---
Interview Setting: BRIDZA Technical Documentation Series — Frequency & Timing Division
Participants:
---
Chief Engineer: Dr. Marchetti, thank you for joining us today. You manage one of the largest multi-disciplinary research labs in the country, and you've overseen the procurement and integration of multiple timing and frequency systems over the past fifteen years. I'd like to tap into that experience for the benefit of our readers — particularly lab managers, engineers, and researchers who are selecting their very first frequency standard.
Dr. Marchetti: Happy to be here. When I started in this role, I'll admit that frequency standards weren't something I thought much about. We had an aging rubidium oscillator bolted into a rack in the corner, and nobody really questioned it. But as our research programs grew — we added a precision laser spectroscopy group, expanded our RF characterization lab, and took on a contract for satellite payload testing — it became painfully clear that "the old rubidium in the corner" wasn't going to cut it. Choosing that first serious frequency standard turned out to be one of the most consequential infrastructure decisions I've made.
Chief Engineer: That's exactly the kind of journey I want to explore. Let's start from the beginning.
---
Chief Engineer: For someone who's never specified a frequency standard before, how would you explain what it does and why it matters?
Dr. Marchetti: At its core, a frequency standard is a device that produces a highly stable, highly accurate electrical signal — typically at 5 MHz or 10 MHz — that other instruments in your lab can use as a reference. Think of it as the heartbeat of your measurement ecosystem. Every spectrum analyzer, every counter, every data converter that relies on a clock signal is only as good as the reference feeding it.
Here's a concrete example. In our RF characterization lab, we routinely measure frequencies in the 1–40 GHz range. A spectrum analyzer with a cheap internal crystal oscillator might have a frequency accuracy specification of ±1 ppm. At 10 GHz, that's ±10 kHz of uncertainty. If you're characterizing a filter with a 3 dB bandwidth of 50 kHz, that ±10 kHz error makes your measurement essentially meaningless. But if you lock that same analyzer to an external reference with a stability of 1 × 10⁻¹⁰, your frequency uncertainty drops to ±1 Hz at 10 GHz. That's a four-order-of-magnitude improvement.
Chief Engineer: And that improvement comes purely from the reference?
Dr. Marchetti: Purely from the reference. The analyzer's internal circuitry is already capable of that precision — it just needs a better clock to drive it. This is why I tell people that a good frequency standard is the single highest-leverage investment you can make in a measurement lab. It doesn't replace your instruments; it unlocks the performance that's already locked inside them.
---
Chief Engineer: What's the first thing a lab manager should consider when choosing a frequency standard?
Dr. Marchetti: Requirements, requirements, requirements. I cannot stress this enough. Before you look at a single datasheet, you need to answer three fundamental questions:
Let me expand on each. For stability, the key metrics are accuracy (how close the frequency is to its nominal value) and stability (how much the frequency drifts over time). These are measured in parts per ten — so 1 × 10⁻⁸, 1 × 10⁻¹⁰, 1 × 10⁻¹², and so on. Different technologies occupy different points on this spectrum, and the cost curve is steep. Going from 10⁻⁹ to 10⁻¹⁰ might cost you an extra $2,000. Going from 10⁻¹¹ to 10⁻¹² might cost $50,000 or more.
Chief Engineer: Can you give us a rough guide to what levels different applications need?
Dr. Marchetti: Sure. Based on our experience and published standards:
| Application Domain | Typical Stability Required | Suggested Technology |
|---|---|---|
| General bench instrumentation | 1 × 10⁻⁸ to 1 × 10⁻⁹ | TCXO, OCXO |
| RF/microwave test & measurement | 1 × 10⁻⁹ to 1 × 10⁻¹¹ | OCXO, Rubidium |
| Precision spectroscopy, atomic physics | 1 × 10⁻¹¹ to 1 × 10⁻¹² | Rubidium, Cesium |
| GPS-disciplined calibration labs | 1 × 10⁻¹¹ to 1 × 10⁻¹² | GPSDO, Cesium |
| Time-domain metrology, VLBI, deep-space tracking | 1 × 10⁻¹² and beyond | Cesium Beam, Hydrogen Maser, Optical |
These are rough guidelines, not gospel. But they give you a starting point. The worst thing you can do is over-specify — I've seen labs spend $80,000 on a cesium beam standard when a $6,000 rubidium with a GPS discipline option would have exceeded their actual needs by a comfortable margin.
---
Chief Engineer: Let's walk through the main technologies a newcomer will encounter.
Dr. Marchetti: Absolutely. There are essentially five categories you'll see in the commercial market:
These are the cheapest option — often under $100 in volume. They use a quartz crystal with electronic compensation for temperature changes. Stability is typically 1 × 10⁻⁶ to 1 × 10⁻⁷. You'll find these inside most bench instruments as their internal reference. They're fine for casual use, but nobody building a serious measurement infrastructure should rely on them as a primary standard.
