Parasitic oscillations in general

[SpreadSpectrum]

I'm looking for information about parasitic oscillations in tubes in general. Anyone know of a good reference or explanation out there on the net? I know there are several mechanisms that can cause oscillations, and I am interested in learning about them all.

I am mainly interested as I would like to run some 6L6s in AB2(triode), so a grid stopper is out. Understanding the mechanisms that cause oscillations would help me prevent them from occurring.

I've done some research on 'snivets' (I think that's the term). Am I immune to the oscillations caused by strange behaviors around the knee of the anode characteristics in beam tubes, since I am operating them in triode? What causes parasitic oscillations in high gm triodes?

A lot of the explanations that I have found out on the net so far don't make a lot of sense and I get the feeling that the authors of the web pages don't really understand. The solution is always, 'put a grid stopper in there'. What other contributing factors and possible solutions are there and how do they work?

[Tom Bavis]

The theory books always showed the "tuned-grid tuned-plate" oscillator - not as a circuit that you'd actually USE, but as a parasitic circuit that will cause you trouble. Put a tuned circuit at the grid, another at the plate, and tweak either one - you'll make it oscillate. In this case, the feedback is via the grid-plate capacitance, and the tuned circuits set the frequency and provide the phase reversal that makes the feedback POSITIVE.

Now look at the 144-432 MHz (VHF) amplifier circuits in a '50s edition of the Radio Amateur's Handbook - the tuned circuits are lengths of copper tubing only a few inches long! The wires in your output stage might be that long... and if the tube still has gain at a few hundred megahertz, it may oscillate. 2-30 MHz (HF) amps almost always have a parasitic suppressor on the plate (50 Ohms with a few inches of wire wound onit). You could use the same for a grid stopper - wind a small choke on a 1K resistor. 10- 20 turns, say will have no affect at audio frequencies.

Paralleled tubes also have a push-pull parasitic circuit and generally require some stoppers to de-Q it.

[kenpeter]

Ferrite beads are made to be lossy at RF, rather than High Q.
They make a good stopper. With little effect at Audio frequency.
Very similar to the coil over resistor suggestion above.

[Miles Prower]

Originally posted by SpreadSpectrum
I'm looking for information about parasitic oscillations in tubes in general. Anyone know of a good reference or explanation out there on the net? I know there are several mechanisms that can cause oscillations, and I am interested in learning about them all.

No, not really.

I am mainly interested as I would like to run some 6L6s in AB2(triode), so a grid stopper is out. Understanding the mechanisms that cause oscillations would help me prevent them from occurring.

The 807/6L6 is one of those types that are susceptable to Barkhausen oscillation at plate current cutoff. I had that very problem come up when doing a PP Class AB1 design with 807s. At the cutoff, I had a 60KHz damped oscillation. The fix here consisted of adding a 1K5 screen stopper. This design also used a plate stopper of a 100R / 2W C-comp resistor paralleled with an RF choke: N= 10; #18 (space wound); ID= 7/16" with the resistor through the center of the coil, and the whole thing connected right at the plate cap connector. This stopped any RF instability. 6L6s in triode mode will probably still require screen stoppers, though the lower gain may make a plate stopper unnecessary (YMMV).

I've done some research on 'snivets' (I think that's the term). Am I immune to the oscillations caused by strange behaviors around the knee of the anode characteristics in beam tubes, since I am operating them in triode?

No. In any pentode, the screen grid is the main anode. The plate simply collects electrons. The tube itself does not know, nor can it care, that it's being operated as a pseudotriode.

What causes parasitic oscillations in high gm triodes?

High gain. Triodes and transistors have always been problematic components since they are all three terminal devices. There is no shielding to reduce the reverse transfer capacitance, and "Miller" oscillations are always a possibility.

A lot of the explanations that I have found out on the net so far don't make a lot of sense and I get the feeling that the authors of the web pages don't really understand. The solution is always, 'put a grid stopper in there'. What other contributing factors and possible solutions are there and how do they work?

The only contributing factor to unwanted oscillation is positive feedback. This can not be eliminated completely, but there are ways to minimize it. Grid and screen stoppers are one way. Make certain that these are C-comp resistors since C-comps have minimal inductance. With metal films, the metal is laid out as a spiral that makes it a coil. Even though this probably doesn't amount to more than a few nanohenries, it can still be problematic at RF, and can make the problem even worse. The aforementioned plate stoppers consisting of a coil de-Q'd with a 100R resistor is another.

