Performance Optimization: Speed, Latency, and Throughput

Imagine you’re in a busy restaurant where the waiter takes too long to take your order, forgets items, and delivers food slowly. Frustrating, right? Similarly, a poorly optimized Modbus network can cause delays, missed data, and unreliable communication. In industrial systems, these issues can lead to production downtime, safety risks, and financial losses.

In this article, we’ll explore techniques to optimize Modbus network performance, covering baud rate calculations, polling strategies, message packing, timeout tuning, and network loading formulas. We’ll also provide rules of thumb for different applications.

Understanding Modbus Performance Metrics

Before diving into optimization, let’s define the key performance metrics for Modbus networks:

1. Speed

Impact: Higher speeds reduce transmission time but can increase noise susceptibility

2. Latency

Components:

– Transmission time (time to send the message)

– Processing time (time for devices to process the request)

– Propagation delay (time for signals to travel over cables)

Impact: High latency causes slow response times, especially in real-time systems

3. Throughput

Impact: Determines how many devices can be supported on a single network

4. Reliability

Impact: Low reliability causes data loss and system instability

Optimization Technique 1: Baud Rate Calculations

Baud rate is the foundation of Modbus performance. Choosing the right baud rate balances speed and reliability.

Understanding Modbus RTU Message Structure

A Modbus RTU message has a fixed structure:

2 bytes: CRC Check

For example, a Read Holding Registers (FC 03) request has 8 bytes:

“`

[Slave ID] [FC 03] [Start Addr Hi] [Start Addr Lo] [Qty Hi] [Qty Lo] [CRC Hi] [CRC Lo]

“`

Calculating Message Transmission Time

The time to transmit a Modbus RTU message depends on:

Serial protocol overhead (start bits, data bits, parity, stop bits)

#### Formula for Transmission Time

“`

Transmission Time (ms) = (Message Length × Bits per Byte + Overhead Bits) × 1000 / Baud Rate

“`

Where:

Overhead Bits: Same for all messages (11 bits total per byte)

#### Example Calculation

For a 8-byte Modbus RTU request at 9600 bps:

“`

Transmission Time = (8 × 11) × 1000 / 9600

= 88 × 1000 / 9600

≈ 9.17 ms per message

“`

Choosing the Right Baud Rate

| Baud Rate | Max Cable Length | Best For |

|—————|———————-|————–|

| 9600 bps | 1,200 m (3,937 ft) | Long distances, noisy environments |

| 19200 bps | 600 m (1,969 ft) | Medium distances, moderate speed |

| 38400 bps | 300 m (984 ft) | Short distances, higher speed |

| 115200 bps | 150 m (492 ft) | Very short distances, maximum speed |

Key Factors to Consider

1. Cable Length: Longer cables require lower baud rates

2. Noise Environment: Noisy industrial environments need lower baud rates

3. Number of Devices: More devices on the network may require lower baud rates

4. Message Frequency: High-frequency polling benefits from higher baud rates

5. Response Time Requirements: Real-time systems need higher baud rates

Optimization Technique 2: Polling Strategies

Modbus uses a single-master architecture, meaning the master must poll each slave sequentially. Efficient polling strategies can significantly improve network performance.

1. Priority-Based Polling

What It Is: Polling devices based on their importance

How It Works:

Priority levels are assigned based on application requirements

Example Schedule:

Low Priority: Poll every 2,000 ms (status indicators, non-critical data)

2. Staggered Polling

What It Is: Spreading out polling requests to avoid network congestion

How It Works:

Reduces peak network load

Benefits:

Improves overall network reliability

3. On-Demand Polling

What It Is: Polling devices only when needed

How It Works:

Use Modbus exception responses to trigger on-demand polling

Example:

The master immediately polls that sensor for detailed data

4. Batch Polling

What It Is: Combining multiple polling requests into a single network transmission

How It Works:

Reduces the number of network packets

Benefit: Reduces network overhead by up to 50%

Optimization Technique 3: Message Packing

Message packing reduces network overhead by combining multiple requests into fewer messages.

What Is Message Packing?

Instead of sending multiple small requests:

“`

Request 1: Read Register 0 from Slave 1

Request 2: Read Register 1 from Slave 1

Request 3: Read Register 2 from Slave 1

“`

Send one larger request:

“`

Request: Read Registers 0-2 from Slave 1

“`

Benefits of Message Packing

Better Reliability: Fewer packets mean fewer opportunities for errors

How to Implement Message Packing

1. Use Multi-Register Function Codes:

– FC 03: Read Holding Registers (1-125 registers)

– FC 04: Read Input Registers (1-125 registers)

– FC 16: Write Multiple Registers (1-123 registers)

2. Calculate Optimal Request Size:

– Balance between maximum register count and message transmission time

– For RTU: Limit to 256 bytes per message (Modbus specification)

– For TCP: Can use larger messages, but consider network MTU

3. Avoid Fragmenting Registers:

– Group contiguous registers in the same request

– Avoid reading non-contiguous registers in separate requests

Example: Impact of Message Packing

|————–|————————|——————–|———————————–|

| Individual Reads | 10 | 10 × 8 bytes = 80 bytes | ~91.7 ms |

| Packed Read | 1 | 1 × 12 bytes = 12 bytes | ~13.7 ms |

Optimization Technique 4: Timeout Tuning

Setting the right timeout is crucial for balancing responsiveness and reliability.

