I have a confusion involves the SNR comparison of DSM simulated in MATLAB and VA. In MATLAB, I can get the theoretical maximum SNR (88dB) using transfer function blocks and quantizer block. But in VA, it gave me 68dB (I used the laplace transfer function in VA to create my loop filters, and quantizers and DAC are 16 levels). I don't understand why they don't match since I don't have any non-idealities in my VA models.
Matlab error is simple 64bit IEEE-754 numeric. If you use some functions in matlab the error control is "eps"
type eps=2.22e-16
That is typical much tighther than what simulator implement. But set reltol to eps of matlab could fail because most spice relevant models are calculated in 64bit floating point but sometimes with hard coded accuracy or the numerical presiscion is lost by some orders by the implicite routines. So failure in convergence is the result.