• 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

Assuming that scaling relationship is reasonably correct, the 330 uF caps I use for B+ filtering in my preamp might be expected, very roughly, to have a resonant frequency in the range of 100 kHz. That is a factor of 4000 lower than the 400MHz capabilities of the valve itself.

It does not make any plausible engineering sense to me, to suggest that a capacitor that stops being a capacitor at 100 kHz, provides sufficient high-frequency bypassing for a 400 MHz-capable valve.

-Gnobuddy

If the capacitor is useless in your mind above 100kHz then it's also useless below 100kHz. The impedance rises in BOTH directions from that minimum. Fortunately for us the impedance rise is very gradual and the exact value of the impedance at any particular frequency is largely irrelevant as long as it's low compared to the impedance of the circuit it's supplying. To a 10 MHz signal the impedance of your 330uF cap would likely be in the 1ohm range, so very very low compared to the high impedances found in tube circuits. If you consider the capacitor useful in the audio band, decades below it's resonant frequency, then you must consider it equally useful decades above it's resonant frequency. It would be completely illogical to think otherwise.

It's also illogical to say that simply because a tube is capable of 400MHz operation that you should be designing audio circuits based on that bandwidth. That would be absolutely terrible practice. If your circuit has gain at 400MHz there's no point in worrying about capacitor minutiae, the design needs to go in the garbage.
 
  • Like
Reactions: 1 user
Gnobuddy said:
Add in the 400 MHz capability of the 6AG5 valve I'm using, and, once again, I am forced to the conclusion that an electrolytic mounted, at a minimum, over an inch away, is not going to provide adequate supply decoupling to guarantee stability for my circuit.
On the contrary, the lossy nature of an electrolytic at 400MHz may help guarantee stability for your circuit. Put a low loss cap across it, and you have created a UHF resonator with the wires; the electrolytic loss will be swamped and no longer effective in providing HF damping for the circuit.

People add low value bypasses when they want to amplify RF (or oscillate at RF). For audio you may be better off with a nice lossy supply rail (at RF frequencies).

We deliberately put lossy elements elsewhere - grid stoppers. Why remove lossy elements too?
 
  • Like
Reactions: 1 user
Exactly, as DF96 said.

I had this problem with a s.s. design; kept getting x-MHz oscillations until I purposely removed poli-caps from the power supply rails. This seemed somehow to be wrong as simulation did not show any problems, until I hit on the thoughts by DF96. (As said elsewhere, simulation does not show resistance/inductance of p.c. tracks and intertrack/component capacitance, however small. Though only audio, one has to keep some r.f. savvy in the mind!)
 
Perhaps I should also have said that if you want to build audio circuits with UHF valves then it helps if you know how to build UHF circuits with UHF valves, and also know the difference between an audio circuit and a UHF circuit.

The thing which people forget is that the valve is not very intelligent; it can't tell which frequency range you wanted it to amplify. It doesn't know whether you intended to build an amplifier or an oscillator, but simply behaves according to its nature in the electrical environment in which you placed it.
 
Fortunately for us the impedance rise is very gradual
It's a first-order filter either way, either plus or minus 6dB/octave.

the exact value of the impedance at any particular frequency is largely irrelevant as long as it's low compared to the impedance of the circuit it's supplying.
Agreed, the object is to make the B+ rail effectively a signal ground.

It's also illogical to say that simply because a tube is capable of 400MHz operation that you should be designing audio circuits based on that bandwidth.
I believe we have a miscommunication here. UHF-capable devices are capable of misbehaving at UHF frequencies, even if you intend to use them at audio frequencies; therefore I want the circuitry around them to behave properly even at those frequencies. In particular, we don't want enough stray inductance to show up to create an unintentional UHF oscillator.

If your circuit has gain at 400MHz there's no point in worrying about capacitor minutiae, the design needs to go in the garbage.
If you use a device capable of 400 MHz, you DO have gain at
400 MHz, whether you want it or not.

