A fun little BJT buffer

[nic]

Recently starting playing with BJT amps , have only ever used BJT in power supplies , CCS etc before so have been getting used to the Vbe thing . Have built a few breadboards and originally intending this experiment to be a diamond buffer but a bit optomistic with the NPN/PNP Vbe matching of the devices at hand . Came up with another solution of feeding the Vbe multiplier from either side of a current mirror and take the buffer's input from across Vbe multiplier . Output stage is CFP/Sziklai pair . Currently running the bias-block/current mirror set to 2.5mA , output stage at approx 80mA and there is small amount of drift so will probably have to increase the emitter resistors and rebias but for the time being having a listen to the thing . 5k pot sets the programming current , 100R pot sets the offset . This used surplus mains transformers (monoblock) , surplus heatsinks and parts from CPC which only cost a few pounds . It's currently bolted to a piece of wood . Approx 5 watts , no measurements taken , sounds pretty decent using a 6197 valve linestage , Raspberry Pi/HifiBerry

buffer.jpg (205.15 KiB) Viewed 2397 times

Pro with old Kef Cadenza speakers . Far from perfection but I'm having fun...

[Mike H]

Interesting! Not sure I would do it like that.

More knowledgeable people might suggest putting the bias preset on the emitter side of your middle transistor, just in case the preset goes high resistance or open-circuit, then the magic smoke won't be released. Unless you've got a fuse.

[nic]

Thanks , no magic smoke yet . I can see why I'm getting a bit of drift. As this started as a Diamond , I bolted the BD139/140 drivers to the heatsink and left these in place when I rejigged into a CFP output stage . I think these need to be thermally isolated from the main heatsink The BD are probably overkill for such a low powered amp , will see what alternatives are suitable , then have a rebuild of the CFP stage after a few more days listening