Another possibility is to borrow a technique from solid state practice: Connect a capacitor from the plate to ground. Size it so that it has a high impedance at the signal frequencies, but presents a heavy load to the oscillation frequency. If done right, it will have minimal effect at audio frequencies. This will be the case if you made sure to keep leads as short as possible to drive up any parasitic frequencies that may appear.

Yet another SS practice is to connect a coil between the plate and control grid. This acts with the reverse transfer capacitance to form a parallel tuned trap to reduce the positive feedback. (This probably won't be practical for AF circuits unless the parasitic is of an unusually high frequency, but it's something to keep in mind.)

When dealing with high gain circuits (pentode voltage amps, cascodes) construct like you were doing an RF project. Keep your leads as short as possible, even if this sacrifices an asthetic layout of components. When installing stopper resistors, mount as close to the socket pin as possible. If your tube sockets (7 pin and 9 pin minis) have a central metal pin, be sure to ground it so that it can operate as an electrostatic shield. When installing screen bypass capacitors, connect the outer foil to ground, and mount the capacitor across the socket between the control grid and plate so that it can do double duty as an electrostatic shield.

If using more than one gain stage, a baffle shield between stages is also a good idea.

When doing cascodes, it's important to make certain that the upper control grid is well bypassed, and that parasitic inductances be reduced as much as possible. The upper VT is operating as a grounded grid, and any grounded grid will oscillate nastily if there is excessive control grid to ground impedance, so you definitely don't want a grid stopper there. You may, however, add a cathode stopper (100R should be good enough for that). If done right, a grid stopper at the lower VT control grid should be all that's necessary to kill off any parasitics.

[wakibaki]

Oscillation is a well-understood phenomenon.

It is sometimes understood in terms of the Barkhausen criteria (google it) and sometimes in terms of negative resistance (voltage decrease associated with an increase in current).

Parasitic oscillations are a result of the inadvertent satisfaction of these conditions as a result of real world component qualities as opposed to idealised component qualities.

The best way to understand these phenomena is to learn how to design an oscillator to meet a phase noise specifiication.

w

[SpreadSpectrum]

Thanks for the replies.

Tom,

Your explanation helped. I've found information on this subject puzzlingly scarce. I guess I would still like to know more about the tuned circuits. Some of it is obvious, like there is a big inductor connected to the plate with a distributed capacitance, etc. I'm more interested in the grid side. Where's the inductor? Is it just the inductance of the lead? Mine are 22ga solid core less than 2" long. An equivalent circuit would probably clear up all of my remaining confusion.

Ken,

What do I look for in a ferrite bead? I guess I don't know how to best select one. This seems like the ideal solution as it won't cause distortion with grid current.

[wakibaki]

30 nanohenries/inch is a common value used for PCB traces

w

[Miles Prower]

Originally posted by SpreadSpectrum
I'm more interested in the grid side. Where's the inductor? Is it just the inductance of the lead? Mine are 22ga solid core less than 2" long. An equivalent circuit would probably clear up all of my remaining confusion.

Stray impedances are reflected into the grid circuit by Miller Effect. In theory, this should be a pure conductance. However, it doesn't work that way IRL. Transit time effects and parasitic impedances in the plate circuit cause the reflection of dirty admittances. These can set up a parasitic resonant circuit that'll need to be damped out with grid stoppers. You can try Googling up Miller's original white papers on the subject.

What do I look for in a ferrite bead? I guess I don't know how to best select one. This seems like the ideal solution as it won't cause distortion with grid current.

The ones I used for a longwave project I did have u= 125; Len= 0.5"; OD= 0.125". So far, haven't needed 'em for audio work.

[SpreadSpectrum]

Thanks, Miles for that very comprehensive response. I've got 270Ohm screen stoppers at the moment, so I can raise that if I have problems. Thanks for the plate stopper recipe as well.

Oscillation is a well-understood phenomenon.

I understand the basics of phase shifts causing feedback to become positive etc. I am just having trouble mapping out all of the parasitic components to understand the more intricate points of how this oscillator works. The responses I have gotten here have gone a long way to clear up the confusion. However, a picture would be worth at least several hundred words, at least for me.

[SpreadSpectrum]

This is nice, usually you can hear a pin drop in the threads that I start...

Miles, when I read your last post, I heard a distinct 'click' in my head. Things are starting to make sense now.