International Telecom Sync Forum (ITSF), November 2014

Holdover Requirement: High Stability OCXO Challenges

© 2014 Rakon. This presentation contains confidential information.

Presenter: Cyril Datin (R&D Engineer – Technical Leader)

International Telecom Sync Forum (ITSF), November 2014

OCXO Solutions for Holdover Requirements

© 2014 Rakon. This presentation contains confidential information.

Presenter: Cyril Datin (R&D Engineer – Technical Leader)

2

What is time and phase alignment?

What is Accuracy or stability?

Holdover mathematical definition

Holdover graphical approach

How to measure holdover?

Phase Noise (PN) over time (long term)

Short/mid/long term stability –mathematical tool

Agenda

Frequency instability contributor ADEV translation

Thermal compensation limitation

Ageing prediction limitation:

Ageing prediction limitation:

Summary

Conclusion: The next generation requirement?

Typical clock stability σy(τ)

ADEV interpretation medium / high stability comparison

OCXO fundamentals

OCXO fundamentals

Telecom system requirement:

High end digital compensation OCXO topology:

3

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0 2 4 6 8 10 12 14

mag

nit

ud

e

time

What is Time and Phase Alignment?

Frequency holdover: Ability to maintain a frequency alignment over a time period (Stratum 2, 3 3E requirement)

Time holdover: Ability to maintain a phase accuracy over time (Time division requirement)

Ideal reference

Signal same frequency AND

phase offset

t0

Signal same frequency without phase offset / time

aligned

4

6

Accuracy or Stability ?

Accurate but…

ACCURATE AND STABLE

Stable but …

12

9 3

6

12

9 3

6

12

9 3

6

12

9 3

6

12

9 3

6

12

9 3 time

time

time

Day 1H1m1

Day NH1m1

NOT STABLE

NOT ACCURATE

5

The frequency drift is cumulating phase error x(t)This is also called time Holdover

Holdover Mathematical Definition

Fractional frequency fluctuation

Phase time

Ideal harmonic signal is represented

6 Holdover Graphical Approach

Phase error is cumulating gradually over time following frequency evolution

Linearif frequency constant (ageing free)

Quadraticif frequency linear (constant ageing)

Cubicif frequency quadratic (linear ageing)

Holdover (phase time)Is the area underneath the curves

7

As an oscillator manufacturer we deal predominantly in the frequency domain, and convert to the time domain…

y(t) fractional frequency deviationx(t) phase time… => time Holdover

How to Measure Holdover?

OscillatorFrequency f0

τ0 counterintegration time

Each frequency sample has a τ0 duration

8 Phase Noise (PN) Over Time (Long Term)

Graphical approach – Phase Noise (PN) over time (long term)

f0 f0

f0 f0’

f0f0

’’

PN L(f)

poor PN

Caesiumclock

Rubidiumclock

OCXOclock

Excellent accuracy10-12

medium PNVery high accuracy

10-11

excellent PN Medium accuracy10-9

Holdover

Day 1 Day 10 Day 100

time

time

time

Absolute frequency slightly shifting

9

Allan Deviation ADEV

τ : counter integration time (frequency sampling)

Most common time domain metric of frequency stability

ADEV remains convergent for most noise types

Another common metric is the modified Allan deviation mod σy(τ) which may be converted easily to TDEV

Short/mid/long term stability – mathematical tool

Short/Mid/Long Term Stability

y(t)M samples

yi yi+1

τ

time

10 Typical clock stability σy(τ)

1.00E-13

1.00E-12

1.00E-11

1.00E-10

10 100 1,000 10,000 100,000

AD

EV σ

y(τ)

τ [s]

Typical Clock Performance – Overlapping ADEV

Caesium clock

Rubidium clock

High stability OCXO

Medium stability OCXO

High stability

High accuracy

11 10MHz OCXO ADEV Interpretation

5.E-14

5.E-13

5.E-12

5.E-11

10 100 1,000 10,000 100,000

Ove

rlap

pin

gA

DEV

σy(

τ)

τ (s)

High stabilityDouble oven equivalent

Medium stability SMD

State of the art USO

10MHz OCXO ADEV Interpretation Medium / High Stability Comparison

12 OCXO Fundamentals

Highest Q resonator oscillator (>1000 000)

Environmental insulation construction

Low transient response (heating element thermal inertia)

But…

Long term stability limited due to inherent ageing effect…

13

Good short term stability

Telecom System Requirement

Excellent time keeping ability

• Need for digital compensation implementation

OCXO Short Term Stability RubidiumMid/Long Term Stability

14 High End Digital Compensation OCXO Topology

Keep very short term stability performance

Upgraded TIME

keeping OCXO

High Stability OCXO

THERMAL Compensation Algorithm

Ageing Compensation Algorithm

Overcome unwanted THERMAL and TIME effect

15 Frequency Instability Contributor ADEV Translation

1.E-12

1.E-11

1.E-10

10 100 1,000 10,000 100,000

Ove

rlap

pin

gA

DEV

σy(

τ)

τ (s)

High stability OCXO

Ageing compensation algorithm

Thermal compensation algorithm

16 Thermal Compensation Limitation

Apply temperature profile

Evaluate frequency variation

-20

-16

-12

-8

-4

0

4

8

12

16

20

-100

-80

-60

-40

-20

0

20

40

60

80

100

0 20 40 60 80 100 120 140

y(t)

[p

pb

]

Tem

pe

ratu

re [

°C]

time

Temperature

-12

-10

-8

-6

-4

-2

0

2

4

6

8

10

12

-50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90

y(t)

[p

pb

]

Temperature [°C]

Frequency Response Over Temperature

Then apply compensation algorithm

Compute the ideal mathematical response

But…Oscillator hysteresis limiting residual error

Frequency vs. Temperature response

17

Ideal ageing behaviour is predominantly logarithmic over time

Ageing prediction limitation

0

10

20

30

40

50

60

70

80

90

0 30 60 90 120 150 180 210 240 270 300 330 360 390 420 450 480 510 540

df/

f [p

pb

]

Number of days

The ideal curve fit may be computed

18 Ageing prediction limitation

Frequency gap between expected and actual ageing response

-1E-10

-8E-11

-6E-11

-4E-11

-2E-11

0

2E-11

4E-11

6E-11

-8E-11

-7E-11

-6E-11

-5E-11

-4E-11

-3E-11

-2E-11

-1E-11

0

1E-11

2E-11

3E-11

Prediction limited to extrapolation accuracy

19 Summary

High stability OCXOs are not able (by themselves) to maintain very tight holdover specification such as 1.5 µs over 24 hours

Additional distinct digital compensation is required to a target specification

Although mathematically easy to implement, actual OCXO behaviour requires a perfect understanding of Piezoelectric crystal phenomena

OCXO manufacturers are able to assess each of oscillator’s behaviour by monitoring over significant time periods and environmental conditions

20

Time keeping is 1.5 µs over 24 hours

Conclusion

What about the next generation requirement?

Time keeping is 1.5 µs over 72 hours

Current Tomorrow

It means 3,5 x 10-11 (35ppt) stability for all causes over 24 hours

A wristwatch disciplined by such a stable clock would be off one second after 913 years!

A wristwatch disciplined by such a stable clock would be off one second after 8219 years!

Holdover is not linear

A clock 3² is needed = 9 times more stable to target this specification

3,8 x 10-12 stability (3,8ppt)