The Telecom Timing Stack: OCXO, TCXO, and IF SAW Filters Explained
Modern telecom hardware has two “invisible” jobs that make or break real-world performance:
Keep frequency and time stable across packet timing (PTP), physical-layer frequency distribution (SyncE), and GNSS-assisted architectures.
Keep receivers selective and linear under blockers and interference—before the ADC and DSP can do anything useful.
This post gives you a practical mental model of the telecom timing stack, explains where OCXOs, TCXOs, and 70–300 MHz IF SAW filters fit, and ends with a lab-ready verification checklist plus a resource roundup you can keep bookmarked.
Reference roundup page: https://www.fujicrystal.com/news_details/telecom-timing-stack-ocxo-tcxo-saw-resource-roundup.html
What people mean by “the timing stack”
A lot of engineering pain comes from treating timing parts as isolated line items:
“We picked a low-noise oscillator, so jitter will be fine.”
“We filter in DSP, so analog selectivity doesn’t matter.”
“We have GNSS, so holdover doesn’t matter.”
In reality, timing quality is a system transfer-function problem:
The servo / PLL bandwidth decides which noise sources dominate.
Close-in phase noise can dominate integrated jitter with narrow loops.
Analog filtering prevents overload/IMD that digital filtering can’t undo once the front end compresses.
So the “timing stack” approach is simple: define the chain from source → conditioning → local flywheel → output interfaces, then verify where it matters instead of relying on bench-only measurements.
Quick glossary: PTP vs SyncE vs “reference clock”
PTP (IEEE 1588)
PTP (Precision Time Protocol) synchronizes clocks over packet networks. It’s widely used across telecom timing profiles and deployment models.
SyncE (ITU-T G.8262 family)
Synchronous Ethernet (SyncE) distributes frequency through the Ethernet physical layer (PHY), with timing characteristics defined in ITU-T recommendations.
PRTC / ePRTC (ITU-T G.8272 / G.8272.1)
In packet-based phase/time distribution, PRTC and ePRTC define requirements for primary reference time clocks.
You don’t need to memorize standards to design well, but you do need to map requirements to measurable KPIs like: time error (TE), MTIE/TDEV, jitter, wander, and holdover drift.
Where OCXO, TCXO, and SAW filters sit in a real telecom design
A useful way to visualize the whole chain is in three layers:
Front-end selectivity & interference control (RF/IF)
Edge/local references (small cells, outdoor nodes, GNSS timing modules)
Operator-grade timing nodes (boundary clocks, aggregation, SSU/SEC, grandmasters)
1) IF SAW filters: protect the chain before it turns nonlinear
Even with fast ADCs and powerful DSP, analog selectivity still matters because it prevents overload and blocker-driven compression in mixers, IF amps, and ADC inputs.
IF SAW filters are still a workhorse because they offer:
Steep skirts (adjacent-channel + blocker suppression)
Repeatable production behavior
Practical selectivity at common telecom IF points (often 70–300 MHz)
The trade-off: tighter selectivity and stronger stopband rejection often cost more insertion loss and can worsen group delay ripple, which can show up in EVM margin depending on the system.
2) TCXOs at the edge: the practical local reference
At the edge, the constraints are brutal: power, size, wide temperature swings, and cost.
A TCXO often wins because it balances:
Stability vs temperature
Power and warm-up time
BOM cost
“Good enough” holdover when disciplined by GNSS or network timing
Typical TCXO roles at the edge:
Synthesizer/baseband reference in small cells and RRU/RRH
Local clocking inside rooftop GNSS timing receivers or outdoor timing modules
Stable reference for TSN/industrial edge gateways
Important mindset: output timing quality isn’t “TCXO alone”—it’s the TCXO + reference quality + servo/PLL transfer function + power integrity.
