I am a newbie to the field and I am kind off intrgued on how people manfucture small electronic devices. That is, say, I have an idea: to make a "calculator" kind of device that can take the amount of time you run as an input and gives the calories burnt as output.
How would I manufacture such a device ? Do I write a computer program, give it to some manufacturer (say, in China) or should I sit and design the hardware blocks ??
Any help would be greatly appreciated (I am kind of really curious :))
Sounds like there is virtually no hardware design to your proposed device, so any work on the design of the hardware you could do (as a neophyte) would not likely be of any help, and the manufacturer might misinterpret your suggestions as constraints, thus increasing the overall cost.
You could ask a manufacturer to produce a finished product which implements pretty much any algorithm you could imagine. They would want a large order paid for by irrevocable L/C at sight or T/T in advance and/or NRE money up front. They would either get their own engineers to do the design or farm it out. You could also get a consultant to write the program and produce a prototype, which might allow you to retain more control over the details of your algorithm (perhaps not reveal it at all), and then deal with a manufacturer (it also helps with market testing and getting funding). You could hire your own industrial designer to produce a housing prototype or go with whatever the manufacturer wanted. There are many, many approaches which may be viable depending on how much money you have to put into the project, and how sure you are of the market.
I'm currently doing a couple of projects like this for established companies, which actually involve very little electronic design (unfortunately). I'm even in the process of getting a engineering post-graduate certification in plastic injection mold design to better deal with (and between) offshore tool and die makers and US and Canadian industrial designers. The process isn't so different from electronic design. A bit of math (not so much), some tricks, some fancy simulation, a fair number of standard components to select, and a lot of documentation. More custom stuff and more 3D CAD such as Solidworks or Catia. The end result is one piece worth maybe $25,000 (which can produce perhaps 1,000,000 parts) rather than, say, 2,500 pieces worth $10 each, but the design process isn't so different. They have this irritating habit of referring to "tenths of an inch" when they mean increments of 0.0001" (tenths of a thousandth of an inch).
Best regards, Spehro Pefhany
--
"it\'s the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Yes, I am asking how do I design the hardware. Assume later I want to manufacture a large quantity of them. Should I figure out what processors are needed (like ARM, FPGA ?) and try to "design" the hardware blocks ? Do I write a C++ code or some Verilog/VHDL code ? How would I figure out how to interact with memory and other devices (this neither looks like "plain" software design, nor "plain" hardware design). Also, since such products would have to be sold cheaply, I assume that the design process has to be a lot simpler than what semiconductor companies would do.
If I do this, as Spehro mentions, how would I know what a manufacturer would actually do ?
It's usually a two step process (at least): Step 1: Prototype your design. You'd typically use off-the-shelf components and hand build it on a breadboard or PCB. This proves that your concept works and gives you an actual hardware platform to write any requied software (microcontroller, FPGA etc) on. For more advanced stuff (high speed and complex stuff like a PC motherboard for example) simulation and real-world PCB design are required here.
Step 2: "Design For Manufacture". This is where you repackage and often redesign your product to use lower cost parts and shape your boards etc to get hte form factor you need. The goal is to make it easy for someone to manufactue for you, and to have the product look you want (important for consumer products for example). It is not uncommon these days for the "design for manufacture" part to take most of the time and effort.
For your exampe of a calorie calculator, it would require a microcontroller and it would work at low speed, so this is something you could easily cobble together a prototype for, even on a breadboard. You'd use say a $10 Flash microcontroller in the prototype, but in the final design you would switch to say a $0.50 OTP or mask micro in a different package and al surface mount parts to lower your volume cost. Then you need a fancy cutom case etc.
Most design engineers hate stuff that requires custom moulded cases and tooling, as then it basically become mechanical engineering, so we like just putting stuff in a standard Jiffy box and be done with it!
I don't have any experience with Catia, though I've heard it's not so hard to learn (seems to be popular with the automotive industry?). Pro/E is supposed to have a steeper learning curve (again, hearsay). Cost-wise Solidworks runs around $6-8K depending on options (more if you want the COSMOS FEA features). I think it's significantly less expensive than Catia (closer to $20K, IIRC).
