Troubleshooting
Fix slow SLC 5/04 DH+ communications to Ignition SCADA
Short answer
Slow updates from an SLC 5/04 to Ignition are a shared-path and servicing problem, not a scan-class problem: every request crosses a DH+ segment at 57.6K baud that two RSView HMIs and any RSLogix session also use, and the processor throttles message servicing when S:2/15 or S:33/7 is set. Clear both bits, add SVC instructions between large logic sections, then swap to a 5/05 when the bridge hardware fails.
On this page
- Why is the SLC 5/04 slow in Ignition while every other PLC is fine?
- How fast is DH+ compared to the Ethernet side of the bridge?
- What should I measure before touching scan classes?
- Should I raise the DH+ baud rate to 230.4K?
- Does a 1 second scan class guarantee 1 second updates?
- What do S:2/15 and S:33/7 do to throughput?
- Where do I put SVC instructions?
- When should I stop tuning and replace the 5/04?
- Questions asked
A 1 s scan class cannot buy updates the path cannot deliver. The 5/04's tags cross a DH+ segment that two RSView HMIs and any RSLogix 500 session also use, and the processor throttles how many communication requests it services when S:2/15 or S:33/7 is set. Fix the path and the servicing before the polling: clear both servicing bits, add SVC instructions between large sections of logic — a reported fix moved updates from about 2.5 s to 1 s — and swap to a 5/05 when the bridge hardware fails.
Why is the SLC 5/04 slow in Ignition while every other PLC is fine? #
Because the 5/04 is the only PLC whose every tag crosses the DH+ segment and the bridge chassis: the other controllers answer over Ethernet, so the shared DH+ path and the 5/04's communication servicing are the two things slowing you down. Three symptoms, one cause family. A window opening creates a burst of new tag subscriptions; RSLogix 500 uploading values is a burst of programming requests on the same wire; RSLogix struggling to connect is contention on a slow shared segment. The device status page in Ignition 7.9.8 that reads "Hello world" confirms the driver can reach the processor and says nothing about throughput — measure tag behavior instead.
| Symptom | First cause to check |
|---|---|
| Tags take seconds to reach good quality when a window opens | Request backlog on the DH+ path or throttled servicing in the processor |
| RSLogix 500 uploads values slowly | Programming traffic competing on the same DH+ segment |
| RSLogix 500 struggles to connect or stay connected | Segment contention plus slow processor response, not the client |
| Other Ignition-connected PLCs stay fast | Confirms the 5/04's path and servicing, not Ignition |
Read them together, because they share the route: Ignition over Ethernet to the bridge chassis (ENBT/A), across the backplane to the DHRIO, then out on DH+ to the 5/04. A delay that shows in RSLogix the way it shows in Ignition is not an Ignition problem.
How fast is DH+ compared to the Ethernet side of the bridge? #
DH+ on the SLC 5/04 selects 57.6K, 115.2K, or 230.4K baud, and even 230.4K is 2.3% of a 10 Mbit/s Ethernet link: the segment, not the bridge's Ethernet side, is your ceiling. The DH+ specification table in the SLC 500 system selection guide (1747-SG001) gives the maximum cable and terminating resistor for each rate, with 3048 m (10,000 ft) at 57.6K. Mixed networks usually end up at 57.6K because every node must run one common rate and support for the higher rates varies by device. The arithmetic that sets expectations: 230.4 kbit/s ÷ 10,000 kbit/s = 2.3%, and 57.6 kbit/s ÷ 10,000 kbit/s = 0.6%. The bridge does not change this — every request and reply still crosses the slow segment.
What should I measure before touching scan classes? #
Measure four things before changing any setting: seconds to good quality when a window opens, the same number with RSLogix 500 offline, the processor scan time, and the state of the two channel 1 servicing bits S:2/15 and S:33/7. Work in this order, one change at a time:
- Record the route as built: which nodes remain on DH+ (the 5/04, the DHRIO, both RSView HMIs) and which clients reach the 5/04 through which segment.
- Record the Ignition scan class and the slow group's interval, then time two numbers: seconds from window open to first good-quality tags, and the steady update period after that.
- Repeat step 2 with RSLogix 500 offline, then during typical programming activity. A large difference between the two convicts the shared segment.
- Read the processor scan time and the channel 1 servicing bits, S:2/15 and S:33/7, in the processor status screen.
- Change one factor at a time — servicing bits, SVC count, baud, scan class — and repeat the same measurements after each change; revert anything that makes the controller less responsive.
Make these changes with the machine in a safe state and the program saved: changing the network rate or downloading logic briefly takes the DH+ HMIs and the SCADA view with it, and that must not be what stops the machine.
Should I raise the DH+ baud rate to 230.4K? #
Raise the rate only when every node on the segment — the 5/04, the DHRIO, and both RSView HMIs — supports 230.4K: every device on a DH+ network runs one common rate, so a single node that tops out at 57.6K decides it for all of them. The decision rule is the common denominator: list every node's supported rates and configure the highest rate all of them share. If everything shares 230.4K, set it — it multiplies segment capacity by four over 57.6K. If anything tops out lower, run that maximum and spend your effort on servicing instead. When you do change the rate, re-terminate the trunk to the resistor size the 1747-SG001 table shows for the new rate, and recheck the maximum cable length against the same table row.