This is where things get interesting. An OCXO places the crystal inside a small, thermostatically controlled oven — usually held at the crystal's "turnover temperature," around 70–80°C. By maintaining the crystal at a constant temperature, you eliminate the dominant source of frequency variation. Good OCXOs achieve 1 × 10⁻⁹ to 1 × 10⁻¹⁰ stability over temperature ranges, with aging rates of 1 × 10⁻⁸ to 1 × 10⁻¹⁰ per month.
I consider the OCXO the sweet spot for many labs. When we upgraded our general RF test bench three years ago, we installed a set of four high-quality OCXOs — specifically, the BRIDZA OCXO-100 series — as distribution masters for our four primary workstations. Total cost was under $12,000, and the improvement in measurement repeatability was dramatic. We saw our cross-correlated measurement uncertainty drop by a factor of fifteen.
Chief Engineer: We're glad to hear those units are performing well. What about the OCXO's weaknesses?
Dr. Marchetti: The main weakness is aging. A crystal oscillator, no matter how well controlled, slowly drifts over time as the crystal lattice undergoes microscopic changes. In a lab where you run the same measurement protocol every day, you might not notice this drift because it's slow and monotonic. But in a calibration lab where you need to certify that a frequency is within a certain window, that aging becomes a real concern. You'll need to periodically calibrate or discipline the OCXO against an external reference — which leads us to the next technologies.
Rubidium frequency standards use the hyperfine transition of the rubidium-87 atom at approximately 6.834 GHz as their physics reference. Because they're locked to an atomic transition, they don't suffer from the crystal aging problem. Typical specifications are:
A new rubidium standard will cost between $2,500 and $8,000 depending on performance tier. These are the workhorses of the timing industry. GPS satellites carry rubidium clocks. Telecommunications base stations use them. And they make outstanding lab references.
One important note: rubidium standards have a finite lamp life. The RF discharge lamp that excites the rubidium atoms degrades over time. Typical lifetime is 10 to 15 years for units like the older LPRO-101 or the modern BRIDZA Rubidium-50 series. This isn't a dealbreaker — it just means you should plan for eventual replacement.
Cesium is the gold standard — literally, since the SI second is defined by the cesium-133 hyperfine transition at 9.192 GHz. Commercial cesium beam standards achieve 1 × 10⁻¹² stability over averaging times of one day and have no lamp to degrade. They are the primary reference for national metrology institutes worldwide.
The downside? Cost. A new commercial cesium standard runs $40,000 to $100,000. Units from vendors like Microsemi (now Microchip) or Symmetricom are superb instruments, but for most university labs, they represent a significant capital expenditure that needs strong justification.
This is actually a hybrid approach, and it's extremely popular in university labs. A GPSDO takes a high-quality OCXO or rubidium oscillator and disciplines it using timing signals from GPS satellites. The GPS constellation effectively gives you a free, globally available cesium reference — because GPS time is steered to UTC(NIST) and UTC(USNO) with sub-nanosecond accuracy.
The beauty of a GPSDO is that you get long-term stability approaching 1 × 10⁻¹² over days and weeks (because GPS corrects for drift), combined with the short-term stability of your local oscillator (because GPS signals have noise and dropouts that the local oscillator bridges over).
When we first built out our satellite payload testing capability, we couldn't justify a dedicated cesium standard. Instead, we procured a GPSDO built around a BRIDZA Rubidium-50 core with GPS steering. Total cost was about $9,500. It's been running for four years with no intervention, and our calibration checks against the national institute's published UTC offsets show we're consistently within 20 nanoseconds. For our application, that's better than we need by an order of magnitude.
---
Chief Engineer: Let's talk about something that newcomers often overlook — how do you actually get the reference signal to all your instruments?
Dr. Marchetti: This is critical, and it's where I see a lot of labs make expensive mistakes. You can buy the best frequency standard in the world, but if you distribute its signal poorly, you'll negate most of the benefit.
There are three main distribution architectures:
Star Distribution: One cable runs from the standard to each instrument. Simple, but requires a lot of cables and a lot of output ports. You'll almost certainly need a distribution amplifier — and the quality of that amplifier matters enormously. A cheap distribution amp can add phase noise that degrades your 10⁻¹¹ reference to 10⁻⁹ or worse.
Daisy-Chain Distribution: Instruments are connected in series with tee connectors and short cables. This is tempting because it uses minimal hardware, but it creates impedance mismatches, reflections, and signal degradation with each tap. I've seen this cause intermittent lock failures in spectrum analyzers that were maddening to diagnose.
Dedicated Distribution Amplifier with Isolated Outputs: This is what I recommend for any lab with more than two or three instruments. A good distribution amplifier provides isolated, buffered, impedance-matched outputs for each load. Look for specifications like phase noise floor better than −155 dBc/Hz at 10 kHz offset, and channel-to-channel isolation of at least 60 dB.
Cable quality also matters more than people think. For runs under 10 meters, a good-quality 50-ohm coaxial cable with BNC or SMA connectors works fine. For longer runs, consider phase-stable cable or, if budget allows, fiber-optic distribution with electro-optic converters at each end. Temperature-induced cable delay changes can introduce apparent frequency offsets — roughly 1 ppm of cable delay change per degree Celsius, which for a 30-meter cable run can amount to several picoseconds per degree.
