14 replies to this topic

[Orpheus]

I just got another call from my much-dreaded paternal ancestor (or so he claims -- I can't believe we're related)

It turns out his car is billowing smoke - not from the tailpipe, but under the hood.

On further interrogation it turns out that it failed inspection -- in April (not that it stopped him from driving it daily) and will cost more than the car is worth to repair. He wanted help finding a new one.

Now I love his dealer (I go to the same one), but I've never seen a dealer who wasn't overly eager to write a car off. It's not greed, it's keeping the customer happy. They don't want to take chances. If the car keeps breaking down after $5000 in repairs, you're going to blame them, and that's not good for business.

On even more interrogation (I tell you, I'm on the verge of beating the man with a bag of lentils), I learn that he has no idea what is wrong with his car. After a harsh round of over-the-phone legume abuse, I finally get him to fax me a copy of the dealer's report. Sure enough, it may not require a full transmission replacement (though that's what the factory recommends) and may be an easy fix. I just depends which of to flavors of engine he has.

The smoke? I'm 99% sure that's just a loose valve cover gasket. It turns out he's been slowly losing oil for six months.

Now, we'll set aside the question of why he couldn't have told me all this in the spring, so I could have worked on it in the lazy warmth of the summer (no, we won't. I hate hate HATE working on cars in the cold, it numbs my hands, freezes the nuts and makes every rough edge sharper). We'll try to make the best of it. This is a good excuse for me to build something I've wanted for ages: a full OBD II/CAN engine computer analyzer.

In Massachusetts, like most states with so-called "smog control inspections", *any* error code will fail your car. They don't even do a real safety inspection anymore. They just give it a visual once-over on their way to plug their computer into the car's computer socket. I don't want this to turn into a winter-long project. I want to fix it all in one weekend and be done -- and that means reading and resetting the error codes, various sensors, and maybe doing some realtime datalogging

OBD? OBD-II? CAN?
OBD (On-Board Diagnostics) was an early system that let mechanics read crude error codes from your onboard car computers (on some early models, it simply flashed an uninformative "Engine Trouble" light)  in cars up to roughly 1996 (depending on make/model) OBD II was an upgraded version used on all cars since 1996 (and some cars before that) [more info] []way too much info on wikipedia]. CAN (Controller Area Network) is a much faster and more sophisticated networking system that is in all 2008 and later models (in the US, I don't know about the rest of the world) The faster bus allows reading of real-time operating data.

CAN is sophisticated enough to create a "glass cockpit" for automobiles -- or even aviation simulators and light aircraft (I haven't checked what fighters/airliners use). General in-the-field automation, and many other industries are starting to use it too. It's been available in the Linux kernel since 2.6. You can see why I want to get some hands-on experience with it

You can buy a simple OBD-II reader for $70-150, but I figure I can build a combined OBD-II/CAN interface, run by free downloaded software for less, and it will be capable of doing considerably more: resetting errors, realtime data colection and debugging, digital analysis and diagnosis. It will work on any car made since 1996, and probably any car I drive for the next ten years. Hacker value aside, it's a good investment.

It should be a heck of a ride. Join me, won't you?

[Orpheus]

Terrifyingly, this project turns out to be much simpler than I thought.

There is an affordable ($32.50 CDN) specialty chip on the market, the ELM327, which handles OBD-II/CAN [6pp Quick Reference PDF] [59 pp full datasheet], and a wide assortment of freeware software that communicates with this chip via USB or RS-232 or BlueTooth. The schematic for the adapter is pretty simple, and I bet I can find a pre-designed Printed Circuit to etch, if I look hard enough. Why re-invent the wheel?

Here are some videos of these solutions in operation so you can see some of the things it can do.
OBD too techie for ya? There are step-by-step OBD for Dummies tutorials. This one is for the free ScanTool
From Chile: A Chinese knockoff running an old version of EasyOBDII -- video is over a year old.
A home brew circuit running OBD-diag (note how tiny and simple the serial adapter is)
CarMan software on a Nokia N800 handheld computer, using Linux, connected by BlueTooth, 1997 (cool ideas: use OBD-II to provide additional dash gauges or modify an LCD for a windshield head-up display)
Using OBD to assess a COSM-II  HHO (hydrogen from electrlysis system
I'm skeptical of HHO, but I seriously applaud his method, and I'll read his blog to see what he has to say
Man spricht hier Deutsch -- weil OBD-II weltweit ist.
A demo/discussion of CAN

There are several commercially manufactured adapters based on the ELM-327, if you don't feel like rolling your own, including some quite cheap Chinese knock-offs (<$30 shipped -- less than the single-unit price of the ELM327 chip).

