Hi Frank,

There are several possible reasons. I will list them below:

1. Your sampling switches at the delta-sigma input are not linear enough. This causes a problem because these switches are tracking a continuously varying input signal and so their Ron needs to be much better than when tracking a previously sampled-and-held signal, or else the Ron needs to be made linear with bootstrapping. Try replacing those switches with ideal ones with comparable but fixed Ron to see if the harmonics disappear. The charge injection from nonlinear switches will also not get cancelled properly in a differential arrangement.

2. Poorly implemented differential arrangement. Merely using a differential circuit does not mean that you have implemented everything differentially. Is this schematic-based simulation, or have you also some layout-based backannotation? In either case, check to see if you have made the circuit really symmetrical. In a schematic, this is easy as long as you are using a fully differential structure. In the layout, it is not so obvious. Remember that symmetry of the diff. signals means symmetry in both amplitude and phase, i.e. the signal delays through both paths must match. So, the two paths must see 100% identical conditions. This also holds true for the source driving the 2 differential inputs and the reference inputs and the output loading on the integrators etc.

Vivek

thanks