Contact Us

STM32H742VGT6

Brand: ST
STM32H742VGT6 is a ARM Cortex-M7 high-performance microcontroller from STMicroelectronics, designed for high-performance drone autopilots (Cube Orange+/Purple), ArduPilot/PX4 flight controllers, and embedded AI edge computing. Superb Automation supplies genuine STM32H742VGT6 with full traceability, competitive pricing, and fast global delivery — supporting prototype to production volumes for drone, robotics, and industrial customers worldwide.
quote now

Product Specifications

STM32H742VGT6 — ARM Cortex-M7 Microcontroller by STMicroelectronics

The STM32H742VGT6 is an STMicroelectronics ARM Cortex-M7 MCU in LQFP-100 for Embedded control and general-purpose MCU applications. Supplied by SUPERB Automation.

Part Number Decoded — Field by Field

FieldValueMeaning
STM32H742VGT6Full Part NumberSTMicroelectronics ARM Cortex-M7 Microcontroller
STM32Product FamilySTMicroelectronics 32-bit ARM-based MCU
H7SeriesARM Cortex-M7 based MCU — H7 series

Quick Specifications

ParameterValue
ManufacturerSTMicroelectronics
Part NumberSTM32H742VGT6
CoreARM Cortex-M7
PackageLQFP-100
Temperature Range−40°C to +85°C
Supply Voltage1.8V–3.6V

Product Overview

The STM32H742VGT6 is an STMicroelectronics ARM Cortex-M7 microcontroller in LQFP-100 package. Part of ST's extensive STM32 portfolio, it delivers a high-performance processing platform with integrated Flash, SRAM, and a rich set of communication peripherals for Embedded control and general-purpose MCU applications. Supplied by SUPERB Automation with full traceability and competitive lead times.

Key Features & Benefits

ARM Cortex-M7 Core — Efficient Processing for Embedded Control

The ARM Cortex-M7 core provides deterministic real-time performance with hardware multiply/divide and optional FPU, executing control loops, communication stacks, and sensor fusion algorithms efficiently in resource-constrained embedded environments.

Rich Peripheral Set — GPIO, Timers, ADC, Communication Interfaces

Features a comprehensive set of on-chip peripherals including multi-channel DMA, 12/16-bit ADCs, advanced PWM timers for motor control, and multiple USART/SPI/I²C interfaces for sensor and communication module connectivity.

ST Ecosystem — CubeMX, HAL, LL, Middleware

Fully supported by ST's STM32Cube ecosystem including CubeMX graphical configuration, HAL/LL driver libraries, FreeRTOS integration, and extensive middleware stacks (USB, TCP/IP, FATFS, graphics) — accelerating firmware development and reducing time-to-market.

Target Applications

Embedded Control & IoT Systems

Functions as the main processing unit in embedded control systems, IoT endpoints, sensor hubs, motor controllers, and human-machine interfaces — leveraging the Cortex-M architecture for efficient, low-power operation.

Industrial Automation & Motor Control

Powers industrial motor drives, PLCs, sensor transmitters, and factory automation equipment — taking advantage of advanced PWM timers, high-speed ADCs, and CAN/RS-485 communication for distributed control architectures.

Consumer & Portable Electronics

Integrates into consumer devices, wearables, smart home products, and portable instruments — benefiting from the STM32 family's low-power modes, small packages, and extensive third-party firmware ecosystem.

Frequently Asked Questions

Is the STM32H742VGT6 supported by STM32CubeIDE?

Yes, the STM32H742VGT6 is fully supported in STM32CubeIDE with graphical pinout configuration, clock tree setup, and automatic peripheral initialization code generation — enabling rapid prototyping and production firmware development.

What firmware/RTOS options are available?

The STM32 family is supported by FreeRTOS, Zephyr, NuttX, ChibiOS, and bare-metal HAL programming. For drone applications, Betaflight, iNav, ArduPilot, and PX4 all have STM32 HAL targets.

What is the minimum order quantity (MOQ)?

Contact SUPERB Automation for a tailored quote. We support quantities from prototype tray to full production volumes.

Request a Quote

SUPERB Automation supplies genuine STMicroelectronics STM32H742VGT6 MCUs with full traceability and batch-level QA documentation. Whether you need a single tray for prototype builds or pallet-level volumes for production, we source through verified channels with competitive lead times. Contact us with your target quantity and delivery date — we'll respond with a detailed quotation within 24 hours.

STM32H742VGT6: a 480 MHz M7 with 1 Mbyte of flash in 100 pins

The V is the 100 pin LQFP at 14 mm by 14 mm, G is 1 Mbyte of flash, T is the leaded body and 6 is the range of minus 40 to plus 85 degrees C. That combination is the entry point of the H7 line for a board that wants the 480 MHz core and the external memory interfaces but cannot use a ball grid array in its assembly process.

What sits behind the code is a Cortex-M7 with a double precision floating point unit, 16 Kbytes of data cache and 16 Kbytes of instruction cache, 1027 DMIPS at 480 MHz, and a memory arrangement of 1 Mbyte of RAM split into 192 Kbytes of tightly coupled memory and up to 864 Kbytes of user SRAM. The figures in this page are read from the STM32H742xI/G data sheet.

