• 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

Member
Joined 2006
Paid Member
Adding a small cap in parallel may create exactly the situation it is presumably intended to avoid: a high frequency peak in impedance. Fortunately most electrolytics have sufficiently high HF losses that the peak is not too big, so the unnecessary cap does less harm than it might.

The reality is that because electrolytics are not too great at HF they can damp down the resonance introduced by the extra cap. Normal quality electrolytics don't need bypassing; very low ESR electrolytics must not be bypassed!

Am really glad you pointed that out!
 
I was chasing some spurious Barkhausen oscillation (while testing to near clipping at 4 ohms) in my new amps and discovered that the driver signal to the power tubes was not balanced at high frequencies, even though with the power tubes removed the signal was balanced. I haven't tracked down the reason for this, my guess is magnetic coupling from the output transformer since one side of the push-pull quad is closer than the other. This side showed much lower output at the power tube grids than the side farther away from the OPT.

Previously I had reduced the oscillations by fiddling with the Zobel values across the plate/B+ and CFB winding (which are returned to driver cathodes) and feedback capacitors across the feedback resistors but was not able to completely cure it. After I discovered the imbalance, I added 130pF total capacitance across the driver plate resistor that showed more gain and the imbalance disappeared, along with the oscillation.
 
This is while McIntosh MC-60 has small chokes in each output tube plate, with different values, to reduce and/or eliminate the feedback that would occur if the resonance frequency were the same in both sides of a push-pull primary (with same value chokes). Also, if you pay attention, the valves are not physically aligned with the plates/ grids in the same orientation in the chassis, they are rotated 180 degrees to try to avoid the external coupling effect. This last one can be tricky to identify with a cold amp or running at civilized power level, but it can be clearly seen when you are prototyping your amplifier with all the wiring flying around and your output tubes are too close and somehow aligned to each other. Keep them at least 3" far from each other if you can. If you are running into problems using a single-ended, maybe the issue is the tube itself due to poor internal construction or it's worn after a long lifetime use.
 
Hi, lately i built one tube amp from audio innovation 500 circuit diagram, it works good without feedback, there is some hiss when volume down to zero but there is no oscillation, but when i apply the feedback it begins to oscillate. i toggled the polarity of output transformer and it makes the oscillation worst. i tried paralleling 330pf with feedback resistor and didn't see any difference, what do you think i should do?
 

Attachments

  • rys12.gif
    rys12.gif
    21.4 KB · Views: 471
...what do you think i should do?

1) Make certain there is no oscillation while running open loop. "There is some hiss when volume down to zero" -- this could be a manifestation of a high frequency oscillation. You'll need to o'scope around to make sure you don't have an oscillation well above audio frequencies.

2) Remove that 450pF capacitor from the plate load of V3a. Whoever did that didn't know what he was doing. The phase shift of a parallel RC connection is:

B(w)= -arctan(wCR)

As w increases, the phase heads towards -90deg. That isn't what you want to see as it can promote instability. The proper method is to parallel the plate load with a Zobel (series RC) network that is phase limited. The usual values are ZZobel= 0.1RP, and whatever C is needed to limit phase shift while not doing excessive damage to the high end response.

3) Get rid of that 1M5 and 0uF01 between the first and second stages. It's yet another unrestricted source of phase shift.

4) Why is the gNFB input connected to the 16R tap? If you're not using 16R speeks, then move it. Connecting to the highest tap on the OPT secondary leads to many feedback headaches. Better to connect to the 8R tap if you're using 8R speeks and then readjust Rfb to compensate for the lost gain.

5) How much gNFB? Since the finals are xfmr matched, you have to be careful not to try too much gNFB. The original Williamson design could use unusually large amounts of gNFB due to the special OPTs it used. Otherwise, it would never be stable.

Also, make certain your ultralinear taps aren't crossed.
 
I agree with Miles Prower,

I did not read his response until after I typed this:

Phase shifts:

1. You have 2 parallel high frequency poles on the input tube:
Plate load.
Coupling rc in series with coupling cap c.

2. You have Miller Effect Capacitance on the output tube control grids.
Ultra Linear screens move in opposition to their control grids.
This is one that most people forget. It is particularly troublesome when the driver impedance is high.