And if you're assuming I've done nothing externally (outside the device in question) to control bandwidth, you're making the wrong assumption.

Nevertheless, I want the B+ rail to behave itself - provide a low AC impedance - over the entire frequency range where the device itself is capable of amplification.

-Gnobuddy
 
Last edited:
Perhaps I should also have said that if you want to build audio circuits with UHF valves then it helps if you know how to build UHF circuits with UHF valves, and also know the difference between an audio circuit and a UHF circuit.
If you know something about UHF that you're actually trying to convey, you would be much more successful if you stopped being condescending.

The thing which people forget is that the valve is not very intelligent; it can't tell which frequency range you wanted it to amplify. It doesn't know whether you intended to build an amplifier or an oscillator, but simply behaves according to its nature in the electrical environment in which you placed it.
Which is exactly why I want the power supply to be well-behaved even at the highest frequency the device is capable of.

Which "people" are you speaking of, exactly?

-Gnobuddy
 
It's a first-order filter either way, either plus or minus 6dB/octave.

Far from the resonant frequency yes, but there's a broad flat area where resistive elements dominate. See the lower left of page 8 here.

If you use a device capable of 400 MHz, you DO have gain at
400 MHz, whether you want it or not.

Sitting there on your desk it's gain is zero. The circuit you put it into determines whether there will be gain, and how much at any given frequency. The device may be capable of functioning in an RF amplifier circuit, but that doesn't mean you should be building one for it. Quite the opposite really. You say you're already taking steps to this effect so I probably don't need to drone on about this.

The inductance of an aluminum capacitor is almost entirely from the loop of it's leads, there's no other significant internal inductance. A radial 330uf/450v cap should have about 15nH of series inductance, that's 2 centimetres of wire. The composition of the cap is irrelevant, to do better you'd need to get your circuit's loop area down under 2cm. Given the size of tubes and sockets it seems unlikely you're going to do significantly better. With miniature components maybe you could turn 2 ohms at 10MHz into 1 ohm, but if that tiny difference is make or break for your circuit (or has any effect whatsoever) then it should never have been built.

You should be less worried about this than you are about the impedance rise in the audio band where you have a lot gain, so try to keep this in perspective. At 500Hz your cap already has higher impedance than it does at 10MHz. At 20Hz it's 25ohms. If you're not obsessing about reducing that by 1ohm then it's illogical to do the same for a band your circuit shouldn't even be amplifying.
 
Last edited:
  • Like
Reactions: 1 user
Gnobuddy said:
Nevertheless, I want the B+ rail to behave itself - provide a low AC impedance - over the entire frequency range where the device itself is capable of amplification.
No you don't, for reasons I explained earlier. You could get better RF stability by allowing some resistance in the supply rail. If a valve sees mainly resistive terminations then it can't oscillate. For oscillation you need the right mix of capacitance and inductance, relatively undamped by resistance. The circuit wiring and component strays provide the C and L, so you have to ensure that there is enough R to damp things down.

As I said, it is not enough to know how to build a UHF circuit; you also need to know the difference between that and an audio circuit.
 
If you use a device capable of 400 MHz, you DO have gain at
400 MHz, whether you want it or not.

Not desiring to belabour this, but I reallly have a problem with this statement, with due rerspect.

A device surely is only one part of the capability of a circuit; this has been stated above in different ways. It needs all elements of a circuit to 'come together' at a particular frequency to have gain there. E.g. nothing on earth will make a 400 MHz circuit work at say 100 Hz, although the active device is fully capable of working at that frequency.

In general, I am somewhat puzzled about all the worries about deep-MHz problems in audio circuits. I must have been singularly lucky/gifted not to have really had such problems. Sure, one takes care of matters which can bother outside the intended frequency band, but much of that comes with basic design procedure. If problems still creep in, it should be simple to 'quench' such by sober investigation/action. One can have power anode leads in a loom with others if the basics regarding undesired proximity and length is observed; many amplifiers do this. This should not be thumb suck or other such guesswork.
 
