From experience, I can tell you that you won't get anywhere with a hardware startup if you don't start prototyping until you've lined up resources to see your product through production. No investor is going to listen if you're asking for millions up front in material, engineering, and manufacturing costs, especially if you don't have a prototype or mock-up to show for it.
Bring in an engineer experienced in the technologies and market you're targeting (USB, embedded, PCB design, CE products). Give them a real stake in the company. Fuel out of YOUR pocket a small prototype run. Once you have something in-hand and with enough functionality to demonstrate your idea, seek an investor to fund everything the parent poster mentioned.
Expect that your first prototype will not be in the final form factor. This has two advantages: Larger components (and I mean TQFP, 603 discretes, etc) can be hand-soldered relatively easily in small quantities. Getting a company to do small batch assembly is expensive, and unless you have more than a few hundred components hand assembly isn't too time consuming. If you must use an IC that's BGA or hard to solder QFN, see if the board house can solder just that part. You can hand solder them if you're daring, but you probably don't want to risk your company on a "reflow skillet". Using a larger board gives room for debug headers and test points, which will give your engineer the ability to fix design bugs with some crafty soldering. Even with an oversized prototype board, budget for at least two respins of the board.
Actually, my UVerse DVR is smart enough to spin the drive down when it's in soft power-off (i.e. not using the disk as for the 1-hour live tv buffer). Every so often, I'll hear the classic spin-up whine and then a few seconds later the record light turns on. Microsoft may be the devil, but their IPTV software stack can at least get this right.
That's right, Power Computing. They thought they could force their way back into the 3rd party Apple market. And they would have done it, too, if it weren't for those meddling Cupertino lawyers.
All of Cray's systems made in the last 15 years or so have been turnkey clusters. The reason people pay the big dollars for them over a few racks of HP 1U systems is their crazy fast interconnect and software environment.
I'm positive I can explain this with a car analogy but I'm sure someone else will do that for me.
The USB standard for transferring driver code from device to host is not the issue. That already exists in the USB mass storage class. The trouble is that it takes both design resources and additional hardware cost to distribute what are usually out of date drivers.
What I could see happen in the future is the use of an existing class (say, usb comm) to transfer some XML data to the host that would direct the host's USB driver stack to a URL to download compatible drivers. Hardware vendors, be it the finished goods manufacturer, the product OEM, or the peripheral chip's manufacturer usually provide drivers on their website, so delivering drivers this way incurs little extra cost (the cost being increased bandwidth usage) and no extra design effort. Of course the USB host stack and the peripheral would both need to support this standard, but given the rising popularity of devices that have no driver CD in the box (Apple, I'm looking at you) this might gain acceptance quickly.
I agree that driver installation is an unnecessary evil, but I don't see that going away any time soon. Standards bodies for interfaces like USB can only come up with so many standard device classes, so inevitably some new and innovative (or old and unpopular as in this example) product will come out that won't fit into one.
But I don't think companies will find it practical to use your method. Once you've paid for a flash chip, a micro with enough oomph and extra I/Os to run USB mass-storage and talk to a flash chip in addition to your intended application, factory programming for the flash, extra board space and testing costs, you might as well have paid the 25 cents or so for a pressed CD.
The only application I can see for something like this is where the need for software's physical portability is part of the usage model. There are Vonage branded USB sticks out that not only have Vonage's app and drivers on the stick's mass storage device, but also have analog audio in/out for a headset.
In USB parlance, "full speed" means that the device supports 12Mbps transactions. It has no bearing on the throughput the device must sustain. And the 2.0 part is just marketing fluff: any full speed device is compatible with the 2.0 spec by way of legacy support for 1.1. For the record, "high speed" is the official term for 480Mbps USB, not 2.0.
Slashdot Mirror
I regret that I do not have the mod point to give, that I may repay you for of the brilliance that was TV's Frank.
I hope this has a low-pass filter, otherwise people with a facial tic might wind up somewhere unexpected.
From experience, I can tell you that you won't get anywhere with a hardware startup if you don't start prototyping until you've lined up resources to see your product through production. No investor is going to listen if you're asking for millions up front in material, engineering, and manufacturing costs, especially if you don't have a prototype or mock-up to show for it.
Bring in an engineer experienced in the technologies and market you're targeting (USB, embedded, PCB design, CE products). Give them a real stake in the company. Fuel out of YOUR pocket a small prototype run. Once you have something in-hand and with enough functionality to demonstrate your idea, seek an investor to fund everything the parent poster mentioned.
