Do Product Design and Engineering From First Principles For Extreme Performance or Lowest Cost Or Lowest Power Consumption or Great User Friendlines

by Sanjay Mishra

We are often called in to redo product design and engineering for products designed by average product design consultancies. These products work somewhat but fail to meet the overall requirements in terms of size or cost or weight or user friendliness. One common thread that we see running through all these failed products is that these product design companies used design blocks that they were familiar with and assembled them into a product. I will call this the "Assemble a Box Approach" to product design development as opposed to a design from "First Principles Approach".

Often these designs revolve around an off the shelf single board computer (x86 or ARM), an off the shelf untweaked generic OS (Linux, Windows CE - for some reason BSD users tend to be more sophisticated :-) flame suit on), and generic network communication TCP/IP on top of 802.11B or Ethernet in a generic enclosure (usually a Hammond or Bud box). These are great choices for doing a proof of concept and validating a concept or for doing limited run production but are often not the best choices when you are mass producing a specialized product or trying to achieve the best performance in a specific area.

Here are some advantages of the "Assemble a Box Approach"

Here are some advantages of the "Assemble a Box Approach"

• Quick design
• Low risk
• Lower cost if you have very small production volume

Advantages of a "First Principle Design Approach"

• Lower manufacturing cost
• More user features
• Lighter
• Lower power consumption
• Better performance

I am not saying that you design every element of the design from first principles. You just examine the entire design, and see what the optimal design would be if you had the freedom to use the most current tools, chipsets and an infinite amount of development time and money. Of course, you don't have any of these :-) but this is an important exercise as it frees your design from the tyranny of the past and the familiar. It is a fun exercise too.

Of course you can only do this exercise if you know the domain and can visualize and architect every aspect of the product from networking to power modules to user interface to packaging.. If you feel that you don't really really know the domains get a product design consultant, an electronic design consultant, a RF consultant or a packaging design consultant. The small amount of money that you spend on an experienced product design consultant at this stage can end up saving you a lot of money down the road.

After doing this exercise you will have several options for each of the design elements - the controller electronics, the display, the software, the power subsystem, the communications. The key is to evaluate each of the optimal design elements against an off the shelf version. The key factors would be development cost and risk versus benefits of lower production cost, lower power consumption or more user features.

Here are some examples of products that we have redesigned to great effect.

Custom wireless data acquisition device: A military customer came to us with a specialized wireless device used to identify faults in and repair military vehicles. The device had a precision Analog to Digital Converter, a controller and a wireless communication interface. The previous vendor had literally shrunk a PC down to fit into a module. It used an x86 chip, a DOS on a chip, FPGAs and was a six board design. We shrunk it down to a single board precision 24 bit ADC based design with a low power micrcontroller with a custom kernel controlling the set up. We also designed a custom mold for the enclosure. The size of the device, its cost and its power consumption all went down by a factor of 5 or more.

Custom RF modulation scheme and protocol: Another customer had licensed frequency bands and were using a generic data modem based on binary Frequency Shift Key(FSK) modulation. They were barely able to attain a throughput of a few hundred bits per second. We redesigned the entire RF module to use a 4FSK modulation with raised cosine filter, wrote a custom, efficient Time Division Multiplexed protocol, and a mesh networking scheme. The client is now achieving throughputs of about 6500 bps in the same 12.5 kHz RF channel with mesh networking turned on. I don't think his throughput is going to be beaten in the market for quite some time :-)