Low power 433 MHz BPSK data transmission system

WaveRider, Mon May 29 2006, 01:44PM

It's been a while since I have done a radio project. So, I thought I would experiment with data transmission using BPSK (Binary Phase Shift Keying) on the 433 MHz European ISM band.

What's so special about BPSK?
Most commercial systems seem to use frequency shift keying (FSK) or some form of on-off keying (OOK). I wanted to explore the possibilities of constructing a simple system for binary phase shift keying (BPSK). Binary phase shift keying is a form of modulation that is very "power efficient." This means that to achieve a certain bit error rate, the required signal to noise ratio is lower than that required by, e.g., frequency shift keying or on-off keying. BPSK (or QPSK) is often the modulation mode of choice for deep space probes where power consumption must be minimised and signal path losses are enormous. See graphic (taken from ) for comparison of error rates....




Why 433.92 MHz?
Besides sitting in the middle of the 433MHz European ISM (industrial, scientific and medical) band, it is also in the amateur 70cm band, so parts were easy to get and it can be modified to operate under amateur radio rules. For those in the US, I think this band also falls under Part 15 of FCC regulations for low-power communications devices. Also, the local oscillators use of-the-shelf SAW resonators instead of a PLL synthesizer. This simplifies the design greatly (at the expense of having to operate on a fixed frequency).

What will the transmitter power be?
Operating below 10mW in this band is permitted without a license. I plan to limit the emissions to somewhat less than this. Many garage door openers and automotive key fobs seem to operate in this band (altho' at much below this power level). Hopefully this won't mean intolerable interference in my experiments!

First step: The receiver design
Since the receiver design is more challenging than the transmitter, I started off there. Here is a block diagram of my proposed receiver.





I am basing the system around the Philips SA639 RF/IF subsystem chip, which is really designed for FSK systems, using a 10.7MHz IF. I managed to find a 500kHz bandwidth ceramic IF filter which should allow upwards of 384kbit/sec communications. BPSK demodulation is not as easy as FSK. Basically, we need to add a carrier regeneration system (a PLL) for demodulating the BPSK signal coherently. I do this with a simple "squaring loop" which doubles the IF carrier frequency. A NE564 PLL/VCO system is the heart of this.

The system
The received signal enters the system (as you would expect ) at the antenna input. A low noise amplifier, based on the BFR540 NPN transistor amplifies the antenna signal. I was a bit torn about whether to put the microstrip-based band/image-reject filter before the low noise amplifier. If I put the filter before the amplifier, the filter insertion loss will kill the noise figure of the system. However, putting the filter after the amplifier increases the chances of amplifier overload by strong out-of-band signals. I opted for putting the filter after the amplifier. I will rely on the antenna system to reduce out-of-band signals. Plus, the BFR540 amplifier as I have designed it should be able to withstand -20dBm on the input terminals without saturating.

The next stage is the downconverting mixer. I found a SAW resonator for 423.22MHz which is exactly 10.7MHz away from 433.92MHz. Hence, the local oscillator (based on a BFR92 transistor) fed to the mixer (part of the SA639 chip) downconverts perfectly to 10.7MHz. The ceramic IF filter passes a sharply defined 500kHz channel to the
first IF amp and limiter.

At this point, we split off the signal to the second mixer as well as a push-push active frequency doubler. The output of the doubler is passed through a 21.4MHz (second harmonic) bandpass filter (with a band-stop notch at 10.7MHz) to the PLL carrier regeneration circuit. The 21.4MHz signal is divided by 2 using a flip-flop. After a phase correction and level adjusting circuit, the regenerated carrier is fed to the final mixer, where it is mixed with the original amplitude limited IF signal. The output will be the original BPSK digital data signal (a 250kHz low-pass filter on the output remives any residual RF). A schmitt trigger squares up the output and produces a TTL compatible signal.

I am hoping for a BER at 384kbit/sec of under 10e-6 with -106dBm input at the antenna terminals (in the absence of other interfering transmitters).

Notes
I am at the stage of making the printed circuit board. I can make the schematics available, if people are interested in replicating this work. I will continue to post modifications, measurements and simulation results as I get them.

My main motivation was to experiment with digital transmission of sound and image data for a remote telemetry and surveillance system. I noticed that many commercial systems tend to use FSK, because of its simplicity of implementation. However, for a given power level and a bit of added system complexity, it should be possible achieve improved reception over longer distances using BPSK than FSK.

