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: Left-2
ad: Subscribe
ad: Left-3
ad: MessiPaoloni-1
ad: L-MFJ
ad: K5AB-Elect-1
  1. AG6K

    AG6K Ham Member QRZ Page

     Would it seem a bit strange to you if someone professed that a 1/4 wavelength vertical antenna has an RF ground somewhere in the middle? As I see it, the grid does not shield the input from the output very well at or above its resonant frequency. In radar equipment 1/4 wavelength conductors are commonly used as RFCs.
    • Rich, ag6k

  2. AG6K

    AG6K Ham Member QRZ Page

     Three SB-220 47Ω, 2w carbon comp R-supp resistors got smoked within days . Since this happened on 7MHz, and the 47Ω R-supp was paralleled by 2.5Ωs of XL, how could 0.8a of 7MHz current have smoked the three resistors? It couldn't of course, but 110MHz current surely could have.

     I could smell burning phenolic when each resistor changed value. This is all somewhat puzzling to me Tom since during our telephone conversation you talked about many of the Heath SB-220s that you repaired having parasitic oscillation damage.

    Rich, ag6k
  3. W1QJ

    W1QJ Ham Member QRZ Page

    With all the SB-220 amps I have used, tested, repaired the single most well behaved band they operated on was 40 meters. The 40 meter band contact was the least involved with being toasted. Sometimes 80 meters and 10 meters went bad due to bad indexing. Most of the time ONLY 10 meter contacts were burned. Often times 10 and 15 both with indication of tracking between them. I suppose I can actually do a comprehensive pictorial study of all this based on MANY burned bandswitches I have. I am not certain I have a digital camera good enough to take close up photos that can show enough detail however. It probably will take quite some time to photograph that many switches. I offer anyone who cares to go through quite a few stock suppressors and cut the coils and measure Rsup to let me know, I will ship them to you for analysis. I'd say I have enough of them to create data that should be considered representative to the discussion as opposed to the study to be isolated cases. Any takers?
  4. G0HZU

    G0HZU QRZ Member

    Hi Rich
    You can assign AC ground anywhere you like in an analysis. Putting it at the gate of a JFET doesn't mean the device has perfect input to output isolation.

    It doesn't change anything about how the circuit works. It just sometimes makes the circuit operation easier to understand.

    You could forget that this circuit is trying to be a grounded grid amplifier/oscillator and analyse it with AC ground at the cathode. You then end up with a negative resistance analysis at the gate/grid. This oscillator analysis is just as valid as nothing has physically changed in the actual circuit but this analysis is far less intuitive (to me at least)

    However, AC ground is really a single point location and with a tiny semiconductor like a JFET at VHF you can confidently assign AC ground to the gate terminal.
    Here we have a much bigger package.

    That's partly why I looked at the multi cell analysis and it's also why I'm keen to find out more about the concept of 'grid resonance'.

    Last edited: Jan 7, 2012
  5. G0HZU

    G0HZU QRZ Member

    If anyone closely studies the latest model circuit below they should be able to spot several things about the feedback path.

    First, there is an optimum 'worst case' grid wire to chassis impedance in terms of loop gain. (i.e. what people refer to as the grid float impedance) This is because if you look at the anode circuit it forms a resonant tank with an inductive tap off point as per a Hartley oscillator. The tap off point is set by the grid/chassis impedance represented by L2.

    The worst case grid impedance for high loop gain is the one that matches the tap point into the feedback path to the cathode. This will give maximum power transfer in the feedback path. So there is an argument that you could SOMETIMES make the instability WORSE by reducing the external wire inductance at the grid to chassis connection.

    A better approach might be to deliberately introduce a slightly lossy RLC network here instead of going for the shortest wire connection. It really depends on the true figure for what the 'grid resonance' really is. I guess there will be no common answer to this as different amplifier models will have different grid grounding techniques. Modify them at your peril....

    The other thing is that any phase shift you can introduce after this Hartley tap point is going to be a good thing wrt spoiling the loop bode plot because you can shift the zero phase crossing point over to the left and away from the main peak in the loop gain. The loop gain that MATTERS is the gain where the phase around the loop is zero.

