Quad II ringing

[Radio Wrangler]

Originally Posted by PJL

I am sceptical about the slowly decaying high frequency resonance having anything to do with open loop gain, if it was as bad as shown in the earlier scope image, the slightest change would have it bursting into oscillation. My suspicion is the ringing is local to the output transformer, keep in mind that the KT66 anodes look relatively high impedance to the transformer.

I agree. There are two possible mechanisms which could be at work, one is instability around the whole amplifier loop including its feedback arrangement. The other is local oscillation in the power amp stage where impedances presented to the valves can by gyrated into negative resistance impedances at other electrodes. This can be a combination of frisky valves, lack of damping resistors (grid stoppers). This can be affected by speaker load impedance, funny speaker cables etc and also fits with the plots in stereophile.

I think that one of these Quad IIs is showing somewhat worse than normal behaviour, though the good examples of this design have issues.The other amplifier looks to have something definitely wrong. Adding stoppers is a good move, it can be reversed as needed.

David

[Gabe001]

Am I correct in thinking that both of you suspect the ringing is due to the OPT resonating (at 2 resonant frequencies) rather than phase shift issues introduced by feedback?

If so, isn't the cure a Zobel across the secondary?

[cathoderay57]

If the output transformer(s) are deemed to be responsible for this symptom then what physically is likely to have gone wrong with them? J

[cathoderay57]

Whilst we engineers like to have kit working perfectly, the fact remains that the HF oscillation is well above audio range and has no impact on audio quality. So, apart from potential damage to tweeters, are there any other impacts that I should be concerned about? I do have a spare Quad II 100v line output transformer that I could slave-in and check the effect on HF oscillation. However, I suspect that would be comparing apples with oranges so maybe not worth the effort? J

[Gabe001]

My thoughts are that it wouldn't matter if the feedback loop didn't include the OPT, but the high frequencies are being shunted back in circuit by the negative feedback loop, which I suppose is undesirable. I think your newly installed lag compensation network has shaped the HF response in such a way to attenuate these frequencies to some degree, but the problem should ideally be addressed at source.

You could perhaps localise the issue if you shunt the output from the power valves of the ringing quad into the OPT of the non ringing quad.

If it's the opt, it may be worth trying a 100nf cap +47ohm resistor in series across your output transformer secondary before doing anything drastic. Values may need adjustment.

[PJL]

Pretty much anything can resonate but I would try swapping valves first. Reality is the fast rise time of your square wave is not something that will occur under normal use.

[Richardgr]

First thing surely is to compare with a control. Maybe you are lucky with one rather than unlucky with the other.

[cathoderay57]

Hi Gabriel, thanks for the tip. This morning I rechecked all the connections to the OPT and I think they went where they are supposed to. I was confused by Keith Snook's notes against the transformer details. According to his circuit Y-Z primary feeds V3 anode and U-V serves V3 cathode. Y-X feeds V4 anode while V-W serves V4 cathode. Here is a copy of Keith's notes:

Primary V4 ~ 115Ω to 125Ω ~ X–Y = 100V {3.6H–4.5H}
Primary V3 ~ 170Ω to 190Ω ~ Y–Z = 100V {3.6H–4.5H}
Primary anode to anode X–Z {13.5H–14.5H}
Cathode V4 ~ 16.8Ω to 17.5Ω ~ U–V = 11V {90mH–110mH}
Cathode V3 ~ 16.5Ω to 18Ω ~ V~W = 11V {90mH–110mH}
I think V3/V4 in the cathode winding details have been transposed as a typo. Mine is correct per circuit. Primary winding dc resistances compare well with Keith's figures:
X-Y 119 Ohms
Y-Z 178 Ohms
U-V 17.5 Ohms
V-W 17.0 Ohms

However the secondaries are higher (Keith's figures in brackets):
Q-S 0.8 Ohms (0.55-0.61)
R-T same as above, paralleled windings for 8 Ohm speaker
P-Q 1.2 Ohms (0.55-0.61)
Allowing for the multimeter lead resistance (about 0.5 Ohms) these are maybe fairly close, considering Keith's measurements appear to have been made with the secondaries linked for 15 Ohms output. Jerry
P.S. I did swap the output valves with the better-performing Quad and it made no difference to either unit's HF oscillation levels.

