Cloudy and Cool, with a Chance of QPSK

Braving the elements is an important part of many communication systems; knowing the practical environmental effects can prevent unpleasant surprises.

By Rob Howald

One of the finest months of the year is upon us. February has always been a personal favorite for several reasons. For one, it is my birthday month, representing yet another chance to try to pull-off the old 29 routine. February is also such a fantastic time of year for weather in the Northeast corridor. Who can argue with a constant 66ýF ambient? Some places it might be 68ýF, depending on how the thermostat is set. Most importantly, February marks the beginning of baseball's spring training - the first real symbol of hope that the season of morning removal of windshield frost is waning. In honor of our shortest month, the discussion for this piece will slide into the realm of the directly practical, describing two of the most common environmental hurdles and their possible effects on digital communications.

Temperature is one obvious environmental element to which designers must specify their equipment. Simply put, a piece of outdoor hardware must work reliably in February in Bismark, ND and do the same in July in Miami, FL. Similarly, a military radio for a soldier must work whether stationed in the Persian Gulf or the Bering Straight. Another common situation to characterize in the field is the vibration and mechanical stress environment. There are many others that deserve attention, some of which are reserved for equipment survival only, and not necessarily requiring performance specifications.

Due to the nature of the topology of systems exposed to the elements, most involve RF or microwave links, and it is these pieces that are often most affected. Indeed, this is one of a digital implementation's great advantages. The topics described here rarely get discussed in textbooks or in literature associated with new design ideas. They are typically out-of-scope in academic journals, and, unfortunately, hardly ever appear where they belong most - in industry-type journals and trade magazines. While it is difficult to generalize about entire communication systems, it is actually quite straightforward to describe environmental impacts on individual hardware pieces of the system and extrapolate how this degrades a system.

What's hot?
High-temperature plateaus vary from the rather benign, such as 40ýC to 50ýC in controlled indoor locations (manned and unmanned equipment sheds), to 60ýC or higher for outdoor equipment specifications. It is important to note that individual circuit pieces, such as integrated circuits (ICs), transistors, and RF circuit modules or subassemblies, are often limited to about +85ýC for guaranteed operation and up to +125ýC for military pieces. However, this includes potentially substantial increases due to local heat dissipation. At the cold end, performance temperatures can be as low as -40ýC. Depending on the application, other common low-temperature plateaus include -30ýC and -20ýC. Benign cold situations may be only 0ýC.

Hardware thermal effects
What happens when things get hot? In general, active circuits become less efficient. For example, currents on amplifiers tend to increase, gain tends to drop, or both, at the hot temperature plateau. On the positive side, increases in current can result in improved compression characteristics, which can improve RF performance of cascaded systems from the standpoint of intermodulation distortion and harmonic performance. By contrast, cold temperatures are often characterized by minor increases in gain, lower current draws, and decreased power dissipation. Because of this, compression characteristics often become worse.

In terms of catastrophic behavior, cold temperatures seem to draw out any instability characteristics that might exist in RF and microwave circuits. Hot temperatures tend to draw out issues associated with circuits or chips on the threshold of run-away thermal problems. These can be internal chip problems or problems aggravated by external temperature issues.

Other temperature-sensitive circuits are oscillators. Oscillator frequencies will drift in temperature. Crystal oscillators, a cornerstone in a large percentage of communication circuits, have predictable drift characteristics that are based on the particular physical cut of the crystal itself. Crystal oscillator stability characteristics are specified in terms of parts-per-million (ppm) relative to the center frequency of the oscillator. For example, a ý50-ppm crystal oscillator with an output frequency of 10 MHz has an absolute range of ý50 x 10 = ý500 Hz. Obviously, the higher in frequency the unit is, the larger the variation caused by a 50-ppm oscillator is. Typically, the manufacturer of the crystal chooses the physical cut of the quartz crystal based on the stability requirements of the user. Some systems require better frequency stability than that achievable by only varying the cut of the crystal. In such cases, a manufacturer may compensate for the temperature variation with a diode-based circuit, reducing the instability by roughly one order of magnitude. For exceptional stability, such as below 1 ppm over temperature, a manufacturer will enclose the crystal and the associated oscillator circuitry within a sealed, ovenized case. High stability is obtained by keeping the circuit at a constant temperature internally.

