by miniature computer I mean things like TI graphing calculator, Palm handheld, dictionary gadgets and similar devices that have a small keyboard, small screen, low end cpu, nonvolatile memory and battery.
When it comes to desktop computers AFAIU components are standardized and available in bulk. So in the highly unlikely case of a new useful design appearing, you could just buy the components, open workshop and crank them out.
So how does it work in the world of handhelds? Is there an off the shelf component called “miniature keyboard for foldable dictionary” and another component “miniature keyboard for devices looking like Palm or TI”? Are variously sized screens and miniature wifi cards commoditized?
Or do newcomers in this field work by ordering design and production of their components from scratch, let’s say ordering a keyboard of particular size from the “keyboard design and manufacturing company”?
Some high tech things will use all custom components, but a lot of gadgets will use off the shelf displays. Lower end CPUs are almost always off the shelf. Some high end CPUs are off the shelf as well. One thing you are starting to see more of is the implementation of CPUs inside of FPGAs for more flexibility and lower cost. A low end calculator could have the keypad scanner, display driver, and CPU all realized inside of a single FPGA chip. You end up with a single chip design, with a few added components to drive the signals that the FPGA’s weak outputs can’t handle.
A lot of embedded CPUs these days have built in RAM and ROM. You can make RAM and ROM inside of an FPGA as well, though the limited space inside of an FPGA may force you to use external chips.
Something like a calculator keyboard will often be a custom design, though it may be based on an off the shelf design to save development money.
If you want to make some new fangled whiz-bang device, an awful lot of displays, keypads, embedded processors, and all sorts of things are commonly available. You can get a small wifi board, but wifi is just a single chip and a few support components these days so it is fairly simple to integrate one onto your custom circuit board. A lot of hardware design is just taking chips that do what you want and integrating them all together. Circuit design has almost moved back to the point of block diagram type of design instead of the old fashioned way of individual component design.
Essentially all high-volume electronic devices use custom cases and keyboards. They may be only slightly different than their competitors, but they won’t be “off-the-shelf.” Small volume products may use standard components, but they are going to look generic, which is not usually a big selling point.
Outside of industrial equipment, almost everything is made from custom parts now. There are some small scale production techniques that can be used, but they will raise the per-piece price of components significantly. For full scale manufacturing, you have to think about a minimum order that will fill a cargo container. For a small device, that would be thousands.
Usually functional and design prototypes are completed, and marketing plans are made with a major player as a partner before manufacturing comes into the picture.
Some of the internal interfaces may subscribe to standards or convention (for example, the method by which the unit addresses the display or keyboard, or RAM), but that’s just another way of saying that technology manufacturers don’t always reinvent the wheel. These aren’t bits you can yank out and transplant.
everybody, thanks for your responses. Let me try to further clarify and understand the statements that seem to be the crux of the matter.
Ok, so engineer_comp_geek says:
whereas TriPolar says:
My own summary of the above, which may be inaccurate both to their intent and to the physical reality, would be “yes, you can make the yet-another-handheld out of more-or-less-off-the-shelf components (possibly except for keyboard), but doing this would be too expensive because these components are more expensive than custom components”.
First of all, is my summary accurate?
Second, so why is it that custom components are so much cheaper or, specifically as put by TriPolar, “small scale production techniques … will raise the per-piece price of components significantly” ? In other fields, e.g. desktops, off the shelf components are cheap and widely available. What makes the handheld field different?
Is it, let’s say, due to the collective decision by all handheld producer companies to always use custom parts and police/prevent sale of these parts as generic components to other manufacturers? In other words, if some factory is mass-producing cheap “custom” components in bulk to sell to TI to make calculators, why can’t I buy small batches of these components from that factory and use them to make my own gadgets?
There is an issue of design and setup costs. If you use off the shelf parts, you save quite a bit on the design and setup costs, but the production costs are about the same.
Let’s say you want to make some whiz bang spiffy new thing. It’s going to cost you about $50,000 to set up the molds to make the case for your spiffy thing. After that, you can crank them out for 50 cents apiece. Your cost per unit varies quite a bit depending on how many you make. If you only make 1,000 units, your cost per unit is $10.50 ($10 for the setup costs spread out per unit and 50 cents for the actual production). If you make 100,000 units, your cost is only $1 per unit (50 cents per unit for setup and 50 cents actual production).