If you're considering a knock-off, due to limited need, you should be aware that some models have poor quality control, or other limitations -- e.g.: they may be based on the older V1.2 instead of v1.3 chip being shipped by ELM Electronics today, or may not work with the latest version of some free software (knock-off detection?) By all accounts, you should *at least* research the specific seller and their product carefully before buying, to assess your risk of getting a lemon. I'm curious to try one -- I can most likely fix one that arrives broken-- but I know I'm going to have to build my own from scratch eventually, if only to have something to experiment with.

I'm off to read...

[Orpheus]

Oops, sorry for not finishing this write-up. I wanted to improve some of the technical drawings, an reshoot some pics. I did this project last year while I was still testing image capture software for the webcam project, so my original pictures had that annoying trialware overlay, which obscured some key details.

Naturally, I forgot, until today, when a note in my project planner warned that Elm Electronics (maker of the IC chip) would be closing for a 10-day vacation. So when better to post this than today, when I have no tech backup?

Below are some drawings for the circuit (you an wire it point to point, if you wish). I have more data, including spec sheets, PDFs for cleaner template printing, and Gerber files for automated production. I'll post a ZIP bundle later.

Someday, I'll do a printed circuit board thread. It easier than you think. Here are the outlines of three methods that I recommend. You can Google for the details, or start a thread here in the TC.  Remember, these are just outlines, designed to show you that, yes, you CAN do it. I taught myself to do it as a broke ~10 year old, using incredibly crude materials (I even had to make my own FeCl). These instruction probably leave out useful details (e.g. FeCl should only contact plastic, glass, unchipped enameled, or other nonreactive materials -- never metal)

1) For small projects with a SINGLE-SIDED board, I use MANUAL DRAWING: rub-on transfers and a Sharpie marker

• Start with a *clean* blank copper-clad board, available in any electronics store or online. It's often much cheaper to to buy larger sheets (say, 8x12" = 20x30cm) and cut small pieces for individual projects; if you pay the premium for having them cut it for you, you'll still be getting boards that don't exactly fit your project, and will waste much more than if you just cut exactly what you need from a larger board. A light burnish with fine steel wool, followed by a wipedown really helps
  
• Use rub-on transfers to mark the positions of the ICs, connector pins and other spacing-critical elements. Draw the other pads, and the traces (electrical roadways) connecting them with a Sharpie pen.
  
• After double checking your work, fill a shallow plastic tray with Ferric Chloride solution (available in any electronic store, and not nearly as corrosive as some people say. I have a third of a century of FeCl stains to prove it). Place the board in the tray, and gently shake it from side to side every minute or two until all the pink color is gone (etched copper turns pink)
  
• Rinse the completed PCB with water. Dry with paper towel. Wipe off ink/rub-ons with nail polish remover.

2) BETTER: For double-sided boards or boards where I will be making multiple copied I use PHOTO MASKING. It may seem more complicated, but its often faster: you save the tedious manual rubbing and drawing of the first method.

• Use a computer drawing program to create the circuit layout. For pre-designed projects, you can often download a drawing file (e.g. a PDF) of the finished board, in a format that preserves the exact spacing. When I was a boy, magazines often printed full-size drawings in their circuit articles, which can also be transferred to a photomask. If you are making a double sided board, make sure the corners are aligned and marked on the drawings for both sides. Consider including any mounting holes as further alignment marks, and text, so you don't put the mask on backwards or upside down
  
• Xerox/laser-print the drawing onto acetate (overhead transparency plastic). Inkjet may work if you let it dry thoroughly, and are careful not to smear it. Pen-plotters definitely work: I often used those for over a decade, before the price of laser printers plummeted. Touch up with a black marker if needed. Finished photomasks can be reused indefinitely.
  
• Photosensitize (one copper side of) a copper clad with a thin layer of photoresist. I prefer spray-on. It requires a wee bit of spray-painting skill, which you really should develop anyway, but it's actually *easier* to make 10-20 boards in the same sweeping strokes, and store the rest in a light-tight box for the rest of the year. The photosensitized boards aren't all *that* sensitive (even a minute of direct sunlight won't ruin them), so you'll still have plenty of working time to cut pieces later -- just store them in the dark, and keep them out of UV-rich direct sunlight.
  
