Microchip Gives Its Latest dsPIC a 200 MHz, High-Speed Upgrade

The new line operates at 200 MHz—the highest dsPIC clock speed to date—and adds a double-precision floating point unit and enhanced DSP engine.

Today, Microchip announced the latest in its digital signal controller (DSC) digital signal processing (DSP) enhanced microcontroller (MCU) line, the dsPIC33A family. Microchip equipped the newest dsPIC MCUs with an improved core that runs faster and delivers higher precision math than prior offerings. The 32-bit MCU operates at 200 MHz, a first for dsPIC processors. It has a double precision floating point unit (DPFPU) as well as a full set of digital signal processing (DSP) instructions and up to 128 K of Flash for program storage.

New, higher performing dsPIC targeted at a broad variety of control applications. 

Unlike AI, where lesser precision math is often used to speed the process along, signal processing greatly benefits from higher precision floating point math. A DSP needs to sample at speeds higher than the frequency being analyzed, and it needs to perform highly precise calculations within the shorter time allowed by faster sampling rates. The 200-MHz clock speed and DPFPU address the need for faster signal processing in an increasingly digitized world. Along with the increased computational capabilities, the chips have faster analog-to-digital converters (ADCs) and higher precision pulse width modulation (PWM) outputs.

We spoke to Joe Thomsen, vice president of Microchip’s digital signal controller business unit, to learn more about the dsPIC33A core family of DSCs. 

The dsPIC33A at a Glance

Microchip will offer the dsPIC33A architecture (datasheet linked) in a family of MCUs, starting with the dsPIC33AK128MC1xx. The packaging is as small as 4 mm x 4 mm, and pin counts range from 28 to 64.

dsPIC33A family architecture. 

Some key features of the dsPIC33A include: 

• 200-MHz clock speed
• 128 K program Flash memory
• Digital signal processing engine
• Double-precision floating point math unit
• 12-bit ADCs with 40 mega samples per second (Msps) sampling rate
• High-speed comparators and op amps
• Eight channels of high-speed PWM
• Four configurable logic cells to reduce the need for external discreet logic

The dsPIC comes with a rich set of built-in peripherals. As with many other Microchip PIC microcontrollers, the peripherals are connected to an interconnect crossbar, affording flexibility when locating the I/O pins. The ability to choose the location of pins can make backward compatibility and PCB routing significantly easier.

Microchip's dsPIC evolution.

“We've been in the dsPIC business for 20 years. When we first started out, you'd be lucky to run a control loop at 50 kilohertz,” Thomsen said. “With the dsPIC33A, we believe that without too much effort, you can run a 2-MHz control loop for the same type of control algorithm that you did back 20 years ago.”

Improved Application Efficiency

Many dsPIC applications involve real-time control. They monitor environmental sensors, process the data, and utilize it in closed-loop or narrow-loop real-time applications. This is where the fast ADCs and math come in handy. Microchip increased the DSP engine data bus, registers, and instruction set from 16 bits to 32 bits for the dsPIC33A line. For math operations, prior generations relied on fixed-point math. The new engine uses a double-precision, floating-point engine.

“We enhanced all the DSP, the instructions, the engine, the multiplier, the accumulators—all for the wider data bus,” Thomsen said. “It gives you a lot more efficiency.”

Having floating-point capability in the processor also speeds up development time. Often, real-time control (RTC) systems are developed in concert with MATLAB, which performs modeling and code generation with floating-point math. Prior generations of dsPIC would require a manual conversion step between MATLAB’s floating point and the DSC’s fixed point. With the dsPIC33A, that conversion step is no longer necessary.

Processing power is not the only factor to consider with real-time applications. Interrupt overhead is one of the greatest challenges for real-time control. Any time spent handling an interrupt is time not monitoring a motor. This includes latency caused by saving registers on a stack and then later recovering them. To combat this issue, the dsPIC33A family has additional working registers. It’s much quicker to just switch register sets than deal with a stack.

“We really worked hard to make sure your interrupt latency is almost zero,”  Thomsen said. “As part of that, we've added a bunch of additional working registers so that when you get an interrupt, instead of pushing stuff onto a stack and then popping them when you get it back off again, you just switch register sets, and it reduces your interrupt latency time by an order of magnitude.”

Security and Software Tools

No advanced processing device today is truly safe from attack, and Microchip acknowledges that in the new line by including a number of security features. The DSC Flash memory is equipped with security features such as secure debugging, restricted memory access, and a hardware immutable root of trust (RoT). 

Automotive applications for the dsPIC33A.

Microchip set its security standards high, meeting the stringent automotive industry requirements (AEC Q-100) for performance, reliability, and manufacturing. All of the parts in the new family meet the same requirements, whether they are sold for automotive, industrial, or commercial applications.

Microchip fully supports the new dsPIC33A family with its MPLAB XC-DSC compiler and MPLAB code configurator (MCC). The MCC helps to configure the peripheral pinouts, peripheral configuration, and device-specific initialization code. Microchip offers the chip in development board form with the EV74H48A Curiosity development board.

“The dsPIC33A is very well suited for many of the wide band gap devices we see now. Gallium nitride and silicon carbide have very fast switching speeds,” Thomsen said. “As you start to look at supplies for AI servers or for onboard charges for electric vehicles, where efficiency really matters, having that very fast control loop makes a big difference.”