1. Introduction

The CryoSnap Thermoelectric Development Platform is an all-in-one evaluation kit designed to eliminate the "second build" in thermal system development. It allows engineers to rapidly assemble, power, and control a complete thermal system on the bench, and then break it apart to embed directly into their final product.

To achieve this flexibility, the CryoSnap platform is built as a five-module snap-apart system:

1. Thermal Module: The physical cooling stack, including a Sheetak CENTUM® series Peltier module, heatsink, fan, and mounting hardware.
2. Power Module: Handles USB Power Delivery (PD) negotiation and input protection, providing a stable high-power DC bus up to 24 V.
3. Micro Module: The system controller, featuring an Arduino-compatible ATmega328P (Nano) and precision analog signal conditioning for thermistors.
4. Interface Module: Provides real-time visual feedback and manual control via addressable LEDs, an optional OLED display, a potentiometer, and tactile switches.
5. Driver Module: The core power electronics stage, featuring a bidirectional H-bridge with precision current and voltage sensing.

This document serves as a deep-dive technical guide into the architecture of these modules, with a primary focus on the Driver Module. It explains the rationale behind the specific component selections, how the modules interact to provide safe and precise thermoelectric control, and how developers can leverage this architecture in their own production designs.

Figure 1: CryoSnap system architecture showing the full power path from USB-C PD input or DC screw terminal through the HUSB238 PD negotiator, TPS55288 buck-boost converter, INA226 power monitor (measurement only, power passes through unchanged), DRV8701E gate driver, and H-bridge to the TEC load. The H-bridge bypass terminal provides a direct connection path from the INA226 to a TEC for unidirectional applications. The ATmega328P MCU coordinates all I²C devices and issues a single GPIO direction signal to the DRV8701E.

2. System Architecture Overview

The CryoSnap platform is designed around a clear separation of concerns. Power enters the system, is negotiated to the correct voltage, is regulated to a safe ceiling, is measured for telemetry, and is finally directed through an H-bridge to the TEC.

The power and control flow follows this path:

1. Input Negotiation (HUSB238): The Power Module negotiates a high-power contract with a USB-C PD source.
2. Voltage Regulation and Current Ceiling (TPS55288): The Driver Module uses a buck-boost converter to provide a regulated voltage to the TEC while strictly enforcing a hardware current ceiling.
3. Current and Voltage Measurement (INA226): A dedicated monitor chip measures the actual power flowing to the load in real time.
4. Direction Control (DRV8701E): A gate driver controls a discrete MOSFET H-bridge to set the polarity (heating or cooling) without requiring complex microcontroller PWM generation.
5. System Control (ATmega328P): The Micro Module orchestrates the entire process, reading temperatures, adjusting the voltage setpoint via I²C, and toggling the H-bridge direction.

This separation of concerns allows the system to achieve exceptional precision in current regulation while maintaining the ability to rapidly switch between heating and cooling modes without risking component damage or requiring complex firmware timing.

3. The Five Modules in Detail

3.1 Power Module: USB-PD Negotiation (HUSB238)

The Power Module is responsible for securing the necessary power to drive the TEC. It utilizes the HUSB238 ^[5], a standalone USB Power Delivery sink controller.

By default, the HUSB238 is hardware-configured via a resistor network on its VSET pin to request 20 V at 3.25 A from any attached USB-C PD supply. This hardware fallback ensures that the board will negotiate a high-power contract immediately upon plug-in, even before the microcontroller boots.

Once the microcontroller initializes, it communicates with the HUSB238 over I²C to verify the contract. If the supply supports a higher current (e.g., 20 V at 5 A), the firmware writes to register 0x08 to select the 20 V PDO and triggers negotiation via register 0x09 to secure the maximum available power budget. If 20 V is not available, the firmware warns the user via the serial console but allows the system to continue operating. The downstream TPS55288 buck-boost converter is capable of driving the TEC as long as the input supply provides at least 12 V.

For applications where USB-C is not desired, the Power Module also includes a fused DC screw terminal input, allowing developers to bypass the HUSB238 entirely and supply up to 24 V directly from a bench supply.

