The modulation scheme 16FSK is pretty unusual and not widely used. Yet it is now available in a new IC from CML Microcircuits. And it offers some unusually good benefits for those of you building low speed wireless telemetry or control devices for industry.
Multilevel modulation methods are widely used because they solve the problem of getting higher data rates in a narrow channel. Speed is proportional to bandwidth so multilevel modulation schemes buy you some extra speed because they pack more bits per symbol or baud. The most widely used multilevel scheme is quadrature amplitude modulation (QAM). This is a combination both phase and amplitude modulation where multiple bits are represented by unique carrier phase-amplitude combinations. Each combination is called a baud or symbol. For example 2QAM is just BPSK. 4QAM represents two bits per symbol and is generally known as QPSK. 16QAM codes 4-bits per symbol. Versions 64QAM and 256QAM are popular and even higher level versions are used in satellite transmissions and cable TV.
While QAM is effective and widely used to cram more data bits per second into a narrow bandwidth, it comes at a price. A higher signal to noise ratio is essential to avoid a high bit error rate (BER). But the real problem is that QAM requires linear amplification if the phase-amplitude symbols are to be preserved and accurately recovered. That means low efficiency power amplifiers (PAs) that use more power and generate more heat. Despite these disadvantages, QAM is the modulation of choice in wireless LANs, cell phones, and other wireless systems especially those using OFDM.
Now along comes FSK. Plain old frequency shift keying or 2FSK uses two closely spaced carrier frequencies to represent binary 0 and 1. Only the frequency changes while the carrier amplitude remains constant. It is a “constant envelope” method. That means you can use non-linear modulation methods like classes C, E or F. These are far more efficient making PAs smaller, simpler, cheaper and cooler. FSK is also more immune to noise. Now you can use 4FSK, 8FSK or 16FSK coding 2, 3 or 4 bits per symbol respectively.
These higher level FSK methods of modulation and demodulation are difficult to implement so are a general pain to use. But CML Microcircuits now has their CMX7164 IC, a modem that includes full 2FSK, 4FSK, 8FSK and 16FSK. It does the modulation and demodulation digitally with DSP so you never have to worry about it. The chip supports symbol rates of 2k to 10k symbols per second. At the full symbol rate using 16FSK you can get 40 kb/s through a 25 kHz wide channel. The chip includes full data filtering before modulation using a root raised cosine filter with an alpha (α) of 0.2. Sinc filtering is optional. The interface to the system microcontroller is by way of a serial C-BUS. The CMX7164 is still just a modem so you need to add the external transmitter and receiver circuits. But it is now easier than ever to use higher level FSK and you can reap the benefits of non-linear PAs and higher data rates.
If you want a complete FSK transceiver, check out Analog Devices’ ADF7021 series of ICs. These ICs come in versions that cover the 80 MHz to 650 MHz range as well as 862 MHz to 950 MHz making them ideal for ISM band products. The ADF7021 chips implement 2FSK, 3FSK, and 4FSK but not 8FSK or 16FSK. 3FSK is an unusual version that uses three frequencies, the carrier center frequency and upper and lower shift frequencies. 3FSK is a bit more bandwidth efficient that 2FSK. The IC also offers raised cosine or Gaussian filtering for the data prior to modulation to squeeze out a few more bits per Hz of bandwidth.
Anyway, FSK is not dead and should be considered as a key modulation choice for new short-range wireless device design.