Pugh Matrix

Here is the Pugh Matrix and Pairwise comparison matrix that ben and i sketched up at Ratsies.


[Pairwise Comparison Matrix]


Pugh Matrix

uC Sourcing/Sinking current discussion on electro-tech-online

See this message posted on electro-tech-online.com.

http://www.electro-tech-online.com/viewtopic.php?p=64273#64273

Good article on the growing market for 8-bit uC's and containts survey results from leading uC vendors with price / feature comparison. http://www.edn.com/contents/images/46248.pdf

Concept #2 Evaluation / Design tradeoffs of using two batteries

After putting together our through-hole alphanumeric concept we're slightly concerned with the brightness.

We kinda assumed that since we were going to use put a resistor for each segment that we would be able to suppy enough current to the display to make it bright.  However, it turns out we are limited by the amount of current the PICs source.

If we raise the incoming voltage to around 4.5V then the PIC is able to supply enough current that the segments look bright enough (as you raise the Vdd the pic is able to supply more current... we think). 

At this point in the development of the parking spot recaller 2000 our research and development time is a major consideration.  If we don't finish this soon it's never going to be sold.  Therefore, we're willing to take some product part cost increase so that we actually do get our product on the market.

One idea we considered is adding a second battery.  If we could add another battery that was smaller.  We would be increasing brightness and potentially decrease the overall size of the unit.  Both of which would improve the concepts design criteria (see pugh matrix). 

Costs of Adding an additional battery:

3V 12.5MM 35mAH Cell = ~0.85
3V 12.5MM w/ Legs = ~$1.50
3V 16MM Cell = ~$0.83
3V 16MM w/ Legs = ~$1.68
3V 20MM 220mAH Cell = ~$0.33
6V 165mAH 23mm w/ Legs = ~$3.10

CR2032 Coin Cell =
Coin Cell Dual Battery Holder = $0.90
Coin Cell 20mm Dual Battery Holder = $0.55 (BH800S-ND)
Retainer Clip which require SMD Pad = ~$0.30

Best Deal is to go with the CR2032 Coin cell which is cheeper then other cells by a factor of 3 since it's the most commonly used coin cell. 

Cost = 2*0.33 + 0.55 = $1.21

So, the only question is will two batteries stacked on top of each other be too tall.  From the datasheet the profile is 8.89mm which is about a mm taller then the LED Display which is 7.80mm tall.  That's a little bigger then i'd like it to be.  We are planning on putting a red gel over the display to block any light from going through other then the red light.  This makes the display easier to read. 

If we can deal with the added 1mm height for adding about $0.70 in direct costs its worth not having to reinvest any more man hours in R&D.  If we sell about 50 of these units, the additional cost is 50*$0.70 = $35 dollars which is worth it.  Also, we have a number of CR2032 batteries from previous projects we could use and if we have a Vdd of 6V we can get away with using the non-low power PIC chips which cost a little bit more and are harder to find.

Using the Alphanumeric display

This alphanumeric display is kind of interesting to work with.  It has 4 different segments that can be switched.  If you wanted to make a 3 on the left digit you would have to switch back and forth to get a 3.

Cathode 1 = low, 2=high, 3=high 4=high
Segments A, D = on

Then switch to

Cathode 1=high, 2=low, 3=high, 4=high
Segments A, B, C, D = on and the rest off.

See the schematic thumbnail below for more info:

Click on image to expand

LED Display Alphabet Image

This font image was sort of hard to find so i'm posting it here for future reference:

Image Taken from:
http://www.maxim-ic.com/appnotes.cfm/appnote_number/3211

Keywords: LCD alphabet, LED alphabet, 14-Segment Display images

Design Updates

Today i found an incredible deal on Alphanumeric Displays from goldmine electronics.  A package of 5 displays can be had for $1.00 which makes them come out to only $0.20 a piece!  This is cheeper then all the regular 7-Segment Leds we've been looking at and with the alphanumeric displays we can make any letter we want. 