3.2 Micro Module: System Control and Signal Conditioning

The Micro Module serves as the brain of the CryoSnap platform. It is built around an ATmega328P ^[6] microcontroller in the standard Arduino Nano footprint. This choice ensures that the firmware is highly accessible, compiles without complex toolchains, and allows developers to easily swap the Nano for a more powerful compatible MCU if their application requires it.

A critical function of the Micro Module is reading the NTC thermistors used for temperature feedback. Thermistors are highly non-linear, and reading them directly with a standard microcontroller ADC often results in poor resolution at the extremes of the temperature range.

To solve this, the Micro Module employs a TLV9004 ^[7] quad operational amplifier. Three of the op-amp channels are used to buffer and condition the signals from the cold-side, hot-side, and ambient NTC thermistors. The circuit utilizes a precision resistor network based on a Texas Instruments linearization topology ^[1], which flattens the NTC response curve before it reaches the ATmega328P's ADC. The fourth channel of the TLV9004 provides a stable 3.3 V analog reference (AREF) to the microcontroller, ensuring that fluctuations in the digital power rail do not corrupt the temperature readings.

Adjusting for Different NTC Resistors: The default linearization network is tuned for standard 10 kΩ NTC thermistors. If a developer needs to use a different thermistor (e.g., 100 kΩ), the resistor values in the voltage divider network must be recalculated to maintain a linear response curve across the desired temperature range. Texas Instruments provides a detailed application note (SBOA323A) outlining the exact mathematical procedure for calculating these values ^[1].

3.3 Interface Module: Real-Time Feedback

The Interface Module provides immediate, PC-free visibility into the system's state. It features a chain of 23 WS2812B ^[8] addressable LEDs. These are logically divided into a 10-LED temperature bar, a 10-LED setpoint bar, and three discrete indicators for mode, enable state, and H-bridge direction.

The module also includes a header for an optional 0.91-inch I²C OLED display. When connected, the OLED provides a dense telemetry readout, showing the firmware version, current mode, temperature setpoint, actual cold-side and hot-side temperatures, and real-time voltage, current, and power draw.

3.4 Thermal Module: The Physical Stack

The Thermal Module contains the physical components under test. The CryoSnap kit includes a Sheetak CENTUM® series Peltier module, mated to an optimized aluminum heatsink and an active cooling fan. The fan speed is controlled via a hardware PWM signal from the Micro Module, and its tachometer output is polled to verify operation.

As an example of the thermal capability, the kit can be equipped with the Sheetak SKHC1-071-06, a compact single-stage TEC designed for efficient thermal performance in tight spaces.

3.5 Driver Module: Precision Power Delivery

The Driver Module is the most complex subsystem in the CryoSnap platform. It is responsible for translating the microcontroller's digital commands into the high-current, bidirectional DC power required by the TEC. The remainder of this document focuses exclusively on the architecture of this module.

4. Why TECs Require Current Control

Thermoelectric coolers are not simple resistive loads. Driving one with a fixed voltage produces unpredictable results, and understanding why is the foundation for understanding the CryoSnap driver architecture.

4.1 The Problem: Back-EMF and the Seebeck Effect

When current flows through a TEC, it pumps heat and builds a temperature differential (ΔT) across its ceramic plates. Because of the Seebeck effect, that ΔT generates a reverse voltage that opposes the drive current. This is back-EMF. As ΔT grows, back-EMF grows with it.

The practical consequence: if you apply a fixed 12 V to a TEC and it develops 1.5 V of back-EMF, the effective driving voltage drops to 10.5 V. Current drops. Cooling power drops. And because ΔT is itself a function of the ambient environment, the back-EMF changes continuously as room temperature shifts, the heat load changes, or the heatsink fan varies in speed. A fixed voltage drive cannot compensate for any of this.

There is a second problem at startup. A TEC at rest has no back-EMF and presents a very low DC resistance. Applying full supply voltage before any ΔT has built up causes a large inrush current spike that stresses the TEC and wastes energy from the supply.