EDIT 20060530
Completed PCB drawings and etched board





The long strips on the left of the board form the three resonators of the 433 MHz interdigital band filter. They will be tuned with trimmer capacitors at one end and earthed at the other.

Also, ideally, all the ICs would look best as SSOP (small surface mount) packages. However, for this first prototype, I'll use the ICs I have on hand (SA639 in TSSOP-24, NE564 in DIL16 and the 74F74 in SIL 14...)

I made this board by printing the mirror image on a transparency and used a photosensitive (positive transfer) board exposed with UV light. The TSSOP pins came out very sharp! I initially tried toner transfer, but it did not work very well.... The photo process seems to be far superior when you have very large black areas or very small features (like the TSSOP pins)...

EDIT 20060615: Managed to get the low noise RF amplifier, band filter and local oscillator to work. It was particularly difficult to get the SAW resonators to...well..resonate! I tried the basic Colpitts oscillator topology, but it just would not work. The section of transmision line that leads to the first mixer would resonate at 1.3 GHz...not what I wanted! So I opted for the so-called SAW stabilised Colpitts oscillator. This uses a bog-standard LC tank in the collector circuit which resonates near 423.2MHz (the LO frequency). The SAW goes in the base circuit, where it looks like a low impedance at 433.2MHz, thereby forcing common-base operation at this frequency. It works like a charm! Rock-steady strong signal at 433.177MHz! 43kHz out, but well within the +-75kHz specified by the SAW data sheet...

The PCB looks like:




Here is the schematic:




The frequency counter was used to verify the oscillator frequency:




The oscillator frequency can be "pulled" several 10's of kHz by tuning the collector tank circuit. The circuit behaviour is a lot like the quartz-crystal oscillators so often used in HF.

The oscillator power output is about -4dBm.




This is a bit high for the mixer input (which needs about -12 to -10 dBm). I'll use a smaller coupling capacitor (presently using 2.2pF... perhaps a 0.47pF will be enuf to give about 6dB reduction to the output signal...)

The other challenge so far is tuning up the three-element band filter (the three strips on the right). The trick is to make sure there is at least 20dB of attenuation at the "image" frequency (at 412.5MHz) and as much attenuation as possible at the LO frequency (to prevent radiation of the LO by the receiving antenna). When I get the receiver assembled, I'll make some definitive measurements of LO radiation and image rejection...


EDIT 20060623:
Got all the major components of the receiver to work (with the exception of the carrier regeneration subsystem). Tested the downconversion with a 384Kb/sec BPSK signal. Here is a look at the partially populated board. The Philips SA639 FM/IF chip is in the top right corner. I am capacitively coupling the oscilloscope through a high resistance (33K) through a 50 ohm cable terminated with a 50 ohm load (this reduces resonances in the coax to the o-scope input) to the output of the IF limiter. See schematic: o-scope is connected to "limiter out" pin.





The resulting signal is at 10.7MHz with a 500KHz bandwidth (as defined by the ceramic IF filter....the white thing under the green tantalum cap in the upper right-hand corner).





A BPSK "transmitter" is simulated using my bench RF signal generator set to 433.92MHz and a doubly balanced mixer module. The modulating signal is just a 192kHz square wave.. The RF output of the signal generator is -60dBm. The mixer suffers 10 to 14dB conversion loss, so the RF level to the receiver is around -74dBm.





Here's a closer view of the mixer... These little modules are great....they can be a bit pricey, if you buy them new....





Here you can see the modulating "data" signal on the top trace and the output of the receiver limiter on the bottom trace. you can just make out the transtions in the IF signal.





Here is an expanded version showing the phase of the 10.7MHz IF signal slewing during the bit transitions.





The waveform is a bit foggy because the data signal and the carrier are not synchronised to a common reference timebase. However, it is clear that the receiver circuit has properly downconverted the RF signal. I can visualise that "fuzziness" in the IF waveform down to -94dBm... The sensitivity is not too bad for the simple circuit. When I finish the demodulator, perhaps I'll see the signal at even lower levels... (I'm hoping for at least -100dBm sensitivity)...

Next step is to assemble the PLL carrier recovery circuit to complete the demodulation of the signal.