We also scraped the SCR circuit.  After we got the idea to use LCD displays we ran into a new problem which was that we needed 3.0V and the SCR dropped our operating voltage from 3.0V to 2.1V.  I posted some questions on a forum about having microcontrollers wake from an external interrupt pin and other low power configurations and sure enough their is a new revision of the microcontroller we were using that make the SCR circuit obsolete.  First of all it has an internal 37kHz osc. in addition to the internal 4Mhz osc.  Since our project requires practically no processing the 37Khz osc is suitable.  As a rule of thumb the higher the clock used the more power the pic consumes. 

I'm currently not sure how much of a difference this will make.  I'm not sure if the PIC shuts down the oscillator when it's in sleep mode.  I remember reading somewhere before that when using the internal oscillator even in sleep mode the PIC consumed something like 50uA.  Then today i think i read in the datasheet that when the PIC is operating at 3.0V with the 37kHz clock in only consumes 12uA.  So on our particular battery which has a capacity of 215mAh.  The PIC never even going to sleep would last: 215mAh *  ( 1000 uAh / 1maH ) (1/12uA) = 17916 hours (1 day / 24 hrs ) = 746 days = 2.04 years!

LCD Displays

The LED seven segment displays have one big disadvantage in that you can't see an LED in flurescent lights or in sunlight.

For this reason we explored using a seven segment LCD.

Findings:

Lumex: makes a .5" character height 2.5 digit LCD which can be had for 1.78 from digikey

Things which still need to be worked out.
1) need a higher AC voltage
2) need 14 (7*2 for segments) + 2 (switches) + 1 (COM pin)
3) doesn't save us any money
4) Package takes up an additional .1"

Possible workarounds:
1) perhaps not having the scr in the circuit and having the PIC go to sleep.

Update on Friday, October 15, 2004 at 10:49AM by Blake Robertson

I ran across this article looking for something else and i'm posting it here in case i want to revisit this idea.

http://www.reed-electronics.com/ednmag/index.asp?layout=article&articleId=CA268383

Evaluation of Prototype after 3 months use

I've had the parking spot recaller prototype attached to my keychain for 3 months now.  Battery is still doing fine. The buttons were broken by a women with sharp fingernails.  The buttons have to be pressed by the pads of your fingertips or the metal dome can get bent.  I looked into some other way of doing the buttons but all other options seem like they would complicate things too much.

Software Improvements:
- Have the unit shut itself off after 2 minutes even if a button is always being pressed.
- Improve button debounce algorithm.

Design changes from here we are planning on creating a new model using all through-hole components.  The Through-hole version will allow us to make a board that is single-sided and make the assembly process quicker. 

Reducing costs

In an effort to reduce costs more we investigated possibly doing a single sided board with all through hole components and then using our lab's cnc milling machine to actually route the boards rather then having them fabbed.

Cost of singled sided copper pcb per square inch = 5 / (6x12) = 7 cents/inch^2

If we panelized the (6x12) board we could probably get 2x10 = 20 boards

So our effective (material) pcb costs would be 5/20 = 0.25 cents/pcb

Using our CNC machine as a plotter we could etch the boards which would be faster and wouldn't have to worry about router bit costs which are ~10 a piece.

So for more time spent preparing the boards we could get the costs down by about $2.00 a board.

Estimated Costs:
PCB - $0.50 (once you factor in boards that will not etch properly and board cutout and drilling bits)
Switches - $0.29 x 3 = $0.87
Battery - $0.25
PIC - $2.00
SCR - $0.30
DISPLAY - $1.30
Resistors - $0.09 (for all)
------------------------
Total = $5.22

So, we're almost to our desired cost of $5.00.  I think we can find a LED display at a surplus electronics store for about $0.50. 

Single Sided Layout:

Update on Tuesday, May 18, 2004 at 03:43PM by Blake Robertson

On Digikey:
Found some other 10-Pin DIP Packaged displays that would save us $0.23 - $0.37 depending on whether or not we could get away with using a regular green display opposed to a high efficiency green display. 

