Sony HP-511 - Blowing Fuses

[dave cox]

David, good point on 'current limit' nature of that Vbe.

I should have chosen a transistor with higher gain, at low current, rather that the one I chose on the basis of being within reach The ZTX450 gain is ~80 at a few mA.

I will find something better and give it a try ...

dc

[dave cox]

Restored the driver Q508 from my trial BD140 to original 2SA706 (tested OK on curve tracer) and now some return of oscillation. This occurs only at 'high volume control settings', somewhat independent of actual output power. For this test the input stage is purely passive through tone, loudness and volume. It seems oscillation is a function of source impedance and input level ?

For example, at 50mV input I can advance the volume control until I get 5V p-p output and oscillation starts and at 100mV input I can get to 10V p-p before any problems.

dc

[Radio Wrangler]

Originally Posted by dave cox

It seems oscillation is a function of source impedance and input level ?

That fits. It seems like there is slew-rate limiting going on. Something which is current limited is driving capacitance, so instead of getting a rolled-off amplitude of higher frequency sine components, the thing tries to slew as a straight-line ramp.

In terms of loop stability this means that there is a pole in the open-loop response whose frequency moves with larger signal amplitudes. The circuit design tries to use dominant-pole compensation and needs all other poles kept high in frequency so they don't cause any significant phase shift until the VAS miller capacitor has brought the loop gain down below unity. If any transistor is a bit slow, then the stability margin is reduced, and oscillation will break out first at times of high dv/dt.

A quicker device and a C across the Vbe multiplier may help the margin enough, but it seems this design was rather close to the edge... hey close to the edge was an LP contemporary to this set

David
.

[dave cox]

Thanks for the info David.

I should probably pop the Miller cap out to check it is not too far off spec, maybe try adding a few pF's as well.

dc

[dave cox]

I've made some more progress with this.
So far ...

(1) Q505 C / E is now shunted with a tiny 6u8 SMC ceramic cap between the legs. This greatly reduces the amplitude of oscillation but it doesn't stop it ...

(2) I found some BC546B's and used them to replace Q505 / Q506 which were originally 2SC633A / 2SC634A. The gain of these is quite a bit higher than what I had tried before and probably closer to the originals.

(3) I checked the miller cap, C552 which was spot on. However, as Q506 has no degeneration the miller 'multiplication factor' depends on the gain of Q506 (correct?) so I tried increasing C552 to compensate for somewhat reduced gain compared to the original (probably). I found increasing from 5p to 12p has improved overall loop stability without effecting frequency response.

(4) I refitted the BD140 (to replace Q508 - 2SA706) and also a matching BD139 (to replace Q507 - 2SC1124) which again gave measurable improvement. I suspect the 44H11 outputs are faster than the original drivers can cope with but the BD139/140 are typically faster still. Also added 10R stopper resistors very close to the bases of Q509 / 510 as the tracks are long and snaking - some improvement but this is probably just compensating for the faster transistors used

(5) There was still a problem when the volume control was advanced with no input signal, at about ~80% it would start oscillating at about 6MHz but eventually it would trip my PSU out. This really puzzled me, too high a frequency for the whole amp but it must be related to the impedance seen by the input transistor. I swapped Q504 (still original) for a BC556B and it made no difference but I did notice 2 extra holes in the PCB close to Q504 (between B/C). They are too close together for a resistor so I tried a 10p ceramic cap. Now rock solid to 100% volume. I figure this is outside the main amplifier loop and is not really a 'miller cap', since the collector will only be moving a tiny amount of voltage. At 6MHz this has ~3K impedance so is shunting to ground via the 10K. Lucky I noticed the holes as I was not thinking of trying a cap there!

I now have the amp operating into an 8R dummy load at appreciable power, no weird effects from the position of the volume control, similar levels of distortion for the 2 channels (@ 1KHz)and only 1 remaining gremlin ...

dc

[dave cox]

A final gremlin.
Grrrrrrrrrrrrrr

This seems to be purely related to out voltage swing / load when with judicious adjustment it will break into oscillation on the negative going swing (A little hoot.jpg). Measured at R567 it is really going like the clappers! (More than 10MHz.jpg). I tried to track it down with my homemade H field probe (Spectra.jpg) but it seems to be coming from everywhere. I need to make a better probe!

dc

[Diabolical Artificer]

Nice work Dave, I'm glad to see your getting somewhere with this amplifier. It's doesn't want to come to heel does it? : ) Any idea why it blew in the first place?