[GrimJosef]

Originally Posted by Gabe001

If it's the opt, it may be worth trying a 100nf cap +47ohm resistor in series across your output transformer secondary before doing anything drastic.

Originally Posted by PJL

I would try swapping valves first.

Originally Posted by Richardgr

First thing surely is to compare with a control.

There are lots of interesting experimental possibilities. Since we're listing them I would add "Try disconnecting the negative feedback". That should tell you whether it's causing the problem or trying to suppress it.

Cheers,

GJ

[Radford Revival]

The shape and long decay constant of the ring definitely looks feedback induced to me. The Quad II appears to rely on the OPT and circuit strays to roll off the HF gain, which is a recipe for ringing, seeing you have at least 2 or 3 poles from the OPT and even the miller capacitance from the output stage, while small is driven from a very high impedance. Not having measured one before I don't know what their HF behaviour is supposed to look like but on consideration, if I'd just seen the circuit for the first time I'd be surprised if it didn't ring.

It does make a lot of sense that it is commonly found to rely on the stray-C from the coupling caps to keep it stable.

[cathoderay57]

Hello again. Well I took GJ's advice and disconnected neg feedback. Unsurprisingly a massive increase in gain. However, no evidence of ringing whatsoever. So what do you suggest I try next? If no other ideas I will reconnect neg feedback then try a Zobel filter across the speaker output as suggested by Gabriel.

[Radford Revival]

Can you estimate roughly what frequency the ringing was? It's possible a small RC across the 470 ohm feedback network may help, though this will either help, or push the oscillation further up in frequency!

[Gabe001]

Yes it's increasingly looking like a lead and lag compensation exercise

[Radio Wrangler]

The lack of ringing open loop is a dead give-away that the issue relates to the stability of the complete feedback loop.

The aim is to get the gain of the amplifier without feedback to roll off at higher frequencies, BUT without letting the phase lag of it build up to the point where negative feedback has rotated to become positive feedback.

Nyquist's stability criterion says that the gain around the loop (that is the amplifier forwards gain, times the feedback network transfer function) must have fallen below unity before the phase lag has reached 180 degrees.

What makes this difficult is that the things which roll off the amplifier gain inevitably introduce phase lag. And the things which you can use to apply compensating phase lead with act to roll-UP the amplifier gain with,increasing frequency.

Nyquist's stability criterion (Same Harry Nyquist that has another criterion named after him in sampling theory) doesn't tell the whole story. If you only just meet the criterion, you have a feedback system which is only just stable. Hit it and it will ring. Do a frequency sweep and you will see a damned great spike of a resonance.

So you want to pass the unity gain frequency with some phase margin in hand. Or you could view it as passing the 180 degree lag frequency with some gain margin inhand.

Doesn't look too easy?

The places in a traditional valve amplifier where the poles (roll-offs) live are places there the signal is at a high-ish impedance and there is stray (or deliberate) capacitance to ground. Think of the anode circuits of the input valve and of phase splitters. The Miller effect of stray/internal C from anode to grid of a valve magnifies this effect by local feedback. On top of it all, you have the effects of stray C, stray L and leakage inductance in the output transformer.

This is why good output transformers have complicated winding recipes, to minimise leakage inductance and to an extent use stray capacitance to oppose it a bit.

Pentodes are less susceptible to Miller effects, being essentially screened inside. Cascode circuits help too, which is where the Bailey modified long-tailed pair phase splitter that he developed for Radford comes in.

With two amplifiers, each showing different ringing. I'd have both open loop and very carefully plot the gain versus frequency and phase versus frequency of both. You should see a difference, and then you can go hunting the root cause of the difference. This is a scientific approach rather than just trying changing things to see what happens. The difference id that you can see what you're doing.

The other bit of bad news is that it looks like Quad IIs are a bit marginal on stability and tend to show some ringing. It looks like they never found a really good solution to getting as much feedback as they wanted (in order to hit distortion and flatness goals) and played it a bit close to the edge.

The people who designed these things were not gods, just ordinary human beings and had to find acceptable compromises.

The amplifiers for all their reputation, glamour, and resale values aren't perfect.