    I also suspect that some of the characteristics of this instability are set by the proximity of the nearest shunt impedance to the cathode. i.e. how close the nearest 'cathode to chassis' shunt capacitor is.

    If it is close to the cathode then it's BAD for stability as there is minimal phase degradation around the loop.

    Basically, the worst case scenario is a matched Hartley tap point and very little phase shift after the tap point because you end up with a bode plot that has the zero phase point closest to the VHF gain peak (dominated by the resonance of L1, L2 and Cpg)

    Basically, anyone who is tempted to modify these amplifiers without knowledge of the loop response could be making the system WORSE even though the grid grounding appears BETTER and the decoupling CLOSER to the cathode. Attempts to improve stability by adding small shunt caps from cathode to chassis will usually make the feedback path more robust.

    At this point of the modelling process we need to start comparing different amplifier designs because the areas outlined above are KEY to stability and it would be interesting to see what various manufacturers do with respect to grid connections and cathode shunt caps.

    Attached Files:

    Last edited: Jan 7, 2012
  6. G0HZU

    G0HZU QRZ Member

    With clever design, it may be possible to exploit the internal cathode to grid capacitance at VHF to introduce extra phase shift that spoils the bode plot.
    i.e. it could be beneficial to deliberately have some stray series inductance in series with the cathode input (eg a shortish wire).

    Hopefully by now even the most ardent 'anti modellers' can see there could be potential benefits to applying modern system analysis techniques to tube equipment. I'm not implying that this system is in any way mastered yet, but maybe some useful progress has been made :)
  7. WA4OTD

    WA4OTD Ham Member QRZ Page

    Hello Jeremy;

    What is your latest and best model? I would like a starting point if you have it to look at this.

    Which spice do you use? I use LTC spice.

  8. G0HZU

    G0HZU QRZ Member

    Hi Leroy
    At the moment I'm not really using spice as this is a small signal issue so can be initially modelled reasonably well on a decent RF based linear simulator.
    I still think the external components and layout dominate the characteristics of this oscillation mode and a simple amplifier model will suffice for the tube for initial analysis.

    The CAD tools I'm using for the simulations above are the ones I use at work and I can use them at home too at evenings/weekends for company work or the odd bit of personal research.

    I'm using Agilent Genesys 2010.05 SP1 for most of the work and I'm editing simple mathmatical models that are within its library.

    At the moment I think the best tool is a linear simulator combined with a simple AC model of an amplifier 'similar' to a tube. I'm using a Genesys model based on a JFET for the basic model.
    This model is really simple and heavily based on a VCCS.

    The big multicell model I have is also on Genesys.

    Last edited: Jan 8, 2012
  9. G0HZU

    G0HZU QRZ Member

    I did also have a brief play with a dual amplifier version as there were images of amps with two tubes in parallel.

    What I found with this case was that the suppressors aren't as effective especially when you start playing with the inductor that models the tank network at it's 'worst' series equivalent (a series inductor)

    This is because the net inductance of the tank network (L1) heavily influences the most likely frequency for oscillation and it appears harder to tame the dual amplifier model across 80-180MHz.

    What did seem to work well was a series (100nH in parallel with 470R) suppressor at the cathode input as it shifts the phase to the left and also reduces loop gain. But a lot depends on the location of the first shunt cathode element as the suppressor has to be fitted after any shunt reactance. i.e. between the shunt cap and the cathode. Adding a few pF across the suppressor can help in places at the expense of others.

    Are suppressors OK to fit at the input? I'm unsure as to the large signal behaviour of a GG tube when cathode driven. Not sure if the suppressor would cause issues here at HF.

    Just having a length of wire feeding into the cathode appears to be benificial.

    I do suggest that the time has come to declare the current system model as having gone as far as it can. This is because in order to progress any further the system model will have to attempt to model a specific amplifier with specific locations of cathode components and also specific models of the tuning tank circuit.
  10. G0HZU

    G0HZU QRZ Member

    Bit of a typo above (sorry) I should have written 80-130MHz for the dual PA above. Where it really wants to oscillate is around 100-110MHz as this is where the bode plot aligns gain and phase the best.
Thread Status:
Not open for further replies.

Share This Page