Home / Tools / Process bus network
IEC 61850 process bus network
Work out the load of the busiest link with sampled values and bursting GOOSE, the worst-case latency of a trip GOOSE against its time class, how many merging units fit, how long the network takes to recover and the accumulated PTP time error.
IEC 61850-9-2sampled value frameIEC 61869-9preferred merging unit ratesIEC 61850-8-1GOOSE frame and its repetitionIEC 62439-3PRP and HSRIEC 61850-5transfer time classesIEC/IEEE 61850-9-3§ 7.2PTP time budget
Results
— %
—
—
—
— ns
— Mbit/s
Assumptions
Want an engineer to review this case? Our team reviews your data and results and quotes the study or the settings work you need.
Charts
- Utilization
- Design threshold
- 100 % of the link
- Network budget
- Full time class
Calculation and details
Step-by-step calculation
What it means for your network
Work out the load of the busiest link with sampled values and bursting GOOSE, the worst-case latency of a trip GOOSE against its time class, how many merging units fit, how long the network takes to recover and the accumulated PTP time error.
IEC 61850-9-2sampled value frameIEC 61869-9preferred merging unit ratesIEC 61850-8-1GOOSE frame and its repetitionIEC 62439-3PRP and HSRIEC 61850-5transfer time classesIEC/IEEE 61850-9-3§ 7.2PTP time budget
What it is for
When moving to a digital substation, the network design questions are always the same: whether 100 Mbit/s is enough or 1 Gbit/s is needed, how many merging units fit in an HSR ring, how many switches a trip GOOSE can cross without leaving its transfer time class, what happens when a fibre is cut and how many PTP clocks can be chained. This tool does those sums with the real frame structure and the worst case of every queue, to review an architecture before buying switches or to defend it in an engineering meeting.
The MAC, APPID and VLAN addresses of each stream are checked separately, with the GOOSE and SV addresses tool.
How it is calculated
IEC 61850-9-2savPdu in ASN.1, BER encodingIEC 61869-9F4800S2 and F14400S6
Sampled value frame
ASDU = svID + smpCnt (2) + confRev (4) + smpSynch (1) + seqData (8 · channels) · frame = 12 + 4 + 2 + 8 + savPdu(n · ASDU) + 4 [+ 6]Each field carries its tag and its BER length (2 bytes below 128 bytes of content, 3 or 4 above). With 8 channels and a 10-character svID the ASDU is 93 bytes and the frame 130: 12 of MAC addresses, 4 of the IEEE 802.1Q tag, 2 of EtherType 0x88BA, 8 of APPID, length and reserved fields, 100 of savPdu and 4 of FCS. PRP appends a 6-byte redundancy control trailer and HSR inserts a 6-byte tag after the source MAC (IEC 62439-3).
Frames per second: 9-2LE publishes 80 samples per cycle with one ASDU per frame (4000 at 50 Hz, 4800 at 60 Hz) or 256 with 8 ASDU; IEC 61869-9 sets 4800 samples per second with 2 ASDU and 14,400 with 6, that is 2400 frames per second at both 50 and 60 Hz.
IEC 61850-8-1goosePdu and repetitionIEC 61850-6GSE MinTime and MaxTime
GOOSE frame
goosePdu = gocbRef + TAL + datSet + goID + t (8) + stNum + sqNum + simulation + confRev + ndsCom + numDatSetEntries + allDataThe time allowed to live (TAL) is taken as 2 · MaxTime, stNum and sqNum at their 32-bit worst case (5 bytes) and each signal as value plus quality. With 10 signals and 66 reference characters the GOOSE is 227 bytes.
IEEE 802.3preamble and interframe gap
Load of the busiest link
B = NMU · fframes · (LSV + 20) · 8 + NGOOSE · (LGOOSE + 20) · 8 / MinTime + backgroundThe 20 bytes are the preamble with its delimiter (8) and the minimum interframe gap (12). GOOSE is counted in burst: after an event each publisher repeats at MinTime and the interval grows up to MaxTime. The load applies to the link that carries every stream: the trunk of a star or of each PRP LAN, any link of an RSTP ring, and each direction of each link of an HSR ring, because there every frame travels the whole ring in both directions until it returns to its source. Merging units that fit: ⌊(threshold · C − GOOSE − background) / BMU⌋.