• Assemble the copper-clad board and photomask in an ordinary picture frame, so light passes through the glass and photomask onto the photosensitized copper. Double check the alignment and orientation, then place outside in direct sunlight for 10-20 minutes (The first time you do this, you should probable do a test board, consisting of rows of various types of IC pads and parallel traces of various sizes/spacings, running the length of the board. After 5-10 minutes, put a piece of cardboard over one end, and slide it 1-2" up the glass every 1-2 minutes. Develop and etch as below to determine the exposure range that removes your photoresist while preserving smooth sharp details. The middle exposure of that range will likely work for direct sunlight in your area, year round) You can easily build a UV lamp instead of using sunlight.
  
• After the exposure, develop the board by sloshing it in a shallow tray filled with a few mm of with photoresist developer (it's not a picky process: "developer" is just a solvent that dissolves photoresist that has been exposed to UV). If making a double-sided board, clean and dry thoroughly, drill any alignment holes, then photosensitize/expose the other side.
  
• Etch the completed board as above in FeCl solution. For 2-sided boards, flip over and check both sides.
  
• Remove the unexposed photoresist from he board with photoresist remover (another solvent)

3) BETTER STILL: Glossy paper. You should probably Google this method to get some tips & tricks. It's easy and fast, but may take a practice run or two to get right.  It also requires some exact brand of paper that has been proven to work well. Though the mask is destroyed in the process (unlike photomasks), it's trivial to print off another mask.

• Either create or obtain a drawing or drawing file that is exactly to scale.
  
• Xerox or laserprint the drawing onto a brand of glossy paper that has been proven to work (I use Staples Photo Basic, Gloss, Item #471865 -- a single package has lasted me a decade so far)
  
• Carefully align the print, face down, on a clean copper-clad board. Cover with a cotton sheet and iron with a hot iron.
  
• Let the fused board/paper combination cool, then soak thoroughly in water and gently rub off the paper.
  
• When only black toner remains on the copper, dry and inspect. Touch up any scrapes with black marker if needed.
  
• Etch as above.
  
• Clean toner off the board with nail polish remover or other suitable solvent

4) BEST: Press-N-Peel film
I've never really used this (I think I tried it one long ago). I was already well-experienced in a panoply of approaches, and the sheets seemed expensive and wasteful for the size of most PCBs. However, since the price is currently ~$3/sheet, and most laser printers have "special paper" (envelope) slots, it now seems a very cost-effective alternative. Just pre-cut the 8.5x11 sheet into (4x) 8.5"x2.85" strips (for both sides of a double sided board), (6x) 4.25x3.8,  (8x) 4.25x2.85, etc.

• Create or obtain image files for the PCB
  
• Print/xerox the images onto precut Press-N-Play film (don't run a full sheet through a xerox/laser printer repeatedly)
  
• Align the film on a well-cleaned copperclad board. Cover with a piece of cotton/towel. Iron to transfer the image.
  
• Allow to cool and peel of mylar backing
  
• Etch the copperclad in FeCl, as in all the other approaches

SCHEMATIC:


PCB LAYOUT: (both sides)



TEST VOLTAGES WITH and WITHOUT IC: (to test that your wiriing/traces are correct, and IC is working)


As I said, I'll be providing a package with better building details/drawings. These are just for study/info purposes.

NEXT UP: A look at one of those cheap Chinese knockoffs
Under $5 to buy, over $20 to ship, yet still a bargain for parts alone. Some say this kills Western business, but I say we tinkerers help our nations far more by developing our skills to invent/design. If our nations fall, it will be because we made ourselves incapable of anything but monkeywork and passive consumerism --and then couldn't compete on monkeywork. Did 30 years of blind "buy American" flag-waving help Detroit be competitive?

It's better to make ourselves the next Henry Ford than the next Ford Motor Co -- before WE are the ones making knockoffs

One branch of this project will modify/upgrade the knock-off with "genuine" parts to make a latest-generation unit while putting MORE cash in the Canadian designer/inventor's pocket than a pricey unit from a manufacturer (with his big volume discount). It's no more unYourCountryan than fixing you own car and depriving a pricey mechanic/dealership of business.

[ETA: outlined Press-N-Peel method, which I don't personally use, but is probably the best way for most people]

[Nick]

I actually bought a usb OBD-II interface as you describe.  Pre assembled with shipping, I beleive it was around $35.00.  Let me know what software you find you like the best.  It's been a while since I've played with the diagnostic stuff.

