HF oscillation with no load, but only on one channel.

[bigwill]

Hi there,

Just put together an amplifier, which is working great, giving a healthy 37W at 50Hz into 10 ohms (8 ohm tap) before clipping.

The circuit built is a Radford STA25, built exactly down to the component (with the correct transformers).

However, there is an issue: One channel, as soon as the load is disconnected, instantly oscillates at some ultrasonic frequency. Obviously this is not something I want to keep triggering for the sake of the output stage, but the other channel is dead stable, and this circuit/transformer combination simply shouldn't do this - Radfords are stable with no load.

As far as I can tell, both channels are identical.

The only omission is I've left the heaters floating, but this is simply due to the hurriedness of the construction (though the signal wiring is fairly neat). Grounding the heater winding of the unstable channel does not cure it, unfortunately.

I've applied a signal generator to the primary of both OPTs each in series with its own 1k resistor (75 ohm src) to roughly simulate an imperfect voltage source, and simultaneously scoping the secondaries of each. They both track fairly well in amplitude and very well in phase, diverging slightly above 10khz but until then there is no discernible difference. With both traces overlaid in dual channel mode, it simply looks like one.

I think/HOPE the output transformers aren't the culprit. I have yet to swap them to see if the problem follows them due to the hassle.

As far as I can tell, the primary leads to the OPT don't get excessively close to any sensitive signal circuitry wiring - but they do pass under the PCB where the components reside.

The input stage on the faulty channel begins to roll off at about 10Khz by design, which I have confirmed by measurement, so the driver circuitry doesn't have excessive bandwidth.

All in all I'm quite stumped!

If anyone has any insight into this I would be very appreciative

[DF96]

Running with no load connected is a good way to ruin an OPT, so be careful. HF stability is not just determined by the circuit and the components but also by the exact layout of everything. The 'well-behaved' channel might still have an HF peak, but you won't see it unless you plot a wide frequency response. But you should not do this, unless you can afford to buy a new set of OPTs.

The solution is simple: never disconnect the load. If you can't guarantee this, add a Zobel network.

[artosalo]

As far as I can tell, both channels are identical.

According to the circuit diagram I found a lucky find, the NFB-circuitry is quite complicated. You could first check that the amount of NFB is same at both sides.
Secondly you could reduce it few dB and see if it helps.

[bigwill]

Originally Posted by DF96

Running with no load connected is a good way to ruin an OPT, so be careful. HF stability is not just determined by the circuit and the components but also by the exact layout of everything. The 'well-behaved' channel might still have an HF peak, but you won't see it unless you plot a wide frequency response. But you should not do this, unless you can afford to buy a new set of OPTs.

The solution is simple: never disconnect the load. If you can't guarantee this, add a Zobel network.

Hi DF96,

The fault was only discovered by chance when my dad changed some speakers over with the amp still on (though without signal). A loud pop resulted when the speakers were connected, along with some cone movement as the circuit re-established DC balance. If it was me I probably would never have discovered it!

For various reasons I can't guarantee a load - I wont necessarily be the only one using it. What happens under heavy signal conditions (with no load) is unavoidable though.

I will look into a zobel. Can you recommend any values? 100nF + 10 ohms sounds about usual...

[bigwill]

Originally Posted by artosalo

According to the circuit diagram I found a lucky find, the NFB-circuitry is quite complicated. You could first check that the amount of NFB is same at both sides.
Secondly you could reduce it few dB and see if it helps.

Ideally I'd like to not mess with the circuit too much - this particular one usually work just fine, although to be fair I really have never been in the habit of running valve amps without a load - this was discovered by accident!

[bigwill]

When I say I can't guarantee a load, I mean I can't guarantee the user making good speaker connections. The user knows not to actively run an amp without a load, but it would be nice if the amp didn't destroy its output transformers in the event of a loose connection

[bigwill]

Originally Posted by DF96

HF stability is not just determined by the circuit and the components but also by the exact layout of everything.

To be fair, the layout is the same as every other STA25, which is a bit disconcerting. I may have just stumbled across an unfortunate combination of factors though...

[DF96]

Sudden HF oscillation with no load could be caused by parasitics in the output stage, not instability in the feedback loop. You might find that there are bursts of oscillation at peak signal too. Have you had a look with a scope?

Anyway, a Zobel on the output may help. Alternatively, snubbers on the OPT primary taps (g2-anode).

[bigwill]

Originally Posted by DF96

Sudden HF oscillation with no load could be caused by parasitics in the output stage, not instability in the feedback loop. You might find that there are bursts of oscillation at peak signal too. Have you had a look with a scope?

Anyway, a Zobel on the output may help. Alternatively, snubbers on the OPT primary taps (g2-anode).

The signal looked very healthy into a (mostly) resistive wirewound dummy load - very clean clipping and no sign of HF bursts.

I have 1K screen resistors directly on the output valve sockets, though the screen stoppers (2.2k) are located on the PCBs, which always worried me slightly.

[bigwill]

It's probably prudent to point out that these are transformers I've had made and not original Radford ones... (though they have been wound to the exact specification)