Simulating tube amplifiers

What do you mean higher gain? I've never seen that.

There is so much bad information put out about tubes and tube operation that constructive forward movement is very difficult.

For example, instability in the FL2100, where Svetlana tubes allow it to oscillate on standby, is blamed on higher mu.

Higher mu causes lower quiescent current, in a given tube design, for a given HV and bias. When a newer 572B is plugged into a FL2100, or any other amp, idle current increases. This indicates lower mu, yet all over the Internet the claim is mu is higher.

The real problem of the instability appearing is simply that the tubes are not biased into cutoff, which actually indicates lower mu. As a matter of fact, if anyone had taken the time to install original Cetron tubes and lower the cutoff bias, they oscillate as well as, or better than, the supposedly "higher gain higher mu" tubes.

This totally dispels the popular claim mu is higher, and anyone knowing the difference between a 3CX3000F7 and a 3CX3000F1 also knows how lower mu in the same geometry means more quiescent current at the same bias, this is all common sense, yet everyone leaped to the conclusion the oscillation is from higher gain.

The problem with moving forward is unlearning the bad information that keeps getting put out, and thinking through the system instead of leaping to conclusions.

I don't think many people do understand how a suppressor works. Many people think Q of the suppressor as an isolated component is important, rather than its effect as a part of a large more complex system. I think some people have a good idea. G3RZP certainly has a good understanding, as do engineers who work regularly with large RF systems, but I think very few other people who don't regularly work with oscillator and amplifier systems have a grasp.

There are a dozen different ways to do suppression, if required. All of them require either loading the system, changing resonant frequency in the system, or changing feedback, or a combination.

What we have is an unreasonable fixation on dip meters and Q, because there was a vacuum in education, and it was incorrectly filled.

All of this is logical, provable, and can done in steps once we get past the urge to get all answers from almost no knowledge of real system behavior.

73 Tom

 

Thanks for that resource, Mike. I'll look for it.

Many books, because they only deal with low frequency operation, ignore transit time. We generally need to get into VHF operation to cover things applicable to VHF parasitics, and when most of that work was done with tubes what we know of as VHF was called ultra high frequency.



Ultra high frequency Techniques (D. Van Nostrand Company Inc 1942) is one example.

At about 100 MHz transit time starts to become important.

 

For what particular amplifier on what band?

 

I was really just looking for a general example but I suspect if we pursue this then we could end up with a very long thread about suppressors (again) so maybe my question wasn't a wise one...


So maybe best to go back to topic.

I did look again at your web page Tom

(I hope you don't mind me pasting your web content here)

As you know my experience of valves is very limited.

But I'd like to try to quantify typical values of the above because it would help with (crude) modelling.

The one I am struggling with a bit is a value for C2 and where it connects in relation to the grid and chassis ground.

My first impression is that C2 will be connected close to the 'chassis ground' end of the internal grid wires reducing its significance in the model somewhat.

There will also be some capacitance across the internal wires but I don't know the grid to gnd spacing or grid area to calculate C2.

I suppose I'd really like to break C2 up into a more complex network.

But I only base this on my interpretation of a typical layout and possible socket arrangement. Can you clarify?


I know you work with lots of valve types but maybe if we looked at the 3-500Z as an example?

 

I wouldn't say it's IMPOSSIBLE to properly model a tube, but there are a lot of parasitic components that generally AREN'T properly modeled, the way most hams do it.

In the microwave industry, for instance, modeling is used extensively (bun not EXCLUSIVELY....very important distinction!) for very complex devices. In fact, at microwave frequencies, even a simple MOSFET can be much more difficult to model than a multigrid tube, because so many of the parameters are non-linear and frequency dependent. Every semiconductor junction, for instance, exhibits some VARACTOR effects, which can become DOMINANT at superhigh frequencies. This is something you don't even have to think about in tube amplifiers.

It's a fact that most MMIC devices would never even get ONTO the drawing board without a lot of modeling right at the beginning. Of course, these models are always tested with real world devices....but I think the technology is at the level that modeling is necessary to even have a workable prototype.

Eric

 

That schematic is far too simplified for a SPICE representation. At 100-200 MHZ transit time becomes import, and many strays are omitted.

The only thing I can think of doing, and I have been considering it, is cutting holes in a tube to gain access to the actual elements with a network analyzer. The problem I keep rolling over in my head is how to get a negligible impedance ground for the probe, and what future use it really all has.

What might be easier is to measure where oscillation starts and stops as the band switch is changed and capacitors rotated, and see what the impedances are looking out toward the tank at the oscillation frequency. That's up at, or over, 150 MHz in the AL80B.

By the way, if I drop the voltage to 2000-2500 volts, I cannot make it oscillate without any suppressor at all using any tube I have.

73 Tom

 

To me that would suggest the conditions that exist that enable oscillation are marginal and far less from ideal. Since by lowering the voltage you are in effect lowering the gain of the amplifier.

 

Originally Posted by K1ZJH
"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.

What do you mean higher gain? I've never seen that.


For one example, I believe K7RDX posted some test results comparing the original Eimac tubes to the newer
Chinese imports on the Amps board for the Contesting forum thread several years back.
He claimed the Chinese tubes exhibited higher gain.
You may have been participating in that thread? I've also read where you've mentioned Amperex 3-500Z had higher
mu than the Eimac tubes. Even if we discount the Chinese tubes as having the same or lower gain, that still leaves
the Amperex tubes as the odd ducks out.

Pete

 

A Swan Mark I running at 2500vdc anode has no parasitic suppressor using a pair of 3-400 tubes with no bias. It only has a small RF bead choke on the anode line. Later as Swan moved the voltage up and added 3-500's in the Swan Mark II the suppressors appeared. This is an example of what Tom was talking about when lowerng the anode V will automatically keep the tube from oscillating.

 

This is kind of what one does in the plasma lab. The Langmuir probe can detect the actual space charge in any region in a volume. You can actually measure the difference in space charge with a good set of Langmuir probes, and derive the transit time and other cool stuff like that. Unfortunately, they don't make many vacuum tubes with probe ports.

Despite the practical limitations, however, the ancient RCA Radiotron Handbook fairly completely characterized a whole lot of tubes, and quite well.