• 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

Unhook the negative feedback resistor network. If it is an output tube it will probably happen with the nfb unhooked. And they can show up at any point in the wave but usually halfway or higher up to the peak.

Grid resistors usually tame the beast. A snubber across the primary of the opt can help but that is usually feedback related.

Sometimes it takes some capacitance between tube elements to stop it or an r-c series network. Small 100pf 1kv caps are a way to start.

These gremlins are usually layout related or due to lack of grid resistors. You can also use ferrite beads on a screen grid or plate lead.

Trial and error. Don't ask me how I know.
 
Yes, I know what you mean even though I don't think it's called a grid stopper. The idea is the same. The resistor has to be right on the base/gate of the device for it to work. Particular MOSFET, they die a sudden death when oscillates. Die quietly, with no pop, heat, just die stone cold with source shorted to drain.
 
I have designed numerous class d and class AB mosfet amps successfully.
I recently designed my first bipolar transistor amplifier.
It oscillated badly. I hadn't noticed at first and the first I knew the amplifier had fried and blown a PNP transistor.
I tried my old mosfet amp tricks like increasing VAS and feedback capacitor values to fix it. Neither worked. In the end I resorted to the mosfet gate resistor trick and added 10ohms to each base connection and that fixed it.
 
Yeh, those oscillation has nothing to do with the amp global NFB and all. They just oscillate on it's own as an independent stage.

It seems like the tubes are much easier to layout and wire as they are not very prone to oscillation. It seems like their frequency response is a lot lower than SS. You can get away with sloppier layout and wiring with tubes than transistors.
 
I've been thinking recently (the past few years) that it makes a lot of sense to put an RC damper on the secondary of the OT. Just to provide some reasonable HF load rather than the pure (more or less) inductance from the tweeter or whatever driver.

Have not tried it. Surprised that this is not seen more in tube amps. I have only seen a few commercial designs where it is done.

Pete B.

i have seen this done on many japanese tube amps...i also put them in my amps...
an RC network across the OPT primary also, but this is harder to do, but i have seen this done....
 
I'm currently chasing Snivets in a 6P41S amp that is laid out on a breadboard similar to the way George builds this prototypes.

The oscillations take place during zero crossings, and only with a speaker attached as a load. They do not show up with a resistive dummy load.

The oscillation is more likely when cranking up the plate current (makes sense as gm increases with current).

Oscillation frequency is 166KHz, and is audible as harshness on S and F, particularly on female voices.

I have ferrite beads at all control socket terminals as well as Grid-Stop (4.7K) and Screen-Stop (270R) resistors.

A temporary fix of a 500pF cap in series with a 1K resistor from anode to cathode seems to quench the oscillation.
 
Could be a resonance in the effective network formed by the OT and speaker reactances.

Moving the series R-C network to across the OT primary may spoil its Q enough to suppress oscillation.

Maybe 500pF (1600V MKP) + 3.3K 3W: this should do very little harm to the sound.
 
  • Like
Reactions: 1 users
That was pretty much my thought Rod. I could not get the oscillation with a resistive load, but when connected to the FF125WK ( single Full range speaker) it would.

I'll try increasing the resistor value to see how large I can go before it starts to oscillate.

There is a good chance that it is in part because of the breadboard layout. Once I do a proper layout in a chassis it may not be an issue, however I will leave room for Zobels in case they are needed.

This begs the question of where the Zobel is best applied. (1) across the output tube anode to cathode, or (2) across the reactive load (speaker terminals).

In either case the reflected impedance is seen by the output of the tube, dampening the oscillation response. So it seems reasonable to me to apply it at the anode of the tube.
 
In my experience, oscillation is usually a function of layout and proper placement of grid resistors, which should always be soldered directly to the tube socket.

When wiring from the rca jack hot signal to the grid, if I am not mistaken it is best to have the grid resistor directly wired to the tube socket, then on the other side of the resistor, run the signal cable to the tip of the jack with a shielded cable.

Where is the best place to solder the shield of the signal cable? I'm assuming the closest ground point nearest to the socket?
 
