Cause of ultrasonic oscillation in a push-pull output stage ?

[Radio Wrangler]

That box looks to be about the usual sensitivity of feedback amplifiers to high frequency capacitive loading. G-J was having trouble with resistive loads close by, so it was fairly definitely not the usual issue raising its head.

David

[PJL]

Assuming it was not unstable when new, either the transformer has changed from age/abuse or the design relied on particular characteristics of the original components. Electrolytics spring to mind particularly the cathode decoupling capacitor but valves are another possibility.

[GrimJosef]

A range of valves was tried and most of the testing was done with old-stock, little-used vintage Mullards and GECs. The problem seemed essentially independent of valve type.

I did try adding a bit of resistance to the odd electrolytic, but that made no detectable difference.

Swapping the amp's original output transformer for one from another TL/50 Plus fixed the problem. There is some indication that the original transformer might have been over-run (GZ34 was replaced with solid-state rectifiers) and it did measure differently in terms of its HF phase shift from the well-behaved transformer in the other amp and also from the transformer in Terry's (Valvepower's) well-behaved amp. So my money's on a transformer issue. I do still find that result surprising though.

Cheers,

GJ

[Radio Wrangler]

The lingering mystery is just what is wrong with that transformer.

One possibility is that something has changed. Perhaps the creation of partially-shorted turns due to insulation degradation by time or over-voltage (solid state rectifier or running off load). Perhaps a change in the core due to moisture affecting the interfaces between laminations. Perhaps moisture affecting inter-winding capacitances? There are many possibilities.

One other possibility is that that transformer was badly made in the first place... maybe the wrong winding stack-up, maybe a dropped turn affecting coupling, maybe a length of wire with dodgy varnish. There could have been a new operator in the coil winding department.

Another possibility is that the original transformer failed or was damaged and this one is a re-wind by someone who got the recipe wrong.

Another possibility is that there is nothing really 'wrong' with that transformer and that it is undamaged, built the right way and is just an outlier on the statistical distribution of parameters.

Measurement of individual winding inductances and Q's would be interesting. mutual inductances, too.

There is plenty of evidence that ultra-linear stages are a bit critical on stability in general, and some families of valves can be very touchy about impedances presented to their screens in any circuit. The 4CX250B for example is notorious at RF, and even for lower frequency use needs a spacial base with built-in ceramic decouplers for the screen. It even has a characteristic where under some conditions it presents current from its screen connection and screen supplies need shunt regulation! This may well be the worst problem child of all the tetrodes.

Things get harder at RF, but audio doesn't get a free ride.

Valves are not easier than transistors, they both have some subtle issues. The circuits might look simpler on paper, but circuit diagrams don't show strays or component selection.

There's more to simplicity than component count.

This is a really good mystery. Agatha Christie, eat your heart out.

David

[Valvepower]

Hi,

I know this isn’t that scientific, but to some degree it fits the cooked output transformer syndrome we encountered back in the 70’s. I was still a rookie cutting my teeth and I was yet to grasp the concept of leakage inductances and feedback in the screen grid circuit and such like, but you could hear something “discordant” when playing a guitar through the amplifier.

I know from a previous experience with another Haddon transformer they can miss something, but in their defence the volume of transformers going through production would have been enormous in the late 50’s.

There is also the concept it was cheaper to reject a “marginal” transformer and just fit another – given the low quantities this amplifier was made, it could have been a case of “Dick, we’ve got a marginal one here, and Smithy replies, yeah, just swap out the output transformer and that’ll sort the problem and that'll keep production going"

I know to fulfil orders Leak in the latter days of the TL50 fitted annular ring “TV dropper sections” from RS in place of the Standee cathode resistor to fulfil orders.

Maybe it was combination of marginal design, production spread, and a cooking and it just tilted (for want of a better phrase) the transformer parameters and it started to hoot.

It would be interesting to know the ratio of 100V line, standard 4/8/16ohm and modulator (and such like) units.

Anyway, this is an insight into the world of commercial mass production of amplifiers, be it valve or transistor, or whatever, as you never know when the unit being made will bite you on the ass in production Every engineer dreads the call half way through the week “we’ve got a marginal batch in test and we’re having problems getting them tested and to make it worse the order has to go out by the end of the week”

Terry