[Orpheus]

The casing is extruded aluminum, with pretty good fit and finish. (It was actually better when I received it, before six months of occasional disassembly and reassembly) Boy, this brings back some memories. This is what we used to impress customers back in the early 80s when I ran a small design/consulting house, and most shops were submitting prototypes in phenolic boxes with sheet aluminum faceplates. We'd stick a slab of iron in the second slot: the Pointy Headed Bosses swooned over the "precision engineering" (We'd tell the techs it was for rigidity in "space reserved for future expansion)

Take a tip from Scotty: even today, when everyone seems to want to trim every ounce, adding weight to a prototype will make your competitors' prototypes feel like toys, and then make your version without it seem like genius.



Sometimes we actually *did* use that second slot for expansion. For example, we could add a wireless board here, or flash memory for data logging and logic for debugging really pesky intermittent events--or maybe a microphone and the new inexpensive Arduino accelerometer, for a gadget that could help diagnose mystery noises in the car as you drove. GPS has a lot of interesting potential for the true geek, since CAN/OBDC knows *exactly* how much fuel is being metered into the engine at any instant. You measure a lot of former imponderables, like the exact effect of hills or sharp turns on mileage.

Note that the casing consists of two identically asymmetric extrusion (one edge has a ridge, the other has a groove), and that the PCB is in the bottom half. This will help you with speedy reassembly.

The CAN/OBDC cable is a nice bonus. At one point last year, these were running almost as much as an entire assembled unit. Though overseas competition spurred some vendors to bring the price down, it's the second costliest "necessary component" for a home-built unit, second only to a legitimate ELM chip. Since it has a DE-9 connector on the other end (also known, incorrectly, as a "serial" or "RS-232" connector), we can reuse the cable when we build a full CAN unit.



I'm including larger shots of the top and bottom of the PCB, because we may need them to determine why some newer versions of various software won't work with knockoffs, if upgrading to the latest real ELM doesn't clear that up. Though it isn't evident in these shots, build and soldering quality was pretty darn good on this model, as shipped, unlike horror stories I've read about some other models -- too good, I think, to be just "the luck of the draw" without some real production QC.

[Nick]

Mine has an ELM327 1.2, but I can replace the chip if I want.  It's socketed in rather than soldered, so it wouldn't be difficult to do.  Of course, I don't see any features in 1.3 that jump out at me as "must haves" right now.

I've been using Digimoto Light, but I'm hoping I get ahold of the author of a custom piece of software at the MP3Car forums.  I haven't found anything Linux based that I liked yet.

Have you done any tinkering with your OBD2 interface yet, Orph?

Also, I've heard it's possible to reprogram the PRAM with the interface.  Any idea what kind of things that could do?

[Orpheus]

I only had a chance to play with software packages briefly, back when I started the project, and didn't take good notes.

Some of the software was borrowed, and the controller computer I'd been using is currently being used on another project. I hope to get the band together this summer to write up some sort of proper software reviews.

I will say this much: the ones I tried ran fine on a Virgin Webplayer that long ago modified for automotive dashboard mounting and 12v operation. Since that's a 90s-era 200MHz Pentium-class LCD web appliance (given away free to subscribers to Virgin's failed dial-up ISP -- see my thread on replacing broken LCDs), almost any computer that's still in your closet will be powerful enough to run these programs under Win98se (for USB support) or Win2K or *nix.

The one reason for upgrading your chip that leaps immediately to mind is the resolution of some buffer lockup issues, but there were other workarounds for those, and not everyone had problems to begin with.

[Orpheus]

Since I am a big believer in hands-on building (if you've never done it, it'll open a whole new world), and since making the board is by far the hardest part of the project, you might want to consider this kit which has a simpler design with a lower parts count (i.e. easier assembly) and is selling the PC board separately for just $6 +S/H (which is just $1 to the US, $3 to Canada) A full kit with all parts (except cable) is a pricier $45



It doesn't come with a cable, but it lists sources for the OBDC/CAN connector (and all the other parts as well). If I were building it myself, I'd either spend a few more bucks for a fully premade cable with both ends assembled, plus a hood (I've seen them for as little as $7 lately) or if expense was an issue, skip the D-connector end, and solder those wires directly into the corresponding holes on the board -- that will save you buying both the male and female D connectors, which are kind of going out of style anyway. On the other hand, since they are going out of style, you may be able to salvage them from an old serial cable extension (or dead motherboard, modem cable, etc)

Since the ELM327 is the real brains of both designs, and both interface with the same OBDC programs, they should be functionally equivalent. Any modifications we make to our design won't directly apply, but can very likely be adapted.

