John Phillips
pfm Member
That was interesting even if a bit hyperbolic. I'm also not sure R-2R aligns well with what is said in the video. More consistently (IMHO), the 32 "flip-flops" could be driving 32 nominally identical resistors that generate 32 voltage levels when current-summed through an I-V converter. A discrete 5-bit D/A converter, in fact.Linn call it a "discrete conversion stage" and looking at the circuit board here, https://www.linn.co.uk/uk/technology/organik ,it sure looks like a R2R design... so I'm not sure I agree with "there is nothing from Linn to suggest..."
That converter sits in the middle of a noise shaper to raise fast 5-bit conversion to slower 24 bit conversion. Two-level conversion at this point with higher levels of noise shaping is used but is known to trigger mathematically chaotic behaviour and undesirable low-level "idle tones". Some DAC makers like ESS have secret sauce to eliminate this behaviour but others prefer to use multi-bit converters here to the same end.
And my guess is that they also use dynamic element matching, where the discrete DAC is arranged to work even faster so that each output voltage step derives from the average of all 32 discrete current steps. Thus you reduce the need to match the current source resistors very precisely. That's much nicer than R-2R in production. I think that's how the ring DAC works (and the likes of the Wolfson WM8740 series). A very good DAC architecture IMHO.