4.2 The Solution: Control the Current, Not the Voltage

If the current through the TEC is held constant, the thermal pumping power is consistent regardless of back-EMF. The driver simply adjusts its output voltage upward as back-EMF builds, maintaining the same current. Startup inrush is eliminated because the current is clamped from the first moment power is applied.

Figure 2 shows the difference in practice. The dashed line is constant voltage drive: an overcurrent spike at startup followed by a long drift downward as back-EMF builds. The solid line is current-ceiling drive: the current is clamped immediately and the output holds at the target for the life of the run.

Figure 2: Conceptual comparison of constant voltage drive (dashed) versus current-ceiling drive (solid). Constant voltage produces an overcurrent spike at startup and a gradual loss of cooling power as back-EMF builds with growing ΔT. Current-ceiling drive clamps the current from the first instant and holds it at the target throughout the run. How the CryoSnap driver achieves this is described in Section 5.

5. TEC Driver: TPS55288 Buck-Boost Converter

Section 4 established that a TEC driver must hold current constant and prevent inrush at startup. This section explains how the CryoSnap Driver Module implements that using the TPS55288 ^[2], a high-efficiency synchronous buck-boost converter. The TPS55288 was designed for USB Power Delivery and programmable power supply applications, but its architecture maps directly onto the requirements of a TEC driver.

5.1 How the TPS55288 Implements Current-Ceiling Drive

The TPS55288 is fundamentally a programmable constant-voltage source. The microcontroller writes a target output voltage over I²C, and the chip holds that voltage. What makes it suitable for TEC drive is its hardware current ceiling: a precision shunt resistor in series with the output produces a small voltage drop proportional to current, and the TPS55288 monitors that drop continuously.

When the TEC tries to draw more current than the programmed ceiling, the chip immediately throttles its output voltage downward until the current falls back to the limit. It does not shut down or latch off. It simply reduces voltage until current is back in bounds, then holds there. As back-EMF builds and the TEC's effective resistance rises, the chip raises its output voltage to compensate, keeping the current steady. The result is the flat solid line in Figure 2: constant current from startup to steady state, with no firmware PID loop required.

This is also why the TPS55288 handles startup inrush automatically. At power-on, the TEC has zero back-EMF and very low resistance. The chip starts from zero volts and ramps up, hitting the current ceiling almost immediately. It then holds at the ceiling as voltage climbs slowly, never allowing the inrush spike that a fixed-voltage source would produce.

5.2 Selecting the Sense Resistor

The current ceiling is enforced by measuring the voltage drop across a precision shunt resistor (R_sense) placed in series with the output. The TPS55288 uses a fixed internal reference threshold of 50 mV for its maximum current limit.

The maximum possible current ceiling is determined by the physical resistor chosen:

The microcontroller can program the ceiling to any value below this maximum in 256 discrete steps via the I²C interface. The resolution per step is:

This provides a critical design advantage: Scalability via Sense Resistor. The current range and resolution can be hardware-scaled simply by swapping the external current-sense resistor. The default 10 mΩ resistor provides a broad range suitable for large TECs. Swapping this for a higher value resistor increases the resolution at the low end, allowing the exact same architecture to precisely drive micro-TECs.

Figure 3: Complete Driver Module schematic. The TPS55288 (U9, lower left) provides regulated voltage with a hardware current ceiling via sense resistors R26 and R32 (0.01 Ω each, effective 5 mΩ in parallel). The DRV8701E (U8, upper center) drives the four 40N06D N-channel MOSFETs (Q5–Q8, upper right) that form the H-bridge. The Diode-OR circuit (D30, D31, upper center) ensures the DRV8701E supply is always at least 12 V, even when the TPS55288 is outputting a low voltage for micro-TEC applications. The INA226 (U7, lower right) monitors current and voltage on the V_TEC rail via shunt resistors R26 and R32. The H-bridge bypass screw terminal (J29, right) connects directly to the V_TEC rail, allowing the TPS55288 output to drive a TEC directly without the H-bridge while still being monitored by the INA226. Connectors J26 (left) carry power and I²C signals to and from the Power and Micro modules.

6. Real-Time Telemetry: INA226

While the TPS55288 enforces the safety ceiling, the microcontroller still needs to know how much power the TEC is actually consuming to implement advanced control algorithms or provide user telemetry.

