Know your parameters
Designing a radio system requires knowledge of some fundamental parameters to understand what will influence link performance and reliability. In Part 1, I have already discussed perhaps the most fundamental parameters: output power and receiver sensitivity.

The others that are important are listed here:

• Receiver dynamic range
• Co-channel rejection
• Adjacent channel selectivity
• Reference frequency stability
• Mirror image attenuation
• Modulation principle

To answer that question, the designer must consider each of the parameters in turn. The receiver dynamic range is the maximum power variation at the receiver input pins that still results in a correctly demodulated signal. This means that to be received properly, the power of the signal of interest must be between the sensitivity limit and the sensitivity limit plus the dynamic range. The co-channel rejection (CCR) is a measure of the capability of the receiver to demodulate a "wanted" modulated signal without exceeding a given degradation due to the presence of an "unwanted" modulated signal, when both signals are at the same nominal frequency. This parameter is specified in decibels. For example, a co-channel rejection of 10dB would mean that if the wanted signal is 10dB (or higher) than the unwanted signal, then correct demodulation will occur (with a typical BER of less than 10^-3).

If this parameter is not specified in the datasheet, then a designer should assume 12 to 14dB co-channel rejection, the typical frequency shift keying (FSK) demodulator threshold (see below). As an example, consider the system in Figure 5, with two adjacent transmitters operating at the same frequency. Let's calculate how far away the unwanted signal source needs to be so as not to interfere with the demodulation of the wanted signal.

The equation for received power is :

(P_out1. G_ant1. G_{ant_RX})/Path_loss₁ ≤ ((P_out2. G_ant2. G_{ant_RX})/Path_loss₁). CCR

Can be reduced to:

P_out1 - P_out2 +20. log (range₂/range₁) ≤CCR

Assuming a CCR of 12dB, and that the transmitters have equal output power and antenna gain, the ratio between range2 and range1 must be at least 4 in order for the receiver to demodulate the wanted TX₁ signal correctly without interference from TX₂.

The adjacent channel selectivity (ACS) of the receiver is defined by the European Telecommunications Standards Institute (ETSI), for example, as the ability of a receiver to demodulate a received signal at the sensitivity limit, in the presence of a sine component centered in the adjacent channel (see Figure 6). (I.e. if the ACS for a 25kHz channel system is given as 30dB, demodulation of the wanted signal at the sensitivity limit may be performed with a sine component of 30dB higher power than the received signal present in the adjacent channel.)

Note that the ETSI definition is merely provided as a comparative figure. The system ACS is typically lower, as the signal in the adjacent channel is unlikely to be a sine component, rather a modulated spectrum.

Reference frequency stability is a parameter that influences ACS. Deviation from the ideal crystal reference frequency will result in a corresponding deviation of the transmitted frequency, and an intermediate frequency (IF) offset/deviation for a superheterodyne receiver (one that shifts the received signal to an IF) appearing as an offset in the IF-filter centre frequency (see Figure 7).

Receivers use the superheterodyne principle for its excellent channel filtering performance, but suppression is needed to avoid mirror image interference. Mirror image attenuation (MIA) is a measure of the extent of this suppression.

In Figure 7, the mirror image is shown positioned at the local oscillator (LO)-frequency minus the IF, but will also appear at the IF together with the wanted signal. Consequently the mirror image frequency must be attenuated to avoid destructive disturbance or loss of sensitivity.

MIA has traditionally been performed with an external filter at the antenna input, or more recently, with on-chip cancellation techniques. As the mirror image appears inside the IF-filter bandwidth after mixing, MIA minus CCR is the maximum allowable power difference between the two signals to ensure demodulation. For example, if the mirror image attenuation is 35dB and the co-channel rejection is 12dB, the received mirror image frequency power may not be higher than 23dB (i.e. 35dB -12dB) compared to the wanted signal.

The modulation principle is the final critical parameter to be considered for the transceiver. In the early days (particularly in the 433MHz band), amplitude shift keying (ASK, also known as on-off keying or OOK) dominated the license-free low power radio bands. Unfortunately, although ASK-based transceivers are simple and reasonably priced, they suffer from relatively poor reliability when under the influence of in-band interference. In ASK systems, logic '1' is represented by a carrier frequency, while logic '0' is represented by no carrier. Consequently, the presence of a very weak unwanted signal in the channel can be interpreted as logic '1' if the receiver is sufficiently sensitive.