If you've only used 2D CAD like Autocad, using a 3D parametric modelling package is like night and day. You can set up equations and/or Excel spreadsheets to determine relationships and configurations. Change a dimension of a feature anywhere in the tree and the model rebuilds with the new size. You can create a plane and generate an instant section view, rotate it, make it partially transparent etc. and see what's going on everywhere- helps to see if you've inadvertently created too thick wall sections that will result in sink marks in the molded product. The mold related features are perhaps still a bit weak, but getting significantly better every year. I think it's faster to learn than most 2D packages- just work through some tutorials and if you're like me you'll take to it like a fish to water. It has a Windows look-and-feel. AFAIUI, Solidworks is not really set up to help you generate complex compound curves (NURBS),
Best regards, Spehro Pefhany
--
"it\'s the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
It's low-end embedded system design. It would most likely use an 8-bit microcontroller and an LCD display. Firmware would most likely be written in assembler or possibly C. Probably from a manufacturer you've never heard of who mostly sells parts in die form.
If you invest a bit of time and money and take apart some technically similar products you'll quickly get an idea of what's what.
Interesting assumption.
Best regards, Spehro Pefhany
--
"it\'s the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Yes, you need to figure out what the requirements of your design are and choose a suitable processor or hardware solution. For your calorie calculator you would almost certainly use an 8bit microcontroller like an AVR or PIC. An FPGA is not the right solution in this case.
For high volume manufacture you would pick a microcontroller family which had (alomst) seamless transition from Flash (low volume, most expensive), to OPT (medium volume, cheap), to factory mask ROM (high volume, very cheap).
For a high volume display you might go for a custom LCD instead of say a standard dot matrix LCD.
Sounds almost entirely software to me. You would typically use C on a microcontroller, or assembler if you want to waste some extra development time.
Component cost may be a small part of it, you might find that your fancy case is the most expensive part, and has a large NRE setup cost.
1 - You may want do a google search to make sure something that does exactly the same thing is not available. This sounds like a pedometer with extra features like a "time mode". It also sounds like a "runner's watch".
2 - Make sure you define your spec closely. The number of calories burned depends on several factors:
a - weight b - gait c - pace d - terrain e - body temperature
3 - Perhaps you could do all of these things with a program written for a cell phone:
formatting link
4 - Or better yet you could just make it a subscription service that users called on their cell phone and then typed in the particulars using hot keys. Once it owrked you could sell it to Virgin or T Mobile and start on your next project.
In addition to the custom case, the most expensive NRE cost is the custom LCD. They won't talk to you for less than 1K. To reduce the cost, you probably want to mount the chip on the glass anyway. It might be easier to find an existing case with LCD of the right size, then design your LCD (w/ chip). We are currently work for a customer using an AVR with LCD, unfortunately, the AVR's concept of multiplexing is not the same as most LCDs. I guess there is really no standards in LCDs.
I've always figured that -- even though one has probably never heard of the manufacturer -- the CPU in question is usually still compatible with ubiquitous microcontroller cores such as 8031s, PICs, etc. True? Or do you get some "home spun" CPU designs?
It's a good idea to do a little market research first. See what already is out there, what their features are, and if you can compete.
Also a little bizness plan-- how much $ do you have, how much will staff, development, marketing, sales, distribution and customer service cost.
I suspect there are dozens of these devices already in existence. Last year our company gave those away as part for a walk-for-health program. They're unlikely to have cost more than 50 cents each in quantity.
Interesting. I never knew it can be made so inexpensively.
I accept that before I make the device, I will have to explore a lot of existing devices and make a business plan. Also, as Rich mentioned, may be I would have to be rich to take the risk and manufacture the device, which is currently not the case :)
Also remember that some ridiculous number like about 90% of new businesses fail within the first year or two. Luck has an awful lot to do with it, being in the right place at the right time with the right product and the right marketing.
Unless you hit some niche spot, you are unlikely to make any money from such a venture with a consumer level product. If you are lucky you'll come out of it without going bankrupt :->
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.