What Is Modbus Timeout?

The timeout is the maximum time the master waits for a response from a slave before considering the request failed.

Consequences of Incorrect Timeouts

Too Long: Delayed error detection, slower system response, reduced throughput

Calculating Optimal Timeout

#### Formula for Maximum Expected Response Time

“`

Timeout (ms) = Transmission Time (Request) + Processing Time + Transmission Time (Response) + Buffer Time

“`

Where:

Buffer Time: Add 20-50% to account for network delays and retries

#### Example Calculation

For a 8-byte request and 8-byte response at 9600 bps, with 50 ms processing time:

“`

Timeout = 9.17 ms (request) + 50 ms (processing) + 9.17 ms (response) + 20% buffer

= 68.34 ms + 13.67 ms

≈ 82 ms

“`

Best Practices for Timeout Setting

1. Test with Actual Devices: Device processing times vary significantly

2. Add Buffer Time: Always add 20-50% to calculated time

3. Implement Exponential Backoff: Increase timeout for retries to avoid overwhelming the network

4. Use Device-Specific Timeouts: Critical devices may need shorter timeouts

5. Monitor Timeout Errors: Track timeout rates to identify problematic devices

Network Loading Estimation

To ensure your Modbus network can handle the load, you need to estimate the maximum number of devices it can support.

Formula for Network Loading

“`

Network Loading (%) = (Total Polling Time per Cycle / Polling Cycle Time) × 100

“`

Where:

Polling Cycle Time: Desired time to complete one full poll of all devices

Formula for Maximum Devices

“`

Maximum Devices = (Polling Cycle Time × (1 – Desired Load)) / (Transmission Time per Device)

“`

Where:

Desired Load: Typically 50-70% to leave headroom for retries and unexpected traffic

Example: Calculating Maximum Devices

For a Modbus RTU network with:

Transmission Time per Device: 20 ms (request + response + processing)

“`

Maximum Devices = 500 ms / 20 ms per device = 25 devices

“`

Network Loading Guidelines

| Network Type | Maximum Loading | Recommended Loading |

|——————|———————|————————-|

| Modbus RTU | 80% | 50-60% |

| Modbus TCP | 70% | 40-50% |

| Mixed RTU/TCP | 60% | 30-40% |

Rules of Thumb for Different Applications

1. Factory Automation

Timeouts: 100-200 ms

2. Building Management Systems

Timeouts: 200-500 ms

3. Power Distribution Systems

Timeouts: 150-300 ms

4. Oil and Gas Pipelines

Timeouts: 300-500 ms

Practical Implementation: Optimizing a Modbus Network

Let’s walk through a real-world optimization example:

Scenario: A factory with 25 Modbus RTU devices on a single RS-485 network, experiencing slow response times and occasional timeouts.

Current Configuration:

Network Loading: 90% (high)

Optimization Steps:

1. Increase Baud Rate: Change to 19200 bps (cable length is 300 m, within limits)

2. Implement Message Packing: Read 10 registers per request instead of 1

3. Adopt Priority Polling: Poll critical devices every 200 ms, non-critical every 1000 ms

4. Tune Timeouts: Reduce timeout to 200 ms based on new baud rate

5. Segment Network: Split into two networks with 13 devices each

Results:

Reliability: Increased from 95% to 99.9%

Monitoring and Measuring Performance

To verify your optimizations, you need to monitor key metrics:

1. Tools for Monitoring Modbus Performance

Custom Python Scripts: Using pymodbus to log performance data

2. Key Metrics to Track

Timeout Count: Number of timeout errors per hour

Conclusion

Optimizing Modbus network performance requires a systematic approach, balancing speed, reliability, and network capacity. By carefully calculating baud rates, implementing efficient polling strategies, using message packing, tuning timeouts, and estimating network loading, you can significantly improve the performance of your Modbus network.

Remember that every network is unique. Start with the rules of thumb for your application, then measure and adjust based on actual performance data. With these optimization techniques, you can ensure your Modbus network meets the demands of modern industrial systems.

The key to successful optimization is continuous monitoring and iterative improvement. Regularly check performance metrics, identify bottlenecks, and implement targeted fixes. By doing so, you’ll maintain a fast, reliable, and efficient Modbus network that supports your industrial operations for years to come.