Have you ever looked at a GPSDO datasheet and seen "Holdover accuracy: ±1.5μs/24h" — and wondered whether that's good, bad, or just meaningless marketing numbers?
Many engineers see that figure and think "1.5 microseconds? That's tiny, right?" But this is one of the most commonly misunderstood specs in timing. Let's break it down properly.
A GPSDO works in two modes:
Normal mode: It has a GPS lock. It uses the 1PPS signal from GPS satellites to discipline its internal OCXO to UTC time.
Holdover mode: GPS signal is lost (indoors, underground, shielded, interference). The GPSDO switches to its internal OCXO and keeps generating timing signals on its own, without any external reference.
The problem: OCXOs drift. Temperature changes shift the frequency slightly. Aging shifts it slightly more. Power supply noise adds a little more. Over time, this drift accumulates.
It means: after losing GPS signal for 24 hours, the timing error accumulated relative to true UTC will not exceed ±1.5 microseconds.
Sounds tiny? Let's put it in perspective:
The speed of light is 3×10⁸ m/s. A 1 microsecond timing error translates to 300 meters of position error in GPS ranging. So 1.5 microseconds sounds like it would throw your position off by 450 meters?
Don't panic. There's a critical distinction here. The GPSDO is the local clock reference for your receiver — it's not on the satellite. If the local clock drifts, the receiver's ranging calculations will be off, but only while it's operating without GPS reference. That's the whole point of holdover performance: how long can you run without GPS before your timing/position accuracy becomes unacceptable.
Different applications need wildly different holdover performance. Let's compare:
| Application | Required Holdover | Typical Hold Time |
|---|---|---|
| Consumer GPS receiver | ±1000ns (1μs) / 24h | A few hours |
| 5G small cell | ±1.5μs / 24h | 24 hours |
| Financial trading system | ±100ns / 24h | 72 hours+ |
| Telecom core network | ±1μs / 24h | 30 days+ |
| Metrology / lab reference | ±10ns / day | 90 days+ |
Notice the range? We're talking three orders of magnitude difference between consumer and metrology applications. Many engineers buy a GPSDO based on phase noise specs alone, but if your device regularly operates without GPS, holdover is the spec that actually matters.
1. The OCXO itself
This is the foundation. A standard OCXO has temperature stability around ±50ppb. A high-grade OCXO can do ±5ppb or better. Better stability means slower drift, which means longer holdover time. But the OCXO is only part of the story.
2. Temperature variation
Most OCXO drift comes from temperature. If your device is outdoors and temperature swings 20°C in a day, the OCXO frequency will drift. High-end GPSDOs use double-oven OCXOs or advanced temperature compensation algorithms to minimize this.
3. The holdover algorithm
Here's where manufacturers differentiate. A good GPSDO doesn't just "discipline and forget." While GPS is locked, it continuously records the OCXO's aging rate, temperature coefficients, and drift behavior. When it enters holdover, it uses this historical data to predict and compensate for drift. This is why two GPSDOs using the exact same OCXO can have 3-5x different holdover performance.
Don't just pick the best number on the spec sheet — better holdover costs real money. Ask yourself three questions:
1. How long can GPS actually be lost? If your device is always outdoors with clear sky view, you don't need great holdover. If it's indoors, underground, or in a shielded enclosure, holdover matters a lot.
2. How much error can your system tolerate? A 5G small cell needs ±1.5μs/24h. A financial timestamping system might need ±100ns. Know your system requirement before you start shopping.
3. What's your budget? Holdover performance from ±10μs/day down to ±10ns/day spans a 100x price range. There's no reason to buy a metrology-grade GPSDO for a consumer tracking device.
μs/day is not a "smaller is always better" number — it's a spec that's directly tied to your application. Before buying a GPSDO, ask: how long will GPS be unavailable? How much timing error can my system tolerate during that outage?
Answer those two questions, and you'll know exactly what holdover grade you need.
Got questions about GPSDO holdover for your specific application? Contact us — we've designed GPSDOs with holdover performance from ±10μs/day to sub-100ns/day.