What intermittent fault might be burning these power supply resistors ?

[GrimJosef]

I have a stereo amp in for service. The design is simple. Each channel has a small-signal gain stage using a 6C45P triode followed by a single 300B triode as the output stage. There are three HV supplies - an HT one for the 300Bs, an HT one for the 6C45Ps and a third one (-100V) for biasing the 300B control grids.

I first saw this amp (it belonged to a friend) 6 years ago when I traced the circuit and took some voltage and audio measurements. My friend has since sold it on and its current owner has now brought it back saying it's developed a hum and is now running physically hot. In particular the smoothing choke in the 300B HT supply becomes too hot to touch and the GZ37 rectifier there is red-plating.

Attached is a circuit of part of the small-signal HT supply and all of the 300B bias supply. I suspect the amp may be some kind of prototype. The chassis and wiring are rather untidy and circuit modifications have clearly been made. For example, the EZ80 in the bias supply has been bypassed by a BY127. Some components are also being over-run e.g. the reservoir capacitor in the bias supply, which is rated at 450V but is running at 560V !

The bias supply had in fact failed, which would account for the over-running of the components in the 300B HT supply. Specifically the two resistors circled in red were open-circuit. When I first saw this amp, 6 years ago, these resistors were 'cement-block' ones (perhaps 3W ?) and showed clear signs of overheating (labelling all but burnt away, values drifted high). I replaced them with a pair of Welwyn W22 vitreous enamelled ones (nominal 7W). The second picture shows them now. One, at least, has run so hot that the vitreous enamel on it has melted and run into the gap between them, 'glueing' them together ! This must have happened before their open-circuit failure.

I've modelled the bias supply using PSU Designer II by Duncans Amps and it shows, once things have reached the steady-state, the pair of resistors passing a train of 50Hz current pulses with a foot-to-foot duration of ~3.2ms and a peak current of 92mA (so a peak voltage of 77V or so). Even if those were square pulses the average power dissipated in the resistors would only be 1.1W or so. In practice their truncated sine nature means the average power should be well under a watt. I have fitted replacements and sure enough they are running cool to the touch. All of the other bias supply components have survived whatever it was that blew the resistors (despite some of them being routinely pushed beyond their ratings), and the fault must have persisted for long enough to melt the vitreous glaze.

For completeness I should add that the HT winding on the mains transformer delivers 455V-0V-455V off load and each half of it has a resistance of 52 ohms.

I confess I can't think of a recoverable fault that could have caused the resistors to overheat. But I have seen it happen twice now. I would be really grateful for any thoughts ...

Cheers,

GJ

[Robert Gribnau]

Maybe a short to ground of the heater winding of the lower EZ80 (this possibility depends a lot on the wattage of the 1K8 resistor between the heater winding and the cathode of the lower EZ80)?

Or a combination of such a short and the failing of the heater to cathode insulation of the lower EZ80?

[frsimen]

The BY127 is running very close to its maximum repetitive peak reverse voltage in this circuit.

With 560V on the anode and the cathode possibly seeing positive voltages in excess of 643V, that doesn't leave much margin for a 1250V max diode. Could it be breaking down intermittently if the mains voltage goes higher than expected? That would lead to the loss of the bias supply, very hot resistors etc.

Paula

[Diabolical Artificer]

Bias pot wiper lifting? That's an odd way of configuring a bias supply, the pot is usually wired as a potential divider with a pull up resistor to stop bias going to 0v or more usually minus whatever. The bypass C's usually go to ground too, to couple AC to ground, why wire them like that? Lastly on my bias supply's I put two 1m resistors from wiper to each end of the pot to prevent a no bias situation.

Then and again it could have been an instability issue, the amp reacting to a change in speakers or some other issue.

Andy.

[G.Castle]

Intermittent short in the right hand electrolytic in the bias circuit.
I'd guess that the BY127 was included as an afterthought to try to make sure the bias was present before the HT.

Greg

[PJC58-Hythe]

Breakdown in transformer insulation between windings or between windings and the core might be the cause. I once saw this in a jukebox amp where it had burned out the hum balance pot and once that was open circuit the amp continued to work albeit with the heaters at a very high potential!

