Accelerator as time master or time slave
The accelerator can be either master or slave. If the accelerator is master, it does not synchronize to an external incoming signal on one of the time synchronization connectors. If the accelerator is slave, it synchronizes to an incoming time synchronization signal.
The NT time synchronization protocol
Time synchronization connectors can be configured for PPS (pulse-per-second) / TTL (transistor-transistor logic) or for the NT time synchronization protocol. This protocol contains embedded PPS and time information.
Daisy-chaining of NT accelerators
The NT accelerators each have two internal time synchronization connectors, which can be used for daisy-chaining time synchronization between multiple accelerators installed in the same server/chassis. The NT time synchronization protocol and the PPS protocol can be daisy-chained between accelerators.
Time source for an accelerator
Control of the internal accelerator time depends on whether the accelerator is time master or time slave:
- Time master/host: The host provides the source for the internal accelerator time. This source is the host time, which is usually synchronized to the time from an external network time server.
- Time master/accelerator: The accelerator runs on its free-running clock and is never adjusted.
- Time master/PPS: The time source is a PPS signal and optionally the UTC time from an external time device.
- Time slave: The time source is provided from the external time signal, and the internal accelerator time is synchronized to the provided time. The source can be a time master accelerator.
Sliding internal accelerator time adjust
Minor adjustments to the internal accelerator time is done with a sliding adjust, where the time stamp update rate is increased or decreased by a small amount. It is thereby ensured that the time stamp values for received frames show a nice increasing time, without any sudden forward or backward changes in time.
This sliding adjust applies both when the accelerator is time slave and time master/host. Sliding adjust can be enabled or disabled from the application.
Initial setting of the internal accelerator time can be done by giving a specific value.
Loss of source for synchronization
A bit indicates the current status of synchronization with GPS or another accelerator, and a bit indicates if the synchronization is lost. In this case, an entry in the driver log is generated as well.
External and accelerator time samples for Napatech equipment
When configured as master and connected to another NT accelerator, an NT accelerator can provide a synchronized set of external time and accelerator time every second. The external time is the time of the other accelerator.
The synchronized set of time values can be used as input to application time algorithms.
PPS-triggered accelerator time samples in connection with 3rd-party equipment
When a PPS (pulse-per-second) signal from a time device, for instance a GPS, is connected to a time synchronization connector on a master accelerator, the PPS signal from the time device can trigger the sampling of the accelerator time.
For every PPS pulse (one per second) a 64-bit accelerator time is sampled. This means that the application has one second to read a sample value.
The PPS-synchronized accelerator time values can be used as input to application time algorithms.
For details on PPS time synchronization, see DN-0985.
Synchronization relative to a PPS signal
The accelerator time stamp clock rate can be synchronized relatively to any external time source with a PPS output at TTL levels.
Synchronization to absolute UTC time
If absolute time information is available, the accelerator time stamp clock can be synchronized to the absolute UTC time. The pps_endrun tool (see DN-0449) illustrates use with an EndRun Præcis Cf CDMA receiver or an EndRun Præcis II CDMA receiver, the pps_oregano tool (see DN-0449) illustrates use with an Oregano SYN1588 PCIe NIC (IEEE 1588 / PTP), and the pps_symmetricom tool (see DN-0449) illustrates use with a Microsemi SyncServer®.
Synchronized transfer of block statistics
When two or more accelerators are synchronized with respect to absolute time, block statistics are transferred synchronously from the accelerators to the host. This ensures consistent counter sets when handling data from multiple accelerators in absolute synchronization.