If you use a device capable of 400 MHz, you DO have gain at 400 MHz, whether you want it or not.

I agree with this, unexpected parasitic circuits can & are a real problem to be aware of. Terman & others cover these in some detail for those of us willing to take the time to look.:)
 
If you use a device capable of 400 MHz, you DO have gain at 400 MHz, whether you want it or not.

I agree with this, unexpected parasitic circuits can & are a real problem to be aware of. Terman & others cover these in some detail for those of us willing to take the time to look.:)

Please see my post #130; perhaps I worded it poorly.

One is well aware of the threat of parasitic oscillations; such awareness should be standard practice in all designs. But I consider the reference to 400MHz per sé misleading. Parasitic oscilations can occur at any (high) frequency, where a device as well as circuit elements are favourable to such.

Again: It may be considered as semantics, but devices are not capable of a certain frequency (per sé) although there is obviously a frequency limit for each. Circuit elements can cause unwanted oscillations whatever the device, etc. - I don't think we disagree, but it can be important to have the correct approach regarding this. For prevention one concentrates on the circuit and layout, the device is of secondary importance (in the right context, naturally).

Whichever way; if we both succeeded in emphasising the importance of possible parasitic oscillations, the purpose is served.
 
Defective electrolytic filter/decoupling capacitors will also cause oscillation or "motor boating"...

Electrolytic capacitors are best checked by substitution or wiring a good unit across the suspected cap. If the capacitor is shorted or leaky, the defective capacitor must be replaced.

The older multisection can capacitors can be tested by substitution. The older can capacitors are expensive or hard to find; either the entire cap has to be replaced if a defective section is found, or a suitable capacitor wired across the defective section if space is available. If any section of the multisection can capacitor is leaky, that section cannot remain connected.

If you cannot find a replacement for the original defective multisection can capacitor and want to preserve the integrity of the unit, CAREFULLY remove the capacitor as to not break the phenolic chassis mount. Then put on gloves, remove the bottom seal with the terminals carefully and save it, then remove the "innards" of the capacitor (goo, etc.) and save the can. Do not remove any internal grounding jumper from the can. Discard the "goo", etc.; then clean the can thoroughly, and install replacement capacitors inside the can, using insulating tubing to prevent shorts when reconnecting to the bottom seal terminals. The negative side of the capacitors is the can, so the can and the negative side of the replacement capacitors must be connected together, preferably at or as close as possible to the common negative terminal. Take your time and do not rush this.

WARNING - HIGH VOLTAGE - Make certain the unit is unplugged and the capacitors are discharged before working on the unit. The voltages used in tube equipment can be lethal!
 
Clicking Noise

As for clicking noises, I have found that many are caused by leaky caps resulting in what has also been called popcorn noise. It can sometimes be found by doing a simple dc resistance measurement across each cap... a leaky cap can show a relatively low resistance (<80kohms or so).
 
give em a push

Lead dress is a BIG issue....Iv'e found out the hard way but its helped me build lovely quiet amps...

motorboating and clicks.....sounds like caps....are you using new ?

Get a stick and push the wires around and see how they react while monitoring...you may discover a few sleepers laying around in loops

:D
 
over a year has passed between the last two posts. Does this problem still exist?

Regarding sorting out feedback, I would respectfully warn against just switching leads regardlessly. Oscillation caused that way can overload the amplifier, thus causing extra peaks/spikes, which can spike through/blow an OPT.

It should be easy to wire tubes up to an OPT the correct way. Before application of global NFB, first use a feedback resistor of say 50 - 100 times the given value. Feedback would then be too low to cause oscillation, but one will notice the correct connection from gain either increasing (pos. fb) or decreasing with the feedback resistor in place .... When the right phase relationship has been established one can then wire in the given NFB resistor.