Troubleshooting
Why MELSECNET II data reads zeros after adding a Modbus card
Short answer
The zeros are not a network fault. The QJ71MB91 occupies 32 I/O points, and inserting it ahead of the MELSECNET II module moved that module's start I/O number, so the access code reads the wrong position. Diff the I/O assignment against the map from before the change, re-address the assignment so the MELSECNET II module sits at 0070, and update every instruction that targets it.
On this page
Your MELSECNET II B and W data reads zeros because the QJ71MB91 shifted the rack's I/O assignment, not because the network failed. The QJ71MB91 is documented at 32 occupied I/O points in its own specification, and an intelligent module that size, inserted ahead of the A-series MELSECNET II module, moves that module's start I/O number. The converted code keeps calling the old position, so the CPU transfers what it finds there, zeros, instead of the network data. Ethernet SCADA and the Modbus link to the DCS keep working because neither touches that address. Read the I/O assignment in GX Works 2, diff it against the map from before the change, and re-address it: in the installation this fault came from, the resolving assignment put the MELSECNET II module at start I/O 0070.
What each symptom tells you #
B (link relays) and W (link registers) are the devices the MELSECNET II loop shares between stations, so all zeros means the CPU stopped collecting them, not that the loop slowed down.
- All B and W values go to zeros with the card installed and return when it is removed: the fault travels with the hardware change, so it lives in the rack's addressing, not on the network.
- Ethernet SCADA and the Modbus path to the DCS keep communicating: both outgoing links are fine, so work the incoming collection path first.
- No errors on the CPU, the QJ71MB91 or the network module, on the front panels or in the software diagnostics: module health says nothing about whether the program addresses the right slot.
- Moving the QJ71MB91 to different slots changes nothing: every slot ahead of the network module shifts it by the same amount, so the fault follows the card.
- The old GX Developer code behaves exactly like the converted GX Works 2 code: the conversion is ruled out, so stop re-running it.
The decision from that list: go straight to the I/O assignment page in the PLC parameter settings and compare it with the pre-change map.
How Q series start numbers move #
Q series start I/O numbers are hexadecimal and advance in 16-point blocks: a 16-point module consumes 0x10 (16 points) of address space, a 32-point intelligent module consumes 0x20. Inserting the QJ71MB91 ahead of the existing modules adds 0x20 to the start number of every module after its slot. What your calling code ends up out by is the difference between your before and after start numbers for the network module, so read both maps and subtract; do not infer the displacement from the card's size alone, because the previous allocation at that position decides how the numbers land. For the underlying device-to-slot mapping, see how Q series PLC addressing works for I/O and analog modules.
The assignment that cleared this fault #
The recorded resolving map for this rack, hexadecimal start numbers and decimal point counts:
| Position | Module type | Points | Start I/O |
|---|---|---|---|
| 0 | Intelligent (QJ71MB91) | 32 | 0000 |
| 1 | 16-point | 16 | 0020 |
| 2 | 16-point | 16 | 0030 |
| 3 | 16-point | 16 | 0040 |
| 4 | 16-point | 16 | 0050 |
| 5 | 16-point | 16 | 0060 |
| 6 | none | 0 | 0000 |
| 7 | Intelligent (MELSECNET II) | 32 | 0070 |
This is the recorded map of the rack that had the fault, not a template: positions 1 through 6 carried no identified module types in the record, so read your own hardware and parameter entries before applying the numbers. The deciding entry is position 7. The network module sits at 0070, and every instruction or configured transfer that touches it must resolve to that same start number.
Where 0070 comes from #
The arithmetic, so you can reproduce it on your own map:
- The QJ71MB91 takes 32 points, and 32 decimal = 0x20, so it occupies 0000 through 001F and the next module starts at 0000 + 0x20 = 0020.
- Each 16-point module takes 0x10: the five of them step 0020 → 0030 → 0040 → 0050 → 0060 and cover 0020 through 006F.
- Position 6 is empty and consumes nothing.
