The load on a process bus link is estimated by adding up the streams that cross it: each sampled value stream contributes its frames per second times the on-wire size of each frame, and GOOSE contributes little at rest but a lot in a burst. Under the 9-2LE guideline each merging unit publishes thousands of frames per second, so a few dozen streams fill a 100 Mbit/s link; in an HSR ring every link carries all the ring's traffic. That is why the process bus is designed with 1 Gbit/s, VLANs and filtering.
Why bandwidth matters on the process bus
On the process bus, merging units digitize the currents and voltages of the instrument transformers and publish them as Sampled Values (SV): layer 2 multicast frames sent thousands of times per second, without pause. Unlike the station bus, where traffic is moderate and arrives in bursts, the process bus has a continuous, high load. If a link saturates, samples are lost, and the relay using them blocks its protection, as explained in lost sampled values on the process bus.
That is why the load on each link is estimated during engineering, before choosing port speeds, topology and segmentation.
Sampled Values traffic
Each SV stream is characterized by two numbers: how many frames per second it publishes and how many bytes each frame takes.
Frame rate. It depends on the profile:
- 9-2LE guideline (UCA, 2004): 80 samples per cycle for protection, one sample per frame, with a fixed set of four currents and four voltages. That is 4,000 frames per second at 50 Hz and 4,800 at 60 Hz. The guideline also defined a 256 samples per cycle profile for metering.
- IEC 61869-9: the preferred configuration for protection is 4,800 samples per second with two samples per frame, that is 2,400 frames per second, regardless of the system frequency. The standard provides higher rates for power quality.
Frame size. It is made up of the Ethernet header, the VLAN tag, the SV header, one or more samples (each with its channels and quality) and, on redundant networks, the PRP or HSR tag. Added to that is what the frame takes on the wire even though it is not content: the preamble and the minimum interframe gap, which in Ethernet add 20 bytes per frame.
The load of a stream is then:
load (bit/s) = frames per second × (frame bytes + 20) × 8
The size of a 9-2LE frame is not fixed: it grows with the length of the svID, with the VLAN tag (4 bytes) and, on PRP or HSR networks, with the redundancy tag (6 bytes). With a 10-character svID and no redundancy tag it is around 130 bytes, and an example published by a manufacturer gives 146 bytes. With those 146 bytes, at 4,800 frames per second the stream uses about 5.6 Mbit/s without counting preamble and gap, and about 6.4 Mbit/s counting them, so some 15 to 18 streams fill a 100 Mbit/s link with no room for anything else. You can run the calculation with your svID, rate and redundancy in our free process bus network tool.
GOOSE traffic
GOOSE behaves differently. At rest, each publisher repeats its message at a long interval, and the load is small. On a change of state, the publisher sends a burst at very short intervals that gradually lengthen. If a fault makes many IEDs change at once (pickups, trips, position changes), the bursts coincide.
That is why GOOSE is sized for its worst case: all relevant publishers bursting at the same time. Its load is still lower than that of SV, but it arrives just when the network must deliver trips without delay, and it competes with SV for the switch queues. This is where VLANs and IEEE 802.1Q priority come in: trip traffic must not wait behind less important frames.
The effect of redundancy
Redundancy multiplies the load differently depending on the protocol, as explained in PRP and HSR:
| Topology | What each link carries |
|---|---|
| Star with filtering | Only the streams destined for the devices behind that port |
| PRP (two LANs) | Each LAN carries all the traffic, once; frames carry a redundancy trailer |
| HSR ring | Each link carries, in each direction, all the multicast traffic injected into the ring |
| RSTP ring | The ring traffic over the active path; after a failure, over the alternate path |
The HSR case is the one that surprises most. Each node sends each frame in both directions, and each copy travels the whole ring until the source node removes it. The result is that every link in the ring carries the sum of the streams of all nodes. If a 100 Mbit/s HSR ring has several 9-2LE merging units, it fills up with only a few of them. That is why process bus HSR rings usually run at 1 Gbit/s or are split into several rings.
- 01Stream inventoryMerging units, SV profile, GOOSE publishers and background traffic
- 02Load per streamFrames per second times on-wire size
- 03PathWhich streams cross each link, given topology and filtering
- 04Worst caseContinuous SV plus simultaneous GOOSE burst plus background
- 05ComparisonUtilization against the design criterion and link speed
What adds to the background
The process bus does not carry only SV and GOOSE. PTP messages for merging unit synchronization also flow, along with network equipment management and, in architectures that merge process bus and station bus, MMS traffic. Their load is small, but it is included in the calculation with a margin.
Design criteria
- Link speed: 1 Gbit/s on trunks and rings when there are several merging units; 100 Mbit/s may be enough on access links carrying few streams.
- Segmentation: VLANs per bay or per function and multicast filtering on the switches, so each port receives only the streams it needs.
- Margin: utilization is compared against a conservative design criterion that engineering sets and documents; nothing is designed to the limit.
- Latency: each hop adds the store-and-forward time of the frame plus the switching time, and a trip frame may wait behind a long frame. With many hops, the sum is compared with the required transfer time class.
- Testing: the estimate is confirmed with measurements at the FAT, with all streams active.
The MAC addresses, APPIDs and VLANs of each stream can be assigned with our free GOOSE and SV addresses tool, which helps organize the segmentation.
How we help
Sizing the process bus is part of the engineering of a digital substation, and it is cheaper to correct on paper than at the FAT. At Tiza Group we do the communications engineering of high and extra-high voltage digital substations: process and station bus architecture, redundancy, segmentation, and load and latency estimates. We do this work remotely. Tell us about your project.
Related guides
- Process bus
- IEC 61850 Sampled Values
- Merging unit
- PRP and HSR: redundancy without frame loss
- IEC 61850 digital substation
- PRP and HSR RedBox
- Substation network redundancy: RSTP, PRP, HSR or MRP
- Transfer time in IEC 61850: classes and measurement
- Centralized protection and control (CPC) in substations
- Substation communications by voltage level
- Substation network topologies: star, ring and more
References
- IEC 61850-9-2, Specific communication service mapping (SCSM) – Sampled values over ISO/IEC 8802-3.
- IEC 61869-9, Instrument transformers – Part 9: Digital interface for instrument transformers.
- UCA International Users Group, Implementation guideline for digital interface to instrument transformers using IEC 61850-9-2 (9-2LE), 2004.
- IEC 62439-3, Parallel Redundancy Protocol (PRP) and High-availability Seamless Redundancy (HSR).
- IEC TR 61850-90-4:2020, Network engineering guidelines.
- IEC 61850-5, Communication requirements for functions and device models.
- IEEE 802.3, Standard for Ethernet.
Frequently asked questions
How many Sampled Values streams fit on a 100 Mbit/s link?
It depends on the frame size, the rate and the design margin. In an example published by a manufacturer, a 9-2LE stream at 4,800 frames per second takes about 5.6 Mbit/s, so about 18 streams saturate 100 Mbit/s. In practice a wide margin is left and 1 Gbit/s is used on trunks.
Why does an HSR ring fill up sooner than a star network?
Because in HSR each node sends each frame in both directions of the ring and each copy travels the whole ring until it returns to its source. So each link carries, in each direction, all the multicast traffic injected into the ring, and the number of merging units that fit in a ring is limited by the link speed.
Does GOOSE take a lot of bandwidth?
At rest, little: each publisher repeats its message at a long interval. On an event, the publisher sends a burst at short intervals, and if many IEDs change at once the load jumps. Sizing considers that case, not the average.