IEC 61850-5TT6 = 3 ms, TT5 = 10 ms, TT4 = 20 ms, TT3 = 100 ms
Worst-case GOOSE latency
tb = (S + 1) · [(L + 8) · 8 / C + (1542 [+ 6]) · 8 / C] + S · tsw + NMU · (LSV + 20) · 8 / C + (NGOOSE − 1) · (LGOOSE + 20) · 8 / CS is the number of switches on the longest path: the ones given for star and PRP, the N switches of an RSTP ring and the N − 2 intermediate nodes of an HSR ring opened next to the publisher. At each of the S + 1 egress ports the GOOSE is transmitted in full (store and forward; in cut-through only the header at each switch) and may wait for a lower-priority 1522-byte frame to finish, because IEEE 802.1Q does not interrupt a frame in progress. If GOOSE and SV share a priority, it also waits once for the coherent burst of the merging units, which publish at the same instant because they are synchronized, and for the other GOOSE of the same event: Ingram et al. (IEEE, 2013) measured that effect, and that once queued the frames leave in line and later switches add little. The result is compared with the share of the TT class given to the network, because the class also includes ta and tc, the IED time.
IEC 62439-3PRP and HSR, zero recovery timeIEEE 802.1DRSTP
Recovery after a link failure
PRP, HSR: 0 ms · RSTP ring: T = TL + (N − 3) · TPA (N even) or TL + (N − 2) · TPA (N odd)PRP and HSR send every frame over two paths and the receiver discards the duplicate, so a link failure does not interrupt the traffic. The RSTP formula is the worst case of the Siemens RUGGEDCOM white paper «Performance of the Rapid Spanning Tree Protocol in Ring Network Topology» (2007): TL is the link-loss detection and TPA the proposal-agreement exchange per pair of switches. Sampled values are lost during that time, which is why the process bus is designed with seamless redundancy (IEC TR 61850-90-4).
IEC/IEEE 61850-9-3§ 7.2network inaccuracyIEEE C37.238power profile
Accumulated PTP time error
ε = εGM + S · εTC (or S · εBC) ≤ 1 µsProfile limits: 250 ns for the grandmaster (7.4.1), 50 ns for each transparent clock (7.5) and 200 ns for each boundary clock (7.6.1); with them the profile aims at less than 1 µs after 15 transparent clocks. 1 µs is class T5 of IEC 61850-5, the one sampled values need. In an HSR ring each intermediate node counts as one clock.
Limits of the calculation
- This is a design estimate, not a network study: confirm it with measurements at FAT and SAT, with the real traffic and the chosen switches.
- It assumes the link evaluated carries every stream. With VLANs and multicast filtering the access ports carry less; the trunk and the rings still carry the sum.
- The latency is a worst case: a long frame ahead at every port, all merging units and all GOOSE bursting at the same instant. It leaves out fibre propagation (about 5 µs per kilometre) and the IED time.
- RSTP recovery uses one manufacturer's formula for a ring; in meshes or with a badly placed root it may be longer. Use the time your manufacturer guarantees for your topology.
- The PTP error adds the limits of each device according to the profile; link asymmetry and media converters add separately.
- The typical values of switch latency, RSTP and thresholds come from cited sources or are Tiza Group criteria labelled as such: change them in Advanced mode with the data of your equipment.
Frequently asked questions
How many merging units fit on a 100 Mbit/s link?
A 9-2LE frame with a 10-character svID is 130 bytes and takes 150 on the wire with the preamble and the interframe gap. At 4800 frames per second (60 Hz) each merging unit uses 5.76 Mbit/s, and at 4000 (50 Hz) 4.8 Mbit/s. With no other traffic a 100 Mbit/s link is full with 17 streams at 60 Hz or 20 at 50 Hz; with a 50 % design criterion, 8 and 10 remain. In the tests by Ingram et al. (IEEE, 2013) the switches lost frames as soon as the load went past 100 %.
Why does an HSR ring fill up before a star network?
Because in HSR every node sends each frame through both ports and each copy travels the whole ring until it returns to the node that sent it, which removes it (IEC 62439-3). So every ring link carries, in each direction, all the multicast traffic of all nodes. In a star with VLANs and multicast filtering, only the trunk or the port of the subscriber that receives every stream carries the full sum.
How many switches can a trip GOOSE cross?
It depends on the speed and on the queues. In the worst case, at every egress port the GOOSE may wait for a lower-priority 1522-byte frame to finish, and with store and forward it is transmitted again at every switch. At 100 Mbit/s, with a 227-byte GOOSE and 5 µs of switching, each switch adds about 147 µs: if the network gets half of the TT6 class (1.5 ms), about 9 switches fit before counting bursts. At 1 Gbit/s each switch adds about 19 µs and dozens fit. This tool does the sum with your topology and your bursts.
We design your process bus network
We size the digital substation network: PRP or HSR architecture, speed of each link, VLANs and priorities, latency budget and PTP synchronization, with the calculation report and the acceptance tests. We do it remotely, with no travel.
