Modbus Scaling & Data Conversion Guide

Overview

Modbus registers are 16-bit integers. They don’t inherently carry unit information, decimal points, or scaling factors. A register value of 2500 could mean 25.00°C, 250.0 PSI, 2500 RPM, or just the number 2500 — the register map documentation defines the interpretation.

Converting raw register values to meaningful engineering units is one of the most error-prone steps in Modbus integration. This guide covers the common scaling methods, how to identify them in device documentation, and how to configure them in gateway devices.

Modbus scaling pipeline — raw register value through scale factor and offset to engineering units

Common Scaling Methods

Method 1: Fixed Multiplier / Divisor (Most Common)

The raw register value is divided (or multiplied) by a fixed factor to produce the engineering value:

Engineering Value = Raw Value × Multiplier

or equivalently:

Engineering Value = Raw Value ÷ Divisor

Raw Value	Divisor	Scale Factor	Engineering Value	Unit
2500	100	×0.01	25.00	°C
2500	10	×0.1	250.0	PSI
2500	1	×1	2500	RPM
1234	10	×0.1	123.4	kW
456	100	×0.01	4.56	pH

[!TIP] The most common divisors in building automation are 10 and 100. When a register map says “resolution: 0.1” or “decimal places: 1,” divide by 10. When it says “resolution: 0.01” or “decimal places: 2,” divide by 100.

How to Identify the Scale Factor

Look for these clues in the device register map documentation:

Documentation Says	Scale Factor	Example
”Resolution: 0.1”	÷10	Raw 250 → 25.0
”Resolution: 0.01”	÷100	Raw 2500 → 25.00
”×10” or “Multiplier: 10”	÷10 (to get back to real value)	Raw 250 → 25.0
”Scaled by 100”	÷100	Raw 2500 → 25.00
”Integer, no decimal”	÷1 (no scaling)	Raw 72 → 72
”Units: 0.1°F”	÷10, value is in °F	Raw 720 → 72.0°F

Writing Scaled Values

When writing setpoints back to a device with scaling, apply the inverse operation:

Raw Value = Engineering Value × Divisor

Example: To set a temperature setpoint of 23.5°C on a device with divisor 10:

Raw Value = 23.5 × 10 = 235

Write 235 (0x00EB) to the setpoint register via FC06 or FC16.

[!CAUTION] Always verify the acceptable range before writing. If the register accepts 0–500 (representing 0.0–50.0°C), writing a raw value of 1000 (100.0°C) may cause exception code 03 (Illegal Data Value) or damage the controlled equipment.

Method 2: IEEE 754 Floating Point (32-Bit)

Some devices store values as 32-bit IEEE 754 floating-point numbers, spread across two consecutive 16-bit holding registers:

Byte	Content
Bits 31–23	Sign (1 bit) + Exponent (8 bits)
Bits 22–0	Mantissa (23 bits)

A 32-bit float spans 2 Modbus registers. The register order varies by manufacturer — see Modbus Data Types & Byte Order Reference for byte/word order details.

Register Order	Description	Also Called
Big-endian (AB CD)	High word first, high byte first	”Normal”
Little-endian (CD AB)	Low word first, high byte first	”Word-swapped”
Byte-swapped (BA DC)	High word first, low byte first	Uncommon
Fully reversed (DC BA)	Low word first, low byte first	Some Asian manufacturers

Example: 25.5°C as IEEE 754 Float

25.5 in IEEE 754 = 0x41CC0000

Register Order	Register N	Register N+1
Big-endian (AB CD)	`0x41CC`	`0x0000`
Little-endian (CD AB)	`0x0000`	`0x41CC`

[!WARNING] If you read two registers and the resulting float is an impossibly large or small number (like 1.2e+38 or 3.4e-43), the register order is almost certainly wrong. Try swapping the two register values.

Method 3: Signed Integer (Two’s Complement)

Modbus registers are unsigned by default (0–65,535). For negative values (e.g., below-zero temperatures), devices use two’s complement signed representation:

Raw Value (unsigned)	Raw Value (hex)	Signed Interpretation
0	0x0000	0
100	0x0064	100
32,767	0x7FFF	32,767 (maximum positive)
32,768	0x8000	-32,768 (minimum negative)
65,435	0xFFAB	-85
65,535	0xFFFF	-1

How to Detect Signed Values

If a register represents temperature, pressure, position, or any value that can be negative, it’s likely signed. If you read a value around 65,500 when expecting a small negative number, the value needs signed interpretation.

