|
|
|
|
KYLAND Synchronous Ethernet (SyncE) Timing Solution White Paper Link
The IEEE
1588 standard may be applied to industrial automation
systems to satisfy time-based production environments. Here is what you
need to know to improve your operations.
Manufacturing always trends towards faster, more
precise part production, but some attempts to achieve this goal can
stretch traditional control capabilities to their limits. In traditional
sequential-control systems, input sensors, output actuators, and
industrial controllers are distributed over a local area network.
Some systems use change-of-state or event triggered techniques to make
more precise parts, but many others use a combination of scan-based and
event-based control mechanisms. While suitable for numerous
applications, sometimes these control algorithms can create processing
jitter.
Usually, jitter does not matter as long as the required
response times are satisfied. For applications that require more
precision and have a low tolerance to jitter, the use of time-based
control is preferred. Time-based control techniques may
enable faster, more precise production of parts. Refer to the IEEE 1588
standard for guidance on implementing time-based control in new
equipment and systems.
In a time-based approach, an association
is made between input and output events and time. It becomes an integral
function of the control system and its algorithms. In such a system,
the input events are time-stamped and output events are scheduled. The
control system knows when the input was sampled and can determine when
the output should be actuated. The output device can
schedule the output to actuate at the predetermined time. The only
jitter sources for this system are those associated with accurately
time-stamping inputs and outputs.
Case in point
For
example, in a non-time-based, high-speed conveyor diverter application,
individually manufactured parts travel along a conveyor at a constant
rate of speed. Ideally, the system identifies individual parts as they
move down the conveyor, performs on-the-fly analysis of the parts to
determine if they are defective, and then triggers actuation downstream
to reject the defective parts. However, if the resolution of the control
system does not match the speed of the conveyor system, the
wrong part or more than one part will be rejected.
Here is how
the application would work using time-based control. When the
photoelectric eye detects a part, a time stamp is recorded indicating
the time that the part was detected. The controller sends the inspection
system a time-stamped schedule to signal when the part should be
inspected. The controller then sends the diverter system a time-stamped
schedule to signal when a defective part should be diverted. For this
case, the maximum speed of the system is limited by the delays through
the system from input to output. The maximum speed and parts per minute
are calculated as follows:
Part resolution = 12.4 msec jitter
Maximum
speed = 1/12.4 = ~80 parts per second
Maximum ppm = 80 * 60 = 4800
parts per minute
Jitter delays are confined to the electrical and
mechanical delays for input detection and output actuation devices.
Only the transport delays, the time it takes a data packet to go from
point A to point B, become a factor. System precision is limited to the
transfer delays, and the time-stamping and scheduling accuracies of the
respective input and output
devices. A time-based control mechanism can improve system
operation performance ten-fold because it eliminates jitter sources: the
processing speeds of input, control, and output devices.
1588
for industrial control
The IEEE 1588 standard is based on the
Open Device Net Vendor Association (ODVA) standard and defines a
point-to-point (PTP) configuration for devices compliant with ODVA
supported networks. The ODVA adoption of 1588 is referred to as the common
industrial protocol (CIP) Sync.
CIP Sync supports
version 2 of the 1588 standard and specifies the same set of defaults as
the Delay Request/Response Default PTP Profile in the 1588 standard.
These defaults satisfy most industrial application requirements for
distributing time throughout the control system. The mechanisms suitable
for implementing clocks over industrial networks include multicast
messaging, sync updates in the one-second range, and the delay
request-response path measurement mechanism.
Many I/O devices will implement the 1588 hybrid
clock. This clock satisfies long linear distributed applications where
daisy-chained connections are preferred over star connections, and
reduces wiring and switch requirements. These applications may be as
simple as distributing devices along a robot arm or as complex as
distributing devices along a quarter-mile paper mill application.
Global
positioning system (GPS) clocks are best applied where
precise time of day is required or subsystems are distributed over a
large geographical area. However, a controller or a switch with a hand
set clock is sufficient for many applications.
Time-based control is
preferred for applications that require precision and a low tolerance to
jitter.
For switches and routers that implement
transparent clocks, the end-to-end (E2E) method is preferred over the
peer-to-peer (P2P) for best interoperability in non-homogenous systems.
P2P works only with P2P devices. Most systems will contain both 1588 and
non-1588 compliant switches.
The CIP Sync profile supports PTP
clock domain zero for simple device and system implementation and
operation. Grandmaster devices such as controllers and GPS clocks should
use the PTP Alternate Time Scale option. This option distributes local
time zone and daylight
savings time throughout the factory floor. For clock backup
and redundancy, the normal operation of the Best Master Control
Algorithm (BMCA) will suffice for most applications. Backup grandmasters
may use the PTP grandmaster clusters option for faster switch over
time.
