Game Instance

With the current drive solved and a headphones relay adaptation, I was now ready to make a first listening on my new/old Technics SU-VX920 but as I was gradually cranking-up the volume, I noticed something off. The left channel had a distortion-like noise that didn't come from the potentiometers and wouldn't go after re-adjustments. It suddenly became concernedly clear that this was the reason why this amplifier was serviced in the first place. Whoever checked it left visible traces of interest on nearly all the components on the signal path of that channel and, while handling the unit boards, inadvertently knocked-off the grounding which led to its demise.
Ladies and gentlemen, grab your armrest for this will be a high-g experience.

Warning:
Do not do this at all if you don't know what you're doing!
This amplifier is more than capable of harming you ... really bad.

The noise

was present over a certain volume level which, bizarrely, was not always the same. As an observation, it wasn't manifesting on a cold start but was appearing after a few seconds (warm-up?). Checking it with a scope revealed some interesting patterns: it was triggered on the negative side of the signal and had an oscillating frequency proportional with the signal amplitude.

Tests helped eliminating everything before the actual amplification circuitry. Also, the current limited supply setup was not causing this either as it performed well for the right-side channel. On the other hand, the current draw was increasing dramatically once the distortion occurred on the left channel. If I learned anything from my previous brushes with this amp, was that it serves no purpose trying to understand when and why something weird happens.

I thus chose to focus my energy on isolating the problem and, optionally, trying to understand it afterwards. Sure, some things are intuitive but this one wasn't. The noise was manifesting everywhere on the signal amplification path, including on the negative feedback. In a power amplifier such as this one things are strongly correlated: the output of one amplification stage represents the input of another, with the feedback closing the loop. The issue is somewhere on that path, it gets picked-up and propagated along until it reaches back as negative feedback on the inverting input, at which point the difference with the input causes corrective countermeasures, likely exacerbating the problem. Boy, that was a long phrase!

Divide et impera

When dealing with a complex problem like this one, most often a divide-et-impera approach reduces it significantly. If there would be a way to fold the problem and gradually eliminate good parts of the circuit, it would be peachy. For this reason manufacturers put test points on their circuits and provide values for them. However, that card has been played-out for the current drive fix and holds no value now because I'm no longer dealing with a static behavior. The next thing that can be used to isolate troublesome points is the previously validated components. In this particular case, the right channel was functional so every component along that signal path was OK too.

With a set of jumper wires I swapped the channels at the bridging point between the main PCB and the operation one, namely at the CN401A/B connector. Running the test again in this configuration revealed that the noise jumped channels as well. Relying on the right channel, this meant that the problem was originating upstream of the pivoting point, from the operation PCB. That was a relief because the main PCB has already been abused by others and didn't take it very well.

The operation PCB contains, among other things, the voltage amplifier AN7062 with its adjacent circuitry. The scan quality of the service manual isn't great but the schematic is intelligible. What makes this amplifier stage interesting is the fact that it is powered in such a way that its output has a positive offset. Pins 8 and 11 are some form of a common-collector giving an output just few volts below the positive rail (60.3V). Further down, the Q401 and Q402 are interfacing said output with the 2.1V offset required by the main circuit board.

Should anything upset this balance, say a bad ribbon cable interconnecting the boards, a busted part or a ground issue, the entire system falls into a stable failure mode which, thanks to the AN7073 protection chip, is also safe.

Best guess

With every part double-checked in-circuit and apparently in working condition it was time for another problem folding. Unlike before, I had no choice but to start de-soldering parts and the only part that could not be properly tested was the AN7062 IC.

Being the only probable culprit I chose to replace it, so I carefully sucked-up all solder alloy around its pins and removed it.

With the same care I soldered back the new chip. I must say that I don't appreciate how Technics made these PCBs. Working with them is pretty nerve wracking. While I was carefully heating-up the IC pads I felt one sliding on the board surface. Luckily it didn't break the trace.

No luck

With that out of the way, I went ahead and tested the system with a test sine. To my absolute stupefaction, the noise was still there. By now I knew there's no such thing as a ghost issue. If anything, it is just something perfidiously wrong.

I find it interesting how I'm tempted to re-check components in such situations but as I was running out of options I began doubting my process and actions. Either the new chip was bad in a similar fashion, which was unlikely, or it wasn't the chip after-all. I started suspecting some aging ceramic capacitor for the oscillating behavior. Tapping and blowing hot air onto various components led to no observable change. With all traces intact and no more cold joins, I decided to backtrack my steps and create myself an advantage while at it.

Divide et impera 2.0

I removed the newly installed AN7062 chip and soldered a DIP IC socket instead. Then I inserted the old IC in it and tested the system again to make sure my intervention didn't cause other damage. I was back to square one, which was a good thing. The right channel remained unaffected by these swaps. I knew further troubleshooting will revolve around that chip, so I chose the DIP socket to limit the wear on that fragile PCB.

The socket gave me the opportunity to run jumper wires from all 18 pins to a breadboard hosting the AN7062. I was then able to swap the two channels again, a bit more convoluted this time. Turned on the power, fed the test sine and, surprise, the noise was still on the left channel, thus excluding the chip. But if it wasn't the chip, what else could it be?! All parts, passive and active, were checked in-circuit, some even out of the circuit.

I was glad that the expensive parts were in fact alright. Surely something interesting was going on with this circuit. Assuming the Q411's collector resistor has gone bizarrely bad, I replaced it. In doing so, I slightly lifted another PCB pad. F#ck!!! Obviously, this maneuver had no effect and I felt like slowly destroying the operation board every time I touched it. Nothing else on the signal path could possibly cause this behavior but the Q411. With all fingers pointing at it, I removed and tested it with my DMM to find that, obviously, it was good. In fact it was showing a nice h_FE, consistent with its datasheet. Parts OK, problem persisting. This is nuts!

Then I remembered the noise manifested only over a certain volume level, which made me think that the DMM might not be capable of stressing the transistor enough to reproduce the issue. Next thing I did was to replace it with a BD244, the only one I had that was rated to withstand a high V_CE, though still under the value of the original Q401. As such, I had to slightly reduce the controlled power supply voltage to avoid its premature destruction. I then powered the system, fed it the test sine and ... Eureka! The noise was gone. It works!

For the first time since the headphones relay repair, I had both channels playing at generous volumes. I listened and listened for hours and despite using a different bipolar transistor, there was no perceivable difference between channels. I was already enjoying this amp, watching with a silly grin the supply current meters jump with the music.

Aftermath

Leaving the party in the background for a second, I'd like to iterate the aspects I found pivotal for this fix. The first is patience because things can go wrong in so many ways without it. Every time I felt about to loose my sh!t, I went doing something else and in the process of handling some redundant, day-to-day activity, ideas would pop-up.

The second would be working smart and organized. I personally like the problem folding approach as it helped pin-point the issue significantly faster - the geeky dev in me sees it as O(logN) performance. Keeping notes on actions and observations goes a long way while organizing loose parts, components and screws is very helpful especially when the fix/restoration spawns over several days. Also, a clean and uncluttered work space improves reasoning and boosts the morale.

Never again will I blindly rely on testing parts for common failure modes. Instead I'd use what's known: service manuals, validated components to help identify the bad. Of course, sometimes a defect on auxiliary circuitry may cause indirect effects on the signal. A good understanding of the system and its components is critical to succeed in such cases.

To be continued

This SU-VX920 is still far from ready. I need to order and then replace the faulty 2SA1123R transistor. Also, as I was working with the operation PCB I noticed another resistor pad lifted around a weird acting relay. They don't seem to be on the critical signal path but need to be checked-out anyway.