Frequency Shift Keying (FSK) is a different approach in which each the two logic levels correspond to a frequency value:

DATA_FSK = "1" ' f'1' = center + Δf
DATA_FSK = "0" ' f'0' = center - Δf

The technique ensures that the receiver always sees a strong signal swamping any unwanted signal in the channel.

Gaussian Frequency Shift Keying (GFSK) and Gaussian Minimum Shift Keying (GMSK - the term used for a GFSK signal in which the bit rate is four times the frequency deviation) modulation are enhanced versions of FSK implemented to optimize modulation bandwidth efficiency (i.e. the maximum transmitted number of bits per Hertz of channel bandwidth).

GFSK applies Gaussian filtering to the modulated baseband signal before it is applied to the carrier. This results in a "dampened" or gentler frequency swing between the high ("1") and low ("0") levels. The result is a narrower and "cleaner" spectrum for the transmitted signal compared with the straightforward approach of FSK. Figure 8 shows the general principle.

Avoiding interference
As more companies produce products that use the license-free portion of the low power radio spectrum, designers have to deal with the possibility of interference from radio signals from other sources operating on an identical (or at least very close to) their own chosen operating frequency. In fact, regulations governing the ISM parts of the spectrum state "a device must expect interference". There are three basic techniques for minimizing the impact of interference for devices operating in the 2.4GHz band. These are time domain multiple access (TDMA), direct sequence spread spectrum (DSSS) and frequency hopping spread spectrum (FHSS).

TDMA works by subdividing narrowband allocations within the ISM portion of the spectrum into a number of timeslots allowing several users to share the same single frequency without danger of clashing. Each transceiver waits for its own clear timeslot before transmitting, thus avoiding interference. DSSS and FHSS both rely on modulation of the carrier signal. In addition, FHSS allocates a number of channels within the band, the exact width of which depends on the technology (see below).

In simple terms, DSSS transmissions multiply the data being transmitted by a "noise" component. This noise signal is a pseudorandom sequence at a frequency much higher than that of the original 2.4GHz signal, thereby spreading the energy of the original signal across a much wider band. The noise is filtered out at the receiving end to recover the original data, by again multiplying the same pseudorandom sequence by the received signal.

For de-spreading to work correctly, transmit and receive sequences must be synchronized. This requires the receiver to 'lock' its sequence with the transmitter's via a timing search process. DSSS operates at the cost of transmitting excessive data packets, incurring extra bandwidth usage and current consumption overheads. Using FHSS, a wireless technology periodically retunes to a different channel in a pre-defined (determined by a look up table) or pseudo-random sequence.

Bluetooth wireless technology, for example, uses FHSS in conjunction with GFSK, splitting the 2.4GHz ISM band into 79 x 1MHz channels (with a 1MHz guard channel at the lower end of the band and a 2MHz guard channel at the higher end). Transmitting and receiving Bluetooth wireless technology devices then hop between the 79 channels 1600 times per second in a pseudo-random pattern.

Bluetooth 1.2 uses a revised form of frequency hopping dubbed adaptive frequency hopping (AFH). This algorithm allows Bluetooth wireless technology devices to mark channels as good, bad, or unknown. Bad channels in the frequency-hopping pattern are then replaced with good channels via a look-up table. Nordic's Frequency Agility Protocol (FAP) uses a simplified frequency hopping scheme in that the transmitting and receiving pair establish communication on a particular frequency and then only hop to a different frequency should interference be experienced. The channel on which the interference was experienced is marked and not reused during that particular communication cycle.

In its 2.4GHz variant, ZigBee uses 16 channels at 5MHz spacing; each channel occupies 3MHz, giving a 2MHz gap between pairs of channels. ZigBee then uses a simple DSSS scheme for data transmission.