3. You have an unknown output transformer, with its particular frequency response and particular phase response.

4. Which one of these might be the cause of the oscillation when you apply negative feedback?
I do not know, sorry.

But you seem to have solved the problem.

Happy listening!
 
Last edited:
I have not gone through the 17 pages regarding this subject. I do not recall seeing a discussion on adjusting the feedback loop for stability during the years I have been visiting this forum. I may well have missed it and apologies if I restate the obvious.

On RADIO and BROADCAST HISTORY library with thousands of books and magazines is a pdf of the Magazine Audio dated January 1960 that on page 20 and 21 describes in great detail to adjust the feedback loopback on a PP amplifier to ensure stability with an unknown output transformer.

In a Dutch book dealing with practical construction of amplifiers it is suggested to use something like 22k inseries with 5nF across the primary of an output transformer that is not of "the highest quality". In the Radiotron Designers Handbook Langford-Smith suugests when using UL to put a 47 ~ 200 Ohm resistor in series with 1 ~ 2 nF between plate and g2. This in addition to stopper resistors and perhaps even a small capacitor on g1 to ground.

Most problems can be avoided by carefull layout but you are still stuck with the unknown output transformer hence the suggestion to carefully tune the feedback loop.

AM
 
I was chasing some spurious Barkhausen oscillation (while testing to near clipping at 4 ohms) in my new amps and discovered that the driver signal to the power tubes was not balanced at high frequencies, even though with the power tubes removed the signal was balanced. I haven't tracked down the reason for this, my guess is magnetic coupling from the output transformer since one side of the push-pull quad is closer than the other. This side showed much lower output at the power tube grids than the side farther away from the OPT.

Previously I had reduced the oscillations by fiddling with the Zobel values across the plate/B+ and CFB winding (which are returned to driver cathodes) and feedback capacitors across the feedback resistors but was not able to completely cure it. After I discovered the imbalance, I added 130pF total capacitance across the driver plate resistor that showed more gain and the imbalance disappeared, along with the oscillation.


If you are using a split load phase inverter, the source impedance of the cathode side is much less than the source impedance looking back into the plate side. So might look OK with the power tubes removed but will not be with those replaced in their sockets. Morgan Jones makes that point in his book 'Valve Amplifiers'. OTOH, I've built quite a few amps using that kind of phase inverter & never had a problem.:)
 
Hi, lately i built one tube amp from audio innovation 500 circuit diagram, it works good without feedback, there is some hiss when volume down to zero but there is no oscillation, but when i apply the feedback it begins to oscillate. i toggled the polarity of output transformer and it makes the oscillation worst. i tried paralleling 330pf with feedback resistor and didn't see any difference, what do you think i should do?


Looking at your cct rys12, both V3a & V4b are missing grid stoppers. Grid stoppers should be 1/2 watt CC, mounted as close to the tube socked as they can be soldered. Any value from 1K to 10K is OK.
The noise from these resisters should be not audible, if it is they are defective. I've never had noise problems due to CC resisters & built many amps using them. Most were Allen-Bradley.
Short leads are a must & arranged such that any high signal level such as the output plates are not close to any low level conductor, especially the input.
Try adjusting the NFB resister with the input cct connected to common/ground. The cct as shewn appears foolproof & should work OK with no problems if properly configured.:)
 
jhstewart9,

I think the impedance issue of a split load phase invertor is more complex than it seems.

I believe the cathode impedance is 1/Gm + (RL/(1 + u)).

Lets use a tube with u = 20, Gm = 2600micromhos, rp = 7700 Ohms, and RL = 22k
(recognize the tube type?)

1/GM = 385 Ohms
1 + u = 21
22k/21 = 1,048 Ohms
The cathode impedance is 1433 Ohms.

OK, so now we know that the cathode impedance in parallel with the cathode resistor, is much lower impedance than the plate impedance in parallel with RL.
So far, so good; all the bad-mouthing of a split load invertor seems to be true, right?
Read on . . .

But, now lets operate the split load invertor using a drive signal that does not draw grid current.

If the high impedance of the parallel plate and RL are loaded by the next stage, any current into or out of that next stage will also be coming out of or into the splitter cathode's next stage circuit. That current will come from the other phase of the next stage too.

The change of current in the cathode of the phase invertor, will be exactly the same as the change of current of the plate of the phase invertor.
If that is not true, then the difference of those two currents can only be coming from the grid of the phase invertor.

