lowest cost mcu and how to mass produce

[Newz2000]

I'm working on a potentially cool product. I'm going to start out making 100 devices to see what the interest is.

I know that my market is very cost sensitive, so I'm trying to figure out ways to bring the cost down. The application is simple enough that I should be able to port my code to any mCU platform relatively easy, so here's what I need:

</font>

• <font size="2" face="Verdana, Helvetica, sans-serif">low power and able to run at 2.5 - 3v (2 cells)</font></li>
• <font size="2" face="Verdana, Helvetica, sans-serif">ability to source or sink 20ma (for an LED)</font></li>
• <font size="2" face="Verdana, Helvetica, sans-serif">6 outputs w/ some outputs able to be multiplexed as inputs - however at least 6 i/o must be capable of becoming an output</font></li>
• <font size="2" face="Verdana, Helvetica, sans-serif">small package that can be hand soldered, for example soic</font></li>
• <font size="2" face="Verdana, Helvetica, sans-serif">few/no external components needed</font></li>

<font size="2" face="Verdana, Helvetica, sans-serif">
Here's what I've found so far:
the Microchip PIC 12F629 can use all 6 i/o as an output, has integrated OSC (precision not necessary for me) can sink/source 25ma, can sleep and costs $1.02 when 100 are purchased at a time.

Question 1
Can anyone suggest another platform that I should investigate?
Question 2
If this were your product and you started out producing these by hand in batches of about 30 at a time and all of a sudden demand increased dramatically, what would you do?

</font>

• <font size="2" face="Verdana, Helvetica, sans-serif">What if the demand reached hundreds per month, or</font></li>
• <font size="2" face="Verdana, Helvetica, sans-serif">thousands per month, or</font></li>
• <font size="2" face="Verdana, Helvetica, sans-serif">someone said, "give me 10,000 quickly."</font></li>

<font size="2" face="Verdana, Helvetica, sans-serif">
What do you do? I realize this is wishful thinking, but I don't want to make mistakes early on that could limit the potential later.

When you outsource production, does your cost per unit usually go down because of the economy of scale, or does it go up?

[Engineer1138]

Q1: Atmel Tiny11/12. Starts out at around $1.50 and price at Digikey drops to (I think) about $0.50 at qty 100. Digikey is fairly expensive, should be cheaper elsewhere.

Q2: The biggest problem in this scenario is cash if you don't already have credit with a bunch of vendors. Assuming you have the money problem licked, there are many outfits that will assemble boards for you. I'm sure the guy who posts here from ezpcb.com in China can hook you up with assembly costs. Hundreds or low thousands a month is not difficult. I could probably refer you to someone who could do that. Once you get beyond that, you're out of my league ... and that's a good problem to have

If you seriously think you may make many thousands of these, then you should design with that in mind: ease of assembly, low cost parts, low parts count, etc. Much easier to do that up front than have to change the design later.

Best of luck. Go forth and make your millions!

[ljbeng]

Look at www.holtek.com and www.holmate.com You develop on an HT-ICE and they have OTP or they will do a mask where your program is factory-programmed. There is an up-front charge for this. This can get your huge quantities down to .50 a chip or less.

[philba]

the 12F508 is about $.85/100

I used a 12F629 for an application that sounds similar. One issue i had was getting 20 mA when running at 3V. I was unable to get it to deliver more than about 6 mA (iirc). I wound up using a 2N3904 and got plenty of current. fairly cheap solution ($.06 all in) but not ideal.

[Engineer1138]

I checked the price on the Tiny11. Digikey has them for $0.54 in singles, $0.28 at qty 95. Hard to beat that.

[Newz2000]

The AT Tiny 11 only has 5 outputs. The AT Tiny 12 would work, however it comes with lots of extra features I don't need, which drives up the cost.

It's the same problem I ran into with the 12F508/12F509 --- that chip looks like it'd work, but in the datasheet you read that one of it's 6 i/o pins is only capable of being an input, meaning it has only 5 outputs and I need no fewer than 6.

I apreciate you looking for me, and the $0.28 price tag is awesome. So awesome, in fact, that I'm going to try to think of a some way I can get by with 5 outputs.

By the way, Phil - Thanks for the hint about the limited current output. It didn't even occurr to me that it would be a problem. I will look out for that.

I've got the circuit complete and I'm just finishing up the code to make it work. I ran into an interesting (but logical) limitation...

I thought I'd be smart and lower the power usage by putting the chip to sleep and interupting it every so often to wake up and do something. Unfortunately, Timer0 and Timer1 turn off when the chip sleeps, so it never genrates an interupt and therefore never wakes up. The solution is to use the watch dog timer. I guess on the PIC if the chip is asleep, WDT wakes it up (it doesn't do a reset). So that is my last challenge to address before I get to see it working.

[philba]

yeah, I really wish microchip would move some of the 18F series power mgt features down to the 16 and 12 lines. (or produce low pin count 18Fs...) They have much better power management. You can run timers with the rest of the chip asleep.

Atmel does a much better job, by the way.

<small>[ February 28, 2006, 11:12 AM: Message edited by: philba ]</small>

[Newz2000]

ARGH!!!!!

So, I program the 12F629 and everything is working... except that for some reason, GPIO3 isn't doing what it's supposed to. As a matter of fact, it's not doing anything.

So, check the datasheet... it is NOT A GPIO, it is input only! Apparently, if you don't count MCLR, all of the 8 pin microcontrollers from Microchip have 6 inputs or 5 outputs. GPIO3 cannot be a general purpose output. How did I get so confused?

The good news is, I found that the 16F505 is cheaper than the 12F629 ($0.67 each vs $1.08 each qty 100, soic package, direct from Microchip) and even better yet, I have one in my drawer.

Unfortunately, it's much bigger (8.9mm vs 4.9mm), so I'll have to do a major redesign on my board. Oh well. I wonder if I can teach my wife to solder tssop chips. It's actually smaller than the 8 pin soic.

[Engineer1138]

So when your wife is done building all 100 are you going to tell us what the device does

[Newz2000]

No, and I'm sorry to say, at this point I doubt it will ever make it to market.

The reason:

I just demonstrated my prototype and the first response was, "Cool. My mother in law had a bunch of those at Christmas that she bought at Walmart!"

Retail price at Walmart is about the same as the price that I can manufacture them for. Oh well.

Forgive me for being disappointed, but let me quote a passage:

Ecclesiastes 1:10:
Is there anything of which one can say,
"Look! This is something new"?
It was here already, long ago;
it was here before our time.

[KamPutty]

Matt,

Don't give up! So what if Walmart has it too. It just validated your idea! You learnt alot from this project I'm sure, move to the next one, or just make yours better!

But keep on trying.

We all need a cheerleading section...

~Kam (^8*

[jimandy]

OK, don't tell us what your idea was but DO tell us what your mother-in-law bought at Walmart.

[NV185864]

Don't forget fiducials on you board layout. And do keep trying. If something is cool and useful you can sell it. I can get a pint of potato salad from Wal-Mart for 50 cents. I'd rather have the better tasting potato salad from King Soopers for 90 cents per pint.

[Newz2000]

Yeah, you guys have some good points.

I went and purchased one for myself and the thing looks like crap. And it caters to a completely different consumer than I had intended.

I can't say what it is. It's likely not patentable, but just in case, if I discuss it here I'll ruin my chance.

Don't worry, it's nothing too exciting. It certainly won't make me rich. But it may serve as a nice side business.

To be general, it's an electronic circuit that artists can incorporate into their work to give it some pizaz without needing to use a soldering iron or understand electronics. (My wife is an artist and she inspired the idea)

I think I am going to try and make 100.