Integrator non-idealities
Finite DC gain A0, gain-bandwidth (GBW) and slew rate model the departure from ideal integration and settling.
A second-order, 4-bit sigma-delta modulator in Cadence. Study a 250 Hz signal band at a 256 kHz sampling rate, then see what finite amplifier gain, settling, sampling noise and DAC mismatch do to the spectrum.
This is a separate model-file product—not the MATLAB App and not a transistor-level silicon IP release.
DESIGN OPERATING POINT
The supplied design uses VDD = 1.8 V, VREF = 1.2 V and VCM = 0.9 V. Its input range is ±0.75 VREF, or ±0.9 V. The nominal signal bandwidth is fs / (2 × OSR) = 250 Hz.
CIRCUIT & NON-IDEALITIES
The model brings the CIFF signal paths into a Cadence circuit-simulation testbench, with two integrators, a multi-bit quantizer and DAC feedback. Use it to examine implementation effects before moving to a transistor-level design.
Two integrator stages, feedforward summation and multi-bit DAC feedback. Click to enlarge the original capture.
Finite DC gain A0, gain-bandwidth (GBW) and slew rate model the departure from ideal integration and settling.
Input sampling thermal noise and switch non-idealities are included, so the model is not limited to an ideal discrete-time signal loop.
DAC capacitor mismatch and data-weighted averaging are modeled. Compare the spectra with and without DWA rather than judging mismatch from the ideal FFT alone.
CADENCE SIMULATION RESULTS
The reported model design achieves ENOB above 18 bits. These four existing Cadence captures show the ideal baseline and individual noise/mismatch checks; the sampling-noise case reports 18.85 bits. They are the same circuit-model results shown on the App page, not MATLAB App FFT outputs.
Ideal circuit-model FFT: a baseline for comparing the non-ideal cases below.
With input sampling thermal noise enabled, the displayed ENOB remains above 18 bits.
DAC mismatch without DWA. The displayed setup uses a 35.6 fF unit capacitor and 0.45% mismatch sigma.
The DAC mismatch case with data-weighted averaging enabled. Compare the in-band spectrum with the mismatch-only capture.
FFT captures: 8192 samples, 256 kHz sampling rate, Hanning window and a displayed analysis band of 31.25–250 Hz. These are separate simulation conditions, not evidence that all non-idealities are enabled simultaneously. In particular, the 20.57-bit DWA result is not a combined thermal-noise-plus-mismatch result. ENOB depends on the input, enabled effects and FFT settings; these are model simulations, not measured silicon.
ABOUT THIS PRODUCT
The CIFF architecture uses feedforward paths from the integrator states to the quantizer input. This listing is for a second-order sigma-delta modulator model in Cadence, rather than a complete incremental ADC with a reset controller and digital decimation filter.
The model-file release is independent of Sigma-Delta ADC Designer. Buying the MATLAB App does not include this model, and this model does not include the App.
A circuit model is a starting point for design and verification. It is not a guarantee of transistor-level, post-layout or measured silicon performance.
The ZIP includes the simulation model files. Import the supplied library into Cadence and run the included testbench directly. A compatible Virtuoso/Spectre installation is required; Cadence software and EDA licenses are not included.
PACKAGE & SETUP
The ZIP contains the CIFF model library, schematic and symbol views, simulation model files, ADE/Maestro simulation setup and Verilog-A support cells. The model files needed for simulation are already included; import the package to run the supplied testbench directly. This is the second-order, 4-bit model package; it does not include the MATLAB App, a reset controller or a digital decimation filter.
Simulation results depend on your model settings and environment; no measured-silicon or ENOB guarantee is offered. The purchaser license and redistribution restrictions in our terms apply to this model.
Interested in using the model? Contact sigmadeltastudio@163.com to discuss your simulation environment.