First some stupids question :
- Did I need two blocks of thoses, one per channel ?
The schematic is of one channel.
- Can I assume the IN where the + of the RCA ? And I wire the - to any ground
Yes.
Electrically it will function no matter what as long as all the grounds are connected together. This can cause hum or noise problems, or it can be completely silent.
In a balanced amp where the PSU is not in the picture this is much much less critical for obvious reasons. In a SE amp you have to think in terms of current loops.
A useful and practical approach is using local grounds as I believe somebody already pointed out. For example make each gain stage or buffer stage have a local star ground, and then connect all these star grounds to the last PSU cap ground (never to the first cap ground, it's the noisiest ground in the whole circuit). This is then your master star ground.
You could also go for a ground plane approach, and just connect every ground straight to the chassis. This works really well in guitar amps and also in high frequency circuits like CRT circuits, radios etc.
I myself use the local ground method.
- Rod Coleman : ... hum any schems/plan of this thing ???
Well he sells it, so no. Other systems that work on the exact same principle have schems floating around the technical DIY forums. Basically it's a voltage control (gyrator) on one leg, and constant current source on the other leg. High impedance on both legs; keeps the audio signal out of the PSU.
- About outpout transformer : can I choose it depends of the load ? 4Ohms / 8Ohms / 50Ohms / 300Ohms headphones ?
Depends completely on which kind of topology you go for.
The design that this thread is about uses a 10k : 4R OT that can drive any speakers or headphones no second thoughts about it. I don't see any point in matching impedances, I'm not in the output power game at all. Overkill low impedance drive is where I'm at.
You can ask Sonic if this gross impedance mismatch causes any problems for his headphones.
Now that is a super overkill gyrator right there. First class work. In my experience and opinion it will however not bring any audio benefits.
There has been a pretty large consensus for years that 50x rp impedance for the plate load is about the upper limit where any audio benefits are had. Many would argue it's much lower, 20x or 30x.
This 50x is easily had with a single FET gyrator for almost all useful audio tubes.