EDIT 20060628:
Got the PLL carrier recovery circuit to work. Here's a schematic. (This is the basic circuit.. I've replaced the bulky TO92 2N2222's with SOT (surface mount) BC848 NPN transistors... works fine! This means that the receiver is ready for experiments!





After adjusting the level and the phase of the regenerated 10.7MHz (IF) carrier, we see the data signal (192kbit/sec).





The ringing is the result of the coax I am using to conect to the o-scope. The receiver is producing a clean 96kHz "square" output with the "transmitter" power set to 10uW with random wire antennas on both receiver and transmitter spaced a couple of meters apart. The top trace is the signal generator (transmitted) "data" and the lower trace is the received "data."

What happens if we simulate a long string of "0s" or "1s"... Setting the signal generator to burst mode allows us to test this.





Again, upper trace is the transmitted "data" burst and the lower is the received "data". We simulate a long string of zeros here, as seen by the flat-line part of the signal. PLL lock seems to be well maintained during bit transition times, the output data signal is rock solid.

By cycling the transmitter RF on and off a few times, we can observe the 180 degree ambiguity in the regenerated carrier (which is characteristic of Costas loop and squaring loop carrier regeneration systems, like this one). Notice that the string of "zeros" now produces a "high" output signal during the flat-line stage, i.e. the received data signal is inverted..





This is a shot of the completed receiver board.





This shot shows the key component of the frequency doubler: the 1:1:1 toroidal transformer that feeds the push-push frequency doubler.





This photo shows the completed PLL-based carrier regeneration subsystem. The PLL is on the right and a 74H74 D flip-flop (frequency divide-by-2) is the chip seen on the left. The two white trimmer pots are for adjusting the phase and amplitude of the regenerated 10.7MHz carrier to be fed to the secong receiver mixer (for synchronously recovering the BPSK signal).






Lastly, we have the receiver "random wire" antenna....a white alligator clip lead. (Next step will be to put together a couple of nice 50-ohm matched antennas for 433.92MHz...this should improve low-level signal performance significantly..)






EDIT 20060629

Made an antenna... Will make another one tomorrow!





The elements ar 15.6 cm long.. Simulations indicate nearly a 50 ohm fully resistive feedpoint impedance...


EDIT 20060703
Tried to do some sensitivity tests, but it is very difficult here in the center of a major urban environment. The interference level is very high. In fact, the power meter connected to the 433 MHz (filtered) output from the LNA with the antenna attached indicates -45 to -40 dBm.. this is very high, buit expected given the noisy environment of garage door openers, electronic car keys as well as possible amateur 70cm service... However, with a Tx power of around 500uW (-3 dBm) the carrier regeneration PLL holds its lock (with occasional fading) and a 192kbit/sec signal can be recovered just about everywhere in my flat. The "tripod" antennas work like a charm...10-20dB better response than my "random wire" experiment early on...


EDIT 20060706
Modified the carrier recovery PLL loop filter. Can now stably demodulate 384kbit/sec BPSK data stream (using the full bandwidth of the ceramic IF filter). I'm happy!! Need to add a simple circuit to modify demodulated data stream to TTL levels to feed to a 5V RS-232 serial port input of a microcontroller for processing (parity check, error correction, etc.)... Still need to build the transmitter!!!


EDIT 20060710
Improvements to receiver: matching at RF input to SA639 will improve sensitivity. Would like to study effect of noise in received signal on carrier recovery PLL...

EDIT 20060830
Simplified carrier recovery circuit. Frequency doubler is not needed. Use output from flip-flop and do phase comparison on 10.7MHz IF signal like this.




EDIT 20060921
Transmitted message across my flat! Very happy..





I used two Rabbit2000 processor boards. One to generate the "Hello World" message at 96kb/sec and another to receive. I used Biphase-mark Manchester encoding to synchronously transmit data (the bit clock is easily regenerated at the receive end at the cost of halving the bit rate wrt NRZ RS232 type asynchronous links). The radio link is capable of 192kbit/set using Manchester encoding, but my Rabbit processors are not fast enough and drop bits when the baud rate exceeds 115kbit/sec. Using assembly code and interrupt routines are a must (especially for the receiver).

Next step... build 3 proper transceiver boards with faster processors and start working on simple network.