Selection Criteria for Rubidium Frequency Standards

W.J. Riley Hamilton Technical Services Beaufort, SC 29907 USA [email protected]

• Abstract fications are nevertheless dominated by factors that limit their frequency stability. Those fac- This paper covers the criteria for selecting a ru- tors, associated mainly with internal noise, envi- bidium frequency standard (RFS) or closely-re- ronmental sensitivities, and temporal variations, lated CPT and CSAC gas cell atomic clock de- along with practical considerations regarding 1 vices that use either Rb or Cs as their working their manufacture, are the subject of this paper . element. There are several classes of RFS designed for commercial, military and space appli- While there are many references that describe cations, each with corresponding performance the physics and design of rubidium frequency features and limitations, and the specifications standards and related devices (see Reference for an RFS mainly cover the latter. The primary [1], there are relatively few papers and little function of any RFS is to supply a stable fre- technical literature that describe their selection quency regardless of its operating conditions, criteria (see References [2], [3], [12] and [13]). which include a number of common and application-specific considerations. The paper pro- • Frequency Standards, Oscillators, vides a comprehensive list of those items and and Clocks describes how they relate to a particular RFS usage. It includes examples of specifications for In most contexts, little distinction is paid be- current commercial, military, and space RFS tween the terms frequency standard, oscillator products from several leading manufacturers. and clock. Rubidium frequency standard is the generally-preferred term; rubidium oscillator is • Introduction widely used but strictly-speaking incorrect because an RFS is a form of passive atomic fre- The primary purpose of a rubidium atomic frequency standard, not an active oscillator. The quency standard (RFS, RAFS or AFS) is to al- word clock implies that the RFS is used, along ways produce the same frequency. While its with counting hardware, to keep time, but again, product description may emphasize its special it is widely used as a shorter term. features, the specifications of a typical RFS are mainly a list of its limitations because there are RFS units employed as frequency references many mechanisms that can cause it to depart emphasize absolute frequency calibration, and somewhat from the objective of producing a may include a means such as GPS disciplining. perfectly invariant frequency. And, like any RFS units intended only as a clock source may real active electronic device, it has finite size have relaxed requirements on the phase noise and weight, consumes power, costs money, and and purity. RFS units used as RF sources often will eventually stop working. have more demanding requirements for their

An RFS design can offer distinctive features such as an adjustable output frequency, multiple 1 We use “RFS” herein to include CPT and CSAC atomic outputs, GPS syntonization, etc., but their speci- frequency standard devices that utilize either Cs or Rb gas cells. 1 phase noise and spectral purity, which is mainly ness, and space units emphasize performance a function of their internal crystal oscillators. and reliability.

RFS units intended for dynamic environments RFS products can also be divided into standard that include vibration and rapid temperature and custom designs, where the latter are associ- change can benefit from a low g sensitivity crys- ated with a particular set of requirements for a tal oscillator and a fast frequency lock servo. specific application.

• Types of Rubidium Frequency Stan- Another distinction can be made between modu- dards lar and more complete RFS products, where the latter are complete instruments with such features as enclosures, mains power, metering and Most atomic clocks utilize the ground state hy- monitoring, frequency adjustment, multiple out- perfine resonance of an alkali element, either ru- puts and frequencies, frequency synthesis, low bidium or cesium, as their frequency reference. phase noise, clock circuitry, GPS disciplining, The classic RFS uses an Rb lamp for intensity etc. optical pumping of either an Rb-87 absorption cell in a microwave cavity along with a discrete In all cases, the quality of the crystal oscillator Rb-85 isotopic filter cell, or an integrated reso- and the bandwidth of the servo that locks it to nance cell containing both Rb isotopes. More the atomic reference have an important impact recently, a somewhat different arrangement em- on the RFS performance, particularly with re- ploying coherent population trapping (CPT) has gard to its phase noise. become an attractive alternative because it uses a modulated semiconductor laser diode to avoid the need for the filter cell and microwave cavity. • Stability Terminology CPT is also the basis of chip-scale atomic clocks (CSAC). The CPT devices can use either Cs or Let us first clarify some of the terminology used Rb as their working element. to describe frequency stability (or perhaps more correctly, its instability). As a general statement, one can associate the conventional RFS with a broad range of size, Fractional Frequency Deviation performance and cost, including the highest-stability GPS satellite clocks (see Appendix A), Ignoring adjustability, an RFS will have a cer- while CPT devices have medium-level perfor- tain nominal output frequency, f0, and an actual mance in small, lower-cost products, and CSAC output frequency, f, as measured with adequate devices offer much smaller and lower power but precision against a suitable absolute standard via somewhat lower performance. a traceable process. The output frequency is generally denoted by its fractional frequency This paper, while emphasizing the classic RFS, deviation, Δf/f0 = (f – f0) / f0. For example, for also applies to newer CPT and CSAC devices. f0 = 10 MHz, if the actual frequency is f = In fact, most of this material also applies to oth- 10.000001 MHz, the frequency deviation is Δf er atomic clock technologies such as trapped Hg = 1 Hz, and the fractional frequency offset is 1 ion devices such as the deep space atomic clock ppm, denoted as 1x10-6, 1pp106, or 1e-6, which (DSAC). are dimensionless quantities (not Hz) and generally written simply as Δf/f and denoted as a Distinctions can also be made between RFS function of time by the symbol y(t). This termi- units tailored for either commercial, military or nology is used for most stability specifications, space applications. While no sharp lines can be e.g., a temperature coefficient of 1x10-12/C. drawn between these classes, commercial units Values like that are used to quantify most RFS emphasize cost, MIL units emphasize rugged- stability parameters and environmental sensitivi- 2 ties. Note that the corresponding phase value as RFS frequency is set to within an initial accura- a function of time is denoted as x(t) and has cy specification at the factory before shipment. units of seconds. But the RFS frequency is also a temporal process because it changes over time, an effect Phase and Frequency Noise called aging when due to internal mechanisms and drift as observed from all causes. Frequency is the rate of change of phase, and that relationship describes the connection be- Frequency Measurements tween phase and frequency noise. Phase noise is usually expressed in the frequency domain as Frequency measurements are always made with a spectral density while frequency noise is ex- respect to some reference source. Frequency ac- pressed in the time domain as a statistical quan- curacy is the degree to which it equals its nomi- tity. The most common measure of phase noise nal value, while precision denotes the degree to is the logarithmic form of the power spectral which the measurements consistently resolve density (PSD) of the phase fluctuations, the SSB the frequency variations. The measurement sys- phase noise, £(f), in units of dBc/Hz. It is gen- tem determines that resolution and its scatter is erally expressed as a plot versus log sideband determined by the combined noise of the source, frequency in Hz.2 See Reference [7] for more reference and system. Thus it is vital that any information about these quantities and how they such frequency measurements made to verify appear in RFS specifications. RFS performance are supported by an appropriate system and reference, and that the RFS un- Stability Statistics der test have proper operating conditions (e.g., thermovac for space clocks). Phase measure- RFS frequency stability is actually a stochastic ments are preferable. Reference [7] shows some (noise) process, and the size of the frequency commonly-used measuring systems for preci- fluctuations vary as a function of their observa- sion frequency sources. An investment in a tion or averaging time as characterized by cer- suitable clock measuring system mat not be fea- tain statistics such as the Allan deviation3, sible for an RFS user, who will then have to de-