A.

[Radio Wrangler]

That sounds like the quasi-pnp pair is hooting. Sony appear to have had this problem too. R661 in the PNP driver base looks like an added stopper. Maybe a few more Ohms might help?

David

[PJL]

Somewhere way back in this thread there was a posting in support of transistor amplifiers with a comment that poor designs had given them a bad reputation. Are we ready to retract that posting yet? At least with a valve design you don't need a broadcast license.

[dave cox]

Any idea why it blew in the first place?

Andy, one of the characteristics I see with the 'full-on transmit mode' is that the current pull leaps up (probably both output devices conducting at the same time) and I guess the only thing stopping another self-destruct is a current limited PSU. That is a guess as I have different transistors now but it sure seems prone to taking off at multiple MHz!

I think trying to get 'original' transistors was doomed to expensive failure as nobody has made anything like that (the outputs, at least) for ~ 30 years and will be territory for the fakers. I choose something that is still in production (Fairchild 44H11, but similar to TIP42C) but inevitably it will have quite different characteristics.

Thanks David, will try increasing the stopper …

dc

[Radio Wrangler]

There are plenty of pieces of equipment out there where you can no longer get devices poor enough for them to work properly. Device specs at the time were open ended about just how fast that type could go. Designers relied a bit too much on it.

If you go back to valve era equipment you have damageable and irreplaceable output transformers, unaffordable PX4s etc, and silly prices for period-correct NOS valves. There are difficulties wherever your interests take you.

Back when that Sony was made, we'd got away from thermal runaway of germanium and we thought transistor stuff would last forever... Ah......

David

[Diabolical Artificer]

So the problem here is the contemporary trannies are faster and prone to self oscillate. Still learning about trannies, I'd never really considered how fast a Q can switch as being a spec worth considering in an audio amplifier where they're supposed to operating in the linear region, apart from the class AB or B OP stage that is. So what threshold are we talking about here, how much faster does a new Q have to be compared to the old Q to cause a problem in certain amp designs? Was this due to manufacturing processes at the time or less than truthfull specs?

Sorry for all the question's, I have loads. If we have a Q wanted to be a bird instead of a linear amplifier is it Miller capacitance that causes it to self oscillate or local FB? And how do you stop them doing this?

A.

[Radio Wrangler]

A transistor has one spec called 'Ft' the transition frequency. The frequency at which its current gain has fallen to unity. You can use a transistor to make an intentional oscillator at any frequency up to this limit (and with care to a bit above it due to special circumstances) So it goes without saying that yoou could unintentionally make an oscillator up to these frequencies.

Go look at the current gain specs (hfe) and the transition frequency specs (Ft) for lots of period devices and you'll find specified minima, which tells you that you can be certain to be able to use a device when you need certain frequencies and gains. You'll find a lack of specified maxima. The devices you buy will beat their specs, but by how much?

Over time, semiconductor processing has got more precise. Thicknesses of base regions are better controlled, junctions are, well , more crisply defined. Hfe's have gone up, Ft's have gone up. How far? Who knows?

Imagine if classic valve types were still in continued production after all these years, but the gm values had increased by a factor of ten or more and the frequencies familiar audio types could worked at had gone into the high UHF region?

Imagine the trouble.

Motorola made an oscillator chip as part of their ECL logic family, MC1648. It got used in all sorts of things, but not domestic stuff because of power and price. Over the years they 'improved' their processing and dear old MC1648 got faster and faster. In original circuits instead of oscillating at the frequency set by the tank circuit you connected to it, it found the resonances of the tracks on your PCB a much more attractive proposition. Hell broke out!

In a transistor audio amplifier there are three sorts of instability problem:

Firstly, transistor amplifiers need feedback loops to control their overall behaviour, to control DC conditions, to set the gain they provide, and to reduce impairments. This loop needs careful design, and that takes knowledge of some appropriate areas of maths. To keep the designing process manageable, you put in some time constants to govern the response, and you make sure that all the stray timeconstants are pushed up to high enough frequencies that the planned ones have the loop gain dead and buried before the strays can interfere by contributing phase shifts. Their are standard recipies for doing this, so it's fairly easy to do, but you have to have gone through it the hard way once in your life to appreciate what it all means.