While crystal oscillators form the backbone of many RF communication channels, as well as digital clocks for data transmission, generalized voltage-controlled oscillators (VCOs) used in RF and microwave links also move in temperature. This movement is not as predictable without fabrication and test, mainly because of the many circuits and resonant structures available in oscillator implementation. Since resonators can often be mechanical in nature, they may follow recognizable laws of physics. For example, cavity resonators increase in frequency at cold temperatures due to contraction of the cavity, which determines the wavelength of the oscillation. This relationship is often followed in more common oscillator circuits, such as lumped element LC-type oscillators or in microwave dielectric-resonator oscillators (DROs).

Just as resonant circuits move for active circuits, they can also move in passive circuits, such as filters. The result is a shifting of the center frequency in bandpass filters and an altered rejection characteristic. This also occurs with lumped element parts. Performance variation is sensitive to the quality factors (Q) used in the filters. High-Q filters, represented by narrow bandwidths relative to center frequency, are more sensitive to temperature or any other variation. Filters designed with just inductors (Ls) and capacitors (Cs) can be evaluated for performance effects, as the temperature coefficients of the various materials are often available for common piece parts. Most quality design tools nowadays include the ability to perform Monte Carlo runs that account for parts variation to determine circuit performance effects. At lower frequencies, active filters may be used, which must be carefully chosen in topology due to exceedingly sensitive performance to part variations.1 Even a plain old low-pass filter, like an anti-alias filter in front of an A/D converter, will see its part values vary in temperature. The results can be an increase in the cutoff frequency, less rejection of folded-over frequency bands, and potential for increased in-band interference. Increases in temperature typically degrade Qs in filters and resonators, resulting in poorer filter roll-off characteristics, which determine how sharply the filter rejects frequency components outside of its bandwidth. By contrast, filter "skirts" may sharpen at cold temperature.

Communication system thermal aspects
When compression characteristics of amplifiers are affected negatively, this means that more distortion will occur for a constant signal level. For broadband systems, the result is that many carriers sharing an amplifier will be more likely to disturb one another, resulting in a higher intermodulation distortion and a degraded signal-to-interference ratio (S/I). The error rate performance of systems is obviously related to available signal-to-noise ratio (SNR). However, broadband systems, particularly in high SNR systems, also can be degraded by S/I. For a single signal, the distortion aspect is more likely to manifest as limiting amplification, resulting in time domain squaring up of the signal. More importantly, it creates higher spectral regrowth characteristics in filtered quadrature phase-shift keying (QPSK), such as the kind implemented in wireless systems and used to avoid adjacent channel interference. With regard to gain variation, automatic level control (ALC) will likely be somewhere in the link receive side. Each dB of variation affects the range this circuit must operate over and possibly the noise figure at the receiver. Without ALC, a similar loss of SNR can occur due to an underdriven A/D converter.

Considering the oscillator circuits, crystal oscillators that move tax the ability of the receiver to acquire and track the transmit signal. While this sounds simple, the synchronization aspects are often anything but uncomplicated when the various noise contributors that further aggravate the process are considered.2 A further complication exists when one of the pieces, as in an aircraft, is moving relative to another. This creates an additional phase-locked loop (PLL) dynamic, requiring a more sophisticated PLL in order to provide the necessary tracking.3 The situation is not generally as troublesome for the RF VCOs, although they vary much more in frequency. It is usually the case in coherent digital communications that these oscillators are stabilized using PLLs with reference crystal oscillators. In such a case, the only requirement is to assure that they do not drift beyond the PLL-lock range, which is not difficult to achieve for most designs. In noncoherent systems, such as frequency shift keying (FSK), or in differential signaling, the stability of oscillators may be important in a different way, requiring more careful consideration of any accumulation of frequency instability along the link.