I don't know if the kit comes with a chart of test voltages, as ours does. I'd have found that very useful when I was starting out

[Nick]

Digimoto Lite is so far my (by and large) favorite because it has some really good data logging options.  I wish there was a way to customize gauges so I can get instantaneous MPG and average MPG without having to export the data to something else, but as far as covering the basics, it does so fairly well.  I've tried the paid 4.0 version and didn't see enough useful features to justify paying money for it, so I stick w/ lite.  There's a version 5.0, but development appears to have stalled.  It's also DRMed and requires activation.  I refuse to pay for DRMed software that has a risk of becoming abandonware and end up locked out of what I paid for.

[Orpheus]

As many of you may have experienced, car manufacturers have been withholding key command and data access codes from the auto repair industry, to force car owners to go to dealers for servicing. A warning light may come on, due to a hazardous condition, but an independent car repairman may not be able to reset it, after the repair is completed, leaving the car owner unable to pass inspection or know if any further emergency conditions arise. In other cases, independent repairmen may not be able to interpret error codes or retrieve data necessary to further diagnose or make repairs.

Worse, I think that most of us know that dealer mechanics are not automatically the best diagnosticians or repairment. For example, Volvo dealer technicians are forbidden to work on transmissions. They can only perform replacements, and are required to send the old transmission back to Sweden. If you need such a replacement, you will be given the official line that "Volvo doesn't sell rebuilt transmissions", but there can be little question that the nominally "new" transmission Volvo does sell was rebuilt from a returned unit, especially if you have an older model. Many states, like Massachusetts, have laws saying that you are entitled to keep the broken part for inspection, but manufacturer imposed "core charges" that can run thousands of dollars make that option useless for most owners. What would you do with the broken transmission, even if you kept it? And with so few cores available, there is little expertise in rebuilding them.

Last year I fixed a red warning light problem on my father's Volvo transmission. Far from being impossible to fix, as the dealer said, it was a simple electrical short in the wire to the S2 solenoid. Repair cost? Under two hours and two bucks for an amateur like me. A good auto shop should have been able to do it in an hour, were the resources available.

The Massachusetts legislature is currently considering a "right to repair" law that would force the manufacturers to release the OBD/CAN codes and other onboard computer information. Needless to say, the auto manufacturers are fighting it hard, because manufacturer secrecy has created a thriving industry in third-party OBD/CAN code databases, with subscriptions costing thousands of dollars, which would immediately distribute information released in Massachusetts to the rest of the world.

I'll be following this law, and anyone who votes against it or tries to block it from a vote, because there is no substantial auto manufacturer presence in Massachusetts, nor any benefit to consumers from not having access to critical data about their own car. IMHO, it would clearly highlight state legislators who are "excessively susceptible to lobbying interests"

[Orpheus]

UPDATE:
I'll be buying a "light use" truck/van/cargo minivan at Interstate Auctions soon. I just want some convenient cargo capacity, though I won't foreseeably put many miles on it for many years. I can therefore tailor my usage to how many trouble-free miles it seems to have left in it over time. It's a reality of "convenience" utility vehicles: lifetime reg/ins will likely exceed the purchase price, since you aren't pressed to buy immediately and can wait for a bargain. Since my personal inspection/test-drive window will just a few hours before any auction series, I'm looking into the current state of OBD-II/CAN readers. Sometimes the mere fact that the error codes were reset a few miles ago can tell me not to waste my time. (My favored auction house doesn't routinely reset codes, but beware: some do.)

You can get an OBD-II Bluetooth unit on Amazon for about $20, similar to the price on eBay, DealExtreme or other oriental-origin site, and only several dollars more than the cheapest hardwire USB OBD-II converter. I recommend that even if you feel strongly about using an authentic ELM327 chip or want to build your own unit, you're still better off buying one of these for parts: a bare OBD-II connector still runs $13 or more. My old wired unit has a lot of parts (like the housing and connectors, useful for other projects) that are worth almost what I paid for it, years ago, so one way or another I'm going Bluetooth.

Bottom line: despite the logging capabilities of most current OBD-II apps for laptops or handhelds, It's foolish to do an OBD-II test drive alone. You should have someone else drive while you monitor the OBD-II from the passenger seat -- even if a 3-5 ft (1-1.6 m) hard cable will reach, it's often very awkward.