Secondly, in some amp designs there are local feedback loops. You might not notice them until you know what you're looking for. In this Sony example there is a quasi-PNP power transistor made from a PNP driver and an NPN power transistor. They are in a little loop. There are no deliberate time constant components to control the loop's response. It relies on the relative speeds of the two devices, The gains and Ft's of the devices change with the current they are carrying. In this case, there is a burst of oscillation when this pair are operating and carrying a fair bit of current. There are no R's and C's to fiddle with. Adding ones can create further troubles and still not solve the original problem. If you think this sounds like a problem, the 'Quad Triplets' in their classic 303 power amplifier are much worse.

Thirdly, you can get an individual transistor taking off at RF frequencies entirely on its own. Internal base-collector, base-emittter, and collector-emitter capacitances are enough with the stray inductance of PCB connections (god help you with devices connected through wiring!). Resistors in the base lead or ferrite beads are the tools for preventing this. Analyse the circuit and put in all the strays and you can see classic oscillator topologies.Look at the circuit diagram and you see no feedback loops. But they are there. Having parasitic stoppers is part of life for RF designers. It's just that now you really need an RF designer to design your audio amp.

Can these problems be handled - yes, though it's easier in fresh designs than to try to add RF savvy to an existing thing.

Can people learn how to handle these issues? yes, those who now can once learned.

Do people know that they have to handle these things? Yes and no. Those coming from the RF direction are familiar with these types of problems, it's slowly dawning on the rest.

RF design is no longer a specialisation. Faster devices mean that all design has become RFy

You can't just use audio-savvy design because you also need to make sure that your thing doesn't mis-behave at all other frequencies.

A good design must not only do all the things it is supposed to do, it must also NOT do a huge catalogue of things it is not supposed to do.

Don't think you can retreat into the world of valves to escape this. There is a thread going on right now about a horrid mish-mash of multiple random feedback paths in a Pye Black Box. The shortage of NOS bottles that audiophiles would love to cuddle has led them to RF variants like the TT21 and 807. Not only do these things have lethal voltages on top caps (Darwin award for audio?) but they also need parasitic stopping quite carefully. These people have moved into a world of much friskier beasties. I keep trying to sell audiophiles on the 3-500Z not only is it a triode (and indirectly heated to boot) but anything which looks like that HAS to sound amazing. evin the matching top cap connectors are finned! 2kV might also thin their ranks a bit....

David

[dave cox]

Just to prove the point, increasing R561 from 100R to 200R did the trick - no more hooting - Yeaaaaaa !

The original was holding only about 20mV (200uA) under quiesent conditions, so I figured doubling to 200R wasn't going to change things a lot.


dc

[dave cox]

All done on the power amplifier now.

This will now power up on the normal mains supply. Gareth came round and has taken the chassis back to finish his restoration work. Fingers crossed it will be a relatively straight forward re-assembly job as this unit has caused enough trouble for several pieces of equipment and will now repay by providing many more years of listening pleasure

Think I can now hand my back for smoking a set of transistor in a public forum, but thanks should go to David (Radio Wrangler) for salient advice and super explanations.

dc

[RojDW48]

Well done Dave et al - something of an epic!!!

[gazrawly]

Well, I've had the set running for the last couple of hours and it hasn't turned itself into a fireball! Sounds great, spun a few vinyl's on it. Awesome!! I think i'll be upgrading the speakers pretty soon but the set is working very nicely.

Massive thank you to David, Roger, and everyone else that has helped out on this thread... and of course, big thank you to Dave for finally getting the set to power off the mains (I still have flash-backs every time I turn the set on).

Dave had to modify the amp to stop oscillation from the modern Q's... Dave if you could explain it as I was lost as soon as I looked at the schematic to tell you the truth! Haha

Great forum, great people, great outcome!

Gareth

[Diabolical Artificer]

Ah, we got there in the end, nice to see a happy ending to a thread instead of it petering out and going silent.

Nice one Cyrill, I mean Dave. : )

Andy.

[Station X]

Seconded. It's nice to see a repair thread with so many posts reach a successful conclusion.

Well done to all who contributed to the thread.