IMDHMIControls LibraryShop main store

How-to

How to commission five VFD drives on one Modbus gateway

Drives and motion Published 9 min read

Short answer

Prove the link in three passes: open the Modbus TCP connection to the gateway, verify one read from one drive against its own display, then run one serial transaction at a time across all five. A TCP-to-RTU gateway must serialize every RS-485 request, so schedule run commands ahead of power reads instead of polling all five at once.

On this page
  1. What do I configure first: the PLC, the gateway, or the drives?
  2. How does the PLC choose which of the five drives answers?
  3. What serial settings must match before any drive will answer?
  4. Which number goes in the register field of the read instruction?
  5. How do I prove the first read before wiring all five drives?
  6. How fast can five drives really share one RS-485 wire?
  7. How do I schedule run commands and power reads on the same wire?
  8. When is it safe to drop the hardwired start and stop?
  9. Questions asked

Commission the link in three passes: open the Ethernet connection to the gateway, prove one read from one drive, then run one serial transaction at a time across all five. The order is fixed by the hardware: Modbus TCP can carry several requests at once, but the RS-485 side of a TCP-to-RTU gateway is one shared wire that must serialize every request and reply. Five drives through a gateway differ from one drive on a PLC serial port — say, controlling an ATV12 drive from a TM200 PLC over Modbus RTU — in only two places: which drive the unit ID selects, and who talks when.

What do I configure first: the PLC, the gateway, or the drives? #

Configure the Ethernet connection first: the PLC has to open a Modbus TCP session to the gateway's IP address on its service port — 502 unless the gateway is set otherwise — before any drive is reachable. Bring the layers up in that order and write down what you set in each, because every layer fails with its own signature:

Layer Record these settings Failure if you confuse it
Ethernet connection Adapter model from its label, protocol mode, IP address, service port The PLC never opens the connection
Gateway routing Unit-ID-to-slave mapping, selected serial channel The request reaches the adapter but the wrong drive answers
Serial network Baud rate, parity, data format, unique drive addresses Timeouts, or two drives answering one poll
Drive application Function code, register offset, data width, scaling, command source Clean data, wrong values, or no commanded action

Before anything else, confirm what the AutomationDirect adapter actually does. An Ethernet-to-RS-485 product translates Modbus TCP requests into Modbus RTU transactions only in its gateway mode; the same hardware can also move serial data without translating, and that arrangement needs different PLC settings. Read the model off the adapter's label, get the manual from the manufacturer, and set the mode the manual documents.

How does the PLC choose which of the five drives answers? #

The unit ID in the TCP request selects the drive: on a Modbus TCP-to-RTU gateway the unit ID normally maps straight through to the RTU slave address, so drives at addresses 1 through 5 are polled with unit IDs 1 through 5. Keep unit ID equal to slave address as the commissioning default, because it is the one routing decision you will never have to second-guess.

Two checks keep routing boring. Confirm the gateway does not reserve a unit ID for its own configuration — if it does, keep the drives off that number. And change one thing at a time: if a poll fails, vary the unit ID or the drive's address, never both, or you will not know which change fixed it.

A duplicate address is the classic commissioning trap. Two slaves at one address both answer the same poll, their replies collide on the pair, and the gateway reports timeouts on every drive — the one-drive-works, five-drives-dead signature.

What serial settings must match before any drive will answer? #

Every device on the RS-485 segment must use the same baud rate, parity, data bits and stop bits, and each drive a different address — mismatched framing shows up as timeouts on all five, not error messages. The multidrop board distributes the pair; it assigns no addresses and sets no framing. Set each drive's communication parameters from its keypad before it goes on the network, and check the drive manual for the apply rule — some drives take new communication settings only after a power cycle.

Terminate only the two physical ends of the segment, and follow the drive and adapter instructions for biasing, shielding and topology. Wire by the documented signal names, not by terminal letters: RS-485 polarity labels are not consistent between manufacturers, and a reversed pair gives clean-looking wiring with zero replies.

Which number goes in the register field of the read instruction? #

Use the protocol address — the zero-based offset from the drive's register table — not the reference-style notation (4xxxx for holding registers) some documentation prints, and the two differ by exactly 1. Both conventions appear in the same manuals, so translate once on paper before typing: write the drive's register, its protocol address and the instruction field value on one line, and you stop chasing off-by-one errors in the dark.

