Between Arduino and Raspberry Pi, are we approaching a point where a hardware startup can bootstrap? There's a lot of room for disruption in the home appliance sector that has been out of reach for the typical internet startup. Think garage door openers, doorbells, anything in your house that currently has a chip in it. Those companies are firmly entrenched in their way of doing business.
Building working prototypes is no longer a major problem. Nowadays, (experienced) individuals can design portable game consoles [1], alarm clocks [2], mobile phones [3], and pretty much everything else.
However, you need a lot of money, time and experience for manufacturing, testing, shipping, and supporting the final products.
Unlike software startups, hw startups may need to apply for FCC and international certifications, supervise contract manufacturers, and need to think carefully about liability issues.
I was going to reply but you've done most of it for me. The big costs are going to be in the manufacturing and quality control post manufacturing.
Also, if your product contains anything to do with RF you are in a whole other world where you'll need certifications of various sorts and lots more testing.
Aren't those phase 2 problems though? As a simple test to see if a market exists, why isn't it viable to sell your (polished) prototypes? One could make, (just spitballing) a doorbell that sends you an email as a hobby project, then throw it up on whatever the electronics equivalent of etsy is, and build them on build-to-order basis for the first few hundred units and then worry about investors only if your idea floats.
For the products we're talking about here, no competition exists. So I can sell my gimmicky doorbell for $80 or $90 just to see if there's any interest at all, then get investors and start building my own arduino boards from parts (since the specs are open) at scale.
Actually, there are a number of online shops that you can ask to manufacture and/or sell your design, e.g. [1,2].
However, manufacturing electronics in low quantities is quite expensive. Additionally, in order to legally sell this stuff in the EU, you need to get a WEEE certificate (i.e. someone needs to take care of recycling or disposing of your product) and a CE certificate (compliance to all regulations).
And you cannot easily fix a bug in your board design or even the firmware, once the product has been shipped to the customer. You have to get it right on the first try.
I agree that it has become relatively easy to start a hardware company. However, compared to a software startup, you need to invest a lot of money upfront. And so much more can go wrong that you have little control over.
I don't know a lot about startup economics but I have closely followed the fates of a number of hardware startups. For example, the Pandora guys got screwed over by different manufacturers, costing them a lot of time, money, and goodwill. WakeMate shipped dangerously defective USB chargers they had to replace. OpenMoko just did not find enough potential buyers for their phones (and screwed up in other ways).
Hardware startups do not fail because of a lack of prototyping tools.
why isn't it viable to sell your (polished) prototypes?
From the electronics stackexchange:
Certification with the FCC can cost around 10 to 20K. In the US, all products containing electronics that oscillate above 9 kHz must be certified.
I have been thinking the same thing for a while now.
I imagine they're great to create prototypes with and start up with, but as someone posted to my inquiry yesterday ( http://news.ycombinator.com/item?id=3011227 ), they're too expensive to go into production line.
Hmm, I'm not sure whether an ARM board is really a good idea.
For most applications, a normal Arduino is easily sufficient.
For anything more complicated, you can buy an ARM board running Linux (e.g the upcoming Raspberry Pi [1] for $25-$35).
Arduinos are primarily used for education and prototyping.
Programming raw ARM chips makes sense if you build a small series of specialized devices and want to keep costs down.
Given that the Arduino Due will certainly cost significantly more than $30, I see no market for such a board.
I'm more interested in the Leonardo with the 32u4 chip myself. Much smaller than the Arduino, simpler circuit, built-in USB for higher data transfer bandwidth and comes with lots of A/D's. The 32u4 has been _the_ chip for low-cost projects that involve communication with a computer for some time now (Adafruit and PJRC [Teensy] make one), but Arduino is going to bring it to a bigger audience. The Teensy already worked with Arduino libraries, but the boot loader wasn't free and it wasn't 'officially supported'.
There is a big difference between a Cortex-M3 and whatever is powering the Raspberry Pi. It's intended as an 8-bit microcontroller replacement. I've recently switched from TI MSP430 processors to this same Cortex-M3 from Atmel and performance has increased while per unit cost has gone down.
The Cortex-M3 microcontrollers are awesome. Much more powerful than the older 8-bit architectures available at the same price, and amenable to industry-standard tools instead of just the compilers and IDE sold by the manufacturer. I think this move just mirrors the general industry trend toward the ARM architecture. It's become difficult to justify using one of the old 8-bit architectures anymore, except at the lowest extremes of price or power consumption.
Correct. The ARM M series doesn't even have floating point hardware. The M3 is intended for real-time, low power applications, and is indeed a fantastic replacement for 8 or 16bit microcontrollers.
I want one, and I know several others that will as well. The 32-bit Cortex-M3 core at 96 MHz leaves an AVR in the dust; it has low-power modes that make power consumption comparable; and the peripheral set advantage of the AT91SAM3U over an ATmega is huge.
Also, ARM Thumb2 is a much nicer instruction set than AVR in my humble opinion. Cortex-M3 also has some great instructions to improve the efficiency of real-time preemptive kernels (e.g. RBIT, CLZ and automatic pushing of registers on interrupt). I've written my own tiny kernel for this microcontroller, and I'll definitely be putting it up on github since this release increases the likelihood that others will contribute to it.
Also, Cortex Microcontroller Software Interface Standard (CMSIS) [1] makes porting between vendor's offerings (e.g. STM32 or Stellaris) much easier than going from say AVR to MSP430.
As far as I know, it's institutional, which is unfortunate. Since they didn't take the opportunity to fix it with the Arduino Uno I don't think they ever will.
Would you be kind enough to explain what the "pin header defect" is, for an Arduino noob? I'm just starting to play around with Arduino hacking and haven't heard of this yet (until now).
The distance between the two rows of pin headers on each side is not a multiple of 0.1".
This means that every shield that you attach to the Arduino needs to have this same strange spacing between the rows of pin headers. You cannot just use a normal perfboard.
As all Arduino shields have the same weird spacing, fixing this design flaw would require everyone to buy new shields.
There are Arduino clones that add correctly aligned rows of pin headers [1], however.
One can get that microcontroller around $5 a piece with larger quantities (e.g. 10k). Also, the AT91SAM3U has a high-speed USB peripheral. So, the extra circuitry for the serial-to-USB bridge is no longer necessary.
Also, the specs are a bit misleading. The 16 ADC are broken down into half 12-bit and half 10-bit inputs. The article makes it sound like all 16 ADC lines are 12 bits while only 8 of them are.
Has a schematic or block diagram been released for the Arduino Due? I haven't found one with a quick search. I'd also be very interested to take a look at the bootloader source code if that is available.
If this is the first ARM-based Arduino, what was the Arduino chip based on that Google used at I/O for that accessory API thing? Was it a regular (but smaller) Arduino?
All of the official Arduino hardware designs have used an AVR to the best of my knowledge. Netduino [1] is a related project that runs a stripped down .NET framework on an ARM7 core. However, that's not what Google used. It may be the reference for Windows Phone add-ons.