The memory map is the reason to choose this part

CoreCortex-M7 at up to 480 MHz, double precision FPU, 1027 DMIPS
Cache16 Kbytes of instruction cache and 16 Kbytes of data cache
Flash1 Mbyte on the G density, with read-while-write support
TCM RAM192 Kbytes, 64 Kbytes of ITCM and 128 Kbytes of DTCM
User SRAMUp to 864 Kbytes, plus 4 Kbytes in the backup domain
External memoryQuad SPI to 133 MHz, flexible controller to 32 bit with SDRAM at 100 MHz
Supply1.62 V to 3.6 V for the application and the I/O
RegulatorInternal LDO with a configurable output, six voltage scaling ranges
BackupSeparate regulator at about 0.9 V, battery charging on VBAT
USBDedicated 3.3 V regulator for the internal PHYs
PackageLQFP100, 14 mm by 14 mm
TemperatureMinus 40 to plus 85 degrees C
Mission profile optionA, advanced profile at VOS1 and 480 MHz

What tightly coupled memory buys a control loop

The 64 Kbytes of ITCM and 128 Kbytes of DTCM are not cache. They are memory at the core clock with a fixed access time, so a current loop, an encoder decode or an interpolation table placed there does not depend on a cache hit to meet its deadline. On a servo drive running a 20 kHz loop, that is what turns a worst case latency from a distribution into a number, and it is the reason the H7 shows up in motion control rather than only in instruments.

The user SRAM is where the rest lives: buffers for a network stack, a frame store for a display, a scatter table for a data log. Up to 864 Kbytes is enough that a 480 by 272 display at 16 bits per pixel fits in internal memory, which removes an external SDRAM from a product that would otherwise need one. Where the buffer is larger than that, the flexible memory controller takes a 32-bit SDRAM at up to 100 MHz, and the quad SPI takes a serial device at up to 133 MHz for the cases where pin count matters more than latency.

Getting the split right is a decision made at the first board review, not at the end of firmware development.

Two caches bring a class of bug that does not exist on a part without them. A buffer written by a DMA engine is not in the cache, so a read through the cached address returns stale data until the line is invalidated, and a buffer the core has just written may still sit in the cache when a peripheral reads memory. The fix lives in the firmware and in the memory protection unit, which marks the buffer region as non cacheable or write through. On a board with Ethernet, USB and an SD card all moving data by DMA, that configuration belongs in the initialisation and is not an optimisation.

The density decision is about images rather than about code size. One megabyte holds a bootloader, an application, a parameter region and one update slot with tight margins. Two megabytes holds them with room for a second image and a log. Where the update comes over a cable from a host, one megabyte is enough; where the device must keep the previous image for a rollback it cannot otherwise perform, the larger array is the cheaper answer in engineering time. Our MCU cross reference page sets out the memory configuration of the neighbouring H7 members, because the pin compatible parts do not all carry the same RAM.

Power, and what the regulator does about it

The core runs from 1.62 V to 3.6 V and the internal LDO has six voltage scaling ranges, which is how the part trades frequency for current in a way the application can control at run time. A product that spends most of its life at 100 MHz and bursts to 480 MHz for a computation does not need the same scaling as one that runs flat out, and switching ranges is a register write rather than a board change.

The current at 480 MHz is not a single number, because it depends on the scaling range, the active peripherals and the temperature. What the document gives is the table, and that table is what the thermal budget is built on: the LQFP100 body sits at roughly 40 degrees per watt junction to ambient, so a part dissipating 0.6 W runs 24 degrees above the air inside the enclosure before any margin is applied. On a sealed housing that margin decides whether the top frequency is available in summer.

There is a second thermal path worth measuring on this part. The core is not the only heat source: the internal LDO dissipates the difference between the input and the scaled core voltage, and a board running from 3.3 V with the regulator in linear mode turns that drop into heat inside the package. Moving to an external buck for the core rail is a common change on a sealed product, and it is worth costing before the enclosure is tooled rather than after the first thermal chamber run.

Boot time is the hidden cost of a large image. The core starts from internal flash, the bootloader brings up the external memory, and only then does the application begin executing from that device. On a product that must be ready 200 milliseconds after power is applied, the external device initialisation and the cache warm up are tens of milliseconds each, and the workable answer is to keep the first stage internal and bring the external memory up in the background.

The backup domain has its own regulator at about 0.9 V and VBAT supports battery charging, so an RTC and a set of backup registers survive a main supply loss without an external charger circuit. The USB block has a dedicated 3.3 V regulator for the internal PHYs, which means a board that runs the core at 1.8 V can still present a compliant USB port.

Questions that come with a first order

Is the plain code a 480 MHz part? The document lists an A option tied to the advanced mission profile at VOS1 and 480 MHz, and the plain string runs the standard profile. The clock the core can hold depends on the voltage scaling range and the temperature, so the number to quote is the one for the profile on the order, and the A option is the one to specify when the application assumes the top frequency at the top of the range.

Is the 100 pin body enough for the external bus? It brings out the memory controller and the quad SPI, but not every byte lane of a 32-bit SDRAM, so a board that needs the full width has to move to the 144 pin or the ball grid array. Our component obsolescence management page is where we record the migration path for a family like this, since a board that is built around the widest body in a line has somewhere to go when a density is discontinued.

That step is worth knowing before the board drawing is frozen, and the pin to pin cross reference page lists which H7 bodies keep the same land pattern and which only keep the same pin count. The difference shows up at reflow, not at the quotation.