Rare spurs? What's that?

Since 1-bit DAC has just 2 possible output level, it's output doesn't have non-linear distortion products.

N-bit DAC output has such non-linear distortion products, since it has some non-linearity due to resistors value spread.

This is why 1-bit DAC spectrum is more clean than N-bit DAC (if we don't take into account harmonics level).

BTW just noticed that simulated DAC spectrum does not have sampling 1KHz spur, but two symmetric spurs around 1KHz instead. Some kind of unwanted FM-modulation artefact? Could you please check your simulation and if possible - make 8-bit DAC with 1% R2R ladder?

I applied just quantization and resistors value spread effects. It doesn't include all possible DAC issues.

The result is sampled with 1 MHz and processed with FFT at RBW = 1 Hz with float32 precision.

Here we talk about filtering fundamental frequency out of square wave, right?

Yes, exactly.

The square wave is a sine wave representation on 1-bit DAC output.

So, technically we are compare 1-bit DAC with 10-bit DAC, but we're compare it for fixed sine wave

And yes, we can propose topicstarter with a third way to produce sine wave

He can use 1-bit DAC with oversampling and sigma-delta modulator in order to produce arbitrary waveform.

So, he will not be limited with just a sine wave and will be able to produce any kind of waveform with no need for DAC.

But I think it will be overkill, because topic starter needs just a fixed sine wave, so there is no need for oversampling.

It can be done with a simple fixed square wave and apply low-pass filter with a proper cut-off.

slope between 100Hz and 150Hz does not differ much...Still huge win for filtered DDS, 35 dB.

It may be win or lose. It depends on requirements and selected filter. Harmonics level may not be an issue at all. But more non-linear distortions products may be an issue.