ADEV, y(), which is usually shown on a log- pend on data provided by the clock vendor. log plot versus averaging time, . The slope of an ADEV plot indicates the type of power law • Time noise; it generally improves with longer averaging time until reaching a minimum (the “flicker Although an RFS is often called a clock and is floor”) after which it increases because of drift used for timekeeping, it has no intrinsic sense of and environmental sensitivity. Examples of time, nor does it necessarily have clock hard- RFS ADEV plots are show in the Appendices. ware. Some RFS units do, however, include The ADEV is usually calculated after removal settable counting circuits, 1 pps outputs, and of frequency drift (which is expressed separate- other clock features, and their specifications ly). Information about the statistics used for fre- should describe these in detail. After initial syn- quency stability analysis will be found in Refer- chronization to a specified accuracy, the subse- ence [7]. quent time error is a function of the unit’s fre- Initial Accuracy and Temporal Stability quency offset and stability, and the elapsed time. 2 Do not be unduly alarmed if the phase noise plot shows discrete components related to the power mains frequency since those are most often due to the measuring system. 3 The Hadamard deviation, HDEV, is also used to charac- terize RFS stability because it ignores linear frequency drift. The Time deviation, TDEV, is another statistic that estimates the time error of the clock caused by its noise. 3 • Frequency Offset example, after 1 month (30 days) the quadratic term becomes 38.88 s, greater than the 25.92 RFS frequency offset is generally stated as the s caused by the above frequency offset. fractional frequency difference between it and a suitable reference (e.g., a primary frequency RFS aging is generally specified on a per day standard such as a cesium beam tube device or a basis after a certain period of undisturbed opera- GPS-disciplined source). The RFS will normal- tion, with equivalently lower values expected ly have some means for performing a frequency per month, per year and “forever” because the calibration (i.e., syntonization) by a mechanical, aging of an RFS typically diminishes as it ma- electrical or digital adjustment. That adjustment tures. For a very well-behaved unit with one will not materially affect its subsequent stabili- dominant aging mechanism, it is possible to suc- ty, and may need to be repeated occasionally to cessfully model its aging by log or diffusion law compensate for subsequent RFS drift. Any fits. In general, however, such modeling is residual frequency offset not only affects the problematic. A strategy for dealing with aging RFS accuracy but integrates to a time error is an important consideration in any RFS appli- when the unit is used as a clock: cation, and may require system-level means or periodic recalibration. ΔT = Δf/f  t • Frequency Drift For example, a frequency offset of 1x10-11 would cause a time error of 0.864 s over 1 day RFS frequency drift is the combination of both (86,400 s). internal and external effects that cause it to change over time. In a well-controlled environ- A change in environmental conditions (e.g., ment, aging and drift are equivalent. In fact, the temperature) is likely to change the frequency term “frequency drift” is commonly used to de- offset, and set a practical limit on syntonization note either. effectiveness. • Frequency Adjustment • Frequency Aging The RFS frequency can be adjusted over a range The frequency of an RFS is not perfectly con- of a few pp109 by means of the internal physics stant, even ignoring environmental and other ex- package DC magnetic bias field (“C-field”), or, ternal effects, but rather displays frequency in many modern designs, by setting an internal change due to internal mechanisms associated synthesizer. The C-field adjustment can be via with both its Rb physics package and (to a lesser an RFS control potentiometer and/or an external extent) its electronics. tuning voltage. The latter can be used in an analog control loop to discipline the RFS frequency Linear frequency aging causes an initially per- with respect to an external reference (e.g., GPS). fectly syntonized rubidium clock to experience a That can also be accomplished to within a speci- time error that integrates quadratically as a func- fied resolution via the RFS synthesizer and a tion of time: digital control interface. In cases where the host system has means for making frequency adjust- ΔT = ½ [(Δf/f)/Δt]  t2 ments (e.g., some GPS satellite payloads), it is desirable to operate the RFS without tuning at a For example, if the RFS frequency changes lin- fixed minimum C-field for best stability4. early by 1x10-12/day (1.157x10-17/s), that causes a time error of 43.2 ns over 1 day. In the long run this term will dominate the clock error. For 4 Low C-field reduces RFS sensitivity to both internal and external effects but limits its tuning range. 4 The RFS specifications should include informa- Most commercial RFS applications are satisfied tion about the analog (C-field) tuning adjust- by the vendor’s catalog environmental specifi- ment, if any (range, sensitivity, etc.) and/or digi- cations, but special cases may need to be negoti- tal tuning interface (protocol, resolution, etc.). ated. Military RFS applications often impose a wider range of environmental conditions and/or • Environmental Sensitivity platform-specific requirements. Space RFS applications almost always have program-specific There are quite a few environmental sensitivities needs (e.g., launch vibration, radiation) for that apply to an RFS (e.g., temperature, baro- which the RFS design may need tailoring. metric pressure, etc.) and these are discussed separately below. Information about their phys- • Noise ical basis will be found in Reference [6]. Each application determines those factors that are im- All frequency sources have finite signal-to-noise portant, and an RFS for a military vehicle can (S/N) ratios and noise-induced phase and fre- require a very different design than for a labora- quency fluctuations that are a fundamental limi- tory standard. For example, in a benign lab, tation to their stability. This is commonly re- barometric sensitivity can be a dominate consid- ferred to as short-term stability (STS), although eration where is doesn’t matter for a space clock its effects can extend out to days. STS is a time- in vacuum. Fortunately there is little interaction domain measure that is closely related to phase between most RFS environmental specifica- noise in the frequency domain. These noise tions, and most are quite consistent for a given mechanisms are stochastic processes that are design. That being said, temperature sensitivity specified with statistical quantities such as the is often the largest consideration and exhibits Allan deviation (ADEV). See Reference [7] for significant unit-to-unit variation, probably be- more information about frequency stability, cause it can have multiple contributors of vary- phase noise, power-law noise processes and fre- ing sign and magnitude. quency stability analysis