The issue that TriPolar brings up is that there are things that you can do if you are making small scale production, but you’ll drive the per unit cost up. Let’s say you want to make 1,000 units of something and your total cost needs to be under $15, and you already have $10 worth of electronics in it. Making a custom case for $10.50 as above would make the unit un-profitable. Let’s say you can’t find an off the shelf case that exactly fits you needs, but you can find one that can be modified to fit your needs. The place that makes these cases charges more per unit for smaller lots, so they won’t sell them to you for 50 cents each. Instead they cost $1.50 each. You spend an additional $0.50 per unit modifying that case to fit your needs. Your cost is now $2 per unit, but there’s no huge setup cost. So in a small production run you save money, but compared to large scale production your case costs more ($2 each compared to $0.50 each). If you suddenly get an order for 100,000 units (people really like your whiz bang thing), you may not have time to go back and make a custom case to meet your delivery, so you could end up spending more money.
That’s not the way it always works out, though. Let’s say you want a display for your whiz bang device. You can buy a standard LCD for $5 each. Or you can spend $200,000 developing a custom one, and then the company will make them for you, for $5 each. In this case there aren’t any savings at all from using a custom part, no matter how many you make. The more you make, the closer the cost comes to the standard off the shelf cost (since the per unit setup costs become smaller and smaller when spread out over more units) but you never beat the off the shelf price.
I have here on my desk a little pedometer that the company I work for handed out to all of its employees earlier this week (it’s not that they care about us and want us to be healthier - they are trying to reduce insurance costs). From the looks of things, it has a standard off the shelf LCD display and several standard off the shelf buttons on it that you use to program it (it will tell you how far you’ve walked but you have to tell it how long your stride is first, for example). The case looks to be a custom job. I’m sure they would have used an off the shelf case if one were available, but I doubt that they could find a case that would have looked pedometer-ish enough. A custom LCD and custom buttons would have cost more, and in this case unmodified off the shelf versions looked good enough and were cheaper to buy.
If you look at your typical cell phone though (like the one in my pocket) it has a custom keypad. This was done not to save money, but instead to make the cell phone look good. If the cell phone doesn’t look slick, no one would buy it, and generic off the shelf keypads generally don’t look slick. So, something like a cell phone is going to have a custom case and a custom keypad just so it doesn’t look clunky. What you will often see is that a cell phone manufacturer will use the same keypad in several different models of phone, just to spread out the development and setup costs.
There are a lot of different issues involved, and there’s a big difference between large and small volume production.
I personally am involved in low volume production (a few hundred units per year) which only works for very expensive things. Even so, a lot of effort goes into the producibility of everything that we make.
There are also a lot of “semi custom” parts. You can buy generic keypads with blank keys and have the silk-screened however you want. Most microcontrollers are Flash based these days, but you used to be able to send intel a .hex file and have them make you a batch of 8051s with your program built into the onboard ROM…not cost effective for small runs, but doable for medium scale stuff.
A lot of case type things are “catalog but built to order” Basically the catalog shows what has been done before, and tooling exists for, but there is no stock. Sort of like when you order a car with exactly the options you want.
Basically something that is built with off the shelf components will be bigger, heavier, and uglier than a full custom design. It will be cheaper in small lots, and probably more expensive in big lots, and the turn around time will be shorter.
Custom components are cheaper when they are made in very large quantities. There is high initial cost to setting up manufacturing of the component, but if that cost is spread over a million units can become insignificant. Also, the cost of raw materials is much lower when you make a million units because you buy the materials in very high quantities.
‘Small scale production techniques’ are means of making custom components in small quantities. For a molded part you could use low cost molds with liquid casting resins to make the pieces in small quantities. High volume manufacturing would use steel molds and high temp injection to make the same thing. Your molds will have a limited life, and the cost of your casting resins will be higher, and there will be more labor involved. For electronic circuits it would be soldering up a card with a number of discreet components instead of having a single custom chip made.
The handheld world is different because most handheld devices were made from custom components in the first place. The desktop world emerged out of a world where many one-off and small quantity products were developed and there was a market for generic components. Even there, except for low end brands, when generic components were used, cases and housings were usually customized. Your desk top may have exactly the same keyboard as many others, but the case it is mounted in may be custom made to match the color, texture, and style of the rest of the system. Also, desktop systems may use a lot of off-the-shelf cables, which handhelds won’t have. Handheld battery charging cables and adaptors are often generic components though.
