1% 8 bit R2R DAC @ 44KHz = Joke?

[Peter Heitzer]

I'm trying to make a stereo audio playback device with a frequency
response up to 20khz. I am feeding 16 bits of digital PCM audio into a
16bit shift register which has a storage latch, and then latching the
output into an 8-bit R2R DAC in 100ns. The R2R dac is discrete, i.e. a
bunch of resistors tied together with a buffer opamp. I'm wondering how
imprecision in the resistor values (they will be 1% tolerance) affects
the output? I.e. if I looked at the FFT of the original and the FFT of
the output, what differences would I see? (ignoring quantization
effects)

If the resistors are from the same tape or reel changes are good that
they match to 0.2% or even 0.1%. That would be suitable for a 10bit dac.
10bit audio can sound quite good provided you filter off the noise.