Does a 1 second scan class guarantee 1 second updates? #
No: a scan class is a request rate, not a delivery promise — when the DH+ segment or the processor's servicing cannot complete the request volume each cycle, requests queue and your tags lag seconds behind the class you configured. Demand has to be compared to a ceiling, not to a setting. When a window opens, the client asks for every tag on that window at once; the burst queues behind whatever is in flight, and drain time is backlog ÷ service rate. RSLogix online adds requests on the same segment, which is why the offline and online measurements in the list above matter. How the Ignition side of the connection is set up is covered in connecting SLC 500 processors to Ignition SCADA.
What do S:2/15 and S:33/7 do to throughput? #
S:2/15 (Comms Servicing Selection) and S:33/7 (MSG Servicing Selection) throttle channel 1 message servicing: with either bit set the processor completes one communication request per scan, so your scan time becomes your response ceiling. On a processor whose job includes answering a SCADA client all day, clear both bits. The bit functions are defined in the SLC 500 instruction set reference manual (1747-RM001); the same addresses were used on both a 5/05 and a 5/04 in the field reports, so confirm the status file map for your processor when you set them.
Worked example with illustrative inputs: a 5/04 with a 50 ms scan and one servicing bit set services 1 ÷ 0.050 s = 20 requests per second. A 1 s direct scan class needing 20 requests per cycle demands 100% of that ceiling. Open a new window whose tags add a 40-request burst and the backlog takes 40 ÷ 20 = 2 s to drain — before any RSLogix 500 session adds its own requests. That is the same order of delay reported before the fix. Field reports put the gain from clearing the bits alone as minor, so pair the change with SVC instructions.
Where do I put SVC instructions? #
Put one after every large, contiguous section of rungs: the reported fix started with a single SVC instruction, improved further as more were added between sections, and ended at about 1 s updates from about 2.5 s without overload. An SVC instruction makes the processor service pending communication requests at that point in the scan instead of waiting for the scan's servicing point, so more service points shorten the queue.
The reported numbers: on the 5/04, updates moved from about 2.5 s to 1.0 s — the lag cut by (2.5 − 1.0) ÷ 2.5 = 60%. A separate 5/05 report measured driver overload falling from 170% to 60% on a 600 ms tag group — (170 − 60) ÷ 170 ≈ 65% — after adding SVC, with the bit clearing alone worth only a minor gain. Those are field results from those configurations, not a guaranteed ratio; the method is what transfers. There is no universal count: start with one per large program section, measure, and add where the lag persists. Inserting SVC instructions is a logic change — save the program, download it in a controlled way, and check scan time and machine response after each insertion; if a placement degrades anything, move or remove it rather than restoring the throttling bits.
When should I stop tuning and replace the 5/04? #
Stop when 1 s updates are the best the DH+ segment delivers for your tag count, or when bridge hardware fails: replace the 5/04 with a 5/05 on native Ethernet and keep the DHRIO in the chassis only while nodes still need DH+. The 5/05 takes Ignition's requests off DH+ entirely, and the DHRIO keeps messaging alive for what remains — in this network, the two RSView HMIs. Routing messages through that chassis is its own task, covered in CompactLogix MSG to SLC 5/04 through a DH+ gateway. When the last RSView node retires, take the DH+ segment and the bridge out with it. The reported end state did exactly that: the ENBT/A failed, and the answer was a 5/05 plus removal of DH+ and the ControlLogix bridge, not rebuilding them.
Questions asked
Do the terminating resistors change when I raise the DH+ baud rate?
Yes. The DH+ specification table in the SLC 500 system selection guide lists a different terminating resistor size for each baud rate, and the maximum cable length changes with it — only the 57.6K row carries the 3048 m (10,000 ft) figure. When you change the network rate, re-terminate both ends of the trunk to the size the table shows for the new rate and confirm every node, including the old RSView HMIs, is configured for that same rate.
Can the old RSView HMIs stay on DH+ while Ignition takes over?
They can, but they keep a share of the segment. DH+ is a token-passing network: each node waits for the token, so every HMI poll cycle consumes time the Ignition path does not get back. Keep them during the migration and count them in the load assessment, but the DH+ ceiling only disappears when the last node leaves the network.
How do I pick a scan class the 5/04 can actually serve?
Size the class from the measured service ceiling, not from what you want on the screen. If a 50 ms scan with a servicing bit set yields 20 requests per second, a class should demand no more than about half that — 10 requests per cycle — or window bursts and programming traffic will queue behind it. After clearing the bits and adding SVC instructions, re-measure and shorten the class only as far as the steady update period still matches the class rate.
Does clearing the servicing bits or adding SVC instructions cost me scan time?
Yes. Servicing communication between scans is processor time that comes out of the scan, so measure the scan time before and after each change and use SVC placement to choose where in the program that cost lands. If the scan grows past what the machine response needs, remove or move an SVC instruction rather than restoring the throttling bits.