Understanding Embedded System Architecture Quickly

Typical Embedded System Block Diagram

Typical AP System Block Diagram

Let’s first introduce what an AP system is. AP stands for Application Processor. Our modern devices like smartphones, TVs, and set-top boxes are basically built on this architecture, which is very similar to PC systems. In the two diagrams above, the solid parts are essential components of a system, while the hollow parts are peripheral functional devices that are not essential and can be replaced based on functionality.

So, if we only consider the system composed of the solid parts in the diagram, we call it a “minimum system.” The minimum system is the simplest form that ensures the system can operate; of course, the minimum system generally has no functionality, just that the software can run on the MCU or AP. For example, one of the greatest physicists of the 20th century, Stephen William Hawking, became a “minimum system” after he became paralyzed due to amyotrophic lateral sclerosis. However, with his will and love for theoretical physics, he continued to use his brilliant mind (powerful MCU) to ponder the most profound questions in the universe. Maybe this is not the most appropriate example, but I believe that anyone who has graduated in engineering can understand it.

Let’s look at the differences in the minimum system; they are actually very obvious. It seems that the embedded system’s minimum system does not have external program memory and RAM (both SRAM and DDR belong to RAM). So, that’s all?

On the surface, that’s all! However, training engineers needs to be a bit more detailed. You must know the difference between RAM and program memory. RAM stands for Random Access Memory; don’t be fooled by the term “random access,” it merely expresses that you can read and write any storage unit (like a byte) in RAM at will. It’s like a hotel; you can freely enter any room on any floor (of course, provided you are the boss and the guests won’t complain).

In contrast, program memory (like Flash) cannot access individual storage units randomly; it can only access them in bulk, often reading and writing in pages or blocks. It’s still like a hotel; you can only access one floor at a time (at this point, you might be the cleaner). There are many types of program memory, commonly Nor-Flash, NAND-Flash, hard drives, etc. I will use Flash and program memory interchangeably later, so please don’t be confused.

RAM also has many types, like SRAM, SDRAM, DDR, etc.; I won’t elaborate too much here. Unless otherwise specified in this book, RAM refers to SRAM (Static Random-Access Memory). Besides the difference in access methods, there is another important physical characteristic: the content in RAM will be lost when the system is powered off; in contrast, the content in Flash will not be lost. Therefore, these two types of memory are also called: volatile memory (like RAM) and non-volatile memory (like Flash).

At this point, you must be feeling conflicted about why it’s so complicated. Can’t we just use one? Let’s see what our options are. If we use all Flash, due to the slow access speed of Flash, especially when writing data, it is incredibly slow compared to the MCU’s running speed, which will slow down the entire system. What if we use all RAM?

The data cannot be saved when powered off, and there’s another fatal problem—cost! You can compare the prices of RAM and hard drives in your computer; if you compare the price per unit capacity, you will find they differ by hundreds of times. So, combining these two is the most cost-effective solution. Life is all about constant trade-offs!

Layering of Different Memories

What content must be preserved after power loss? Of course, it is the program, which is why I equate Flash with program memory (some data also cannot be lost, but we care more about programs). So what does RAM hold? Naturally, it holds frequently accessed variables and “time-sensitive” programs. Later, we will encounter related content, but we won’t go into depth here.

Returning to the previous topic, from the perspective of the minimum system, embedded systems and AP systems seem to differ only in whether the memory is placed outside the chip. Is it necessary to differentiate so much? Well, I mentioned earlier that system architecture relates not only to hardware but also to software. Now let’s look at the differences in software. Because embedded systems generally have built-in memory (some MCUs, to save costs, will place Flash externally), the memory capacity is usually quite small.

So, they generally do not adopt most operating systems that people commonly encounter, like Android, Linux, or Windows. Operating systems consume a lot of system resources; besides the requirements for CPU clock speed, the most important is the requirement for memory capacity. I remember that with the continuous upgrading of Windows, our computer hardware also keeps upgrading.

Generally, embedded system software has two modes: one mode runs without any operating system, which saves system resources the most, but when the software becomes a bit more complex, running bare-metal becomes difficult. So, there is another mode that uses RTOS (Real-Time Operating System), which is specially designed for embedded systems. Its characteristics are simplicity, low resource consumption, fast response, and strong scalability.

At this point, you should have a vague sense of why we need to distinguish between embedded systems and AP systems; it’s just a small change in the minimum system components that contains the wisdom (compromise) of countless people, which relates to both hardware and software. Here, I want to emphasize that in real projects, the choice of system architecture involves not just the above issues; many factors are involved, including technical and cost aspects. Here, I will only outline the work to give everyone a preliminary impression and basis.

More Exciting Content