MANC3440-ND

LED 7-SEG GREEN .3"

10-DIP

0.3"

0.86000 25

MAN3480A-ND

LED 7-SEG GREEN RHDP .3"

10-DIP

0.3"

1.07760

On ChipCatalog.com:
http://www.chipcatalog.com/Fairchild/MAN3440A.htm a similar display can be had for $0.39  which would save us nearly a dollar!  I'm waiting to hear back from them the quanity required to purchase them.

Black Tinting Experiments

Ahh finish all my finals and i decieded to give the colormatch 325 another go. 

I tried removing the air from part 'b' before mixing it with part a and i measured with a tablespoon carefully the amount of each.  The results didn't come how that great.  I'm going to try using the scale on the next attempt so i can be sure that i did in fact use equal amounts.  If i do and it doesn't come out right again it could be not letting the mold release dry enough before putting plastic into the mold or using a wax coated cup or possibly the chopstick as the stirring rod.

While i was at it i deceided to make some more casts with the castin resin.  I have it a black tint.  In order to get an idea of what using a barebone pcb would look like i cut up one of my extra pcbs from my frequency counter project from a few years ago.  It's hard to judge the amount of tint.  In the cup it looked pretty dark but there the amount in the cup was thicker so it looked darker.

After i knew it wasn't enough tint i quickly added more tint to the remaining casting resin, broke a piece of scrap pcb i had laying around and tossed it in a project box close by.  I wish i hadn't cause i had to destroy the box to get it out.  But, i only had about 2 minutes until the resin would casted and i think i had inhaled too much of the resin's fumes.

Note: The castin resin casts came out more orangish then i expected.  I think that this was because i let them cure on my window sill with direct sunlight.  I don't think you're supposed to let resin that's curing be exposed to sunlight.  But, i didn't want to smell the fumes so i didn't worry about it.

Using the vacuum sealer and indentation mold

After the partial success of using the syringe to create a vacuum, we ordered a $20 "Vacuum food saver".  We were hoping that this would work good enough because real vacuum chambers cost in the hundreds of dollars.

Here's some pictures of reoflex in the vacuum food saver.

   

Notice the level of the material in the middle and the right picture.  As the pressure decreases in the vacuum chamber the air bubbles get pulled out of the cup.  As the pressure decreases the volume of air increases.  (PV=nRT) So, as you can tell by the level of the cup the air inside the reoflex is taking up more volume but it's still in the reoflex plastic.  I think this is an indication that the pressure low enough to pull all the air bubbles out of the reoflex.  If it was all solid material in the cup then there would be no expansion.

It definently helps but it doesn't solve the problem completely.  One more idea we had for the mold was to engrave our names and some product information on the back of the keychain.  To do this the cnc machine would first put our names onto the the material we would make a mold of.  Then when the mold was made of the mold inverse it had raised areas where the text was.  So finally, when the actual mold is made it's an indentation again.

To do a tryout of the concept we engraved the info into a piece of wood we had around the lab.  The text didn't look very good on the piece of wood.  We engraving was too deep.  Since it was just a test we didn't want to spend more time on it so we went with it.  In the picture you can't really tell on the mold but the text is raised on it.  The text looks really bad on the mold.  It stuck to the wood a little when we were trying to take the wood off the mold.  I think if the grooves aren't as deep this will not be as much of an issue.

Using Syringe to remove air bubbles

Since we were getting a lot of air bubbles in our casts we tried a technique to create a vacuum using a syringe to remove some of the air bubbles from the cast.

Here are the results:

Smooth-on 325 Colormatch

Since the smooth-on 300 which is a white liquid plastic worked well we bought some smooth-on 325 which is a clear liquid plastic designed to be colored using dyes.

Unfortunently, the results we've been getting with the smooth-on 325 hasn't been as good as our results were with the smooth-on 300.  We've been getting a lot of air bubbles and sometimes our casts have just turned out really sticky and other times we've had white spots in the mix.

We're still working out the problems with it and i'll post what we were doing wrong or if we end up sending it back to smooth-on cause we think that the actual stuff is bad.

Smooth-on 300 Results

The liquid plastic is very nice to work with. 

Advantages:

• No harsh fumes which give you a headache
• Cures in 10 minutes opposed to overnight
• It's not as rigid as the casting resin so it doesn't crack when it's pounded on something

Disadvantages:

• It's not as thick so the pcb doesn't float on the resin which can lead to the pcb not being completely encapsulated.

