What burned out these BJT's in a Sansui stereo ?

[MathWizard]

I have this 1980's Sansui R-7 2-CH stereo. I was told it had an intermittent audio distortion type problem, I should have listened to it for a while, in it's initial state. But I took it apart, mapped most of the circuits, checked most of the electrolytic caps for ESR and DF, and reflowed some joints and used new thermal grease on the heatsink. All the resistors are perfect.

With an auto-transformer, isolation transformer and 100W bulb, powering the stereo's transformer and audio section, I have had some large HF oscillations of the main push-pull outputs, maybe not rail not rail, but near it I think. And sometimes the main rails have seemed slow to raise up, and so have the +/-19V rails from zener's. Now some of that is from using the 100W series bulb, a few times it's been lighting up on startup, and I've been keeping the mains voltage mainly from 70-100V. Usually upping the voltage some would get rid of the oscillations, and get the zener's fully on, for the op-amp's rails, and the 100W bulb goes out.

The 1st time I had an MP3 player hooked up to it, the PSU seemed to stabilize, but after about 10-15min, both channels got all distorted. But I might have had the voltage too low.

Since then besides some startup oscillations (IDK how long they could last on their own), it seemed to work fine from 70-100VAC. The channels seemed fairly equal. I can see some oscillations around 45kHz that follows the 60Hz ripples, the input is pretty sensitive and none of this is shielded, and I have 10x probes right on the L/R push-pull outputs, and on the main split rails, which are up near or over +/-40V (when full mains AC in). The main 2 Nichicon filter caps are only rated 50WV.

But overall it sounded great for 1-2 days, and the output filters block most this HF stuff, even on startup when it's osc. I never heard anything. I'm only using 10 or 15W speakers, 4 or 8ohms.


Then just now, WITHOUT the 100W bulb, and the speakers turned off, and the input cable floating in the air, picking up a ton of 60Hz,  I start it up at 60-70VAC, and it did osc. a bit, and I upped the voltage some. I had the scope rolling, maybe 1-2s/DIV and the main +/-V rails seemed to have reached the correct DC level, but there was a bunch of HF stuff on the split HV rails, over lapping the 60Hz. The outputs looked like they had +/-10V osc. on them. IDK if it was just my scope making it look bad while rolling, but after zooming in it seemed to go away.

After 1-2 minutes I turned up the auto-transformer to full 118V. And I was being paranoid, being afraid the caps were going to blow up. And a couple of minutes later when I had my back turned to it, I swear I saw a flash of light, but thought it was my monitor. Then a minute later I notice a bit of cooking smell and see no voltages on the scope. And indeed the mains input fuse, 250V 3.5A, burned out, and the big heatsink, for the 2 push-pull pairs was pretty hot.

After checking those main P_P pairs, the RIGHT-CH complementary pair have both burned short their C-E junctions, and partially shorted the base too. I'm guessing they were oscillating, or the full voltage rails got them. All the other BJT's near by seem ok still.

Or maybe I cooked them reflowing their joints, while off the heatsink, or maybe I used too much good quality CPU TIM. But the only input was 60Hz hum on the cable.

I can see the op-amp barely has any supply filtering, and I'm guessing it and the other diff-amp, and the output loop, and with scope probes, and maybe old zener's or old rectifier diodes, maybe that's a recipe for osc. ??



I want to add an NTC before each main bulk cap. What do you folks think of this circuit, any suggestions on what to change or to add, like small filter caps on the op-amp ?

Here's pretty much the audio section schematic, and 2 working models with mostly correct BJT's. They osc. too, but so far I haven't tried changing caps much. And the Op-Amp is a 4558, and the diff-amp's are some 5-pin dual BJT's.

Here's a working version with just BC547B/557B, and there's 2, 18V zener's for the op-amp rails, I think everything is in default parts list. It's not much different, idling anyways.

[David_AVD]

The 2SA798 5 pin dual transistors can be replaced with matched gain pairs of KSA992. These go bad all the time and I replace them (with 2x KSA992) even if they are not currently playing up. The usual failure mode is crackling (constant or intermittent), but occasionally I'll see one where one transistor in the package has changed gain massively leading to a large DC offset on the output.

