Viewpoint Systems

Reply
This is an open group. Sign in and click the "Join Group" button to become a group member and start posting.
Highlighted

Prototyping with Embedded Systems Part 2: Buy vs. Build Cost Analysis

In Part One, I talked about some of the options for prototyping new product ideas. Remember, this can be a new subsystem of an existing product or a completely new product idea.

Now, we'll analyze a hypothetical situation to illustrate making the product by buying off-the-shelf components or building with custom PCB. This hypothetical example is very useful to explore the many options to consider. Even so, it develops a perhaps surprising conclusion so read on.

As we mentioned in last month’s article, we are looking at two platforms for use in this discussion. The National Instruments RIO (cRIO or sbRIO) platform is the components-off-the-shelf (COTS) platform. The Microchip family of micro-processors is the custom-build approach.

For the purpose of this comparison pricing article, the hypothetical application is a data logging application with alarming, such as might be found in a condition monitoring product. A typical configuration is below. For this comparison, any enclosure, power, mounting, and so on to make this a complete product prototype is assumed to be the same for both approaches.

  • 4 Analog Channels, Simultaneous Sampling @ 10 kHz
  • 4 DI and 4 DO for situational monitoring and control
  • Temperature shutdown capability
  • No UI other than event status (OK, Warning, Error) via 8 DO status bits
  • Data logging for last 10 events (40000*2 bytes per event)
  • USB for data storage
  • Ethernet for email/FTP

This list is very basic and will serve to capture the discussion points.

NI RIO Platform

The cRIO platform is chosen for this implementation rather than the sbRIO. The sbRIO cost can be less than the cRIO, so any breakpoint in moving from COTS to custom is at a lower unit level than sbRIO. The main components of this system would be the (new!) cRIO-9076 4-module chassis, integrated with a controller and FPGA, the NI 9222 4-channel Analog Input module, and a couple of NI 9401 DIO modules plus cables, powers supplies and other miscellaneous parts. The cost associated with the hardware selected is on the order of $4k. Writing the LabVIEW software to meet the requirements would be on the order of $8k. This solution would also require the usual custom harnesses and connectors to interface to the sensors. The custom harnesses might be as simple as pig-tails on both ends through glands/grommets in the enclosure or as complicated as harnesses using bulkhead connectors in the enclosure wall. A rough cost might be about $1k per unit. These costs total to a unit cost of about $13k.

Microchip Platform

The Microchip PIC processor-based custom solution requires complete hardware design including component selection, schematic creation, and PCB layout design. The software would be written in “C” rather than LabVIEW. Note that some of the lower level interface software taken for granted in LabVIEW, such as handling email, may or may not need to be written for this solution, depending on the availability of free tools from Microchip or elsewhere. (Microchip often grants a non-transferable, non-exclusive license for their tools as long as they are used on Microchip processors.)For this application, the hardware design, schematic and PCB design would be on the order of $10k. Software design for this hardware would be on the order of $15k. The hardware component costs would be on the order of $165 (10 Pc) and lower for higher quantities. This gives a total cost of the solution on the order of $25K for the total custom solution.

Solution Cost Comparison

The table below is an estimate of costs per unit over a wide range of quantities.

Solution Cost Comparison

Or graphically for just the unit price rows:

Unit PriceThe breakpoint where the Custom approach is less per unit cost than the cRIO approach is at about 8 pieces.

Factors to Consider

High Level Observations

The data above shows for a simple application that the cRIO approach is less cost per unit for small quantities. Note that we have not included any discounting the NI might provide for cRIO in large volume, whereas we have for the custom approach since we have a good sense about Microchip’s volume discounting. But, even if we did include discounting from NI, they could never match the pricing of the custom approach: NI does not intend to compete with this large volume market (yet?).Nevertheless, the cRIO approach is very favorable for low volume. You could quibble about the estimation costs provided here, but I think it is obvious that there will be a breakpoint below which the cRIO would cost less and above which the custom route is less.The bullet lists below are meant to capture some of the considerations that will move that breakpoint. It is not trivial to know exactly where that breakpoint is at the start of a project, but guidelines are helpful.

NI RIO

  • The I/O may require special signal conditioning which may not be available with the DAQ modules. Some customization might be needed here regardless.
  • The connection schemes for the DAQ modules may not be adequate because of signal speed, channel-to-channel isolation or connector requirements.
  • This example doesn’t require the use of the cRIO’s FPGA for additional signal performance but having the FPGA with the cRIO is an added bonus if such a requirement exists.
  • LabVIEW software programming and debugging can be faster than other languages.
  • The engineering skillset required includes software LabVIEW Real-time and FPGA programming and signal conditioning knowledge.
  • The specifications offered by the DAQ modules may be a compromise from the specs you would actually like. These specs can include operating temperatures, noise, dynamic range, and so on.
  • The physical size of the cRIO may not work in the required solution.
  • The operating temperature may not be adequate.
  • If total unit volume increases, custom solution development may eventually be required. See the chart below.

Custom

  • The timeline for completion of the solution includes a hardware design phase. This additional step makes the complete design/build timeline longer than the RIO approach, but you are done whether you sell small or large unit quantities. See the chart below.
  • Software design is usually done in “C”. Microchip and others have a lot of example code to help minimize software design time.
  • The engineering skillset required assumes electrical design and PCB design knowledge along with embedded “C” programming.
  • This example does not need an FPGA for specific performance requirements, however adding the device to the PCB is not costly in component cost but may be in software development cost. Programming the FPGA would then be done in VHDL.
  • Since the PCB is completely custom, all signal conditioning and custom interfaces will be included and optimized on the PCB.
  • Other custom features like temperature dependency, noise performance, isolation, A/D or D/A dynamic range, or operating temperatures can be designed into the custom electronics.
  • The single PCB solution can be designed with size in mind to minimize total package size or meet particular package requirements.
  • Rework of hardware for design issues can significantly delay timeline and considerably increase design costs.

Timeline

This graphic shows a possible development timeline. Time increases to the right. Each box contains the effort to deliver the solution.

timeline

With the RIO approach, the solution can be deployed earlier than the custom approach. However, if the unit volumes increase to the point that unit costs need to be reduced and a custom PCB is needed eventually, than the RIO approach will eventually take longer than if you had started with the custom approach.

Conclusion

Solutions for prototyping are more numerous than discussed here. Each solution has advantages and disadvantages from one another and choosing that solution requires a broad sense of what is available. The particular application may sway the prototyping solution in one form versus another and a list of factors for consideration was given. The results between an off-the-shelf solution and a custom solution were analyzed and an estimate of the point to buy verses build was given for a single hypothetical scenario. Each specific application may change that decision point, so careful consideration should be given to your application requirements.

Nevertheless, it’s been shown that the RIO approach gets to market faster at lower unit cost than the custom approach when unit volumes are below some low-quantity breakpoint. The challenge is defining that breakpoint and that takes the experience of working with both approaches. Come talk to us. We can help with that decision.

0 Kudos
Message 1 of 1
(4,527 Views)
Reply
This is an open group. Sign in and click the "Join Group" button to become a group member and start posting.