• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Oscillation in tube amps

Is 20 mS steep enough? Working quite well in a low noise audio pre-amp.

I mention AGAIN, as others also did:

No component governs results on its own. A circuit performs/misperforms because all the elements contribute. A 400 MHz spec. only indicates that said tube is capable of working there, all else being in order at that frequency, including the operating point of said tube.

Occasionally graphs are available showing tube characteristics at other than the tested Ia. 'Steepness' (gm) can easily drop to a fraction of the test value when using a tube at lower current than tested, as is usually the case in audio - then no longer a '400 MHz' tube.
 
Hello Crew,

I have read this thread and many other sources and come to the conclusion that oscillation in a tube amplifier is not a good thing. Oscillation like an accident is the result of many things leading to an undesired outcome. Preventing anyone of those thing is like breaking a link in a chain and we avoid the undesired outcome. A casual observer may come to believe that any one link in the chain is the sole cause of the undesired result.

Even the stray L’s C’s add to resonance, high Q’s and instability. Lossy (Resisitive) things like ESR, grid stoppers and snubbers provide some dampening. Simulation neglects too many of the stray things to make it only partially useful as a tool.

A couple thoughts and rerated questions to the crew:

Why do we not spend more time looking at the power supply?

Can we take a closer look at grid stoppers and why do we always want to use Carbon Composition resistors. Why not use much quieter metal film resistors in parallel? (Parallel resistors reduce the stray inductance.)

Seems that we poke and blame individual parts all the time without knowing; only wishing and hoping we can fix things. Seems odd that we do not see more network analyzers, Bode plots, impedance plots and phase margin plots at work? We could watch the phase margin change as we swap out a low ESR capacitor for a cheap high ESR version capacitor.

DT
 
Dual,

Not to over-expose myself, but instruments to display Bode plots are comparatively expensive. Yes, it would be a design-advantage (when at the CSIR there was one for us), but with some experience one can get away by looking at a frequency response plot, or even a square wave only. Tutorials are available on the internet.

It is also very convenient to analyse frequency response with the aid of Spice to detect frequencies of possible instability. Add-on ancillaries for use with a PC programme is available. I myself have never used those; with some experience and background one can quite easily design/solve in situ. (That would need a stache of capacitors and resistors, like some of the more active designers among us have assembled for convenience.)
 
Ah Dual,

I have not explained some of your questions above. Let me try in coloured type after each question in your text.

WARNING: LENGTHY AND TECHNICAL POST. (If such bore you, simply skip the post!)

Hello Crew,

I have read this thread and many other sources and come to the conclusion that oscillation in a tube amplifier is not a good thing. Oscillation like an accident is the result of many things leading to an undesired outcome. Preventing anyone of those thing is like breaking a link in a chain and we avoid the undesired outcome. A casual observer may come to believe that any one link in the chain is the sole cause of the undesired result.

Just about. Add to that 'tendency to oscillate'. Self-oscillation will overload the amp, making normal/linear reaction to signal impossible. It can also come in spurts at certain signal amplitudes, making it impossible to diagnose except with at least an oscilloscope.

Even the stray L’s C’s add to resonance, high Q’s and instability. Lossy (Resisitive) things like ESR, grid stoppers and snubbers provide some dampening (damping!). Simulation neglects too many of the stray things to make it only partially useful as a tool.

Correct, although simulation might reveal a tendency-to/danger-of oscillation, e.g by a sudden increase (peak) in response in a frequency run. Simulation, depending on how good it is, can be a quick (and painless!) way of revealing possible 'danger' areas. To be complete, one also needs to keep component tolerances in mind. Simulated component values are exact. There are sim. programmes available to include component tolerances and examine worst case conditions, but they are quite expensive/complex.

A couple thoughts and rerated questions to the crew:

Why do we not spend more time looking at the power supply?

One should also spend time looking at power supplies, particularly transistor power supplies. They are quite capable of causing h.f. instability at frequencies where their impedances are no longer low. A power supply has a low impedance only at d.c.(if that makes sense!). Examining the impedance over the audio range and higher can reveal quite a horror show! (Also so with I.C. regulators!)

Can we take a closer look at grid stoppers and why do we always want to use Carbon Composition resistors. Why not use much quieter metal film resistors in parallel? (Parallel resistors reduce the stray inductance.)

The accent on resistor inductance is often overrated. Yes, with transistors (low impedance, high frequency response capability) inductance should be considered, but in my > half-century experience I never had problems with say metal film types - neither did I have with wire-wound load resistors. (Inductance of the latter is negligible at way above audio frequency. Talking of normal ones!)

Seems that we poke and blame individual parts all the time without knowing; only wishing and hoping we can fix things. Seems odd that we do not see more network analyzers, Bode plots, impedance plots and phase margin plots at work? We could watch the phase margin change as we swap out a low ESR capacitor for a cheap high ESR version capacitor

Again, very often true and quite informative. But as stated previously such instruments are relatively expensive, and not often seen at hobby level. SPICE programs give amplitude and phase responses; I have often used those, and ESR can be simulated by simply adding a series resistance equivalent to a practical ESR figure (one needs to remember that simulator values are exact and 'inductionless') .... but again as was remarked earlier, also including component tolerances can be complex and expensive - apart from the in-circuit capacitances and inductors not known. Good at tutorial level, but not often used in practice; one simply steers clear of borderline conditions. (I do not know whther such are used in big brand designs)

The effect of capacitor ESR is again overrated to my mind, but it can be a factor particularly in high-power transistor instruments. As said it can be simulated in practice. Often at the frequencies-of-danger, an audio amplifier should no longer be capable of gain - but that is a qualified statement. What with present often high NFB values in practice, anything is possible these days. But then: With my transistor amplifier (80 + 80W), there was a frequency peak at all of 12 MHz :eek: and with medium NFB (for a transostor amplifier) of 29dB! - and not just because of ESR; it also showed up on SPICE where components are 'perfect'.