RFS environmental sensitivity is often defined • Crystal Oscillator in relation to MIL-spec conditions (e.g., those of MIL-STD-202 and MIL-STD-810). See Refer- All5 RFS have an internal crystal oscillator to ence [5] for more details. It is also covered by generate their active output that is locked to the IEEE Standard 1193, IEEE Guide for Measure- Rb atomic resonance by a frequency lock servo, ment of Environmental Sensitivities of Standard The crystal oscillator generally operates at the Frequency Generators [10]. Similar informa- RFS output frequency, and it largely determines tion for crystal oscillators can be found in MIL- its phase noise and sub-second short term stabil- PRF-55310, “Oscillator, crystal controlled, gen- ity. It also influences RFS behavior under vi- eral specification for” [11]. bration. It is therefore important to understand the characteristics of the RFS crystal oscillator It is the responsibility of the RFS vendor to and its influence on the overall unit [9]. For ex- choose a set of environmental specifications ap- ample, the phase noise of the crystal oscillator propriate to the intended product target, or to not only affects the RFS output purity, but it is obtain them from their customer, and to then also an important factor in the quality of the verify and document compliance with those re- physics package microwave excitation. quirements. It is the responsibility of the RFS user to verify that the vendor-supplied set of en- The crystal oscillator may be ovenized or tem- vironmental specification is appropriate to their perature compensated, or may simply operate at application, or to supply them to the RFS vendor, and to confirm satisfactory compliance with 5 I don’t know of any exceptions for an RFS product. An them. RbXO has an external crystal oscillator. 5 the temperature set by a baseplate temperature • Output Waveform and Power controller (BTC, see below). Most RFS units have a sinusoidal RF output sig- The RbXO [8] is a form of RFS where the Rb nal, most commonly at 10 MHz, with a power reference operates intermittently to syntonize level of the order of +7 dBm (5 mW, 0.5 V rms) the crystal oscillator (which has a tuning non- into a 50 ohm load. This is adequate to drive volatile memory), thereby saving energy. most host equipment without compromising stability. Multiple outputs are sometimes provid- Some low phase noise (“LN”) RFS designs have ed, and higher output power can be an advan- a separate phase-locked (usually ovenized) crys- tage if it is to be passively split for multiple tal oscillator that provides the output. Clean-up loads. In the case of multiple RF outputs, inter- OCXO products are also available for that pur- channel isolation may be important and should pose. be specified. Most RFS RF outputs have no DC component, can be either AC or DC coupled, While the design details related to the crystal os- and may be isolated from the RFS case. An iso- cillator are the responsibility of the RFS vendor, lated RF output can help to avoid ground loop the user must determine his system requirements interference. It can be important to know all for phase noise and spectral purity, remember- these (and other, see below) details. ing that subsequent frequency multiplication raises the noise and spurious levels and that vi- Some RFS units have a digital output waveform bration can have a profound effect on the signal such as a 5V p-p “TTL” squarewave or differen- purity (see Vibration). tial LVDS. In that case, the waveform should be fully-specified as to level, waveshape, duty • Choice of Output Frequency cycle, rise and fall times, etc. Such outputs are generally intended for embedded applications Most precision frequency references are distrib- where the RFS is connected directly to host sys- uted at 10 MHz which has proven to be a good tem logic. Sinusoidal RF output is preferred for general choice and has become the de-facto most frequency reference distribution. standard. Use of a non-standard frequency is normally limited to specialized use such as GPS • Harmonics satellite clocks and high-volume embedded applications. For example, some GPS satellite The harmonic content of an RFS RF output is clocks have used the 10.23 MHz pseudo-ran- generally specified as below -30 dBc or so, low dom code rate, and others have used a enough to avoid gross waveform distortion. 13.40134393 MHz “natural frequency” that That can be particularly important when driving minimizes RFS complexity. Binary (e.g., 2.048 a zero-crossing detector since an amplitude MHz) and telecom (e.g., 1.544 MHz) rates are change can cause a phase change. Better purity non uncommon. VHF outputs (e.g., 100 MHz) is seldom required but can be obtain with exter- are used occasionally for low phase noise or di- nal low-pass filtration if necessary. rect digital synthesizer (DDS) clocking applications. DDS in either the RFS or the host system provides great flexibility. Be aware that special • Spurious/Non-Harmonics output frequencies can also complicate RFS testing. Non-harmonic spurious components are present on the RF output of most RFS units and must be An RFS with a 1 PPS pulse output is fairly com- specified. A level of at least -80 dBc is typical. mon, and can be needed for driving clock cir- These unwanted components can occur either at cuitry and enabling GPS time and frequency re- baseband or as sidebands on the desired RF ceiver disciplining. component or its harmonics. Examples of the 6 former include spurious components from • Temperature switching power supplies and RF components from the Rb lamp exciter. Examples of the lat- The temperature environment is an important ter include RF output sidebands separated from consideration for any RFS, both during opera- the carrier by switching power supply and/or tion and under storage. An RFS will have speci- oven controller rates and at the frequency lock fications for its operational and storage tempera- servo modulation rate. Those spurious compo- ture limits. During operation, temperature sen- nents are at well-defined frequencies; it is also sitivity is likely to be the most significant factor possible that there are spurious components at affecting its frequency stability. unexpected frequencies (e.g., UHF parasitics). The upper operating temperature of an RFS is • RF Output Impedance usually determined by the point where the physics package oven with the lowest set-point While most RFS RF outputs are specified for a loses control since performance will deteriorate 50 ohm load, the RF output source impedance above that condition. Newer RFS designs with may or not be controlled or specified. It can be smaller cells generally operate at higher temper- important that the VSWR of the RFS RF output atures and thus may be an advantage. There is be reasonably low (say < 1.5:1) to avoid reflec- no specific determinant for RFS maximum stor- tions from a mismatched load, especially in a age temperature but, because of the possibility situation where it results in low or variable am- of Rb migration under long-term hot storage, a plitude at the load. An RFS RF output imped- recommended limit is imposed lower than that ance specification usually applies only at the where component damage occurs. carrier frequency. The lower operating temperature of an RFS is • Phase versus Temperature usually determined by the point where the physics package oven with the least demand Be aware that any lowpass or bandpass filter power margin goes out of control since perfor- (LPF or BPF) at the RFS RF output introduces mance will deteriorate below that condition. phase shift whose temperature coefficient re- The minimum storage temperature is likely set sults in a pseudo frequency shift when exposed to an arbitrary -55C. to temperature slew. This is especially important with a narrowband BPF such as a crystal There is little actual experience with RFS units filter. Long coaxial cable lines can also show undergoing long-term storage at either tempera- this effect. ture extreme, and the worst situation is probably the thermal stress caused by rapid RFS warm-up • Humidity/Salt Fog/Immersion from cold storage. The operating temperature is usually measured Humidity, especially that resulting in condensa- at the RFS case, in particular its baseplate. tion, can affect certain RFS circuits. Exposure That, in turn, is determined by either its conduc- to salt fog and other ionic contamination can re- tive mounting surface or the ambient air. Con- sult in severe corrosion. It can be an advantage, ductive heat transfer is mandatory for operation or even essential, that RFS circuit boards are in vacuum. In air, without hard mounting, a conformally coated and/or the RFS case be convective heat sink is usually required to main- sealed against such exposure. A sealed unit tain a reasonable temperature Radiative heat may require a means for pressure relief and/or a transfer is normally not a significant factor. Ob- strong, heavy case. It is rare that an RFS is taining proper heat transfer is particularly im- specified for actual immersion. portant for a unit mounted on a vibration isolator. The use of a baseplate temperature con- 7 troller to reduce overall RFS temperature sensi- A sealed RFS eliminates the effect of baromet- tivity is discussed below. ric pressure change.

