If you're curious to see how an anti-imaging filter affects the SQ of your NOS DAC but have been hitherto put off by all the coil winding involved, then this circuit might be just the ticket.
I was curious to see how well just two inductors could do and it turns out not at all badly - there's about 23dB stop-band rejection here and the 10mH inductors are off-the-shelf Fastrons. The capacitors can be all 1nF 0805 C0Gs - just buy a strip of 40 and use parallel and one series combination.
I intended to build one and listen to how it fares against my more complex efforts... :)
Update - a quick listen and its clear on switching back to the original filter that this new filter's noise floor isn't as low. That's not directly perceived as a rise in hiss mind - this is the subconsciously perceived noise floor which translates to more attention given to the music. So I'll put the two inductor filter aside and work on slightly more complex filters next.
Posted 28th May 2013 at 12:24 AM byrjm Updated 28th May 2013 at 12:08 PM byrjm
Experimental : For Research Use Only
It's bring-your-own-voltage-gain. This output stage is a unity gain buffer. A sort of diamond-buffer-meets-sziklai-pair hybrid. It lets the driver pair bias the output pair without the complexity of an additional bias network, but, unlike the basic diamond buffer, the output pair can have a much higher bias current than the drivers.
LTspice file attached, if you'd like to play along.
I did my best to shut my eyes and design this just by messing about in LTSpice from the starting idea of a "level-shifted-complementary-sziklai-pair" (i.e. mirrored J-Mo mk II), but I see now a shout-out to 47 Labs is due as the 0247 Treasure uses the same stage.
Ah well, I guess my neat idea isn't new after all.
The general performance is in the order of 0.01% THD for 50 mW / 16-300 ohms at 100 mA bias though the output pair. I've been trying to...