Bandpass filters react to temperature changes similarly to oscillators. That is, common behavior for lumped element filters, and even surface acoustic wave (SAW) filters, is to increase their center frequency at cold temperatures, and decrease it at high frequencies. Also, cold temperatures result in sharper filter skirts, while hot temperatures cause slower roll-off and rejection characteristics. Lumped-element RF filters are less predictable from design to design, as they rely on multiple-piece parts of various tolerance and temperature coefficient variations. Mechanical-type structures, such as the SAW filter, provide more fixed temperature coefficients. The shifting of center frequency is usually minor and designed to have insignificant impact. However, it can cause increased levels of interfering signals on one side of the offending filter skirts, as well as possible bandlimiting distortion effects on the desired signal on the other. Hot temperature degradation can similarly reduce rejection, increasing interference levels. By contrast, cold temperature can move the center, but sharpen filter characterisitcs. Again, the center frequency shift is generally small. While it may intuitively sound beneficial to sharpen filter roll-off behavior, the unfortunate trade-off of increased rejection is poorer group delay performance (nonlinear phase response) as the band edges are approached. This can degrade performance of single carrier modulation systems and, in broadband systems sharing an available bandwidth, result in poor performing and possibly unusable narrow channels at the edges of the passband. The effect of component Q also shows up in the rejection behavior of low-pass filters, with the magnitude of the effect again related to the relationship between the passband and the stopband edge.

What's shakin'?
It's a rare piece of equipment in the field that doesn't have to deal with some type of microphonic environment. It is easy to pin down various situations that do deal with a microphonic environment: aircraft, smart weapons, car stereos, and various electronics embedded inground in urban areas. Interesting in-between situations are also easy to conjure up - for example, a midsummer hailstorm pounding away on aerial cable television (CATV) hardware (you shouldn't be watching the tube or working at your computer anyway if it's a thunderstorm). A portable radio carried by a soldier may be still when in use most of its adult life. But, in the heat of battle, it may be subject to some wild variations of banging and shaking, and it may be more important to operate at that time than any other. When finally outfitted with my first cable modem, it will be placed in a pretty friendly atmosphere. But, it will be interesting to see how performance stands up to my four-year old using a nearby family room wall as his hockey net.

Hardware vibe effects
A crystal is piezoelectric, meaning it converts mechanical displacement into electrical signals. When a crystal undergoes a constant vibration, the frequency content of the vibrations shows up in a crystal-oscillator output spectrum as either a continuous noise spectrum or discrete components associated with periodicity in the vibration. The response of a crystal to vibration is center frequency and vibration frequency dependent. Vibrations themselves are typically a low-frequency mechanism. The output characteristics of the oscillators under these conditions are very well understood, so much so that a sideband "slide-rule" calculator is available from one of the manufacturers. In contrast to steady shaking, impact-type shock events produce phase transients in crystals and VCOs, and resettling of any PLL involved with transmit and tracking will be of concern.

VCOs also exhibit microphonic behavior, especially the mechanical resonator type. Once again, the frequency content of the shaking is crucial to know. A PLL provides a high-pass net response on a VCO, in contrast to the low-pass filtering associated with its operation on the input-reference crystal oscillator. Other hardware may convert mechanical energy into electrical disturbances. Here, we shall concentrate on the troublesome carrier synthesis aspects.

Communication system vibe aspects
For coherent digital communications, mechanical events can cause phase glitch transients or distributed phase noise that will be manifested as errors in transmission. There are several key elements that need to be understood in order to see the whole picture. The first is understanding the differences in steady-state vibrational effects and impact-type shock events. The second is consideration of our old friend, forward error correction (FEC).4

Error bursts during transient shock events are common, and mitigating such effects can occur in systems employing robust interleaving and burst-correcting error correction. Implementations using these powerful techniques are now more common in even modestly demanding systems, because of the increasingly straightforward implementation capability and the push towards maximum achievable capacity.

