Hardware development department, DANNIE LT, Vilnius, Lithuania
Computer vision department, DANNIE LT, Vilnius, Lithuania
Abstract Using EDGE ML devices in business and everyday life is growing very fast. That’s why an additional value in development such solutions are decreasing a time-to-market stage. We could remember such technological boom in the beginning of IoT devices era. In fact, EDGE ML devices are evolutional step of IoT.
As it was in IoT development different vendors (Quectel, SIMcom etc.) released special modules (with NPU/DSP cores inside) for SMART devices and it is very good opportunity to decrease development project up to 40%.
In this article we will tell all pros and cons of using SMART modules as a base of HW solution for EDGE ML devices. Our experience is based on developed projects and market analyzing for last 2 years.
When it is a good idea. Pros
Every new development project for design houses (DH) has a lot of different requirements. Term is the most important one for customers from any market segment. An important advantage is to have hardware solution which helps to decrease time-to-market time and save money for a product launch stage.
Usually, there is a SoC chip inside SoM with a chipset. It can take a long time for a developer to go through all documents of stand-alone chips. A SoM developments helps to save time on it and decrease connection mistakes and BOM size.
Another important moment is that not every customer is ready to produce huge quantities of device in the first year. It could be a problem for a DH to choose HW platform because vendors from the “top-list” are not ready to waste time on supporting small projects.
According to that, growing popularity of SMART modules shows opportunity to solve both problems at the same time.
A module-based development could be the first step on the way of design customer’s EDGE ML devices, because it would be easier to remake HW using stand-alone chips after having a module-based project.
The simplicity of schematic and layout work using SoM gives another advantage – a development could be done by not very qualified engineer. A development with stand-alone chips means works with a BGA surface-mount packaging, with high-speed interfaces, with big amount of different supply rales. It is not a task even for middle HW developer.
SMART module vendors have a deep expertise in markets of potential clients and provide ready-to-use solutions for a convenience of developers:
- a wide range of wire and wireless interfaces;
- easy HW and SW integration of cameras;
- easy supply connection.
The main advantage for EDGE ML development is powerful SoCs inside modules with specialized cores to compute tasks of computer vision algorithms. For example, the latest versions of SIMcom modules are based on Snapdragon 6th generation which is very good for implementation high performance algorithms.
When it’s not a good idea. Cons
As SMART modules are an electronic assembly it should be based on stand-alone chips inside, that increases a size of final device a lot. As well as that, if the main value of future device is minimization using such modules can be impossible.
Problems with production are caused not only by the size, but also by the price. Without a doubt it should be said that using module increases bill of materials expenses. That could be satisfactorily for middle production orders but is going to be critical for bigger orders. For such projects better to have course on development without SMART module.
Another advantage of a “non-SoM” way is a bigger flexibility while work, because developers can choose every chip and every schematic solution by himself.
To sum up
Development projects are also a compromise between different technical and production aspects. And the right choice of a HW platform in the very beginning is a main key to success.
In DANNIE we are confident that SMART modules-based design has huge benefits for EDGE computer vision development and could be a valuable proposal for different clients.