Consequences of Front-End versus A/D Non-Linearities Hans Hinteregger;
MIT Haystack Observatory, USA
Two kinds of non-linearity, for monotonic versus discontinuous response devices respectively, have different detrimental consequences for a receiver in the presence of interference. Their relative importance also depends on the strength and complexity of the RFI environment and on the application, high-dynamic-range low-frequency wide-field-of-view short-baseline-interferometry being perhaps the most critical.
I will discuss some issues raised for the design of the analog (front-end) & mixed-signal (A/D) parts of receivers:
1. SFDR specification? Is low-frequency-array science (EoR in particular) compromised by SFDR ~ 48 dBc (tested) for candidate 6-bit 800MS/s converter MAX105 in WA weak and sparse RFI environment? Only fast alternative is ADC081000 8-bit 1000MS/s; expect SFDR ~ 60 dBc for $100 part.
2. SFDR can be maximized if necessary at ~90 dBc with an 80 MS/s 12-bit converter; the $20 LTC1749 has 500 MHz input bandwidth and .15 ps aperture jitter.
3. If instantaneous bandwidth is to be traded for maximized SFDR, the least expensive way to tune over the front-end band is to use the standard double conversion scheme in place of the single Nyquist zone filter. Integrated COTS devices, incremental cost <$100. Linearity analysis needed.