G.8266 Wander Tolerance and Transfer

Introduction

G.8266 is the specification for a frequency Grandmaster clock.  It takes in a physical layer frequency , in the form of an E1, T1 or SyncE signal, and generates a PTP signal out, using the G.8265.1 PTP profile for frequency.

In terms of performance, the G.8266 clock is mostly similar to the G.812 clock originally defined for the SDH network back in the 1990s. This was termed the “SSU clock” (Synchronisation Supply Unit), and is equivalent to the N. American Stratum 2 clock.

The wander generation, tolerance and transfer specifications are all identical to that of the G.812 clock, except that the output is PTP instead of an E1/T1 signal.

This document concentrates on how to test wander tolerance and transfer for a G.8266 clock.

Wander Tolerance

The wander tolerance of a clock is the maximum amount of wander it is required to tolerate at its input while operating normally.  In the presence of wander up to the tolerance limit, the clock is required to:

  • Maintain the clock within prescribed performance limits
    (although these limits have never been defined, so there is no requirement to meet)

  • Not cause any alarms.

  • Not switch reference.

  • Not go into hold-over

As stated above, the wander tolerance and transfer requirements of G.8266 follow those defined in G.812.

G.812 defines the tolerance limit in three ways: as an MTIE mask (clause 9.1.1), as a TDEV mask (clause 9.1.2), and a sinusoidal wander tolerance (clause 9.1.3).  The sinusoidal wander tolerance mask in Table 13 and Figure 5 is computed from the MTIE mask defined in clause 9.1.1, and shows the relationship between wander frequencies and amplitudes required in order to meet the MTIE mask.  It can be used in place of the TDEV wander tolerance discussed in clause 9.1.2.

Figure 5/G.812 – Lower limit of maximum tolerable sinusoidal input wander

Wander Transfer

Clocks act as low-pass filters to any wander on their inputs, therefore wander transfer is defined in terms of a bandwidth.  For a Type I clock (Option 1 2048kbit/s hierarchy), the bandwidth is 3mHz.  The clock is expected to filter out the frequency components higher than the specified bandwidth.  

This can be also be verified using sinusoidal tones at different frequencies, and measuring the attenuation of those tones in the clock output.  From this, the frequency response of the filter can be plotted and compared against the expected frequency response.  If the tone frequencies and input amplitudes are carefully chosen, this can be measured simultaneously with the wander tolerance. At least two points in every decade are chosen to show the shape of the filter response.

G.8266 Tone Frequencies for Tolerance and Transfer

The following tables proposes the tone frequencies and amplitudes to be applied at the input, the attenuation due to the filter, and the tone amplitude expected at the output.

This is a single test covering both wander tolerance and wander transfer.  For wander tolerance purposes, the clock must tolerate the input without causing alarms, switch reference or go into holdover.  For wander transfer purposes, the output tone amplitude must be less than the maximum amplitude indicated.  There is no minimum bandwidth defined, therefore there is no minimum amplitude either.

It is assumed that the low-pass filter in the DUT is at least a first-order with a roll-off of 20dB/decade for frequencies above the -3dB point.

Table 1: Tone Frequencies for G.8266 Type I Clocks

Type I clocks belong to the Option 1 hierarchy based on 2048kbit/s. The maximum peak-to-peak amplitude at each frequency is taken from Table 13 of G.812.  This is a translation of the MTIE mask defined in Table 9 of G.812, using the relationship between tone frequency, f, and observation interval, τ, of  f = 1/ πτ .  

The lowest tone frequency specified below is 0.0001Hz (0.1mHz). This is 30 times lower than the bandwidth of a Type I clock.  It has a period of almost 2.8 hours, and hence requires a test time at least that long.   While Table 13 of G.812 extends as low as 0.000012Hz (12µHz), with a period of 23 hours the test time would be extremely long, and being so far into the pass-band it is very unlikely that either the wander tolerance or wander transfer would be any different.

The possible input signals for a Type I G.8266 clock are SyncE Option 1, 2048kHz, or 2048kbit/s (E1).

The parameters used to generate this table were:

  • Maximum bandwidth: 3mHz

  • Maximum gain peaking: 0.2dB

  • Noise generation allowance: ±35ns
    (half the peak-to-peak high-frequency time error defined in clause 8.2)

Frequency

(Hz)

Amplitude

(ns)

Test Time

(Cycles)

Test Time

(s)

Max. Gain

(dB)

Max. Output Amplitude (ns)

1

750

300

300

-50.5

38

0.32

750

96

300

-40.6

43

0.01

750

30

300

-10.8

251

0.043

750

13

302

-23.1

88

0.032

1000

10

313

-20.6

129

0.016

2000

5

313

-14.7

404

0.01

2000

3

300

-10.8

610

0.0032

2000

3

938

-3.3

1403

0.0008

2000

2

2500

0.2

2082

0.00032

5000

1

3125

0.2

5152

0.0001

5000

1

10000

0.2

5152

The filtering characteristic expected from the clock is shown in the following diagram. While there is no minimum gain specified, the frequency response is expected to be fairly close to the maximum. Appendix II of G.812 suggests the result should be within 2% of the mask for long observation intervals.

Tone Frequencies for G.8266 Type II and Type III Clocks

Type II and III clocks belong to the Option 2 hierarchy based on 1544kbit/s. The maximum peak-to-peak amplitude at each frequency is taken from Table 14 of G.812.  This is a translation of the MTIE mask defined in Table 10 of G.812, using the relationship between tone frequency, f, and observation interval, τ, of
f = 1/ 2.5τ .  This is slightly different to the relationship used for the Option 1 hierarchy.

The possible input signals for a Type I G.8266 clock are SyncE Option 2 or 1544kbit/s (T1).

The parameters used to generate this table were:

  • Maximum bandwidth: 1mHz

  • Maximum gain peaking: 0.2dB

  • Noise generation allowance: ±35ns
    (half the peak-to-peak high-frequency time error defined in clause 8.2)

Frequency

(Hz)

Amplitude

(ns)

Test Time

(Cycles)

Test Time

(s)

Max. Gain

(dB)

Max. Output Amplitude (ns)

4

300

1200

300

-72.0

36

1.26

301

377

300

-62.0

36

0.4

303

120

300

-52.0

36

0.126

308

38

302

-42.0

38

0.04

325

12

300

-32.0

44

0.0126

380

4

318

-22.0

66

0.004

550

3

750

-12.3

169

0.00143

1000

3

2100

-4.8

609

0.0004

1007

1

2500

0.2

1066

0.0001

1037

1

10000

0.2

1097

The filtering characteristic expected from the clock is shown in the following diagram. While there is no minimum gain specified, the frequency response is expected to be fairly close to the maximum. Appendix II of G.812 suggests the result should be within 2% of the mask for long observation intervals.