Conversion: If the unsigned value is > 32,767, subtract 65,536 to get the signed value.

Signed = Unsigned - 65536  (when Unsigned > 32767)

Example: Raw register = 0xFFAB (65,451 unsigned)

65451 - 65536 = -85  →  with divisor 10  →  -8.5°C ✅

Method 4: 32-Bit Integer (Long)

Energy meters, flow meters, and counters frequently use 32-bit unsigned or signed integers across two registers for values that exceed the 16-bit range (0–65,535):

Format	Range	Use Case
32-bit unsigned	0 to 4,294,967,295	Energy totals (kWh), flow totals, runtime hours
32-bit signed	-2,147,483,648 to 2,147,483,647	Bidirectional totals (import/export energy)

Like floats, the register order (high word first vs low word first) varies by manufacturer. See Modbus Data Types & Byte Order Reference.

Temperature Conversion

Building automation frequently requires temperature conversion between °C and °F:

Conversion	Formula
°C to °F	`°F = (°C × 9/5) + 32`
°F to °C	`°C = (°F - 32) × 5/9`

Common Temperature Scaling Table

Device Reports	Raw Value	Divisor	Engineering Value
Temperature in 0.1°C	235	10	23.5°C
Temperature in 0.01°C	2350	100	23.50°C
Temperature in 0.1°F	745	10	74.5°F
Temperature in 0.1°C, signed	65501	10	(65501-65536)/10 = -3.5°C

[!NOTE] Some devices report temperature in °F while the BACnet side expects °C (or vice versa). When configuring a protocol conversion gateway, apply both the scaling divisor AND the unit conversion in the point map.

Scaling in Gateway Configuration

QuickServer Scaling

When configuring a QuickServer point map:

Mapping Step	Parameter	Example
1. Read raw register	Source address, FC	FC03, Register 0
2. Apply data type	Signed/unsigned, 16/32-bit	Signed 16-bit
3. Apply scale factor	Multiplier or divisor	÷10
4. Apply unit conversion	°F→°C, PSI→kPa, etc.	(°F-32)×5/9
5. Write to destination	Target BACnet object, register, etc.	BACnet AV, Present Value

Common Pitfalls

Mistake	Result	Fix
Forgot to apply divisor	Value is 10× or 100× too large	Check register map for scale factor
Applied wrong divisor	Value is close but off by a factor	Verify divisor against a known reading
Wrong sign interpretation	Negative values appear as ~65,000	Configure as signed 16-bit
Wrong register order for 32-bit	Value is impossibly large or garbage	Swap register order
Scaling applied twice	Value is 100× or 10,000× too small	Check if the device already provides scaled values

Verifying Scaling

Sanity-Check Workflow

Read the raw register value using CAS Modbus Scanner or equivalent tool
Compare against a known reference — read the device’s display panel, check with a handheld meter, or compare to a known setpoint
Apply the documented scaling and verify the math: Raw ÷ Divisor = Expected Value ± tolerance
Test extremes — verify scaling holds at both high and low ends of the range (not just midrange)
Test negative values (if applicable) — confirm signed interpretation works for below-zero readings

Example Verification

Device documentation: “Register 100 = Room Temperature, 0.1°C resolution”

Test	Raw Value	Calculation	Expected	Verified?
Normal range	235	235 ÷ 10 = 23.5°C	~23.5°C on display	✅
Cold room	55	55 ÷ 10 = 5.5°C	~5.5°C on display	✅
Below zero	65516	(65516-65536) ÷ 10 = -2.0°C	~-2.0°C on display	✅

Modbus Data Types & Byte Order Reference — byte/word order for 32-bit values
How to Read a Modbus Register Map — interpreting manufacturer documentation
Modbus Register Addressing: Modicon, PDU, and Page-Based Conventions — addressing conventions
Modbus Exception Codes & Error Handling Reference — exception 03 for out-of-range writes
Modbus to BACnet Protocol Conversion Guide — scaling in protocol conversion

Chipkin Tools

CAS Modbus Scanner — Read raw registers and verify scaling against known values
QuickServer — Protocol conversion gateway with built-in scaling, offset, and unit conversion
QuickServer — Multi-protocol gateway supporting complex scaling expressions per point
Chipkin Support — Register map analysis and scaling configuration assistance