A timeout and an exception response are different diagnoses. A timeout means no usable reply arrived inside the window — work the connection, routing, framing and wiring list. An exception means a slave answered and refused the request, so the transport is proven and the request is wrong: check the function, the address and the quantity. A read that runs past the end of a mapped block is the textbook case of why a read that crosses a register map gap returns Modbus exception 02.

Two drive-side settings decide whether good data becomes motion. The command-source parameter must select the network, or a write to the run register lands and nothing moves. And the width and scaling of what you read must match the drive's register table — resolve both against the drive's own display before you write anything.

How do I prove the first read before wiring all five drives? #

Disconnect four drives, wire one, and get a single read of a status register whose value you can also see on the drive's own display — that one verified number proves transport, routing and register convention at once. Work the sequence in order:

  1. Record the adapter and multidrop board model numbers from their labels and get the manuals from the manufacturer, with the drive's communication documentation and the PLC instruction help.
  2. Set drive 1's slave address and communication format from its keypad, applying them the way the manual specifies, power cycle included.
  3. Wire drive 1 alone to the adapter with the other four disconnected, following the documented signal polarity and terminating the two ends as instructed.
  4. Match the gateway's serial settings to the drive — baud rate, parity, data bits, stop bits — and note how it maps unit IDs to slave addresses.
  5. In the PLC, configure the Modbus TCP client connection: gateway IP address, service port 502, unit ID 1.
  6. Issue one read of one documented status register and wait for completion; launch nothing else while it runs.
  7. Record the returned value next to what the drive's display shows, and fix offset, width and scaling until they agree.
  8. Reconnect drives 2 through 5, each at its own address, and repeat the single read with each unit ID before building the polling schedule.

In Do-more, the instructions that carry these exchanges are MRX for reads and MWX for writes, pointed at the gateway connection you configured rather than at a serial port. A sample program written for a PLC serial card proves instruction sequencing, not this connection.

How fast can five drives really share one RS-485 wire? #

One transaction at a time, and at 9600 bps with 8E1 framing a two-register read costs about 40 ms per drive, so a full five-drive scan lands near 200 ms — five updates per drive per second. Worked example with illustrative numbers: 9600 bps, 8 data bits, even parity, 1 stop bit means 11 bits per character and a bit time of 104.2 µs.

  • Read request for two registers: 8 bytes × 11 bits = 88 bits → 88 × 104.2 µs = 9.2 ms
  • Reply carrying two registers: 9 bytes × 11 bits = 99 bits → 10.3 ms
  • Modbus RTU silent gap of 3.5 characters before each frame: 38.5 bits ≈ 4.0 ms, so 8.0 ms for the pair
  • Wire time: 9.2 + 10.3 + 8.0 = 27.5 ms
  • Drive answer delay, taken as 12 ms measured on the bench: 27.5 + 12 = 39.5 ms per exchange
  • Five drives in turn: 5 × 39.5 = 197.5 ms per full scan

One scan every 197.5 ms refreshes each drive about five times a second (1000 ÷ 197.5 = 5.06). Baud rate is the lever: 19200 bps halves the wire time to 13.75 ms, giving 25.75 ms per exchange and 128.75 ms per scan, while the 12 ms answer delay stays fixed.

The gateway enforces this serialization whether your code likes it or not: requests that arrive together on Ethernet queue behind each other on the serial side. Launch one request, wait for completion or error, then advance, and your program stays in step with the gateway's queue instead of hiding it. It is the same effect that explains why Modbus RTU drive polling on one serial port is slow before a gateway is even involved.

How do I schedule run commands and power reads on the same wire? #

Give the run-command write the next free slot and poll power only once per second: in a 1 s window you have about 25 slots of 39.5 ms, so four status cycles (20 slots) plus five power reads (5 slots) fits with margin for one command write. The arithmetic: 1000 ms ÷ 39.5 ms = 25.3 slots per second; 20 status slots take 790 ms and 5 power slots take 197.5 ms, totaling 987.5 ms. Each drive gets four status updates and one power value per second — enough for kWh trending, since energy is power integrated over time and a load that changes on a scale of seconds does not need millisecond samples.

A run-command write goes to the head of the queue: worst case it waits out the transaction in flight (39.5 ms) and then runs its own (39.5 ms), about 80 ms from decision to drive. That is fast enough for starting and stopping pumps, fans and conveyors on a VFD, and at 19200 bps it drops to roughly 50 ms.