Chief Engineer: We've seen labs install a $10,000 frequency standard and then feed it through $5 BNC cables from the surplus bin.
Dr. Marchetti: I've been that lab. Early on, we had a beautiful new rubidium standard connected to our time-interval counter through a 15-meter cable run of unknown provenance, routed through a patch panel with corroded connectors. Our measurements showed random wander at the 10⁻¹⁰ level that we couldn't explain. It took us two weeks to trace it to that cable run. Replaced the cables, cleaned the connectors, and the wander disappeared overnight. Lesson learned: the distribution path is part of the frequency standard system, not an afterthought.
---
Chief Engineer: What about the physical environment? Are frequency standards finicky about where they're installed?
Dr. Marchetti: They can be. Here's what matters:
Temperature: Most high-quality OCXOs and atomic standards are designed to operate in a 0–50°C range, but their stability specifications are usually quoted at 25°C ± 5°C. If your lab experiences temperature swings of more than 5–10 degrees — say, because the HVAC is inadequate or the standard is mounted near a heat source — you'll see degradation. For crystal-based standards, this is particularly acute. A good OCXO has an internal temperature stability of 0.001°C or better, but the outer oven can only compensate so fast.
Power Supply: Atomic standards and high-end OCXOs are sensitive to supply voltage variations. Use a regulated, uninterruptible power supply. We had an incident where a voltage sag during a building-wide HVAC startup caused our rubidium standard to momentarily unlock and re-acquire, introducing a 5-nanosecond phase step. For most applications, nobody would notice. But for our time-transfer experiments, it corrupted a week's worth of data. After that, we put all timing infrastructure on a dedicated UPS with line-interactive regulation.
Vibration: This is often forgotten. Crystal oscillators are inherently sensitive to vibration and acoustic noise — it's called vibration-induced phase noise, and it can be severe. If your standard is going to sit on a bench next to a compressor, a cryogenic dewar, or a busy walkway, consider vibration isolation mounts. Atomic standards are generally less sensitive because the physics package is more robust, but it's still good practice.
Magnetic Fields: Rubidium standards in particular are sensitive to external magnetic fields because the atomic transition frequency has a second-order Zeeman dependence. Keep them away from large transformers, motors, and MRI systems (yes, I've seen this happen — a physics lab collocated their rubidium standard within 3 meters of a small MRI system for mouse imaging, and the frequency shifted by 3 × 10⁻¹⁰).
---
Chief Engineer: Let's talk money. How should a lab manager think about budgeting for a frequency standard?
Dr. Marchetti: I recommend thinking in terms of total cost of ownership over a 10-year horizon:
| Item | Typical Cost Range |
|---|---|
| Frequency standard unit | $2,500 – $100,000 |
| Distribution amplifier (8-channel) | $500 – $5,000 |
| Cabling and connectors | $200 – $2,000 |
| UPS and power conditioning | $300 – $1,500 |
| Periodic calibration (annual) | $500 – $3,000 |
| Environmental monitoring | $200 – $1,000 |
| Replacement at end-of-life | Varies |
For a typical university research lab doing RF and microwave measurements, I'd budget between $15,000 and $30,000 all-in for a robust first system. That gets you a high-quality rubidium or GPSDO reference, a proper distribution amplifier, good cabling, a UPS, and your first year of calibration.
Chief Engineer: Any parting advice for someone making this decision for the first time?
Dr. Marchetti: Three things:
First, talk to the people who will actually use the system. The physics group might need different stability than the RF group. Get everyone in a room and agree on a common set of requirements before you start shopping.
Second, buy one tier better than you think you need. Requirements always grow. A lab that starts with one spectrum analyzer will eventually have five, plus a VNA, plus a signal generator, plus a time-interval counter. Your frequency standard needs headroom.
Third, don't neglect the support infrastructure. I've seen too many labs where the frequency standard is a beautiful island of precision sitting in a sea of poor cabling, unregulated power, and uncontrolled temperature. The standard is only one piece of the system.
Chief Engineer: Dr. Marchetti, this has been incredibly thorough. Thank you for sharing your experience so generously.
Dr. Marchetti: My pleasure. If I can save one lab manager the two weeks we spent chasing phantom cable-induced wander, the interview will have been worth it.
---
Choosing your first frequency standard is a foundational decision that impacts the quality and credibility of every measurement your lab produces. As Dr. Marchetti's experience illustrates, the process requires a clear understanding of your actual stability requirements, careful attention to distribution and environmental infrastructure, and realistic lifecycle budgeting. Whether your needs are best served by a rugged OCXO, a versatile rubidium standard, or a GPS-disciplined oscillator, the principles remain the same: define your requirements first, invest in the complete system — not just the oscillator — and plan for growth. The right frequency standard, properly installed and maintained, will serve as the silent, unglamorous backbone of your laboratory's measurement capability for a decade or more.
---
This interview is part of BRIDZA's Technical Education Series. For more information on BRIDZA's frequency standard products, distribution amplifiers, and timing solutions, visit bridza.com/timing.