This visibility is provided by the INA226 ^[4], a precision bidirectional current and power monitor.

The INA226 sits on the V_TEC rail, between the TPS55288 output and the H-bridge input. It measures the voltage drop across a dedicated shunt resistor, as well as the absolute bus voltage. The INA226 communicates with the microcontroller over I²C, providing highly accurate, digitized readings of voltage, current, and total power. The firmware polls these values every 100 milliseconds.

6.1 Dual Sense Resistor Selection Guide

Because the system uses two separate sense resistors (one for the TPS55288 hardware ceiling and one for the INA226 telemetry), developers must select appropriate values for both based on their target TEC.

The INA226 has a maximum shunt voltage range of ±81.92 mV. The maximum measurable current is therefore:

The table below provides recommended resistor values for various Sheetak TECs to optimize both the hardware ceiling and the measurement resolution.

6.2 The H-Bridge Bypass Terminal

The Driver Module includes a dedicated screw terminal (J29) that allows developers to bypass the H-bridge entirely and connect a TEC directly to the regulated output. This is useful for applications that only require cooling and do not need bidirectional control.

Because the INA226 is positioned upstream of both the H-bridge and the bypass terminal, it continues to provide accurate current and voltage telemetry regardless of which connection method the developer chooses.

6.3 Power Budget Management

The INA226 also plays a critical role in system stability. The microcontroller continuously compares the actual current measured by the INA226 against the negotiated power budget from the HUSB238. If the TEC attempts to draw more than 95% of the available USB-PD current, the firmware intervenes to prevent the upstream power supply from collapsing.

7. Polarity Control: DRV8701E H-Bridge

To actively heat and cool, the polarity of the voltage applied to the TEC must be reversible. This is accomplished using an H-bridge composed of four discrete N-channel MOSFETs.

7.1 The Gate Drive Challenge

Driving an H-bridge requires more than just logic signals from a microcontroller. The two "low-side" MOSFETs connect the load to ground, and their gates can be driven by a standard voltage. However, the two "high-side" MOSFETs connect the load to the supply voltage. To turn on an N-channel MOSFET, its gate voltage must be significantly higher than its source voltage.

When a high-side MOSFET is conducting, its source voltage floats up to near the supply voltage. Therefore, the gate must be driven to a voltage higher than the main system supply. A standard microcontroller GPIO pin cannot generate this voltage.

7.2 The DRV8701E Solution

The CryoSnap architecture solves this using the DRV8701E, a dedicated brushed DC motor gate driver ^[3]. The DRV8701E abstracts the complexity of the H-bridge away from the microcontroller.

The selection of the DRV8701E is driven by the need for reliability, simplicity, and cost-effectiveness:

• Integrated Shoot-Through Protection: A critical failure mode in H-bridges is "shoot-through," where both the high-side and low-side MOSFETs on one half of the bridge turn on simultaneously, creating a direct short to ground. Do that once, and the MOSFETs are fried. The DRV8701E features integrated dead-time handshaking, which automatically prevents shoot-through. This makes the system "idiot-proof" and removes the burden of managing precise microsecond delays from the user's firmware.
• Asymmetric Gate Drive for Cost Optimization: The DRV8701E includes an integrated doubler charge pump to generate the high voltage required for the high-side gates. Conversely, it uses a linear regulator to drive the low-side gates at a standard, lower voltage. This asymmetry allows the design to use standard, lower-cost MOSFETs for the low side, rather than requiring expensive, high-voltage-tolerant MOSFETs for all four positions.
• Simplified MCU Interface: By utilizing the hardware (GPIO) control mode, the interface between the microcontroller and the H-bridge is reduced to a single pin for direction control, conserving valuable MCU resources.

The DRV8701E is used in a deliberately simplified configuration. Its Enable (EN) and Sleep (NSLEEP) pins are hardwired high, meaning the driver is always active. The microcontroller controls the TEC polarity using a single GPIO pin connected to the Phase (PH) input.