In my experience, an otherwise apparently stable amplifier can be made to oscillate with the use of excessive amounts of negative feedback. Due to the fact that phase shift occurs in real world amplifiers, at extremes of frequency, the uncompensated feedback loop will carry everything from the output to the input, including what was supposed to be out of phase, but at extreme frequencies, due to phase shift, are now enough in-phase to make the loop positive feedback at those phase-shifted frequencies.
Some tube types, such as 6BG6 and 807, are more prone to oscillation than more common types used such as 6L6 or 6550. Such amplifiers require phase compensated feedback loops (to reduce the feedback amount as the amplifier's phase response starts to shift) and may even require Zobel networks on the outputs to make them completely stable with inductive loads.
Lack of a ground common reference between the output transformer secondary in a negative feedback amplifier will cause the amplifier to oscillate at a wide range of frequencies, usually in the audio band where it is most efficient, when connected to a floating load, such as a loudspeaker.
Of course, before we get to all these issues, we must start with good layout practices. Keep power tubes, output transformer wiring and driver stage signal wiring away from input wiring and put as much distance as possible between these tubes to limit cross coupling. Use of shielded coax cabling for long runs within the chassis may be necessary to ensure good s/n and stability. Use of a common grounding point is also preferred.
I find the best results come from using good RF design practices in audio designs. This makes an inherently stable layout, where audio and power leads are kept short or shielded when longer runs are necessary.
 
I'm thinking that surely hf or vhf oscillations are not created or being sustained by resonance in the OPT which would likely present megohms at those frequencies.
More likely caused by inductive or capacitive coupling in the wiring correct?
 
Last edited:
I'm thinking that surely hf or vhf oscillations are not created or being sustained by resonance in the OPT which would likely present megohms at those frequencies.
More likely caused by inductive or capacitive coupling in the wiring correct?

The bandwith of output transformers at small signal levels can be surprisingly wide. For instance, on an 807 amplifier with more than 20dB of inverse feedback, driving the amp near clipping results in a bubble appearing as it comes out of clipping, on the sinewave output. This bubble contains spurious oscillations as high as 1 mhz when viewed on spectrum analyzer. Some of that could be capacitively coupled between pri and sec windings.
 
Interesting. Hadn't considered capacitive coupling within the OPT. Well this topic has had my interest lately since I modified a SE 300B for a friend using a 3 pin regulator and filtering on the filaments. The mod really quieted down the hum but now the amp ended up motorboating at clipping. I could see it on the scope too, flagging just below the peak of the falling waveform.
Tonight I went looking for it, put some latex exam gloves on for a bit of extra protection and began carefully moving wires around looking for impact on the scope. I found that the wire coming off the negative side of the filament DC supply definitely changed the waveform. Moving the part of the wire that ran parallel to the 300B socket made a huge increase in the spur. Moving it closer made a big spike. I tried using RG58U and grounding the shield but that didn't completely eliminate it. So I made the wire just a tad longer and re-positioned it to make a hard 90 degree turn as it approached the tube socket instead of running parallel. That completely eliminated the spur. Now it clips very nicely with a somewhat rounded waveform peak.
 
  • Like
Reactions: 1 user
For anyone interested, I run my home on an Outback VFX3648 inverter. Well, I bought some Eton silk dome tweeters and installed them on my stereo. They only lasted a couple days and one opened up. Brought the scope home and a Dale 8 ohm non-inductive load resistor(NH250) .I found that there was a ~20khz waveform on the load. If the amp was really cranked (Hafler Pro5000) the waveform got to ~75VPP.

I found it was coming from the inverter. Looking at the inverter waveform there was a little ~20khz oscillation or possibly some kind of computer noise riding on the very top of the 60hz waveform.

It was entering the amp through a defective audio cable between the DVD player and the preamp input. One end of the cable was ungrounded which apparently created a high enough impedance for 20khz to develop across the wire.

I had another tech confirm that he is seeing this waveform on his inverter as well. Replacing the audio cable eliminated the signal but I am going to build a low pass filter and put it in the AC line. Ideally it should go at the inverter output but that would probably require a huge choke.
 
For instance, on an 807 amplifier with more than 20dB of inverse feedback, driving the amp near clipping results in a bubble appearing as it comes out of clipping, on the sinewave output.

I may have run into this recently. At 1 KHz, pushing the input to the point of clipping, on a breadboarded amp that is rather poorly laid out, I got these "bubbles" when I backed off the test input level. This was located right after the phase inverter, and before the final output tubes (not 807s, but pentodes). There is a high level of feedback in this situation. The image shows both "legs" of the inverter output.

Are these the same kind of bubbles that are described for the 807s?

I hope these bubbles will disappear in the final build using better layout and much shorter wires. . . . I also noticed that touching an AC balance pot in the PI section affected the bubbles (in one case well below clipping causing it to appear, and then disappear when not touching the pot), but the bubbles appear on their own at just below clipping. I hope this is just RFI and not due to the OPTs.
 

Attachments

  • 1 kHz near clipping.jpg
    1 kHz near clipping.jpg
    40.4 KB · Views: 1,438