Expect that your first prototype will not be in the final form factor. This has two advantages: Larger components (and I mean TQFP, 603 discretes, etc) can be hand-soldered relatively easily in small quantities. Getting a company to do small batch assembly is expensive, and unless you have more than a few hundred components hand assembly isn't too time consuming. If you must use an IC that's BGA or hard to solder QFN, see if the board house can solder just that part. You can hand solder them if you're daring, but you probably don't want to risk your company on a "reflow skillet". Using a larger board gives room for debug headers and test points, which will give your engineer the ability to fix design bugs with some crafty soldering. Even with an oversized prototype board, budget for at least two respins of the board.
It sucks 'til you realize you're sending files over Ethernet.
IA32 can address up to 64GB with PAE.
Actually, my UVerse DVR is smart enough to spin the drive down when it's in soft power-off (i.e. not using the disk as for the 1-hour live tv buffer). Every so often, I'll hear the classic spin-up whine and then a few seconds later the record light turns on. Microsoft may be the devil, but their IPTV software stack can at least get this right.
or a piece of fairy cake.
That's right, Power Computing. They thought they could force their way back into the 3rd party Apple market. And they would have done it, too, if it weren't for those meddling Cupertino lawyers.
That is exactly why you should never:
a) give friends accounts on a Linux box in your dorm room
b) keep your speakers turned on
c) keep the machine running while you sleep
I speak from experience
All of Cray's systems made in the last 15 years or so have been turnkey clusters. The reason people pay the big dollars for them over a few racks of HP 1U systems is their crazy fast interconnect and software environment.
I'm positive I can explain this with a car analogy but I'm sure someone else will do that for me.
Then why not join the two with Ethernet instead? There's already a free iSCSI initiator (host) app for OSX.
Post his snail mail address. I think last weekend's newspaper had a $5 off coupon on Monster Cable EXTREME HD silver satin phone cable.
Wrong space janitor. Wasn't Mike some sort of temp worker before being kidnapped?
Why is it impossible? Does POTS not run over 24 gauge twisted pair?
The USB standard for transferring driver code from device to host is not the issue. That already exists in the USB mass storage class. The trouble is that it takes both design resources and additional hardware cost to distribute what are usually out of date drivers.
What I could see happen in the future is the use of an existing class (say, usb comm) to transfer some XML data to the host that would direct the host's USB driver stack to a URL to download compatible drivers. Hardware vendors, be it the finished goods manufacturer, the product OEM, or the peripheral chip's manufacturer usually provide drivers on their website, so delivering drivers this way incurs little extra cost (the cost being increased bandwidth usage) and no extra design effort. Of course the USB host stack and the peripheral would both need to support this standard, but given the rising popularity of devices that have no driver CD in the box (Apple, I'm looking at you) this might gain acceptance quickly.
I agree that driver installation is an unnecessary evil, but I don't see that going away any time soon. Standards bodies for interfaces like USB can only come up with so many standard device classes, so inevitably some new and innovative (or old and unpopular as in this example) product will come out that won't fit into one.
But I don't think companies will find it practical to use your method. Once you've paid for a flash chip, a micro with enough oomph and extra I/Os to run USB mass-storage and talk to a flash chip in addition to your intended application, factory programming for the flash, extra board space and testing costs, you might as well have paid the 25 cents or so for a pressed CD.
The only application I can see for something like this is where the need for software's physical portability is part of the usage model. There are Vonage branded USB sticks out that not only have Vonage's app and drivers on the stick's mass storage device, but also have analog audio in/out for a headset.
Got a parallel port? It's even easier. A 44780 based display won't work with a serial port unless you have a separate micro to do translation.
In USB parlance, "full speed" means that the device supports 12Mbps transactions. It has no bearing on the throughput the device must sustain. And the 2.0 part is just marketing fluff: any full speed device is compatible with the 2.0 spec by way of legacy support for 1.1. For the record, "high speed" is the official term for 480Mbps USB, not 2.0.
Standards organizations are weird.
And don't get me started on that unhandled divide-by-zero exception!
Symbian goes back to the late nineties.
She's friends with Lena and, inexplicably, a teapot.
No. DVD and BluRay can have more than one layer. Without servo focusing it would be impossible to select a layer to read.
There are MANY lights!
It took me 9 years of being a broadband customer to realize that the "NO CARRIER" meme is irrelevant. Thank you.
You forgot:
1: Blocks fall
2: Line disappears
3: Repeat
Riveting!