Pictures of some molds made with the white liquid plastic (smooth-on 300)

 
(Front)

 
(Back)

Creating the mold cutout

Ever needed a piece of plastic a certain size and thickness?  I feel like i always do.  Well the good thing about having liquid plastic around is you can just make yourself a piece of plastic.  So that's just what we did.  Using our trusty mold box we made for the rubber mold out of sculpey we just filed it up with with liquid plastic and 10 minutes later we had that white piece of plastic you see below.

To create the rounded edge shape of the product.  I first created a rounded rectangular box in adobe and printed it out until i had a rectangular box that was perfect size and rounding radius.
Then exported the gif to hpgl file format so i could use the CNC isolation milling machine to cut out the piece of plastic which we could make a mold out of.

After fighting with the machine's software to get it to scale it properly we were routing it out.

Smooth-on 300

I was looking around to find something that would seal wood so we wouldn't have to apply wax to the mold each time and the casting resin wouldn't stick to the wood. 

In my search i found out a product called SuperSealer made by smooth-on.  While i was looking around their site i found that they sold the sealer in a liquid plastic and rubber mold making super sampler.  The liquid plastic sounded like it might be a great alternative to the castin resin. 

We just placed an order for a Supersampler from smooth-on which comes with a lot of stuff.  For $25 we're getting:

1. Stuff to make rubber molds
2. liquid plasic
3. release agent
4. sealing agent
5. instructions on how to make molds.

For more information on the super sampler we bought:
http://www.smooth-on.com/Getting%20Started.asp

LED Display Profile

After sampling a few different LED display the only display which was bright enough at our voltage and current requirements is Kingbrights APDC03-41CGKWA additionally we loved this display because of it's low because of it's low profile and it's the best price for a smd display of this type.  (SMD displays are much thinner then through hole versions).

The current displays specs are:
Chip Material: InGaAIP
Common: Cathode
Color: green (570nm)
Iv mcd: 4700 (min), 11000 (typical).
IV Characteristic: @2.1V If=20mA -> good luminous intensity :)

The problem with this display is the cost is about $2 and we'd like to get this to about $1 since TH versions can be had for well under this.  Also a TH version would probably be better suited for assembly and less likely the casting resin will get under the display and when it hardens cause some of the solder joints to become disconnected.

Findings:

current profile is 3.75mm and dimensions are 14.8x12


http://www.us.kingbright.com/data/spec/DC03-11GWA.pdf <-- Profile is 7.20mm
http://rocky.digikey.com/WebLib/Lumex/Web%20Data/LDD-E302NI-RA.pdf <-- profile is 7.2
http://rocky.digikey.com/WebLib/Lite-on/Web%20Data/D2701G.pdf  <-- profile is 6.1 and common cathode

The biggest problems with these displays is the profile.  There are a lot of advantages to them in that they would be easier to assemble and some have properties which would make the PCB layout much cleaner (possibly even able todo single sided boards or route the boards out on our cnc milling machine).

However, we would need to find switches which have longer legs by 3mm.  We could gain another 1.5mm by not putting the switches through the board which would allow us to have a profile of 3.75+1.5 = 5.25.

So, we are very close to the 6.1mm displays.  Perhaps, we could put a socket pin through first and then solder the switches to that.  This will have to be explored more. 

But one thing is time.  Our design works and has a few flaws.  The design process tries to limit those flaws but it's a compromise for research time invested in improving it or just spending the additional time assembling the device. Hmm...

Update on Sunday, May 16, 2004 at 04:20PM by Blake Robertson

MSD349C-ND

0.3"

1.28600 30

Profile is 7.3

PCBFabexpress Green Boards

Tonight i got some great feedback on the appearence of the keychain device.  Specifically, they liked having the look of the green pcb.  To see how much we could get the cost of the pcb to using a PCB board house which applies the green lpi solder mask i priced out several board houses.

PCBFabexpress offers a deal where you get a board for $13 per board (min quanity 5) and the board can be up to 20 sq inches.