David

[Gabe001]

Thank you David for the detailed explanation. I have bookmarked this link in the past which explains how to do the compensation scientifically and how to plot the phase shifts using an oscilloscope in x-y mode. I must have read it 10 times!. It's hard, which is why most people advocate the trial and error approach, using a variable tuning cap and a linear potentiometer.

https://www.angelfire.com/electronic...pensation.html

I hope someone here finds it useful.

[John Caswell]

This is interesting as the Quad IIs I have seen that have been to Quad for repair have a 27pF silver Mica capacitor from both the EF86 anodes to ground.
I now do the same and cannot say I have come across this problem. There is often a slight overshoot on square wave but no ringing.

John

[cathoderay57]

Firstly, to answer Post #32, the oscillation kicks off with a couple of cycles at about 105 kHz then flips to 56 kHz then decays. Many thanks to all who have contributed and particularly for David's Post #34. I don't have a swept audio sig gen so an automated gain and phase v. freq plot using the scope isn't going to happen. I might be able to plot something with pencil and graph paper which will take some time. I have volunteering work tomorrow so I'll report back probably over the weekend by the time I have absorbed and understood what's involved. If nothing else, maybe we've exposed the Quad II warts and all and debunked some of the myth and folklore..... It's still a great amp though and I wouldn't be without one (or even better, 2). Cheers, Jerry

[cathoderay57]

Originally Posted by John Caswell

This is interesting as the Quad IIs I have seen that have been to Quad for repair have a 27pF silver Mica capacitor from both the EF86 anodes to ground.
I now do the same and cannot say I have come across this problem. There is often a slight overshoot on square wave but no ringing.

John

Hi John, yes if you travel back through this thread you'll see I tried 22pF and 30pF caps to ground across the KT66 grids which from an AC perspective is, I assume, identical to putting them across the EF86 driver anodes, excepting the inductance of the interconnecting wiring. It wasn't very effective on one of the amps in particular. Cheers, J

[Radio Wrangler]

Because a lot of the fun things come into play well above the audible frequency range, this isn't something that can be done with an ordinary soundcard and a bit of software.

I have an HP3580 which is a spectrum analyser with tracking generator which goes to 50kHz and this isn't always enough, and it doesn't measure phase. The HP 8903 audio analyser goes to 100kHz and will do stepped sweeps, but again, no phase. The Ideal tool could be the 3577 network analyser.

But usually this sort of work comes down to a broadband millivoltmeter and a scope for looking at phase.

Tedious is the word, but aone you have the broad picture plots, you don't have to keep doing the full things when you are looking for causes.

I ran into these woes in the 1970s, fixing various amplifiers where the problem seemed to be not a fault but a design inadequacy. So I went towards output-transformer-less designs and lower impedance, faster circuitry. Driving normal speakers meant massed parallel sets of output bottles to get the current. Push-pull totem pole architecture and even more output bottles to drive the other direction. Low-z style drivers diluted the effects of strays. The thing was a monster. I turned to transistors, recognising a fundamental limitation on what could be done with valves without things being immense.

David

[GrimJosef]

Harking back to Richardgr’s suggestion at post #27 and the OP’s at post #20 I’ve tried putting a 0.6Vp-p amplitude square wave through one of the Quad IIs I have here (serial no 66913). This one still has its original Erie carbon composition resistors (Erie carbon films for the feedback divider pair). But all of the small capacitors have been replaced with new ones. C4 is a 25uF electrolytic. C1-C3 are polypropylene films. C2 and C3 have 22pF disc ceramics from their KT66 grid sides down to the chassis.

The scope traces show the amp’s output into an 8ohm load (the output transformer is jumpered for this value).

The square wave output is 4.35Vp-p. The initial ringing (first peak to first trough) is 530mV and its frequency is ~108kHz. The component at this frequency decays in a couple of cycles. What follows are 2-3 cycles around ~50kHz and then a superposition whose fundamental component seems to be ~24kHz at perhaps a tenth of the amplitude of the initial ring.

Long story short: this appears to me to be similar to the OP’s ‘better-behaved’ amp. It suggests that this sort of ringing is a characteristic of Quad IIs rather than some exotic feature exhibited only by the OP’s pair.

Cheers,

GJ