The vibration frequency content is critical to evaluation of its performance effects, since the audio-type spectral range also coincides with the type of bandwidths associated with PLLs, which may track such variation, attenuating its effect. Typical vibration envelopes of planes, boats, cars, trains, rockets, weapons, etc., are well-characterized. The content has many effects that must be considered. First, the PLL's ability to acquire lock can be impacted by sufficiently high discrete spurious from any source. Secondly, once acquired, holding lock without cycle slips that require re-acquisition is a function of the SNR within the PLL, which is in part a function of vibration-induced noise from any pieces affected. Finally, the steady-state, operating synchronizing system could still impose some vibrational energy as untracked phase jitter at detection, degrading bit-error rate (BER). Each of these issues can be complex.2

Microphonic situations contribute to some of my favorite "war stories." Maybe someone will benefit from some of these scars. I've seen VCOs in satellite systems disturbed while in operation when it turned out (to the surprise of many) that discrete sidebands would appear on the RF downlinks when the satellite was shifting its spatial orientation. The very low-level vibration of motorized wheels used to induce movement was picked up by very high-Q mechanical resonators. We later found out that our shaker table, designed to simulate much worse, wasn't even capable of reliably imposing such a low level of vibrational energy. In another interesting case in the satellite world, in my last years in that industry (the early 1990's), the new thing was that various nervous satellite buyers wanted more information about the condition of their purchase from the get go. This resulted in some pretty brutal, newly coined "operate-on-ascent" requirements. Part of the solution was to recognize that SC-cut crystals, as opposed to the typical AT-cut type, have performance in this area one order of magnitude better (and price performance two orders of magnitude worse). It turned out that a bipolar phase-shift keying (BPSK) link at 1E-4 was possible, uncorrected. Other application experience suggests that in places with permanent high-dynamics, FSK gets the job done quite nicely, a logical conclusion given the incoherent nature of typical detection. Finally, a nasty trap to be wary of is to recognize that your crystal oscillator will be on a circuit board, in a module, in a box, and in some application environment. The only way to assure that sideband noise will be within desired limits is to apply the vibrational envelope of the application on the box, and calculate, via mechanical moments, what will be going on where the circuit is attached. While layers of mechanical interfaces dampen the unit from shock events, the opposite may occur for vibrational impairments.

All-weather networking
There are all kinds of fascinating things to think about when you start talking environments. When speaking of the space applications discussed here, for example, there are also some interesting insights into particle physics. On second thought, let's just leave it at insights. Back on earth, the basic outdoor commercial electronics dilemma again boils down to the difference between January in Minneapolis, MN and July in Phoenix, AZ. With all this El Niýo business, lets just hope it doesn't add to the pile of environmental criteria already imposed on hardware designers. We already seem to be designing systems dedicated to providing superior digital-quality video, CD-quality sound, and a 600-kbps minimum during any Category IV hurricane or prolonged nuclear winter. One thing I've noticed during about 12 years doing this stuff is that environmental specifications are like taxes - they are never reduced. Despite this, it is true that it is impossible to prepare a good communication systems design without a good understanding of the environment of its deployment. Some of the issues discussed here have a habit of becoming unsuspectingly troublesome during field equipment trials.

Robert Howald is a staff engineer in the transmission network systems group at Next Level Systems (formerly General Instrument's communications division) in Hatboro, PA. He has a BSEE and an MSEE from Villanova University, and is currently a PhD candidate at Drexel University.

References

1. Howald, R., "Op Amps Provide Flexible Active Filter Design," Microwaves & RF, July 1995.
2. Howald, R., "Scary Synchronizers," Building Blocks, Communication Systems Design, October 1997, pp. 20-26.
3. Howald, R., "Give Thanks for Phase-Locked Loops," Building Blocks, Communication Systems Design, November 1997, pp. 14-21.
4. Howald, R., "Springing Forward to High Performance," Building Blocks, Communication Systems Design, March 1997, pp. 14-18.

---