[MathWizard]

Yeah and I see people try stuff like strap/glue BJT's together. Some day soon I'll have to try that, with some pair I match up. So far I've been working out some of the DC biasing, and started on some of the AC analysis. But I haven't really probed anything in detail, besides the voltage rails and outputs, or compared all the L/R voltages. I know the resistors I could read in circuit, measured great, like they were 1% tolerance, but I think they are 5%.

For now, I hooked up a smaller transformer, to lower the voltages. And I soldered in 2 pairs of TIP31C/32C. They should be fine for 2x 15W speakers.

But I'm running into a heatsink problem. The heatsink is supposed to be grounded, and they use Mica insulators, and the 4 main BJT's are TO-3P style with basically a plastic washer built into the package, like the pic below.

The TIP31/32's are TO-220, and I have new Mica insulators and plastic washers on them, and good quality (at least new) CPU thermal grease.

I tried that new transformer, and I didn't have the heastink DC-grounded, and started to over-volt a 220uF 6.3V cap that goes between chassis GND, and the DC GND (they aren't meant to be physically connected on this PCB, it's done on another PCB, even tho the traces/pads are right next to each other).

But what's happening is, as soon as I tighten down the screws through the BJT's, even with the Mica and washers made for TO-220, the + rail get almost shorted, and the negative rail is pretty much shorted. And 2-3 of the screws are doing it.

On ohms with a 5.5D meter, I can read past 100Meg and OL, but then when I tighten the screw down the last little bit, it will start reading 10Meg, then kohms, and then near short or under 1 ohm.

So thankfully I was using the auto-transformer and series lightbulb when I was testing it, but the bulb wasn't lighting up, and I only gave it 75% mains AC.


I haven't worked on that much stuff with heatsinks/THH parts, so this is a new problem to me. The TO-220/washer/Mica look like they should fit, but apparently there's metal on metal contact.


But the original TO-3P's 's couldn't have had that issue. New suitable replacement from Digikey are expensive tho, and I don't need much power, so I'll try another heatsink and smaller screws, or 1 for each set of NPN's /PNP's.

[MathWizard]

I have the whole thing put back together and working with a few Vrms smaller of a transformer, and besides some startup oscillations sometimes still, it should be ok.

But the AM and FM radio seems very quiet. With an MP3 player hooked up, 1/3 volume is plenty. But on radio, I have to have near full volume. But all the signals seem to lock on good, with all the signal strength LED's lighting up as the station is centered, and the FM stereo LED lights up as it should. All the PCB GND's are hooked up. There are 2-3 external antennae connections, but there's strong stations close by.

The stereo has an internal ferrite rod antenna, with a nice finned variable capacitor. Every alarm clock or stereo I've ever owned, with a little speaker, would be louder than this thing's radio powering 2x15W speakers.

There's 3 main RF IC's on this thing, so far I can only find an English datasheet for 1 of them. There's a lot that could go wrong with a radio, and these chips, that I don't know about.

Assuming there really is a problem, what should I be looking for ? It's like some audio pre-amp in one of the chips is not working, affecting AM and FM. I've checked most of the electro caps on the RF board, and the chips get Vcc. I'm just making a schematic of it all, but these chips are pretty complicated.

Any ideas on what to check  ? (there's hardly any BJT's to check, maybe some solder joints could be reflowed) Or anyone know where the English datasheets are, or similar chips ?

[David_AVD]

If it's just the audio level from the tuner that's low, check the coupling capacitors at the audio output of the tuner section.

[MathWizard]

Ok I found a bad solder connection after the RF section, between some switches for the right channel. The other channel was down some from the 100k potentiometers and 10nF caps to gnd I used to feed in an MP3 player. Other than that, the L/R audio output signals look equally strong from the uPC1161, hundreds of mVrms on the chip side, and only 60mVrms on the other side when that 10nF cap is directly connected to the signal..

So I'll just put my MP3 circuits on the aux switch where they should be anyways.

I added some NTC's in series with the main bulk filter caps, just hoping to make them last a bit longer. They run cool, at the audio levels I use, so 4-5ohms. I might swap in some lower voltage zener's for the op-amp rails too. They seem to be 18-19V rated, but with this new transformer, they settle around 16V.

I have to read up more for the startup oscillation problem, it's probably not that uncommon.