RFS temperature sensitivity is usually expressed • Altitude as a static or steady-state value, and is not necessarily linear, and can display significant unit- Altitude exposes an RFS to a combination of to-unit variation, particularly because it is gen- barometric pressure and temperature change, erally the result of multiple internal mechanisms and is generally no different in its effect than having varying signs, magnitudes and time con- those factors individually. stants. In a dynamic environment, an RFS can be exposed to temperature slew and/or shock, and the resulting transient response is not neces- • Supply Voltage sarily the same as that for slower changes. Some hysteresis and lack of retrace is also pos- An RFS generally requires a DC supply voltage sible. rather than being designed to operate from AC mains power, and thus is specified to accept a All those factors make RFS temperature coeffi- certain DC voltage range and polarity, and to re- cient modeling and compensation problematic, quire a certain maximum demand current during but it can nevertheless be effective, especially warm-up and a steady state current that is a for steady-state sensitivity, and is increasingly function of temperature. The variation in power used with RFS units having digital frequency consumption is mainly due to its physics pack- adjustment and microprocessor control. age oven(s). Reverse polarity protection may or may not be • Retrace included within the RFS and, if not, may be needed from the host system. RFS units generally have excellent frequency retrace when they are turned off and on again DC isolation may be required between the RFS under the same environmental conditions. The DC return lead and its case ground, in which duration of the off period is not much of a fac- case a DC/DC converter will likely be needed tor. RFS retrace frequency error is normally either inside or external to the RFS itself. In the non-accumulative. latter instance, consideration must be paid to how the RFS case is electrically insulated while • Barometric Pressure still providing adequate thermal transfer.

Barometric pressure is an often-neglected but Other supply voltage considerations include the potentially significant RFS sensitivity factor, RFS frequency sensitivity to supply voltage, and particularly for units with large absorption cells. to noise, transients and ripple on the supply line. It is mainly caused by drumhead deflection of Conversely, a DC/DC converter inside the RFS, the glass cell, and perhaps because of thermal while providing isolation and/or a wider input effects. voltage range, can cause conducted interference (see EMI below). RFS operation under vacuum or in space is not a problem for units intended for that environment, RFS voltage sensitivity is a combination of elec- but otherwise it can cause internal overheating trical and thermal effects, and is generally small for components that are not conductively heat enough so as not to be a major consideration. sunk.

8 • Power Consumption tion is expected, thermal cycling stress can be an important consideration and may require de- The power consumption of an RFS depends sign verification testing to assure adequate en- mainly on its type. Conventional units and the durance. core RFS of a space unit require around 12-15 watts steady-state at room temperature, CPT • Thermal Interface units 1-2 watts and CSAC units 0.1-0.2 watts. The demand power during oven warm-up is Consideration must be given to the thermal in- considerably higher and determines the required terface between an RFS and the host system that power supply current rating. Power varies with provides a heat sink or other means for dissipat- ambient temperature, especially for a space unit ing the RFS power. Common methods include with a baseplate temperature controller (BTC). a finned heat sink, perhaps with a fan to dissi- pate heat via convection or simply a large • Dynamic Input Impedance enough thermal mass to carry away the heat conductively. Ideally, the RFS would be at- An RFS generally contains filtration on its DC tached to a constant-temperature baseplate. Ra- input supply line, and that can have significant diative heat transfer is seldom a factor, even impact on the impedance that the RFS presents with RFS operation in vacuum. to its DC source as a function of (audio) frequency. High Q resonances can pose a particu- • Baseplate Temperature Control lar problem when used with a host DC/DC converter. Baseplate temperature control is an excellent way to improve RFS temperature stability at the • Warm-up Time expense of additional size, weight, power and complexity. In its simplest form, it can be a Depending on the application, RFS warm-up convective heat sink with a temperature-con- time can be an important consideration. RFS trolled fan. More elaborate arrangements using warm-up is to some extent a design tradeoff be- temperature-controlled heaters have been suc- tween speed versus increased demand power cessfully employed for GPS satellite clocks. and thermal stress, and lockup time can vary from less than 10 seconds to a few minutes. Be- Use of a baseplate temperature controller re- sides physics package and crystal oscillator quires a thermal insulator between the BTC and oven warm-up, a conventional lamp-pumped its mounting surface. Its operating range extends RFS requires time for lamp starting. Additional from the lowest mounting surface temperature time is also needed for frequency stabilization. where it has sufficient demand power to the highest mounting surface temperature where the The RFS RF output is generally present imme- temperature rise from the RFS dissipation reach- diately upon application of DC power, and that es the BTC set point. can pose a problem since it will not be locked to the Rb atomic resonance. In fact, it often • Reliability sweeps back and forth with a large offset until the physics package and crystal oscillator warm An RFS reliability analysis generally follows up. All RFS units have a monitor signal that in- the procedures for electronic equipment pre- dicates lock, and that may have to be utilized by scribed in either MIL-HDBK-217 or the host system. Bellcore/Telecordia SR-332. It is important to use the applicable environment (e.g., Ground In an application where frequent power on-off, Fixed at +40C) for the analysis. The resulting warm-up and large external temperature varia- failure rate is associated mainly with the elec-

9 tronic circuits. The physics package, with the tal interfaces and user software to monitor their possible exception of the Rb lamp6, has been operation. generally found to be very reliable (mostly mechanical, the associated electronic parts (photo- • Storage diode, thermistors, SRD diode, etc.) considered separately). Cell failures are very unusual, prac- There are no significant considerations regard- tically nonexistent, foil oven heaters likewise. ing the storage of an RFS. One should, of Some concern exists for the semiconductor laser course, abide with the unit’s storage temperature diodes used in some conventional and all limits. CPT/CSAC RFS units.

The most important part of an RFS reliability • Aging and Drift assessment may be close examination of the thermal and electrical parts stress that is a pre- Internal aging within the Rb physics package 8 requisite to performing the reliability analysis. causes its frequency to change . That change, along with other effects associated with its oper- Today’s commercial electronic parts, including ating conditions, results in frequency drift over plastic encapsulated microcircuits have achieved time. The aging rate tends to diminish over an outstanding level of reliability. Some mili- time, and, after initial stabilization, for classic -11 tary and most space RFS products further im- units, is typically on the order of 1x10 /day, prove on that with enhanced parts selection and 1x10-10/month and 1x10-9/year (and not much screening. Those activities are generally pro- more than that “forever”). Day-to-day, howev- gram-specific and associated with any required er, environ-mental disturbances are likely to radiation hardening. dominate the aging.

In practice, it is found that most RFS failures are RFS frequency drift can be the largest contribu- associated with bad workmanship. Factory tem- tor toward clock error over time. It can also re- perature cycling and burn-in (Environmental quire that the RFS frequency be adjusted to re- Stress Screening, ESS7) has proven to be impor- store it to agreement with an absolute reference. tant for weeding out such defects. RFS aging can be caused by several mecha- The life of an RFS is typically limited by obso- nisms of varying sign, magnitude and consisten- lesce rather than any specific life-limiting fac- cy, and therefore does not follow any set pattern tor. RFS products are usually covered by a lim- or behavior. It can even flatten, reverse direc- ited warranty than protects against early fail- tion and increase again. This makes its model- ures. ing and removal problematic. That said, some of the highest-performing space RFS units have • Monitors shown highly-modelable aging that fits a log or diffusion (t) model. All RFS units have an indicator that shows that its crystal oscillator is locked to the Rb atomic • Helium Exposure resonance. Most units also have either analog or digital monitor signals that show such param- Atmospheric helium is, to an extent, a “poison eters as the Rb lamp light level and the XO con- gas” for an RFS because it can permeate the ab- trol voltage. Modern RFS units often have digi- sorption cell glass envelope and cause a fre-

