The EXTREME1 utilizes high-speed Sigma-Delta A-to-D conversion, 32-times oversampling, high-order modulation noise shifting, dual digital filters, and a "brute force" analog low-pass filter. Massive oversampling (32-times) noise spreading and shaping, and digital cutoff are digital techniques that we feel are mandatory for useful recordings. Following are simple descriptions of these approaches.
First, a "brute force" analog low-pass filter is used to limit the incoming signal to a bandwidth of 0-230KHz. This reduces noise levels, and makes things simpler for the digital subsystems. Many simple A-to-D systems stop here...the EXTREME1 goes further!
Next, 32-times oversampling is used in the EX1 to increase the "Actual Sample Bandwidth Range" (ASBR) of the EXTREME1 to 7.5MHz. The actual sample range is far greater than the system NYQUIST range, 229.2KHz true useful bandwidth. The ASBR is not used for actual signal, but for digital filtering. Any input content above 230KHz has already been rolled off by the analog "brute force" filter. Though seeming to be overkill, this "extra" bandwidth is successfully utilized for noise reduction techniques.
Quantization noise is an inherent byproduct of the A-to-D process, and can be greatly reduced by spreading the noise across the Actual Sample Bandwidth Range (ASBR). Noise normally appearing in the 0-229.2KHz (useful passband) range is reduced by averaging the quantization noise, which tends to be generated a little more in both ends of the ASBR, across the whole 7.5MHZ bandwidth.
Noise shaping is also used to take advantage of the ASBR. A high-order modulator shifts the remaining noise beyond 229.2KHz, in effect, "rolling" the noise beyond the cutoff point, into the high end of the ASBR.
Finally, dual digital Finite Impulse Response (FIR) filters are used to generate a sharp digital cutoff at 229.2KHz by cascading a series of digital taps, each, in effect, causing a finer and finer cutoff point. If this were attempted with an analog approach, many stages would be required to approximate the results of the FIR filters.