Drive the schedule as a simple state machine:

  • Select the next drive and transaction from the schedule.
  • Launch one request and wait for completion or error — nothing else goes on the wire.
  • On completion, store the data and mark it fresh; on error, record the diagnostic and mark it stale.
  • Advance to the next scheduled transaction.

Cap retries at two, then mark that drive stale and move on. Unlimited retries against a disconnected drive consume the slots the other four need, turning one dead node into five blind ones. Track communication health per drive so the operator sees exactly which drive went quiet.

When is it safe to drop the hardwired start and stop? #

Not until the machine's own validation says so — working Modbus messages qualify the data link, never the start-and-stop function, so keep the existing approved control path in place through commissioning and keep any safety function hardwired regardless. The emergency stop and any safety-rated stop must not depend on the network: a dropped TCP session or a failed gateway has to leave the machine in a defined state, so those functions stay on hardwired inputs, and you set the drive's documented communication-loss behavior if it offers one.

Cutover is a machine decision, not a communications one. Run the old hardwired start and stop in parallel while you commission, validate the network-commanded start and stop under the machine's procedures, and only then remove the old wiring.

Questions asked

Can I copy a sample program written for a PLC serial card?

Borrow the sequencing, not the connection. A sample written for a serial card such as an H2-SERIO-4 module targets a PLC serial port, so its port settings say nothing about a gateway path. Rebuild the connection side as a Modbus TCP client with the gateway IP, service port 502 and the unit ID of each drive, then reuse the MRX and MWX structure from the example.

How do I find a duplicate address when everything times out?

Power the drives down, then bring them up one at a time and re-issue the same single read after each. The network dies on the drive whose address collides with one already on the pair. If all five come up clean, the timeout cause is elsewhere, and termination at the two end devices is the next check.

Does the multidrop board configure anything?

No. It distributes one RS-485 pair to several drives and nothing more. Slave addresses, baud rate and parity live in the drives and in the gateway's serial settings, and the PLC reaches each drive through the unit ID. Wire the same five drives through a terminal block instead and the network behaves identically.

What is the difference between a timeout and an exception response?

A timeout means no usable reply arrived in the window, so work the connection, routing, framing and wiring. An exception means a drive answered and refused the request, so the transport is fine and the request is wrong: exception 01 is an unsupported function, 02 an address the drive does not have, 03 a value it rejects. Fix the request, not the cable.

Is 9600 bps enough for five drives with power monitoring?

Yes, for status and energy trending. At 39.5 ms per transaction you get about 25 slots per second, and four status scans across all five drives plus one power read each fits in 987.5 ms. If you later need faster status, 19200 bps halves the wire time to 13.75 ms per exchange, about 25.75 ms with the drive's answer delay.

Keep reading

  • How-to

    Control an ATV12 drive from a TM200 PLC over Modbus RTU

    The ATV12 takes start/stop from register 8601 (CMD) and speed from register 8602 (LFRD, in rpm) once you set Fr1 to Mdb in the CtL- menu and match address, baud, parity and stop bits at both ends. The ladder then writes both words cyclically, one request at a time, gated on completion. Start from the TVDA sample project in Machine Expert Basic.

    Drives and motion

  • How-to

    Bridging a PLC-5/15 DH+ network to a Kinetix 300 servo drive

    Yes, a DH+-to-EtherNet/IP gateway exchanges command and feedback words between a PLC-5/15 and a Kinetix 300, but it only moves data: the drive must execute the motion profile locally. Replace the stepper controller, pulse expander, and encoder-counter functions with an eight-word map that carries a 32-bit position as two 16-bit words behind a data-valid handshake. If the machine needs trajectory points streamed per scan, migrate the controller instead of bridging it.

    Drives and motion

  • Selection guide

    Running CiA 402 CANopen drives from an S7-1500 through a gateway

    Keep the CANopen drives for axes whose trajectory the drive itself generates — CiA 402 modes 1, 3 and 6 — with the segment's bus load held under about 40% at 250 kbit/s. Replace with PROFINET drives and Technology Objects for any gearing, camming or coordinated axis, because a conformance class B gateway gives cyclic data with no isochronous coupling and no Technology Object can close behind it. This rewinder's axes all generate their profiles drive-side, so the keep path wins.

    Drives and motion