6 Rb lamp life problems have been resolved by lamp de- 8 Several factors can contribute to RFS aging including sign and processing improvements. changes in the Rb lamp intensity/spectrum, 7 ESS is a recipe for applying temperature cycling and chemical/physical effects in the Rb absorption cell, and mild vibration to a unit to precipitate latent defects. changes in the physics package oven temperatures. 10 quency change9. This is not a big issue unless There are no unique vibration vulnerabilities as- the unit is stored or operated in a He-rich envi- sociated with the Rb physics package, but it is ronment for a long time. important that the cells and optical elements be ruggedly attached. • Magnetic Field Performance Under Vibration DC magnetic field sensitivity can be an important RFS factor, but most units have sufficient The importance of RFS vibration sensitivity de- magnetic shielding that this is not a major con- pends very much on its operating environment cern. This sensitivity is a direct result of the and the need for spectral purity. In general, an properties of the Rb hyperfine resonance, and is RFS is less affected by vibration than an ordi- strongest along the axis of the physics package nary crystal oscillator, whose quartz crystal res- internal DC magnetic field. Several techniques onator has a typical acceleration sensitivity of -9 have been developed (e.g., C-field commuta- 1x10 /g and therefore can exhibit strong dis- tion10, Zeeman sensing11) to reduce RFS mag- crete PM sidebands under sinusoidal vibration netic field sensitivity, but passive magnetic or suffer significant phase noise degradation shielding12 is generally adequate for most appli- from random vibration. cations. But an RFS also has a crystal oscillator (XO) that makes it susceptible to the same mechanism • Vibration in a more complicated way. For vibration frequencies well above the bandwidth of the RFS Vibration can have a profound effect on the sta- frequency lock servo, the effect is essentially the bility and even the functioning of an RFS, large- same. Well below that frequency, the servo acts ly because of its quartz crystal oscillator. to eliminate the effect (assuming that the Rb physics package itself is immune). The Rb Survival Under Vibration physics package, if ruggedly constructed, indeed has little vibration sensitivity, mainly associated It is quite remarkable how well properly-de- with deflection of its light path. signed electronic assemblies can survive severe sinusoidal and random vibration. The key, of There’s more to the story however. The Rb ser- course, is “properly designed”. Without proper vo operates by applying deliberate FM to its mi- care regarding circuit board and other structural crowave excitation, using the resultant recov- resonances, proper bonding of components, con- ered AC signal, via synchronous detection, to sideration of fastener strength, and many other steer the XO. Thus the frequency lock loop is aspects of the mechanical design, structural and very sensitive to vibrational modulation at that fatigue failures will occur under the stresses en- particular rate (for the light beam) or twice the countered under rocket launch, aircraft flight modulation rate for the XO. and other such conditions. Correct specification and verification of the requirements is essential. Finally, at low vibration frequencies, the resulting XO modulation can even cause a reduction or loss of the physics package microwave exci- 9 He buffer gas in the absorption cell raises the RFS fre- tation. quency. In normal operation, it reaches equilibrium with the trace atmospheric He tension. Subsequent operation You can see that this is quite a complicated top- in a vacuum can result in negative aging. 10 Periodic C-field reversal reduces the sensitivity to ex- ic. If your RFS application involves significant ternal magnetic field variations to 2nd order. operational vibration, vendor consultation is ad- 11 Interrogation of the field dependent Zeeman hyperfine vised. Besides a rugged unit, a low g-sensitivity lines can measure and stabilize the internal C-field. XO may be required, and/or one with active or 12 It also provides radiation shielding. 11 passive vibration isolation or the entire RFS generally tailored to the particular launch vehi- may need to be mounted on a vibration isolator. cle.

Structural resonance issues have diminished as Shock testing can involve actual drops into a RFS units have become smaller. sandbox or hard benchtop, or shock waveforms applied by an electrodynamic shaker. Vibration requirements can apply either under operating or non-operating conditions, and the • Acceleration former is the more stringent since electrical power is applied and the RFS ovens are hot. Static acceleration is hardly ever an issue for an But there is generally little difference in the RFS and is seldom specified. That sensitivity of severity of those conditions, and, of course, op- the crystal oscillator is eliminated by its fre- eration is required to assess performance under quency lock servo. One possible RFS concern vibration. It is recommended that RFS vibration is movement of molten Rb in the lamp, but that testing be performed with the unit operating be- rarely observed. cause that can better expose problems such as intermittent connections. RFS constant acceleration testing (besides a simple tip-over) is generally done on a large, • Mechanical/Pyro Shock slow centrifuge under operating conditions, and slip ring noise can be problem for making fre- Mechanical shock survival is a fairly common quency measurements. requirement for all RFS applications. Exposure is generally while non-operating, but can be during operation, especially when associated with a • Acoustic Noise time error limit. Acoustic noise is seldom specified as a test con- Internally, the main items are concern include dition for an RFS. The effect is similar to that damage to quartz crystals and displacement of of random vibration but applied as sound waves physics package elements. Dislodgement of through the air. It simulated an environment heavy electronic components, radiation shields such as a rocket launch or jet aircraft engine. and the like are also concerns, as is actual break- Typically this test would be conducted with the age of mounting feet, etc. unit operating and frequency and/or phase noise would be measured. The classic military shock test was to simulate battleship naval gunfire. Other tests simulated • Other Environmental Tests railcar coupling. One seldom sees those requirements anymore. Explosive atmosphere is another item seldom specified as a test condition for an RFS, and Commercial and military RFS units may include would not be expected to pose a problem. a bench handling shock requirement that simu- lates the rough treatment it may encounter dur- Similarly, Rain, Sand and Dust and Fungus tests ing servicing. One hopes that space clocks are are rarely if ever required. handled more carefully!

Space RFS units usually include a pyroshock requirement that describes the units exposure to the pyrotechnics associated with rocket stage and payload separation. Those conditions are

12 • Thermal Cycling Fatigue controllers and those circuits are likely to produce discernable conducted emissions on its DC An RFS that is operated intermittently is subject power line. to thermal cycling fatigue, especially for a unit that has fast warm-up from a cold soak. This is An RFS is often permitted to have radiated arguably the worst operating stress that RFS emissions at its RF output frequency and har- units are routinely exposed to. Thermal cycling monics thereof, but this usually associated more endurance (not just on-off cycles) is an impor- with its connecting cable than the unit itself. An tant design verification test (DVT) requirement, RFS may also have discernable radiated emis- especially for military applications like tactical sions at frequencies associated with its Rb lamp aircraft. In mild form, it also can be the basis of exciter, and perhaps those involved with its RF environmental stress screening (ESS) applied on chain and physics package excitation. The con- an individual basis by the RFS manufacturer to necting cables can often have a significant influ- assure a reliable product. ence on the RFS radiated emissions since they act as antennas.