Explanations were simplified - you can mostly find a more complete treatise of the subjects on internet.
 
Last edited:
DualTriode said:
Can we take a closer look at grid stoppers and why do we always want to use Carbon Composition resistors.
We don't want to use CC, but some people wrongly believe that the grid stopper must have low inductance. In many cases the opposite is true: some inductance helps. You can stop oscillation with an inductive resistor (e.g. metal film) or a resistive inductor (e.g. ferrite bead). Ordinary resistors at typical values are largely resistive up to VHF frequencies, so even if inductance had to be avoided it would not matter very much.

Why do we not spend more time looking at the power supply?
The power supply is rarely an issue in single stage high frequency parasitic oscillation. It may be a factor in low frequency motorboating, or global loop instability.
 
High transductance tubes have enough hoick to overcome the dreaded Miller Effect and have extended bandwidth as a result.

The Miller Theorem tells us that we can add external capacitance and or resistance and reduce the bandwidth. That is what we are talking about we speak grid stopper.

Unfortunately when we add grid stopper resistance we are also adding hiss kind of noise.

It occurs to me that we can add external capacitance to increase the overall Miller Effect. Perhaps a combination of both grid stopper and added external capacitance.

What values of capacitance and resistance do we use? Do we make some assumptions and calculate? Do we use SPICE? We could say yes to both. I recall reading a thing from Jensen Transformers about damping the ringing of a microphone input transformer. The Jensen folks related that things were to variable and complex to calculate. The procedure was to adjust under test. Select a moderate value capacitor and switch in different value resistors to minimize the transformer ringing. When that was done a lower value capacitor is selected and the resistor value is again adjusted to minimize the transformer ringing. This procedure is repeated until the smallest values of resistance and capacitance are identified that achieve the desired results.

Thoughts?

DT
 
A nod to DF96:

.... particularly transistor power supplies

I should have said transistorised power supplies (as in I.C. regulators further down.) (One often sees I.C. regulators used as constant current sources e.g. in long-tailed pairs in amplifiers, being then regarded as such over the whole audio range. Some of them are not!)

Further regarding grid 'stoppers': As DF96 said. I prefer the more exact way of using R.C networks to ground where an exact pole is required for stability purposes. This, what with the spread in tube characteristics often found in modern tube manufacture. However, for general stability purposes where exactitude is not required, a Miller capacitor can certainly be used.

Yes, series resistance need to be considered regarding Johnson noise, particularly in pre-amps. (O.T.: That was the disadvantage of RIAA circuits in early days where the particular stage had a series input grid resistance and the RIAA network connected anode-grid.) One can use SPICE where everything is known or to bring one within ball-park values. As with Jensen, I prefer the practical way if one uses bogey components.
 
Disabled Account
Joined 2018
In the end, how did you resolve this? The photo in your original post is EXACTLY what my amp looks like.

In my case I found that there was a parallel RC network forming part of the feedback loop and reducing capacitance there killed the HF oscillation. In doing so I lost some high end distortion performance. I didn’t spend too much time on it after that but there may be a middle ground where I could have both.
 
DualTriode said:
The Miller Theorem tells us that we can add external capacitance and or resistance and reduce the bandwidth. That is what we are talking about we speak grid stopper.

Unfortunately when we add grid stopper resistance we are also adding hiss kind of noise.
The grid stopper is not necessarily stopping oscillation by limiting bandwidth by working with the Miller capacitance; it may also act directly to damp a resonance in the grid circuit. A typical grid stopper will not add much thermal noise.
 
I read a lot about grid stopper resistors but little about anode stoppers. In a resent build I fixed an oscillation with a resistor between the power tube anode and the OPT. An ordinary cement type wirewound worked best so my conclusion is that here inductance is needed. I've seen others to use the resistor only at one tube of the push pull pair so I did the same. Is that the right thing to do? Also I would like to ask what should be more appropriate to use: higher resistanse with lower inductance or the opposite? Or perhaps something else? I have at hand these inductors SMSC-6R0M-01 Fastron | Mouser Greece Should they work better than the resistor?
 
Has anyone built an amp using uhf mega steep tubes or just using theory?? Can you use the normal ways to control oscillation?

EFP60, not a power tube but developed for TV IF has a very large gain/bandwidth product. So it is able to provide a very low drive point impedance (16K) in this cct to the following grid. And still provide sufficient gain. In an RC coupled cct this results in lower distortion.

Used lots of them in a research project when they were available but never in an audio amp. Another possibility is the 7788. Was used in the HP 1402A vertical amp plug-in for the HP140 Series scopes.

Neither one a project for the inexperienced. Some may have seen this on another thread.
 

Attachments

  • EFP60 Philips.pdf
    127.6 KB · Views: 110
  • 7788 E810F Philips.pdf
    338.3 KB · Views: 107
  • EFP60 Driver Stage Schematic.pdf
    437.8 KB · Views: 133
Triode Plots

FYI found on web
 

Attachments

  • E810F.jpg
    E810F.jpg
    67.9 KB · Views: 456