SI4464 Datasheet, PDF (29/56 Pages) Silicon Laboratories – HIGH-PERFORMANCE

Si4464/63/61/60
5. Internal Functional Blocks
The following sections provide an overview to the key internal blocks and features.
5.1. RX Chain
The internal low-noise amplifier (LNA) is designed to be a wide-band LNA that can be matched with three external
discrete components to cover any common range of frequencies in the sub-GHz band. The LNA has extremely low
noise to suppress the noise of the following stages and achieve optimal sensitivity; so, no external gain or front-end
modules are necessary. The LNA has gain control, which is controlled by the internal automatic gain control (AGC)
algorithm. The LNA is followed by an I-Q mixer, filter, programmable gain amplifier (PGA), and ADC. The I-Q
mixers downconvert the signal to an intermediate frequency. The PGA then boosts the gain to be within dynamic
range of the ADC. The ADC rejects out-of-band blockers and converts the signal to the digital domain where
filtering, demodulation, and processing is performed. Peak detectors are integrated at the output of the LNA and
PGA for use in the AGC algorithm.
The RX and TX pins maybe directly tied externally for output powers less than +17 dBm, see the direct-tie
reference designs on the Silicon Labs web site for more details.
5.1.1. RX Chain Architecture
It is possible to operate the RX chain in different architecture configurations: fixed-IF, zero-IF, scaled-IF, and
modulated IF. There are trade-offs between the architectures in terms of sensitivity, selectivity, and image rejection.
Fixed-IF is the default configuration and is recommended for most applications. With 35 dB native image rejection
and autonomous image calibration to achieve 55 dB, the fixed-IF solution gives the best performance for most
applications. Fixed-IF obtains the best sensitivity, but it has the effect of degraded selectivity at the image frequency.
An autonomous image rejection calibration is included in Si446x devices and described in more detail in "5.2.3.
Image Rejection and Calibration" on page 31. For fixed-IF and zero-IF, the sensitivity is degraded for data rates less
than 100 kbps or bandwidths less than 200 kHz. The reduction in sensitivity is caused by increased flicker noise as
dc is approached. The benefit of zero-IF is that there is no image frequency; so, there is no degradation in the
selectivity curve, but it has the worst sensitivity. Scaled-IF is a trade-off between fixed-IF and zero-IF. In the
scaled-IF architecture, the image frequency is placed or hidden in the adjacent channel where it only slightly
degrades the typical adjacent channel selectivity. The scaled-IF approach has better sensitivity than zero-IF but still
some degradation in selectivity due to the image. In scaled-IF mode, the image frequency is directly proportional to
the channel bandwidth selected. Figure 9 demonstrates the trade-off in sensitivity between the different architecture
options.
1% PER sensitivity vs. data rate (h=1)
-95
-100
-105
Fixed IF
Scaled IF
-110
Zero IF
-115
-120
1
10
100
Data rate (kbps)
Figure 9. RX Architecture vs. Data Rate
Rev 1.2
29