# Simulating tube amplifiers

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

Not open for further replies.
1. ### W8JIHam MemberQRZ Page

I understand what you are saying. I've understood all along.

In order to have a "focused region", the conductor elsewhere has to be electrically isolated though some series impedance that isolates the voltages and currents.

In a transmission line this, in a lumped component analogy, appears as a series of inductances with distributed shunt capacitances. What you are proposing ignores the distribution of fields in the area of the conductor.

If you cannot force the effect you want in an actual physical representation of the structure, you can't magically create it just because you imagine it.

One would think multiple people, with no agenda one way or another to promote a special product or theory, would have noted this effect and documented it if it existed. Especially since it flies in the face of how transmission lines behave. The only documentation of effects like this are where the system is a large fraction of a wavelength long.

For example a microwave signal, over a fairly short distance, can have standing waves that creates localized hot spots. On 160 meters, it would take 100 feet or more. On 150-200 MHz, it would take a foot.

How are you going to do this in fractional inch distances without very low shunt and very high series impedances at the critical frequency?

Anyone can make a model, but the best idea is to understand the limits of how the physical system can behave first. If it can't physically be built, it isn't a good model.

Explain how to build something that heats a uniformly constructed cage of conductors with L/D of about 4 or 5 and a length of about 3 inches on 175 MHz.

73 Tom

2. ### G0HZUQRZ Member

Tom
To clarify, the model predicts that parts of the structure will be stable and parts won't. It isn't a thermal model. The hotspot idea is just an applied guess based on the variation in gain across the structure at VHF and this thermal issue DIDN'T come from the model so the model shouldn't be criticised in this respect.

i.e. the model tells me that different parts of the structure will have different gain at the problem VHF frequency. It manifested itself as the parts near the top going unstable before the bottom.

When multiple parts go unstable they do it at slightly different frequencies. When this happens I see higher RF peak voltages than usually expected.

Do you think the model is wrong?

3. ### G0HZUQRZ Member

I assume you mean unevenly heat a uniformly constructed cage of conductors with L/D of about 4 or 5 and a length of about 3 inches on 175 MHz

Last edited: Jan 4, 2012
4. ### KM1HHam MemberQRZ Page

It is this voltage which causes bandswitch and Tune cap arcing in the SB-220 and certainly not a tube generated parasitic. A simple cure is to place a 10pf 1000V silver mica or equivalent between the 40 and 20 meter switch contacts...at the switch. For a 160-10M amp shift it down a band. This method has been around since at least the very early 60's in the NCL-2000 and Ive used it on uncounted SB-220's as well as amps that Ive had to use the wrong switch wafer configuration in order to get a repair quickly out the door using what was on hand or adding 160 to an old 80-10M Alpha. Many of the more recent troublesome amps get the same treatment. The slight change in tuning in a SB-220 on 10M may require stretching the coil slightly.

http://bama.edebris.com/manuals/national/

Carl

5. ### AF6LJHam MemberQRZ Page

So let me get this straight.......
The idea of adding a small amount of capacitance across that section of the of the tank coil assembly in the SB-220 and other amps is to reduce the voltage developed across those two contacts.
I can see that.....
Do you know the cause?
Is it due to the self resonant frequency of coil itself?
And does this mod have more effect on Eighty meters as apposed to Forty meters??

6. ### AF6LJHam MemberQRZ Page

Oh and Tom;
One of the reasons I haven't said more about the futility of using Spice to model vacuum tubes in UHF and microwave; is because I thought it was obvious to everyone else Spice doesn't have the language to express what needs to be modeled.

7. ### G0HZUQRZ Member

Hi Sue. Two of the models I've used on here are not spice models. I'm not trying to make a spice model of a tube. I'm not the right person to try. I can make crude spice models related to JFETs but that's my limit

I still maintain you don't NEED a complicated tube model to analyse the usual instability mode that requires the suppressor.

The whole point of trying the 'complicated structure' model was to try to capture the 'grid resonance' that several people refer to.
My initial impression of the structure was that I could treat the whole grid as AC ground as per my early models and just stick some inductance between AC ground and chassis ground to model the grid wires. To get grid resonance anywhere near where Tom gets it I have to resonate the grid wires with LOTS of capacitance so I'm doing something wrong.

It would be really useful if Tom could clarify how he gets the VNA plot of grid resonance.
i.e. is it taken with a high impedance probe at the anode with the anode connected up to the tuning tank as normal. Also does it have suppressors fitted for the test.

See post #99 for the 3-500Z grid resonance test plot.

If you could answer that for me Tom it would help a lot

8. ### W8JIHam MemberQRZ Page

If you put a 3-500Z or any other tube inside a shielded test fixture, like a round tube with connectors on the ends, and ground the grid, you can sweep the anode to filament path, either S21 or S12 with a 50 ohm load on the source side to make sure voltage is flat. This will plot transfer though the tube. You will see the response is not at all like a simple capacitor between anode and cathode.

You'll also see a large effect from socket changes, or adding an inch or two of lead length to the grid.

73 Tom

9. ### AF6LJHam MemberQRZ Page

It really is complicated, therefore a complicated model is in order.

10. ### G0HZUQRZ Member

Thanks for the info Tom
I did quickly try and model it but found that the shield may be affecting the result. Especially wrt anode -cathode capacitance.

However, here's my first attempt. This still treats the inner structure as perfect and I found that my interpretation of the test fixture had some effect.

I used 37nH on the external grid to get this (i.e. a LOT more than I wanted to). I treated the grid as a single wire connection on this first stab.

It looks similar to your plot but I moulded it towards your result so my approach/model is suspect as I deliberately tried to make it fit your result. Eg I needed to add 1pF (external) cathode to anode capacitance to get the null near 100MHz.

I'm off to bed now so will reply to any feedback tomorrow

Jeremy
G0HZU.

Last edited: Jan 5, 2012