The IEEE 1588 standard defines the mechanism to distribute
and synchronize time across the system, but does not specify how to
handle perturbations in time that may occur during
normal operation
of the control system. These changes may occur due to one or more of the
following conditions:
The user adjusts the master clock whose
type is “hand set.”
A master with a more accurate clock becomes
available (new grandmaster). This may occur during the system startup or after
the system has been running for some time.
Delta jitter is the
maximum minus the minimum jitter for components. The time-based approach
eliminates jitter sources in the control system.
The time
master is temporarily
disconnected from the slave clock and then reconnected.
Here, given any discrepancy in time between the master and the slave, a
step change will occur.
A photoelectric eye
detects parts and their status on the conveyor belt. Input is
sent to the controller as part of the input scan.
Some
applications need a very stable clock for scheduling periodic control
algorithms. This strategy is especially true for motion-control
applications that run precision control
loops. A significant jump in the system time will impact periodic
control algorithms. Another alternative is using a clock that can
tolerate steps in time and still minimize impact to the control system.
CIP
Sync clock model
The local clock is a frequency-disciplined
clock tuned (synchronized) by the 1588 Precision Time Protocol
each time it receives the PTP Sync or follow-up message. The clock will
tick at some nominal rate and be adjusted by the PTP protocol and
tuning algorithm to match the rate of the master’s clock. This tuning
process may take several seconds during startup and system
initialization
for a default one-second sync rate. The clock may be set to zero at startup, but
the
absolute value is never set thereafter. The accuracy of the
clock will depend on whether the implementation is software or hardware
assisted, the resolution of the clock, the crystal, and all the other
design parameters that go into making an accurate frequency-disciplined
clock.
This table of IEE
1588 clocks shows how they best fit automation devices.
CIP Sync defines a
clock model to address step changes in time.
The
local clock schedules all periodic or cyclic operations on the device.
An offset value (ClockOffset) is maintained between PTP time transmitted
in sync messages and the local time (LocalTime). When a device needs
the current PTP time (PTPTime), it will read the local time and add the
offset value to obtain the current system time. ClockOffset and PTPTime
equations are:
ClockOffset =PTPTime – LocalTime
PTPTime
=LocalTime + ClockOffset
If the Coordinated Universal Time
(CUT) is required, the device will adjust the time value for leap
seconds.
The slave clock limits
the response to the assumed change in frequency of the mater clock. A
small step response is incurred for two tuning cycles.
Tuning
is adjusted based on two conditions: a small frequency offset
form the master clock, or a large frequency offset from
the master clock such as a step change. The magnitude of this offset,
small or large, is determined by the SyncThreshold value. The
SyncThreshold is set based on the synchronization requirements of the
application.
Offset from Master < SyncThreshold: Tune clock
Offset
from Master > SyncThreshold: Tune clock but clamp clock frequency
adjustment
When the offset from the master exceeds the threshold,
the clock
frequency adjustment is clamped. The clamp rate depends on
the requirements of the application. For a step change in time, this
normally results in a small one-time perturbation to the clock. If the
step is a result of a grandmaster change, then the clamping period may
continue for some time as the clock frequency is tuned to the frequency
of the new master. During this time, the application may operate in a
slightly degraded mode. Periodic or cyclic events may then be scheduled
based on the local clock and will not be affected by large step changes
in time provided that the local clock remains synchronized to the PTP
master-clock frequency.
Time-stamp compensation
When a
jump to the master time occurs, not all clocks in the system will see
the jump at the same time. If there is a step change to the master’s
clock, the amount of time it takes to propagate to the Node 1 slave clock and
the Node 2 slave clock will differ by the delay through the PTP boundary
clock. For some implementations, the delay may be as long as one sync
interval. Time stamps generated by Node 1 will differ by
the amount of the time step until all clocks see the same step and
adjust their clocks accordingly.
A time step occurs at
Node 2 between times 310 and 410. Time 310 is adjusted to time 100310 to
account for the step change.
When comparing time stamps
between two nodes or two time stamps within the same node, consider
whether a step has occurred during the time interval between the two
time stamps. Adjusting one or both of the time stamps to account for
this time step is referred to as time-stamp compensation. A similar
process should be followed when comparing two time stamps from the same
node.
One algorithm for compensating time stamps captures the
ClockOffset along with the time stamp. This offset value is carried with
the time-stamp value to be applied to the compensating algorithm. The
offset provides an indication of when the step occurred. Two algorithms
are defined in the CIP Sync profile. There is one for comparing time
stamps between two notes and one for comparing time stamps within a
single node.
Now Kyland has developed a Synchronous Ethernet (SyncE) timing solution accurate to 10 Nano Secornds or better!
|
|
|