- The MELSECNET II module starts at 0070 and occupies 0070 through 008F (0x70 = 112 decimal; 112 + 32 = 144 = 0x90, so the next free start would be 0090).
- Total occupied: 32 + 5 × 16 + 32 = 144 points, exactly the 0000 through 008F range above, so the map is internally consistent.
If the last start your map produces is not the number your code calls, that difference is the whole fault.
Make the code and the map agree #
Work through every reference to the network module:
- Search the program for the instructions and configured transfers that read or write it, and note their destination B and W devices.
- Read each instruction's module-selection operand and how it encodes the start number; an operand can represent the module address differently from the assignment display, so decode it with that instruction's own help.
- Resolve each operand against the actual assignment; anything that does not land on 0070 in this map is a wrong target.
- Cross-reference the destination B and W devices for other writers, initialization rungs or clearing logic that could zero them later in the scan.
If a target is wrong, fix the operand or the assignment, one of the two, then verify. If every target is correct and the values are still zero, walk the data in order: the originating station on the loop, the receiving module's buffer and status, the CPU destination immediately after the transfer executes, and the same destination at end of scan. Present at the module but absent in the CPU points at the transfer's execution conditions; present after the transfer but zero later points at another writer; absent at the module itself points back to loop operation and configuration.
Stop re-testing the conversion and the cable #
The old GX Developer project showed the same zeros with the card installed, and the converted project had run normally before the card arrived, so the conversion is cleared. A dropped Ethernet path shows a different signature, question marks and tag timeouts as in why HMI values show question marks: PLC comms timeouts, not clean zeros from a rack of healthy modules. Every slot swap, cable swap and re-conversion costs machine time while the module operand stays wrong.
Apply the re-addressed assignment as a controlled change #
- Back up the working project and the current online parameters.
- Record the installed module order and each module's points and start number, across the Q rack and the A-series extension rack.
- Set the QJ71MB91 to its documented 32-point intelligent allocation, then set the following starts from the actual hardware so the MELSECNET II module lands where its accesses expect, 0070 in this record.
- Re-check every intelligent-module access and every physical X and Y reference against the resulting map, not just the network read: moving one module moves everything after it.
- Download the parameters and any program edits during an approved machine stop, following the CPU's documented restart requirements.
Changing an I/O assignment can move physical I/O references, so treat the download as a machine-affecting change and verify output states before returning to run. Afterwards, read the assignment back from the CPU and confirm the online parameters match the project.
Questions asked
Does it matter which slot the QJ71MB91 goes in?
Any slot ahead of the MELSECNET II module shifts its start number by the same 0x20, which is why moving the card between slots fixed nothing here. Placing it after the last referenced module would leave those starts untouched, but you would still read the assignment back afterward to confirm. The recorded fix kept the card at position 0 and re-addressed everything after it.
Can I change the code instead of the I/O assignment?
Yes. Either side of the pair works: edit the assignment so the network module keeps the start number the code expects, or leave the assignment and update the module operand in every instruction that touches the card to the new start number. Editing the map once is usually safer than hunting each operand, because one parameter change covers all accesses at once.
Why zeros instead of question marks or an error?
Question marks on an HMI mean the client lost the server, and a hard module fault trips a diagnostic. This is neither: the CPU successfully reads a rack position that no longer holds the network module, and the transfer carries that position's contents, which read as zeros, into the B and W devices. No module reports an error because none of them is faulty.
How do I prevent this at the next hardware change?
Before adding or moving any module, export the project and save the I/O assignment page, including point counts and start numbers across both racks. After the change, diff the two maps before touching code or cables. Any start number that moved means every intelligent-module access to that module moved with it.
Did the GX Developer to GX Works 2 conversion cause it?
No. The conversion was ruled out on site: the original GX Developer code showed the same zeros with the card installed, and the converted project had run normally before the card arrived. If the module code came from a conversion tool, regenerate it once the new assignment is in place, so its embedded module references match the map.