is there something I'm missing in this?
    It seems to me that your description could describe almost anything inside the box,as it is imaginary.
    A network which has more inductive reactance will look very different to one with more capacitive reactance,& if the resistive component varies it will look different again.
    A real "black box" with unknown circuitry inside can be tested,& an equivalent network derived,which may not look anything like the physical internals of the original box,but will be equivalent from the point of view of an external signal applied to the input port.

    I fail to see how you can do the same thing with simulation,without having some idea of the internals of the box.

  3. AE1PT

    AE1PT Ham Member QRZ Page

    As noted in the "why are threads closed" thread, I am going to reopen this on a trial basis. Quite a few have written opinions on this matter--and there is the aforementioned thread.

    Be forewarned however. If this turns into another bully pulpit argument, I will close it--and consider giving the offenders some time in the corner to think things over.

    Oh, Happy New Year...
  4. W8JI

    W8JI Ham Member QRZ Page

    I agree with VK6ZGO.

    G0HZU seems to keep ignoring the fact there is no SPICE model of the tube, or the socket, or the chassis, or the variety of tanks and such.

    Not only do we have to sweep the tank across a wide range, we have to move the PLATE capacitor and sweep again. Then again, then again.

    When we are done with that we move the LOAD, and we start over on the PLATE.

    When we get the LOAD done, we now change bands and repeat the entire process. The result will be the tank path for one particular amplifier.

    I don't think this can be done in a few minutes, but maybe I am wrong.

    So the first problem is tube behavior out through the pins, so we can do a SPICE model representation of the tube. Where do we get that information?

    73 Tom
  5. AC0H

    AC0H Ham Member QRZ Page

    I don't know as much about RF amplifier design as other folks on this board but I do know quite a bit about computers.
    When computers were new and folks started to think of ways to use them, possibly the most important "rule" with respect to computer generated anything was laid down in stone.......GIGO. Garbage IN = Garbage OUT. You had better have a real good idea of the results and run many hours of simulations taking into account ALL variables and conditions before taking computer generated anything as gospel. Computers are just collections of millions of very fast transistors alternating between high and low....1 or 0. They're actually quite stupid machines until somebody tells them how to do something.

    Seems to me since there is no accurate SPICE model of a vacuum tube or it's environment, any simulation which purports to describe how that tube interacts with it's environment, is just guessing.

    How accurate is SPICE when it comes to RF?
    + or - 10%?
  6. G0HZU

    G0HZU QRZ Member

    That's all very true Tom [​IMG]

    But..... as I said earlier you can adopt two approaches. You can take the complicated road and try and produce a monster simulation of a typical commercial valve amplifier complete with a hideously complicated (futile?) 'attempt' to COMPLETELY model a triode valve.... or you can sit back and look at the bigger (actually simpler) picture and analyse a particular behavioural problem of said amplifier and reduce it to sound RF fundamentals.

    SOMETIMES when you do this you hit the jackpot and discover the problem involves extremely SIMPLE analysis to master and you realise that the hideously complicated tube isn't the bad guy.

    When this happens you can replace the tube with a simple model and chase the real 'bad guy' and in the case of suppressor theory the bad guy is the real world physics of the tank circuit.

    Maybe this sounds like I'm ducking the issue here and maybe that's why we aren't appearing to agree. My goal isn't to try and master a comprehensive model of a tube.

    I'm trying to help everybody through using basic RF fundamentals understand WHY the parasitic suppressor is needed for a particular behavioural problem of the whole amplifier SYSTEM.

    Because the physics behind the root cause of the instability is so simple we can use really simple RF modelling to master the issue.

    That's the basis of my position on this subject [​IMG]
  7. W8JI

    W8JI Ham Member QRZ Page

    While there are tolerances that vary greatly because of slight shortcuts or inaccuracies, the result is more like useful or not when the model is wrong.

    Tube are pretty difficult things to represent in a model, because they have emission current limits, series and shunt distributed reactances, a resistances that vary non-linearly with voltage changes on multiple elements, time delays that change with frequency, voltage, and current, and many other things. I have never seen an accurate SPICE representation of a tube yet, beyond DC or low frequencies compared to the cutoff frequency.

    The tank and other things also are complex, because when a control is adjusted or a band changed, the behavior of the tank changes significantly outside the normal operating frequency range in some unusual ways. It is pretty easy to represent things on one frequency with one control setting, but accounting for all the variations in behavior as leads are moved or capacitors tuned is pretty difficult with components a large fraction of a wavelength or very low or high reactances.

    Transistors work OK, as do surface mount or compact parts, but not physically large or distributed components, or dynamically varying things like tubes where changing one voltage changes 15 other things.

    73 Tom
  8. G0HZU

    G0HZU QRZ Member

    I can understand your viewpoint because it is never possible to fully model a real world 'physical' system. Also sometimes the model can be 'bad'. But often the model is good enough to help you analyse a particular behavioural problem of a system. The model can help understand the issue and often it can (correctly) predict other behaviour problems the user has missed.
    Last edited: Dec 31, 2011
  9. G0HZU

    G0HZU QRZ Member

    I think you are looking at dismissing the possibility of ever creating a 'super model' that can comprehensively predict all types of misbehaviour of the tube.

    I agree with you all the way on this and I won't limit the issue to tubes. The same applies to transistors.

    But at the risk of sounding like a stuck record you can often replace the actual device with a simple model if the root cause of the SYSTEM issue lies somewhere else :)
  10. W8JI

    W8JI Ham Member QRZ Page

    Since the tube provides the gain, and the feedback, and primary reactances and phase shifts in the oscillator circuit, how can you model the stability without an accurate representation of the tube?

    It seems to me you are trying to take this back to a point where we only look at impedance OUTSIDE the tube, and somehow decide what impedance outside the tube causes instability. while that is simple, it has nothing to do with real world behavior.

    An anode system impedance that is 100% stable with a certain tube type, wiring style, and socket connection length under all conditions can be totally unstable with a change in tube type, wiring style, and/or socket connection length.

    Some amplifiers, actually many of them, can be 100% stable no matter what is done in the anode path. Others are horrible to stabilize.
Thread Status:
Not open for further replies.

Share This Page

ad: QSKLLC-1