• Orientation An RFS is generally not susceptible to radiated fields. An RFS is, in principle, immune to static g forces and orientation since its crystal oscillator’s acceleration sensitivity and “tip-over” ef- • Nuclear Radiation fect due to its quartz crystal resonator is locked to the atomic resonance. In practice, there can Nuclear radiation hardening requirements apply be some RFS orientation sensitivity caused by to most space RFS usage, some military RFS thermal effects, but these are generally quite applications and a few commercial ones. When small. it is required, radiation hardening influences all aspects of RFS specification, design, manufacture and test. Expert knowledge is needed to as- • EMI sess the environment, define the threats, analyze and tailor the RFS design, devise overall and Most RFS applications impose electromagnetic spot shielding, select and screen its parts, plan interference (EMI) requirements of some sort. and conduct verification tests, and assure manu- These can involve radiated and/or conducted facturing compliance. susceptibility or emissions. EMI requirements are often tailored for a particular application. The radiation environment can be from natural or manmade (e.g., weapons or reactor) sources, An RFS has particular susceptibility to conduct- and it can include total gamma dose, neutron ed interference on its DC power input at audio fluence, proton and other heavy particles, along frequencies associated with its frequency lock with transient X-ray exposure, and single loop servo modulation rate and harmonics there- event/cosmic ray upsets. Each of these radiation of. Under particular (usually contrived) condi- types affects an RFS, mainly its electronic cir- tions, large frequency offsets and even unlock- cuits, in specific ways. ing are possible. One needs to set realistic values for the interfering ripple level, its exact fre- RFS response to total dose radiation involves quency and phase relation with the servo modu- frequency offset and stability degradation (and lation, and the amount of allowable frequency eventually gross failure). disturbance. RFS response to transient radiation involves po- An RFS is likely to employ power switching cir- tential device upset/latchup and either “operate cuits in its power supply and perhaps its oven

13 through” output phase continuity13 with a speci- surement interval. A short measurement inter- fied time error limit or functional recovery with- val provides better resolution of any such distur- in a specified time14. All RFS ports with cables bances. attached may need hardening against system generated EMP (SGEMP). An RFS uses a frequency lock loop (not a PLL) to lock its crystal oscillator to the atomic reso- Fortunately, RFS radiation hardening is well-un- nance. Many mechanisms associated with the derstood and can be successfully implemented Rb physics package and/or the frequency lock without unduly affecting its design or perfor- servo can cause a frequency fluctuation. Those mance. If fact, the detailed circuit analyses frequency changes can take the form of short needed provide valuable information for other transients or permanent offsets, and the resulting aspects of its design. Thus, when applicable, ra- phase change can integrate to a step or a ramp. diation hardening becomes just another item of One such mechanism is a sudden change in the RFS specification. intensity and/or spectrum of the Rb lamp which, via the light shift effect, can cause a frequency • Relativity jump. Generally speaking, only high-performance RFS units intended for demanding appli- Any clock is subject relativistic effects includ- cations (such as GPS satellite clocks) are ing time dilation and gravitational redshift. The screened for such problems because they require former applies to a rapidly moving clock while careful long-term observation. the latter applies to one well above ground level, and thus these effects are significant only for Phase changes at the RFS output can also occur space clocks. For example, GPS satellite clocks without an internal frequency change due to have a frequency that is about +4.4x10-10 higher such effects as crystal oscillator phase “pops”, in their half-synchronous (MEO) orbit. Thus a digital glitches and/or erratic connections. Par- GPS clock must be adjusted to a slightly lower ticular attention should be paid to a phase frequency while on the ground. That RFS spec- change equal to an RF carrier period. ification will depend on its tuning means, if any. • Firmware and Software • Phase/Frequency Jumps Classic analog RFS designs without internal Phase and/or frequency jumps can occur with digital processors do not require firmware, but any RFS, although these disturbances are not most modern implementations do, and while commonly a problem, and are not routinely that is mainly an internal design matter, it intro- specified or tested for. duces additional design review, documentation and configuration control issues. First of all, it is important to understand the relationship between these effects and how they can Externally, an RFS with a digital control and be observed. Frequency is the rate of change monitor interface requires software for its user (the derivative) of phase. Phase is the integral interface. That software is generally supplied of the instantaneous frequency. RFS stability and documented by the RFS vendor. In the case should preferably be measured as phase because of a space RFS, it usually interfaces with the that is the more fundamental parameter from spacecraft control and telemetry system, which which the frequency can be determined by cal- must then be part of the RFS specifications. culating phase differences divided by the mea-

13 A ringing filter at the RFS output can be used to main- tain its output during transient radiation. 14 Perhaps with time maintained by the host system 14 • Sources  Personnel/Expertise  Manufacturing Capacity We will not attempt to comprehensively list or  Support/Warranty compare sources for rubidium frequency stan-  Security Clearances dards and related devices, especially since such  Facilities and Equipment information depends on many factors and changes with time.15 The source of an RFS • Documentation/Data Items product can be either the actual RFS manufacturer or another organization that incorporates it Several types of documentation are associated into their product. Because of the rather wide with RFS specification, design, qualification, range of these products, one likely begins the manufacture, test, procurement and use. process of source selection by comparing the specifications available units against the re- RFS specifications can take the form of standard quirements. As usual, the manufacturer’s repu- product datasheets to custom program-specific tation is a major factor, as may be their national- source control drawings. ity and the potential need for export licenses, security clearances, etc. It may be necessary to RFS design documentation can take the form of examine the vendor’s facilities and verify the proprietary product information to formal con- experience of their personnel. It is relatively tractual data items. These include program ap- rare for even major RFS users to develop their proved parts lists, a drawing package, worst own in-house RFS designs. case circuit analyses, reliability/MTBF analyses, radiation hardening analyses, preliminary Most commercial RFS requirements are satis- (PDR), critical (CDR) and manufacturing readi- fied with standard catalog products, perhaps ness (MRR), software and firmware code re- backed up by a specific SCD. Military RFS re- views, an interface control drawing, qualifica- quirements may require customization of a com- tion and acceptance test plans and procedures, mercial design, or even a custom design to a product user, programming, application notes user SCD. But increasingly, because commer- and service manuals, etc. Certain data items cial electronic parts are so reliable, and most may need review and approval by multiple orga- RFS physics packages are quite rugged, MIL re- nizations. Formal product training may be requirements can be satisfied by commercial units quired in some cases. with little or no modification. Space RFS requirements may need a custom design unless an MIL specs and other such documents should not existing one can be utilized. Some MIL and be listed as applicable to an RFS specification most space RFS programs require extensive unless they are referred to and their specific documentation. conditions provided. If possible, use the same parameter values as given on the standard prod- Some of the considerations for RFS vendor se- uct data sheet so as to avoid the need for special lection include: analysis and test.

 Reputation/Past Relationship • Interface Control Drawing  Nationality/Location  Organizational/Financial Strength The RFS documentation should include an In-  Product Line/Experience terface Control Drawing (ICD) or equivalent 15 At the time of this writing, leading U.S. RFS sources that describes all aspects of the unit’s external are Excelitas, Frequency Electronics, Microsemi, and physical and electrical attributes, including, as Stanford Research Systems. In Europe, the Orolia (Spec- applicable, the following: tracom, Spectratime) supplies those devices, and Ac- cubeat does so in Israel. 15  Case dimensions • GPS/GNSS Disciplining  Weight  Mounting and heat sinking GPS/GNSS disciplining of an RFS is a common  Connector types, mates, pinouts, and func- application that produces a traceable absolute tions time and frequency reference at modest cost and  Power requirements complexity. The RFS frequency is steered to  Output frequency, waveshape and level bring it into phase and frequency alignment  Input signals and functions (synchronized and syntonized) with the  Partial schematics of I/O port circuits GPS/GNSS signal with a fairly long time con-  Monitor/BITE signals stant that optimizes the overall noise level. In  Control communications protocol the event of loss of the GPS/GNSS signal, the RFS provides excellent holdover stability, and that, along with the desired phase noise, deter- • Hazards mines the RFS requirements.