I agree with Andy that is a very odd bias arrangement with no provision for wiper lifting or one of the pots going open circuit.

[PJL]

As well as the issue with the pot arrangement, just how much current does a 300B grid require? In the interest of the climate, how about redesigning to cut back on the wasted heat.

I would also agree that breakdown in the transformer, the unused EZ80 socket, or the electrolytics is the likely cause. Is the EZ80 socket still in circuit?

The EZ80 must have been reduced to molten glass in that circuit.

[ms660]

Regarding the original bias supply rectifier valve, I might be wrong but according to a calculation I just did the total circuit resistance needed to keep the transient hot switching current of the EZ80 bias supply rectifier equal to or lower than the maximum allowable for that valve in that circuit is approx. 656 Ohms or higher, subtracting the secondary resistance of the transformer from that means that the combined rectifier peak transient diode resistance plus the surge limiter resistance needs to be around 600 Ohms or greater.

The combined value of the surge limiter resistors fitted and the peak transient diode resistance would meet that requirement, but the max PIV for the EZ80 is 980 Volts which is much lower than a possible circuit fault PIV of 1,286 Volts.

Max. hot switching transient current for the EZ80 is given as 900mA in the valve data I'm looking at.

Lawrence.

[GrimJosef]

Thanks for the replies everyone.

Yes the EZ80 socket and, indeed, the EZ80 are both still in circuit. But under no-fault circumstances the BY127, with only a volt or so dropped in the forward direction, bypasses it. It doesn't protect the EZ80 from reverse breakdown of course but there seems to have been no gross failure of the EZ80 (shorts in either direction, bits of metal or mica rattling around inside the envelope, loss of vacuum etc).

I did worry about reverse breakdown of the BY127 but the same device has been in place for at least 6 years and, most of the time at least, still works fine. Then again if it could survive a few hundred milliamps of reverse current that would be sufficient to destroy the resistors.

Assuming the huge Al-clad 50k resistor between the two 47uF capacitors is OK then it's hard to see how anything downstream of that could be causing the 830 ohms (1k2 in parallel with 2k7) to burn out. The same average current would have to be flowing through them both and the 50k would limit it to ~10mA.

Breakdown of the EZ80's heater winding insulation inside the mains transformer would explain the symptoms. My limited experience of transformer insulation breakdown is that once it fails it tends to stay failed - the carbonised paper and plastic don't recover their insulating properties. But I suppose it could be arc breakdown across an air gap.

Since the BY127 is doing all the work I'm minded to replace it with something with a higher reverse breakdown voltage - I think I've got some GP10Y-E3/54s around somewhere, or even a couple of UF4007s in series - and then disconnect the high voltage from the EZ80 altogether. It'll still be left glowing orange above deck but it won't be threatening the bias voltage's integrity.

Again my experience of breakdown in electrolytics is that it tends not to be reversible. But simply as a matter of good practice I'm inclined to replace the first 47uF with a couple of 100uF 350V units in series, balanced with a pair of resistors.

As with all infrequent but persistent faults it can take literally years to discover whether or not you've fixed them. But that doesn't mean you shouldn't try.

Thanks again.

Cheers,

GJ

[Maarten]

I think replacing the diode and the electrolytic is a smart move. Electrolytics can be self healing (self forming) to some extent. It has been used well above its rated WV for years, so I imagine it getting tired at some point in time.

[paulsherwin]

Using two diodes in series will reduce the voltage stress.

[Radio Wrangler]

With two (or more) diodes in series, you may need to add resistors and capacitors across them to balance the reverse voltage. The resistors balance the DC component, the capacitors balance the effects of mains transients and spikes.

Otherwise whichever diode is leakier, slower or more capacitive applies most of the stress to the other and you can get a chain of successive failures until the whole string is gone.

You CAN run diodes in series without adding components if they are deliberate controlled-breakdown types which can survive by balancing the stress themselves.

David

[PJL]

The EZ80 will have 560+643 reverse voltage which is well beyond the design specification based on 350VAC and not the 455VAC here. I wouldn't trust anything other than a ceramic valve holder at these voltages either. I suggest completely disconnecting the EZ80 anode and cathode connections. Removing the cathode connection will reduce the stress on the heater winding too.