Vladimir said:
This guy does it without a hot air gun. Just fine tip, steady hands and a zoom lense.
Crammed like
downtown Tokyo...
Panasonic solid state
OS-CONs.
Great pictures. This is the 5 channel version, the same case had two more channels for the 7 channel model. If you look in the pictures you can see the mounting holes for the extra channels on the base. The 5 modules can each be removed separately which makes access easier, but the component density on the boards is high, so there isn't much leeway for a misplaced soldering iron.
Each module has two PCBs, a main board with the power electronics bolted to the heatsink and a piggyback control board mounted on spacers that has signal preconditioning, bias control and the DC servo circuitry (amp is DC coupled).
Again, for those fretting about their 'interconnects' it is worth noting that Krell decided that computer grade ribbon cable is OK to carry the low level input signal to the amp modules.
It is actually this ribbon cable that seems to contribute to the thermal problem. The amp has 6 transistors in parallel in the output circuit, 3 pushing, 3 pulling. To share the load between these transistors, low value 'emitter resitors' are used - these are the 0.68ohm wire wound ceramic white oblong blocks mounted along the bottom of the power PCB. The amp is already hot, but these resistors add to the problem. The way the ribbon cable is folded directs the heat over the piggyback board and fries the components - in particular the three surface mount caps in the centre.
If you look at the first picture, you can see how the middle of the piggy back board has discoloured - I think because of the heat. I had originally assumed this was due to manufacture, but my amp was always very well ventilated and only shows very slight discolouration at this point. The blown amp I bought has piggyback boards that are almost brown in their middle section.
Frankly, not a great thermal design, it would have been far better if the modules had been mounted the other way up. Given that the signal input socket seems to be at the wrong end of the module, I do wonder whether they were designed to be mounted this way originally. It wouldn't be the first time that the electronic design team didn't communicate properly the the team doing the case and mounting design and things got flipped.