When the firmware needs to change the TEC direction, it executes a safe transition sequence: it disables the TPS55288 output, waits 5 milliseconds for the bus voltage to settle, toggles the DRV8701E direction pin, and then re-enables the TPS55288. This ensures that the H-bridge never switches under heavy load.

7.3 Low-Voltage Operation and the Diode-OR Circuit

A unique challenge arises when driving micro-TECs that require very low voltages (e.g., 3 V). The DRV8701E requires a minimum supply voltage to operate its internal charge pump. If the driver were powered solely by the V_TEC rail, a 3 V output from the TPS55288 would cause the DRV8701E to lock out, and the charge pump would fail to generate enough voltage to turn on the high-side MOSFETs.

The Driver Module solves this with a Diode-OR circuit. The DRV8701E's supply pin is fed by both the V_TEC rail and the main 12 V system rail through Schottky diodes. The driver automatically draws power from whichever voltage is higher.

When driving a micro-TEC at 3 V, the 12 V rail takes over. The DRV8701E's charge pump doubles this 12 V to approximately 24 V, providing more than enough gate drive to fully enhance the high-side MOSFETs against the 3 V drain voltage. When driving a large TEC at 20 V, the V_TEC rail takes over, and the system operates normally.

8. Alternative Architectures and Trade-Offs

During the development of the CryoSnap kit, several alternative architectures were evaluated. Understanding why these were rejected highlights the strengths of the chosen design.

1. Basic H-Bridge with Overcurrent Protection (No Voltage Control): Many standard H-bridges feature overcurrent protection that simply cuts power when a threshold is exceeded. Because a TEC is a low-resistance load, applying a high voltage immediately triggers this protection. The bridge shuts off, resets, turns back on, and immediately trips again. This results in the TEC being continuously spammed with the full supply voltage in short bursts, providing no effective control and potentially damaging the module.
2. H-Bridge with Passive Dampening Circuit: A dampening (or snubber) circuit could be added to the output of an H-bridge to smooth the voltage and current application. However, the required component values for a dampening circuit are highly dependent on the specific resistance of the TEC. A circuit tuned for a 5 Ω TEC would not work correctly for a sub-1 Ω TEC. This lack of adjustability contradicts the goal of a universal development kit.
3. MCU-Driven PWM with External ADC Feedback: It is possible to use a basic H-bridge and control the current by having the microcontroller output a PWM signal. This requires the MCU to constantly read the current (via an ADC) and adjust the PWM duty cycle using a PID algorithm. This approach places a heavy processing burden on the MCU, requires an external high-resolution ADC, and forces the user to recalculate PID tuning parameters every time a different TEC is connected.
4. Dual Half-Bridge Drivers: Using two separate half-bridge driver chips instead of a single full H-bridge doubles the required MCU control pins and increases board space. Crucially, it removes the integrated shoot-through protection, requiring the user to perfectly manage the dead-time delays in firmware to avoid destroying the MOSFETs.
5. Center-Tap / Single-Chip TEC Drivers: Specialized, single-chip TEC drivers that utilize a center-tap power supply arrangement typically max out at relatively low currents (e.g., 6 A), making them unsuitable for larger TECs. They often lack granularity at the low end of their current range, making them poor choices for driving micro-TECs. The center-tap requirement also effectively halves the available voltage range.
6. Fully Discrete Custom H-Bridge: Designing an H-bridge entirely from discrete components (MOSFETs, 555 timers for charge pumps, boost circuits, etc.) requires a massive increase in component count and board complexity. It replicates functionality that integrated ICs already perform more reliably and efficiently, while adding significant points of failure.

8.1 Architecture Selection Guide

While the TPS55288 + DRV8701E architecture provides maximum flexibility, simpler or cheaper architectures may be appropriate for specific production use cases.

9. Conclusion

The CryoSnap Driver Module architecture is designed to be the definitive starting point for thermoelectric system development. By separating voltage regulation, hardware current limiting, precision telemetry, and gate drive into dedicated, best-in-class components, it provides a robust and flexible foundation.

Developers can use the CryoSnap platform to prove out their thermal concepts on day one, characterize their exact power requirements using the built-in INA226 telemetry, and then confidently copy the proven driver schematic directly into their production designs.

References