No hazards are uniquely associated with RFS Rb-based GPS disciplined oscillators (GPSDO) manufacture or usage. The operation of an Rb are available as complete units or can be quite atomic clock does not depend on radioactivity. easily implemented by an RFS having a 1pps in- These units do not employ high voltages and put and appropriate firmware. generally do not contain any hazardous chemi- cals. No special handing is needed, nor is any special documentation or labeling required. • Ensembling/Time Scales The Rb-87 isotope is very slightly radioactive, so slightly that its half-life is longer than the age An RFS is rarely used as part of a clock ensem- of the universe. We are fond of saying that an ble or elaborate time scale since higher-perform- RFS has less radioactivity than a bunch of ba- ing clock technologies (e.g., H-masers, Cs or Rb nanas16, and radioactivity can be excluded from fountains) are used in those applications. Nev- an RFS specification. ertheless, certain applications where very high reliability is needed use multiple RFS units in a phase coherent combiner. The main RFS re- • Smart Clock Techniques quirement for such an application would be high-resolution digital tuning. Artificial intelligence has not yet been widely applied to frequency references. Some earlier attempts at Modular Intelligent Frequency, Time • Distinguishing Traits and Time Interval (MIFTTI) and “Smart Clock” devices were explored but did not catch on. A An RFS is nearly always the best choice for a GPS disciplined clock that not only sets its time frequency standard that requires better perfor- and corrects its oscillator frequency but also mance than a crystal oscillator. At the border- learns its aging behavior certainly comes close line between their capabilities where they have to being a smart clock, particularly if one added similar costs, it is more practical to use the low the capability to separate out and learn environ- end of RFS technology rather than the upper end mental sensitivities like temperature and baro- for a frequency standard grade ovenized crystal metric pressure. oscillator.

A GPS/GNSS disciplined Rb oscillator is usually the device of choice for a local time/frequency standard, providing a traceable source of ab- 16 An RFS contains about a milligram of Rb-87 and its ra- solute time and frequency at modest cost and dioactivity is about 20 pCi, about 1/20th of a banana. complexity. 16 An RbXO, or more likely a CPT or CSAC de- variation. Because frequency aging is one of vice is the technology of choice when low pow- those, and its measurement requires a period of er is critical, while the CSAC also offers very time for initial stabilization and data collection, small size. The performance progression of that also provides for a period of burn-in and those choices is in the order given above while screening. Formal ATP processes are a part of the cost varies inversely. military/space RFS procurements. In many cases, the user will conduct his own RFS accep- Distinguishing features between RFS designs tance tests, likely a subset of the vendor’s facto- include: ry tests.

 Component/module/ instrument packaging • Over/Under Specification  Stand alone/GPS disciplined  Crystal oscillator type/low phase noise System engineers are trained to avoid both over  Servo bandwidth/dynamics and under specification. The dangers of the lat-  Output frequency/waveform/1 pps ter are obvious, and can result in serious system  Special features like user interfaces. fre- failures. But over-specification creates prob- quency synthesis, multiple O/Ps, time codes, lems too. In general, it is best to stick with stan- 1 pps input dard RFS products and their catalog specs. Those exceeding the requirements are fine. • Design Documentation Those that are not necessarily will have to be tightened. It may be possible to re-qualify a de- Detailed design documentation beyond that in- sign to the needed conditions. Also, the user cluded on a product data sheet is generally not may have some control over the environment provided for a catalog RFS item, but such infor- that the RFS is exposed to. mation (e.g., a reliability analysis) may be available upon request. Complete design documen- • RFS Selection tation is usually part of a custom design or military/space RFS program. RFS selection logically begins by considering the application and its requirements. Some at- • Qualification Tests tributes may not matter much (e.g., warm-up time for a continuously-running laboratory Manufacturer-conducted qualification tests are clock, aging for a GPS-disciplined RFS) while part of the RFS design process to assure full others may be critical (e.g., power for a battery- compliance with its specifications. This entails powered clock, phase noise for a microwave re- a major effort because it generally includes such ceiver). One can quickly decide on what cate- items as a stress-level reliability analysis, me- gory of RFS is appropriate (e.g., a commercial chanical shock and vibration tests, a suite of unit for low cost, a rugged MIL unit for a tacti- EMI tests, etc. Details about the QTP procedure cal aircraft) and zero in on the applicable ven- and results are generally not provided except for dors and their products, with due consideration custom designs or military/space programs. given to their location and reputation. Even if your application seems to require a custom design, these standard products will give you an • Acceptance Tests idea of what specifications are practical. Some applications (e.g., a GNSS satellite clock) will Manufacturer-conducted acceptance tests are necessarily require a full-up procurement pro- part of every RFS delivery, although the ATP gram and perhaps even a new design. Obvious- procedure and results may not be included with ly, vendor consultation is needed in cases like the unit. The items included in the ATP are that. Organizations that have frequent, major or generally limited to those that show unit-to-unit 17 unique RFS requirements will probably need to COTS Commercial Off The Shelf CPT Coherent Population Trapping establish in-house expertise regarding RFS de- Cs Cesium vices and a close relationship with their clock CSAC Chip Scale Atomic Clock DC Direct Current vendor(s). DDS Direct Digital Synthesizer DSP Digital Signal Processing DSAC Deep Space Atomic Clock • Acknowledgments DVT Design Verification Test EMI Electromagnetic Interference ESS Environmental Stress Screening Many people and organizations have contrib- FCS Frequency Control Symposium FM Frequency Modulation uted to rubidium frequency standard technology FXR Flash X-Ray for over 50 years. Reference [4] contains infor- GNSS Global Navigation Satellite System GPS Global Positioning System mation about their work. Progress continues to GPSDO GPS Disciplined Oscillator be made, particularly with smaller, lower power HDEV Hadamard Deviation He Helium and less expensive CPT and CSAC devices. Hg Mercury ICD Interface Control Drawing IEEE Institute of Electrical and Electronic Engineers • About the Author I/O Input-Output ION Institute of Navigation LPF Low Pass Filter Mr. Riley has worked on rubidium frequency LVDS Low Voltage Differential Signaling MEMS Micro Electro-Mechanical System standard and related time and frequency tech- MEO Medium Earth Orbit nology for most of his professional career. He MIFTTI Modular Intelligent Frequency Time & Time Interval MIL Military began as a Development Engineer at General MTBF Mean Time Between Failures Radio where he pursued that technology includ- MRR Manufacturing Readiness Review NIST National Institute of Standards and Technology ing quartz and rubidium frequency standards. In OCXO Oven Controlled Crystal Oscillator 1980 he became the Engineering Manager of the OCVCXO Oven Controlled Voltage Controlled Crystal Oscillator PDR Preliminary Design Review Rubidium Department at EG&G where he di- PLL Phase Locked Loop rected the design of rubidium frequency stan- PM Phase Modulation PPS Pulse Per Second dards, including those used onboard the GPS PSD Power Spectral Density satellites. His last position before retirement was PTTI Precise Time and Time Interval RAFS Rubidium Atomic Frequency Standard as Manager of Rubidium Technology at Rb Rubidium FTS/Datum/Symmetricom. Mr. Riley is also the RbXO Rubidium Crystal Oscillator RF Radio Frequency proprietor of Hamilton Technical Services RFS Rubidium Frequency Standard where he developed and sold software for fre- SCD Source Control Document SGEMP System Generated Electromagnetic Pulse quency stability and consults in the time & fre- SRD Step Recovery Diode quency field. He is the author of three books, SSB Single Sideband S/N Signal to Noise numerous papers and reports, and holds several STS Short Term Stability patents. Mr. Riley is an IEEE UFFC Life Fel- S/V Space Vehicle SWP Size, Weight and Power low. He has served on several IEEE standards TC Temperature Coefficient committees and the PTTI Advisory Board. He TDEV Time Deviation T&F Time and Frequency received the 2000 IEEE International Frequency TTL Transistor Transistor Logic Control Symposium Rabi Award and the 2011 UFFC Ultrasonics Ferroelectrics and Frequency Control UHF Ultra High Frequency Distinguished PTTI Service Award. VCXO Voltage Controlled Crystal Oscillator VHF Very High Frequency VSWR Voltage Standing Wave Ratio • Acronyms and Abbreviations XO Crystal Oscillator