Equal current in the in-phase load, as the current in the out-of-phase load (equal, but in opposite directions).

Well, there is one other forgotten current path. It is due to the small capacitance from the cathode to the filament.
That cathode to filament current does not go to the next stage, . . . but an equal and opposite phase current will go into the other phase's next stage.
And that is the Only un-balanced current driving the output stages (remember, I defined the operation to be without drawing grid current).

Yes?
If no, please explain.
 
Last edited:
Disabled Account
Joined 2009
I'm quite fond of the floating paraphase spliiter - with current "matching" and quality of tubes not what it used to be the floating paraphase splitetr can be adjusted to take into account the imperfect matching of the output tubes (and perhaps intermediate driver tubes). When I built my Audio note kit 4 clone I adjusted the phase splitter on its own "perfectly". Then when I adjusted it "through all the stages" and measured it at the output I halved the distortion as measured by my distortion facotor meter. Unfortunately the Audio Note kit 4 has too many stages resulting in too much phase shift.

I used a potententiometer and an AM tuning capacitor to work out the optimum feedback loop values.

I've got somewhere a 6BQ5 PP design that is skipping the capacitor between the driver and the phase splitter (12AT7) - helps with stability. I've grown fond of the 5670. I also recently became aware of the 6CW5 / EL86 and just wished I had know earlier about it. Because of the lower B+ requirement and the higher current the OPT can use less impedance resulting in an easier to construct output transformer. Also the OPT are easily to be found since the EL34 OPT's can be used.
 
2. You have Miller Effect Capacitance on the output tube control grids.
Ultra Linear screens move in opposition to their control grids.
This is one that most people forget. It is particularly troublesome when the driver impedance is high.

Very good point, 6A3sUMMER. I also learnt the hard way! This can be particularly bothersome with 43% UL taps, plus when large g! stoppers are used (some use as high as 33K), plus also as you say the feeding impedance is high. [That is why I never use ECC83 as drivers - but let us not start on that! ]
 
John Potgieter,

For UL amps, is it possible that the 33k control grid stopper is the dominant pole of the amp?
Not likely, but if the g1 to g2 capacitance is high enough, the Miller Effect might do that.

Sometime ago I did not like UL circuits at all.
But more recently I have been designing some UL amps, and liking them.
I still have lots to learn about details to make them as good as possible.
I designed and built one SE amp with a 50% UL tap.
And I designed and built a push pull amp with 40% taps. I phase invert with a triode pair, their cathodes tied together to a current sink to ground.

The UL SE and UL PP amps sound good, but I am still exploring ideas to modify them.
 
Last edited:
AM, that Audio Jan 1960 article by Kauder is interesting in its 1960 context. By that stage, voltage meters or even oscilloscopes with frequency response in to the 100's kHz were obviously more commonly available, and so allow tweaking of feedback circuitry for high frequency stability.

The proposed circuit has the basic structure of a Williamson amp, with the recommendation of aiming for about 20dB of feedback, but removed the driver stage to minimise feedback issues. Nice practical comments in the article, like plenty of listening power with a 6W amp, and having 15-20W available for headroom by using recently available more powerful audio output valves. The output stage anode feedback has been deployed by a few, and Kauder illustrates its use to control the high frequency peaking observable on the output transformer primary side.
 
I took a look at that Morgan Jones comment on the Split Load Phase Inverter a couple of years ago. After running some simulations I figured there was no point in going any farther, the HF roll offs were at quite a high frequency. So here are a couple of examples, one with a medium mu triode, the other high mu. The ccts are stripped down to a bare minimum, only the relevant parts are included in the simulations. These are done by Electronic Workbench software.
The 7247 was a solution to the 12AU7 not enough gain & the 12AX7 not enough drive. But now rare & expensive. A good alternative is the 6AN8, the pentode can run as a triode with reasonable gain. Somewhere in my notes I've got some test data that shews the 6U8 & 6AN8 pentodes have a triode MU of ~30.:)
 

Attachments

  • A.JPG
    A.JPG
    47.3 KB · Views: 301
  • B.JPG
    B.JPG
    46.4 KB · Views: 303
  • 7247 GE.pdf
    651 KB · Views: 104