A company like TI may restrict it’s suppliers from selling the components they purchase to others. Even if they are free to do so, the manufacturer may not be interested in making small quantities of extra components. The ones they make for TI are packed and shipped immediatly, and they may not have the space to store extras for potential sale. Overhead increases with each purchase also, so the price to you would be higher than what TI pays anyway, and they may not be interested in assuming the overhead of dealing with small companies at all.
Anyway, you usually can buy these components, they will just cost more. TryDigi-Keyto start with, and search the intertubes for other similar outlets. I haven’t dealt with Digi-Key for decades, but back in the day they would direct you to a source for anything they didn’t carry themselves.
While many elements of handheld systems are standard, the interconnects generally are not, so standard elements end up being integrated into a single custom system-board. It is not like a PC where all the components are standardised.
An example of a miniature computer gadget that was produced for sale is the Pandora Handheld Gaming Console. This was an opensource concept based on a prior design (the GP2X). While this used standard ARM System-on-a-chip and PowerVR Video designs, the case and circuit board were custom, and the entire process from concept and design through prototyping and production took many years (3-5, as I recall). Large scale operations have less lead time due to more resources to throw at the process, but I expect most consumer devices being sold now have had a 2-3 year lead time.
does engineer_comp_geek’s post #7 boil down to “the cost of making the plastic case and keyboard is actually a major cost factor when doing small production runs on low cost devices”? I.e. even if the internal components may be more or less off the shelf, the case itself can easily become a major issue?
If so, would it make sense to make and sell commoditized easily customizable cases, let’s say a kit of panels that can be assembled like Lego into cases of various forms and then glued or “baked” to finish off? Or is it the case that simply nobody in the industry has any desire to make small runs and so there would be no market for such a component?
is there a fundamental reason why that is the case? Why can’t interconnects be standard?
Also, maybe I fundamentally misunderstand the notion of “interconnect”. I always thought that let’s say to connect a screen to cpu you take a couple of cables that come with the screen and solder them to the cpu. So that sounds pretty standard, or at least the interconnect here is just part of the screen. Presumably if we use an off-the-shelf keyboard it will come with similar cables ready to be soldered into place.
So in what case would we have an unusual, nonstandard interconnect? Fight my ignorance, please.
That is done. You can buy cases, covers, and bezels of different sizes and colors with holes and other openings cut by automated equipment. But there aren’t many variations outside of a rectangular shape. The end result is something that looks about as good as it’s Lego representation. These days, more resources may be expended on determining the precise color, corner radius, and surface texture of the case than on designing the electronics. And I don’t think any of these cases are produced and sold in anywhere near the volume of IGizmos.
If you have a unique and useful, new type of device, with broad appeal, the volume manufacturing start up costs are a secondary factor. You need to make a working model based on any available components, then find a way to manufacture piece-meal based on whatever the costs may be so that you can attract venture capital to fund a major marketing and manufacturing effort. If your device has a narrow appeal, you won’t have competition from high volume manufacturers, and your manufacturing costs will be more in line with similar devices.
Also don’t forget that you need more than the device. You need to package it, provide documentation, have a quality control process, deal with returned products, and give it copyright/patent/trademark protection.
With a small device, you are aiming to directly match all the components, to minimise space, power and excess heat. So in a PC, the graphics card is capable of multiple screen resolutions, as is the display itself. This flexibility is great, but adds additional complexity and circuitry. It also requires a functionally flexible interconnect (VGA or DVI or HDMI). For a miniature device, you cannot do this. So you have a specific controller and specific display, and the manufacturers of those devices specify only the bits of interface that are required. Then the physical limitations of your design kick in (clam-shell, or flat, maybe) - you probably need flat ribbons, but you may be limited in length (due to the digital nature of the interfaces - no line drivers to save power etc). So you have to ensure that your overall design meets those physical requirements, and thus you need a customised mainboard to put the connectors exactly where they need to be.
This has been the case for years in the laptop world. While the connection between a laptop system board and screen is probably some sort of DVI interface, you (generally) cannot connect any laptop screen to any laptop, even with the same specifications. You have to have the manufactured specific screen. Back in the late 90’s we had Toshiba laptops at work. There was a design flaw in the screen, and they started to fail. Toshiba only had a certain number of spares, and the screen manufacturer would not tool up to make any more. We had to scrounge around for broken laptops with good screens to get the screens off them for replacements.
Because miniature devices are so specific and limited, and everyone wants to do something different, there are few standards, and customisation is the general rule. Some interfaces are standard I[sup]2[/sup]C is generally used for low throughput devices (keyboards, trackballs, buttons etc) with low power and cabling requirements.