1

I'm about configuring a Meinberg microSync timeserver that features "MRS" (Multi-Reference clock Source), but I could not find a good explanation what it really is (only Multi Reference Clock, but that didn't enlighten me), or how it works exactly.

For the IMS-TCR: IRIG Time Code Receiver and Generator Module Meinberg wrote:

The Meinberg MRS concept supports setting up a prioritized list of input sources that are used to synchronize the internal hardware clock of the IMS-TCR and then generate a large number of different output signals used by IMS I/O compatible cards to provide a user-defined selection of synchronization output signals by adding BPE, CPE or other IMS modules.

However I'm confused: My server has a GPS, and I can configure (among other possibilities that I don't have) one NTP server as additional reference (why not more than one?), and I can configure additional NTP servers that are not a "Reference" (in the MRS sense). As I understand it the "Reference" (MRS) will select the preferred GPS to adjust "the oscillator" when the GPS qualifies as a "sound source of time", and (given no further alternatives) it will select the specified reference NTP server if that qualifies as a "sound source of time" (You can define some criteria with each reference source).

So I wonder:

  • What is the advantage of configuring an additional NTP server as "reference" for MRS, compared to configuring it as a "normal" time server?
  • Normal NTP servers with a local reference clock adjust the local computer's software clock using NTP, while the reference clock has it's own oscillator that might be adjustable by hardware, probably configured by some external signal like DCF-77 or GPS. In contrast the MRS models seems to have only one oscillator (that is shared by the GPS and the "local computer", see photo). That makes it hard to understand how the clock the time server uses is tuned actually, and, maybe more important, how it can be monitored/verified.

Meinberg microSync timeserver showinh GPS module and oscillator

Specifically I'm confused with the timekeeping when the reference is a GPS server on the LAN (the graph shows nanoseconds, obviously, so the adjustment steps are like 50ms):

Statistics example for MRS timeserver

So who can enlighten me?

U. Windl
  • 478

1 Answers1

0

Reference clocks measure physical phenomena and output time codes. Atomic clocks are one implementation, and are a mature technology. It might not take the budget of a national laboratory to run these anymore, but time stamps are already being broadcast for anyone to use.

GNSS receivers, sat nav, is standard on most time appliances due to global availability and inexpensive components. The miracle of turn by turn navigation in a smart phone requires high accuracy time stamps, and of course this can also be used to adjust clocks.

In Meinberg MRS, NTP as a reference source makes that another candidate to discipline the appliance's internal clock. A better than average clock, oscillators are available that are supposed to be accurate within one second per year. Any random public internet NTP server would not hurt as a backup source. However the large latency over the internet introduces error, such that it might only improve accuracy if the internal clock failed.

NTP source could instead be a different appliance on the local network. Even if both were just GNSS receivers, there is redundancy in having another box in a neighboring rack.

Separate from reference clocks, I believe the accompanying software can also monitor a large number of NTP servers. Less about improving your stratum 0 time appliance, and more tracking the performance of a lot of stratum 1s downstream.

You linked Meinberg modular products. Those are a rack mounted blade style form factor with more options than the microSync. Longwave radio, pulses with or without time codes, and of course GNSS. Advanced source selection features are more relevant when you have more to pick from. With just GPS, its mostly about whether you have signal or its free-running in a hold-over state.

Fundamentally, all of this is about determining the difference between the local time, and some authoritative external reference clock. Gradually those offsets are used to estimate error and adjust the frequency of the output. Even with the fanciest appliances, its not like they are self-correcting, more like they are constantly compared to a reference.

John Mahowald
  • 36,071