A Twin-CPU Laptop You Can Rewire With Software program

When the house laptop revolution arrived, it crammed my childhood with fascination and impressed me to check laptop engineering. I needed to design a microcomputer to my very own specs. However in school I used to be by no means taught how an entire laptop system was put collectively. As a substitute we studied varied subsystems and the speculation of issues like digital sign processing and so forth. Someone, some other place, would at all times be liable for assembling the entire system and making the whole lot work collectively.

This was unlucky and unjustified: Placing an entire working laptop collectively isn’t tough, and it may give college students essential early confidence of their means to dwell as much as the label “laptop engineer.” So, having just lately retired from the high-tech business, I made a decision to design a didactical however absolutely purposeful laptop that would function a platform for studying and experimenting with system-level design points—the Cerberus 2100.

I didn’t wish to commit Cerberus to a specific CPU, as doing so would conflate system-level structure ideas with the precise timings and management indicators of that CPU. A lot as a software-engineering course focuses on the construction of an algorithm relatively than the syntax of its implementation in a specific language, I needed Cerberus to give attention to the system-level construction. Cerberus is thus a multi-CPU system, that includes each a Z80 and a W65C02S (6502), two well-known workhorse 8-bit processors that featured prominently within the home-microcomputer period. There’s a wealth of sources accessible for studying find out how to program these processors, that are highly effective sufficient to be helpful and entertaining, but easy sufficient to grasp.

The issue, in fact, is that these two CPUs function with very totally different interfaces to different components of the pc, akin to reminiscence or enter/output gadgets. As an example, the 6502 makes use of a single management line to point whether or not it’s studying or writing to the info bus, whereas the Z80 makes use of two traces. This implies the 6502’s sign must be mixed with the sign from the system clock, through an AND gate, to stop reminiscence miswrites, whereas the Z80 has no such challenge. Additionally, the Z80 has an output line to sign that the worth on the tackle bus is secure, a perform absent within the 6502. And so forth.

These variations imply that I couldn’t use a typical management bus within the Cerberus. As a substitute, I used a big complicated programmable logic machine (CPLD) chip I dubbed “Fats-Spacer” to translate the management indicators of every CPU into an abstraction layer. This layer defines the system structure. Fats-Spacer then interprets the output of the abstraction layer into the suitable enter indicators for every part within the system. These two steps of translation entail each Boolean logic and timing management via flip-flops. I used a CPLD as a substitute of an FPGA (field-programmable gate array) as a result of, not like FPGAs, CPLDs have a hard and fast propagation delay whatever the Boolean logic carried out in them. That is essential as a result of it permits customers to make adjustments to the system structure—by reprogramming the CPLD—with out having to fret that the complexity of their adjustments will take too lengthy to go via a series of logic gates, and so miss the timing home windows imposed by the system clock.

Due to its inside abstraction layer, Cerberus is uniquely appropriate for enlargement: A direct reminiscence entry (DMA) enlargement port can be related to Fats-Spacer. By immediately permitting entry to system reminiscence, I let the person add much more CPUs and microcontrollers to the system through the enlargement port.

One other essential design problem was to decouple the logic of the pc from the timings of the video circuitry

One other essential design problem I confronted was to decouple the system-level logic of the pc from the timings of the video circuitry. Historically, these two are tightly tied collectively in order to coordinate entry to video and character reminiscences by the CPU and show circuitry with out inflicting conflicts or artifacts. However with two CPUs and the DMA enlargement port, this wasn’t an choice.

As a substitute, Cerberus makes use of two dual-ported static RAMs (SRAMS) as video and character reminiscences. Every port permits asynchronous entry to the reminiscence’s contents. One port of every SRAM is related to the pc correct, whereas the opposite is unique to the video circuitry.

The Z80 and 6502 processors use totally different management indicators to interface with reminiscence and interface chips. A reprogrammable logic chip, dubbed Fats-Spacer interprets these indicators as required. One other reprogrammable logic chip handles storage and the keyboard interface, whereas a 3rd generates video indicators. James Provost

Regardless of the dual-ported reminiscences, onscreen glitches may nonetheless occur if the video circuitry was to learn from a given tackle as the pc wrote to that very same tackle. Thankfully, dual-ported SRAMs present a “BUSY” sign to point a battle. This sign is utilized by Fats-Spacer to pause the CPUs at some stage in the battle. The management abstraction layer comes very helpful right here too, because it already has the suitable translation logic for pausing the CPUs.

Fats-Spacer isn’t the one CPLD in Cerberus: Three of them represent the system’s core chipset. Fats-Cavia repeatedly scans the video and character reminiscences, and sends bitmaps to Fats-Scunk, which then generates the suitable RGB indicators and synchs pulses to create a 320-by-240-pixel VGA output. In the meantime, as we’ve seen, Fats-Spacer gives the glue logic. Lastly, there’s an extra chip: Fats-Cat, which is definitely an ATmega328PB microcontroller. That is used to deal with I/O: The microcontroller manages a keyboard, buzzer, the enlargement protocol, and a microSD card for storage. The I/O firmware is held within the ATmega’s reminiscence, which means it leaves no reminiscence footprint within the 64 kilobytes of RAM accessible to the Z80 and 6502.

The Cerberus 2100 is an open {hardware} design accessible to all and full particulars can be found on my web site. However for individuals who don’t wish to construct their very own machine from scratch, I’m working with European electronics firm Olimex for the sale of a totally assembled model shortly. I hope it helps college students and hobbyists to know—and school to show—how an entire, absolutely purposeful laptop will be put collectively, whatever the goal CPU.

This text seems within the November 2023 print challenge as “Software program-Outlined Structure.”