I haven't looked at the current state of the art in laptop apps (Windows or *nix), but if you have a smartphone or tablet, the Torque app seems to be ruling the roost, at least on a first look. Comments? Suggestions?

[Orpheus]

As I've noted OBD-II is electrically the same as OBD-I; It's just more consistent among the various makes, and somewhat more capable *as a standard*. Any specific implemetation of OBD-I may actually be more capable than OBD-II, depending on what the car maker chooses to implement. There's a fair diversity among OBD-II implementations, too, including some broad categories with obscure names, and specific extra features for various makes/models. If you play with OBD, you'll soon learn what settings and capabilities apply to your specific car. The web is full of good sources.

Newer cars use CAN (Controller Area Network) which is just a slightly fancier OBD. Amusingly, CAN was invented for cars by the Society of Automotive Engineers back in 1983, but over the past 20 years this very capable networking system has been used for everything from airplanes (including combat jets) to tanks to factories and power stations, while cars hobbled along with some crippled OBD subset -- usually idiosyncratic and not as easy to use as a full CAN

The Car Kracker project is a useful controller that works with the ISO 9141 "standard" (The standard doesn't deserve ironic quotes, but the implementations do) which is common for most Honda, Chrysler, Toyota, VW, Minis Land Rovers and BMW's from 1995-2005. It can also be used on some other makes/models (especially with all the rebadging that has been going on since the 80s) and as more and more hackers learn to use microcontrollers, I can see it branching out into other standards -- this could be easier than starting a thriving project for those other standards from scratch.

What can it do? Well, I haven't looked into it in depth, but it shows promise for allowing you to reprogram electronic dashboard or radio displays to show what *you* want, to integrate car systems with bluetooth or the web (via cell or wifi), collect running data, add locking and other functions -- it's still in its early days.

This post is just a heads-up so those of you who are interested can research it on the web for your own uses.

[Mark]

Mark: Cool! The Car Kracker is like a combination R2-D2, and C3PO for your vehicle. It negotiates with the onboard electronics and hardware, then it translates the languages like C3PO so you can understand, customize and reconfigure your own Millennium Falcon. Who knows...one day we may be able to activate our FTL drive via our danged smartphones. That device is capable of achieving some really unbelievable results...all things considered.

[Orpheus]

This isn't really an OBD/CAN hack. It's more an OBD/CAN connector hack, suitable for anyone

The official manufacturer repair manuals increasingly recommend disconnecting the battery for more and more repairs. Sometimes it's really warranted, sometimes it's just necessary to get the (+) cable out of the way, and sometimes it's just excessive paranoia about possible static discharge.

Unfortunately, as they recommend it more and more, it's also increasingly inconvenient. Disconnecting the battery allows the car's RAM to discharge, which can lose many settings: the time on the clock, the channels on the radio, custom seat positions -- and on some cars can reset the security system, disable the radio entirely (ostensibly an anti-theft measure) or cause other serious inconveniences.

Mechanics commonly attach a simple jig (two wires connected to a 9v battery clip) across the battery cables before disconnecting the battery. It saves a LOT of time and inconvenience, but it's not ideal. For one thing, 12v is better (but probably not worth the hassle unless you already have a 12v battery handy. On many cars, you have to turn the key for this simple mechanics jig to work, but this activates other circuits, may expose your car electronics unnecessarily and may draw more current than a 9v battery can sustain for very long.

A better solution is to use the OBD connector to feed your cars electronics directly (pin 4 is power ground, Pin 16 is +12v).  I recently saw a $250 OBD battery bypass at my mechanic's garage and couldn't resist taking it apart (with his permission) Inside it had a small 12v SLA battery, a few LEDs, some resistors, and not much else

Small SLAs (Seal Lead Acid) and similar 12v batteries are cheap. or you can use one 9v or 6-8 AAA batteries in series. The rest is optional, and probably in your junkbox. Your biggest expense would be a male ODB plug (which have gotten a lot cheaper in the past year), but you can equally easily use single-pin terminals, micro clips or even small alligator clips wrapped in electrical tape.


The female OBD-II connector on your car, seen head on.
Attach the bypass battery to pin 4(- or GND) and 16 (+)
BEFORE you disconnect the car's battery.

[Orpheus]

Here's a link to Ford's related OpenXC initiative which includes official support, including release of Ford's CAN messages. Anyone who surfs in to read this thread will probably be interested in that one too