Devices such as Nordic's nRF24AP1, used in very low duty cycle applications such as sports-based heart rate monitoring and wireless transmission to a watch-based recorder, employ a proprietary form of TDMA developed by Nordic's partner Dynastream Innovations. The nRF24AP1's TDMA-like collision avoidance approach relies on each transceiver transmitting in a clear timeslot. If there are a number of discrete systems working side-by-side (such as a row of rowing machines all lined up next to each other in a gym) then by "listening" for drifting transmission sources on its frequency the wireless node can determine if there is approaching interference and adapt its transmissions accordingly (see Figure 9).

Interpreting the datasheet
As a systems designer, you know that a key requirement is to have clear information about circuit performance that can be used to make comparative assessments between the various RFIC alternatives. Although datasheets are supposed to help in this respect, this is not always the case. A highly competitive market has led to some ingenious ways of rewriting parameter definitions in order to make circuit performance look better than it actually is.

Consequently, there is a need for caution when reading and using the parameters given in datasheets. Some knowledge of "the not-so-fine art of creative RF-datasheet writing" may be in order. For example, if the measurement conditions of one or more of the key parameters are not given, try to find out why. (The manufacturer may have chosen to be "economical with facts" by omitting the details.)

Start by verifying the data rate. Terms such as "data rate," "chip rate," and "baud rate" are all used to describe the amount of data transferred per unit time by the transceiver. Make sure you understand how the silicon vendor defines this parameter. The transceiver needs to meet your requirement of the data you wish to transfer when employing your designated protocol (this is not necessarily the same as the "raw" data rate). For example, some systems assume Manchester coding when transmitting data so the designer needs to be sure whether the stated data rate does or does not include the extra transitions resulting from Manchester encoding (see Figure 10).

Current consumption is important, particularly for ultra-low power (ULP) applications (for example, those using coin cell batteries such as the 3V, 180 to 220mAh-capacity CR2032 type that can only support a peak load 20mA). Be sure that the current consumption is given for the frequency band in which you intend to use the device. It also important to determine the peak currents when transmitting and receiving, plus currents when the transceiver is in "sleep" or other power down modes. While peak currents are obviously important, battery capacity can be significantly affected if the transceiver spends long periods in other modes.

Next, consider the crystal requirement. This parameter is usually stated as the maximum allowed offset from the nominal frequency in parts per million (ppm). Make sure that the requirement stated in the datasheet is valid for the channel bandwidth and frequency deviation used.

Some transceiver solutions are based upon receiver tracking of the received signal in order to reduce crystal reference requirements. In other words, the receiving frequency is actively adjusted in order to "find" the transmitted signal. Note that the transmitted frequency will still drift according to the transmitter crystal offset.

Although this approach will ensure communication between two units, transmitter frequency drift must still comply with the channel spacing of the system. For example, trying to use a +/-20ppm crystal for a 2.4GHz system with 25kHz channel spacing will result in a worse case transmitter frequency drift of 48kHz. Generally the crystal reference is a significant part of the total system cost. Also note that crystal cost is proportional to the temperature range over which specified performance is guaranteed.

The crystal's load capacitance (CL) will also significantly affect the systems start-up time and start-up power consumption (and important consideration in ultra-low power applications). A lower CL reduces the start-up time, but increases start-up power consumption. Care should be taken to find the best compromise between the two. The crystal manufacturer normally specifies the device's CL value.

Finally, consider the switching time between different operational modes (For example, transmit to receive mode and power-down to receive mode). Remember to add the duration time of "training" - or preamble sequences. Some receiver topologies require lengthy "10101010..."-sequences in order to initialize or synchronize the demodulator. Moreover, the receiver frequency tuning sequences described in the crystal reference section (above), is generally time-consuming compared to the given switching time.

Designing a low power radio wireless link with good range and reliable communication is well within the capabilities of a competent electronics engineer, provided he or she takes time to understand the electronic, physical, and environmental factors that affect RF performance. About the Author
Jay Tyzzer is a senior applications engineer with Nordic Semiconductor based on the US West Coast. This feature is based on a white paper by Frank Karlsen, an RF Designer with Nordic Semiconductor entitled "Guidelines to low cost wireless system design". The white paper can be downloaded from www.nordicsemi.com

Related Articles

• RF Digital: Wireless modules are pre-approved for ISM-band
• RF Basics: Radio Propagation
• Fundamentals of ISM-Band and short range device antennas, Part 1