ADEV Allan Deviation AC Alternating Current AFS Atomic Frequency Standard ATP Acceptance Test Procedure BITE Built In Test Equipment BPF Band Pass Filter BTC Baseplate Temperature Controller CDR Critical Design Review 18 • References 12. M.A. Lombardi, “Selecting a Primary Fre- quency Standard for a Calibration Laborato- 1. W.J. Riley, Rubidium Frequency Standard ry”, Cal Lab, Vol. 15, pp. 33-39, April 2008. Primer, LuLu Press, Inc., ID # 9559425, Oc- 13. ITU Handbook, Selection and Use of Pre- tober 2011. Also available here. cise Frequency and Time Systems, 1997. 2. W.Weidemann, “Application Critical Pa- rameters for Rubidium Standards”, Pro- ceedings of the 1998 IEEE International Frequency Control Symposium, pp. 84-87, May 1998. This is one of the few technical papers that describe the selection of an RFS. 3. Ball Corporation, Efratom Time and Fre- quency Products, Irvine, CA, Precision Time and Frequency Handbook, Ninth Edi- tion, 1993. This handbook contains a wealth of still-valuable information about obsolete RFS products and their specifications and applications. Once commonly available, it is now hard to find. 4. W.J. Riley, “A History of the Rubidium Fre- quency Standard”, IEEE UFFC web site, January 2020. 5. T.J. Lynch, W.J. Riley and J.R. Vaccaro, “The Testing of Rubidium Frequency Stan- dards”, Proceedings of the 43rd Annual Sym- posium on Frequency Control”, pp. 257- 262, May 1989. 6. W.J. Riley, “The Physics of the Environ- mental Sensitivity of Rubidium Frequency Standards”, IEEE Transactions on Ultra- sonics, Ferroelectrics and Frequency Con- trol”, Vol. 39, No. 2, pp. 232-240, March 1992. 7. W.J. Riley, Handbook of Frequency Stabili- ty Analysis, NIST Special Publication 1065, July 2008. 8. W.J. Riley and J. R. Vaccaro, “A Rubidium- Crystal Oscillator (RbXO)”, IEEE Transac- tions on Ultrasonics., Ferroelectrics, and Frequency Control, Vol. UFFC-34, No. 6, pp. 612-618, November 1987. 9. J.R. Vig, “Introduction to Quartz Frequency Standards”, IEEE UFFC web site, October 1992. 10. IEEE Guide for Measurement of Environ- mental Sensitivities of Standard Frequency Generators, IEEE Standard 1193. 11. MIL-PRF-55310, “Oscillator, crystal controlled, general specification for”. 19 Appendix A

Examples of Conventional, CPT and CSAC Commercial Gas Cell Frequency Standard Specifications

These products are current examples of commercial atomic clocks: a Stanford Research Systems PRS- 10 conventional rubidium frequency standard, an AccuBeat NAC1 commercial Rb atomic clock using coherent population trapping (CPT), and a Microsemi SA.45s chip-scale atomic clock that uses a minia- ture Cs gas cell. These datasheets were captured from their respective manufacturer’s web sites in Janu- ary 2020.

Conventional commercial RFS units have 1-second ADEV stabilities on the order of 1x10-11. The Ac- cuBeat CPT unit has a specified STS of 2x10-10, an order of magnitude worse than a typical conventional RFS, but it is smaller and has lower power dissipation.

The CSAC unit has a specified 1-second STS of 3x10-10, slightly worse than the AccuBeat CPT unit. It has a much smaller size and power dissipation than a conventional RFS. Lower phase noise and space qualified versions of this CSAC are available.

Information about past conventional RFS units and their specifications can be found by searching the internet for such products as FRK, FRS, LPRO, and X-72.

Note: The examples of gas cell atomic clock specifications in Appendices A-C include current products from most major U.S. and European manufacturers: AccuBeat, Excelitas, Frequency Electronics, Mi- crosemi, Orolia and Stanford Research Systems.

20 21 22 23 24 25 26 27 28 29 30 Appendix B

Example of a MIL Rubidium Frequency Standard Specification

This example of a current MIL RFS is the Microchip 8200 module for ground tactical, shipboard and airborne applications. Its datasheet was captured from the Microchip web site in January 2020. Distin- guishing features include a wide range of environmental hardening and a sealed case. It offers an excellent combination of package, performance and ruggedness.

Information about past MIL RFS units and their specifications can be found by searching the internet for such products as M-100, RFS-10, RFS-100, RFS-3000, TRFS and 8130A.

31 32 33 34 Appendix C

Examples of High-Performance Space-Qualified Rubidium Clock Specifications

These Rubidium Atomic Frequency Standard Data Sheets from Excelitas, Frequency Electronics, and Spectratime/Orolia were captured in January 2020 from their respective web sites. These units are rep- resentative of the Rb clocks used on-board the GPS and Galileo navigation satellites.

Space RFS designs tend to emphasize performance and reliability, and, of course, must to radiation hardened. Otherwise, space is a relatively benign environment for a Rb clock. Most current space RFS applications are for GNSS systems like GPS and Galileo, and their generic datasheet specifications are usually superseded by program-specific SCDs.

GNSS systems generally involve close control of their onboard clocks and supply daily updates, thereby emphasizing their 1-day stabilities. An RAFS having a pure white FM noise characteristic at a 1x10-12- 1/2 level would have a 24-hour rime deviation (TDEV) of only about 0.1 ns corresponding to a clock-induced navigation error of about 0.03 m, typically small compared with other factors.

GNSS Rb clocks commonly include integral baseplate temperature controllers to reduce their temperature sensitivity to negligibility. That also allows their performance and testing to emphasize that thermal operating condition.

GPS clock stability is commonly expressed using the Hadamard deviation (HDEV) because it ignores the effect of linear frequency drift that is largely removed by the system. RFS drift tends to diminish over time, and its early stabilization is often well modeled by a diffusion ( time) fit. The HDEV stability plot shown for an actual Excelitas GPS III RAFS was from a 1-month factory ATP run where that model was used to remove its early aging.

35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 File: Selection Criteria for Rubidium Frequency Standards.doc W.J Riley Hamilton Technical Services January 23, 2020