First of all, i bought from e-bay a relatively cheap
wirelles module
called XL4463 - D01.The module consists of an Si4463 rfic from Silicon Intergrated (link), a 30 MHZ crystal from a Chinese company (JWT) and fiew other minor components (capacitors and coils).
This module has an SPI connection type with out
world. I used also a module with STM32f103c8t6 medium density microcontroller on it. (link) .After thorow study of both modules, i decided to make the connection between modules, like this:
SDI-------PB15
SDO------PB14
SCLC----PB13
nSEL-----PB12
GPIO0---PB11
GPIO1---PB10
nIRQ-----PB1
SDN------PB0
The final result is visible in the next photo.
Note that this Sony logo is a battery from an old camera which i use it as a power supply (in combination whith its charger) to give power to my device.
The flat cable on the left is a jlink connection with my programming device.
I used c++ to write the code to microcontroller and also used the WDS from Silabs (link) for generate the configuration file for rfic.
Until here all seems ok.
Is it posible to achieve the -126 dbm RX sensitivity from this module? This is a very difficult question to be answered in few lines. Generally not!!!!
But why? Of course the Silicon Labs company has no responsibility for this. But what about the XL4463 module maker?
This rf module has as a base time a crystal oscillator. The crystal has completely unknown characteristics except the (according to ebay seller) +-10ppm accuracy. The long term stability and the behaviour of the crystal in temperature variations are completely unknown.
I observed that in range from -15 deg. Celcius to +80 deg Celsius the stability was about 12 ppm (+-6)ppm.
We have at least 10+6 = +-16ppm (32ppm) inaccuracies in this temperature range.
This inaccuracy means that in an example freq. of 433900000 HZ we have about a 14 KHZ ambiguity about the frequency!! So the RX bandpass filters (and a lot other subsystems) of the receiver section of Si4463 chip must be adjusted at least to pass RXbandwidth + 2*14kHZ (at both sides) .
The sensitivity is inverse to selectivity (bandwidth of the receiver) . According and to the operation manual of rfic, is imposible to achieve the maximum sensitivity with this crystal tolerances!
An obvious solution is to use an OCXO or TCXO but both are very expensive to buy.
The best solution is to use a cheap crystal but with stable characteristics per batch and compensate the frequency over a temperature range whith some algorithm. The chip has the ability to measure the temperature. Near the rfic chip is the crystal. So the temperature of the crystal is almost equal to the rfic's temperature.
In a future post i will show you how can we reach a 0.1ppm stability (not long term) with theese Chinese cheap crystals whithout the use of any extra instrument except the basics (PC, two rfic working modules ) and some minor parts as usb-to-uart adapter etc..
Victor. F.