Each device is 1"x2.5" which means we could panelize them on a boardand get 10 on one board and then we'd just have to use our CNC machine to cut the larger board into smaller boards.  But to take intoconsideration the 1mm router bit width it's safe to say we'd only getabout 9  boards out.

So the cost per board is $13/9 which is $1.44

The minimum order requirements would make it so we'd get 40  boards for $65 dollars + shipping.

I found a much cheeper board house: Olimex they sell you a panel thats 12.6"x7.8" for $104USD.  We could fit at most 35 boards on the panel so it'd cost us per board 104/35 =$2.97

1. In order to get the best deal with them we'd have to try to make the following optimizations:
    reduce the number of holes but most importantly the number of hole sizes.
2. if possible make trace widths greater then 20mils so etch resistsingle silkscreen printing can be used rather then dry film lamination,exposure, developping

Bare Bones PCB:

1000 boards would cost $1.59 a piece but would be precut.  But no solder mask.

Update on Thursday, April 15, 2004 at 07:05PM by Blake Robertson

I emailed olimex to get some price quotes on the how much we could get our board for at production quanities.  Below is the email:

> We are trying to get a quote estimate for our board in quantities of
> 100, 200 and 1000.

FR-4, 1.5 mm, 35/35 um copper, soldermask, component print, 15 working days:
100 pcs $215
200 pcs $350
1000 pcs $850

> We want the board made as cheaply as possible. Our circuit is a
> straightforward low speed board.
> The board is 1" x 2.5".
> Two SMD IC's, and 8 other components.

> The circuit was initially designed for easy assembly because we were
> going to be putting together ourselves.

> We could reroute the board such that only three drills are necessary
> (.8, 1.1, and 4.0).
>

> How many boards would we have to order to get the board cost down to
> about $0.70?

more than 3000 pcs

> How many boards would we have to have assembled such that the assembly cost
> was $1 per board?

machine assembly:

one time setup stencil & programming $200 setup per lot $20 (min.)
1000 pcs assembly $1000
$1.00/ea

manual assembly 100-200 pcs
setup per lot $20
$3.50/ea

EXW no packing, no shipping, no component cost, no programming, no functional test included in above quotes

Fix to tacky molds

I stumbled upon this site: http://www.eti-usa.com/consum/castresn/castips.htm
which gives tips on using the casting resin we've been using for potting our keychain.

Bullet Points:

• "Casting resin's curing process is caused when heat is generated by the chemical reaction that takes place when catalyst is added to casting resin."
  
•  thicker the layer, less catalyst is needed; the thinner the layer, more catalyst is needed.

• The biggest disadvantage to flexible molds is that castings come out tacky and wavy because heat from the resin when curing dissipates quickly through the rubber molds."  <-- this explains why our rubber mold came out tacky and wavy! It's cool finding information which supports your experimental findings

• best molds are made from HDPE (high density polyEthylene plastic).  This is the type of mold which the catalyst amounts were developped for on the label.  Other molds will require either more heat to be applied during the curing process (place in oven ~100F) or doubling the recommended catalyst to intensify the chemical heat. 

• If the cast comes out tacky then place in the oven on low heat until it's completely dry.

Tonight:
1) Placed the completed mold from last night in the oven to let it completely dry.  Hopefully this will make it completely tack free.



2) Potted a test mold using 30 drops of resin instead of the typical 15.  I also placed it under the heat of our incandescent lamp in the living room.  In less then 3 hours the mold was solid but still a bit tacky. 

First Working Prototype Mold

After a few test molds we decieded to mold one of our pretty green pcb's.
Since we were trying to get a nice curved surface on the top we let the casting resin sit before we poured it.

Well... 10 minutes was a little too long.  The resin harded while it was pouring from the cup leaving a stream of resin in between the cup and the mold.

After some quick cleanup to seperate the two we mixed some more resin and finished the rest of the mold.  But, the final product had some imperfections.

Additionally, when the casting resin hardened it must of seperated a solder joint to one of the pads because two of the segments on the right LED display don't work .
But, i suppose the first prototype was destined to get messed up a bit.