3) OCXOs in timing nodes: the flywheel for holdover and close-in noise
When the requirement is operator-grade timing—especially under reference impairment (GNSS loss, packet delay variation, degraded upstream)—an OCXO becomes the system’s flywheel.
OCXO is commonly used in:
PTP grandmasters / boundary clocks
SyncE equipment clocks and line-timing interfaces
SSU/SEC and other critical infrastructure timing nodes that must survive GNSS outages
Why? Because close-in phase noise and short-term stability matter more once your servo bandwidth is narrow and you’re chasing tight TE/MTIE/TDEV targets.
A practical selection logic
Step 1: write system KPIs first
Before you shortlist components, define:
Required holdover duration and allowed TE growth
Jitter/phase noise limits over your real integration band
Expected reference impairments (GNSS loss, PDV for PTP, upstream quality)
Thermal profile (enclosure, airflow, outdoor cabinet)
Blocker environment (for RF/IF chains)
Step 2: pick oscillator tier by risk
Choose OCXO when holdover/close-in noise/servo stability are primary risks.
Choose TCXO when power/size/thermal constraints dominate and holdover needs are moderate (or the node stays disciplined).
Step 3: treat power and VCTRL nodes as timing components
Two recurring field failures:
Great oscillator ruined by supply noise coupling
Great OCXO ruined by dirty control voltage (VCTRL)
If you’re steering an OCXO, treat VCTRL as a precision analog node:
Low-pass filtering aligned to servo dynamics
Keep away from high-speed clocks/SerDes and switchers
Guard routing / clean return path if needed
For TCXOs, don’t underestimate power integrity—test with representative ripple injection (DC/DC harmonics can modulate clocks in surprising ways).
Step 4: for SAW filters, assume the “problem” is the fixture/layout until proven otherwise
Many “filter spread” stories are actually matching/layout/measurement-plane issues. Don’t skip:
Correct reference plane definition (pads/launch)
Proper VNA calibration and fixture discipline
Matching network placement within millimeters
Verification checklist
A) OCXO / TCXO: measure in the final system, not just on a bench
Phase noise + integrated jitter
Measure phase noise at offsets your PLL actually transfers
Integrate jitter over the band your system cares about (not a generic default)
Supply sensitivity
Inject ripple at representative frequencies (switcher fundamentals + harmonics)
Quantify modulation and close-in noise degradation
Thermal behavior
Temperature sweep with realistic ramp rates (outdoor is not a slow oven)
Watch frequency vs temperature and transient effects
Holdover behavior (especially OCXO timing nodes)
Simulate GNSS loss and degraded reference conditions
Measure time error growth vs holdover KPI
B) Packet timing nodes: verify what operators actually care about
Even if you don’t publish plots, internal sign-off usually includes:
TE / MTIE / TDEV under representative impairment scenarios
Servo settling dynamics, stability margins, and sensitivity to reference noise
C) IF SAW filters: validate at the right plane, then validate in system
VNA S-parameters
Confirm S21 skirts / IL, and S11/S22 match
Calibrate and measure at the correct reference plane
Group delay
- Plot group delay and ripple across the modulation-relevant band
System blocker test
- Confirm IF/ADC chain stays linear under worst-case blockers
Temperature sweep
- Ensure drift stays within guard bands for your IF plan
Three “recommended paths” by scenario
1) Boundary clock / aggregation switch
Goal: strong holdover + low close-in noise under packet impairments.
Focus: OCXO selection + servo/VCTRL hygiene + TE/MTIE/TDEV verification.
2) Outdoor small cell / rooftop GNSS timing module
Goal: stable reference under wide temperature + practical power limits.
Focus: TCXO thermal + power integrity + jitter validation; escalate to OCXO only when holdover margin isn’t adequate.
3) Microwave backhaul / satellite receiver IF chain
Goal: keep selectivity and linearity under blockers; protect modulation integrity.
Focus: SAW filter loss vs rejection vs group delay, matching/layout discipline, VNA + in-system validation.
