Handset CAD anyone? :)

Genesis

The Alibre Express package has one big plug in its favor - its free! It also exports to STL files.... (even the free copy!) The "Pro" version appears to compare very well against SolidWorks....

That the free version can do STL export was quite a shock....

Solidworks looks very nice as well, but its crazy expensive.

One thing I like about Alibre is that its UI is very intuitive and easy to work with. Unlike AutoDesk's products that have often led me to "bang-head-against-wall" trying to figure out how to get it to do something....

divetheworld

As always, you get what you pay for I suppose. I never did like Autodesk products, even the Inventor is a bit of a nugget. I create it in solidworks then export to an Autodesk file. Much more civilised!

__________________

Genesis

I'm not a fan of AutoDesk either, but have been more-or-less forced to use it in the past. That "experience" (if you can call it that) is one of the things that has led me to be extremely gunshy about buying something like this for 3D, industry-standard or not (after all, look at how long AutoDesk was THE name in CAD!)

Anyway, the Alibre software is IMHO definitely worth a look. You can buy a new copy of the "Pro" level product for the price of MAINTENANCE (well, ok, within a couple hundred bucks) on the SolidWorks product for one year.... and they include maintenance and support for the first year in the price (unlike both AutoDesk and SolidWorks)

steevke

I've used PTC Pro/E in the past, and now I use Inventor for most of my work. It's a very neat package that (imho) is much more user friendly than Pro/E, even though the sheetmetal part is not 100% yet.

YMMV, as always.

steven

Genesis

Here's the mechanical (no traces or labels, sorry ) on the boards I sent out this evening for fab..... this is a little bigger than "actual size" - the actual boards are just shy of 2.5" high and 1.9" wide (the "double" is 3.8" in width)

The line down the middle is a split line - the left half is the part that goes in the battery can, the right part gets potted and goes in the head. The controller piece has a DB-9 fitting on it for updates and such - I may put a USB bridge chip on there in the future as a mini-USB port is a LOT smaller physically.... The right part currently has connectors laid out on it for testing (bench) purposes but once I'm getting ready to put one I'll just solder wires in place and drill a second strain relief hole.....

Should have the boards back Monday; assuming I didn't farkle anything up the transformation from "breadboard" to a working board should take a couple of hours..... all the components are here to build 'em up.

Scuba_Vixen

Looking Good!


Keep us posted,


Darlene

Genesis

I'm a messy hand-router, but what I route usually works (No autorouting software here...)

I know all the so-called "rules" for PCB layout and routing but sometimes they get difficult to follow - especially the "lay all traces on one layer either horizontal or vertical" rule. Yeah, right. Ok. Now pull the other one. How many vias did you say you wanted to have on the board?

Seriously, this is "Rev A" on the boards; I'm going to build them up with all 1/8w resistors (they're well within power ratings so 1/4s are unnecessarily large), so I can cut the spacing on those. Of course I could go to SMT but rework on that stuff is a royal pain in the tush. The only SMT part on the board right now is the 5V reg as all the "good ones" are showing up only in SMT packages. Such is the price of progress - when I was 20 that tiny stuff and the vision issues were non-events - but I've got 40 year old eyes now!

The board is designed to be fully short-tolerant - I'm still torturing everything to make sure though. I have one potential area of concern in the solenoid drive output - its protected, but whether the protection will act quickly enough to prevent damage (and whether the protection will prevent the V+ rail from getting dragged down far enough to cause a processor reset) remains to be validated under "torture" conditions.

I suspect there will be a "Rev B" on the head board at least. One of the cute things about this particular design is that there are multiple processors that are pin-compatable and will slide right in that big DIP socket. There's a second one I have my eye on that is a LOT faster than the BS2pe that I'm using right now - and is multitasking - but I suspect I may not like its power consumption numbers. We'll see - I'm going to get one and play with it - I've socketed the processor for that reason.

The "gotcha" down the road will likely be a desire to add additional off-CPU storage (e.g. an EEPROM or other persistent storage chip) which may force that board into SMT components due to space issues. The goal is to insure that it will fit down the bore of a 4.5AH light can (fastened to the lid) - makes for a nice inexpensive waterproof enclosure.... Two Fischer bulkhead connectors on the lid make the cable integrity much less important (breaches in the cable do not flood the cannister) but raise cost somewhat. Whether that will be worth it remains to be seen.

Scuba_Vixen

I haven't gotten as far along as you have on the solenoid control aspect, but one of the things I was mulling over, relative to isolating any solenoid problem from the rest of the sytem, was to use a seperate power source for the solenoid/driver, and couple it via opto coupler to the PIC control circuitry.

There's an interesting coprocessor chip for the Stamps that's popular in robotics apps, you may find it usefull to offload some AD and timer functions to.

http://www.bluebelldesign.com/CoProcessorFeatures.htm

I haven't played with it yet myself.


Darlene

Genesis

Actually the A/D part is simple - any serial A/D works very well using only three wires - chip select, data and clock. No problem there.

The issue becomes twofold with respect to the solenoid:

1. Protecting against destruction of the output transistor in the event of a short. "Destruction" usually means smoke, and that's REALLY BAD in a high FO2 environment. Even if potted, destruction means the board is junk as you can't replace the transistor in a potted assembly. Thus, the trick is to either (1) sense the short quickly enough to prevent it from trashing the output stage, or (2) protecting the part so that it can drive into a short and not let the smoke out. The fast way out and what I've done for other applications (which are accessible and not potted) would be to use a picofuse, but that's a one-shot deal and once blown you get to buy a new board due to the potting. I'm trying like heck to avoid that.

2. Protecting against the V+ rail getting dragged down too far during the duration of a short until your protective circuit can take care of it. This one can also be difficult, because a V+ rail drag means a potential processor reset. A separate battery supply can solve that - but I'm not sure I want to go there.

One potential solution is to put a current sense resistor (very low value) inline and then sense the voltage drop across it, thus giving you a current draw. Turn the solenoid on, immediately look at the output draw, and if its out of spec (either open or shorted) immediately shut it off and alarm. This is a potential solution but needs to be backed up with hard protection because you can't risk the CPU not reacting correctly "in duress", given that the risk you're running is that of a fire in a hyperbaric environment or (if primary battery protection opens) loss of ALL controller power.

What I've done tentatively is put a resettable thermal fuse inline with the solenoid, which in theory should prevent the output transistor from melting down even with a solid wire-level short across the solenoid terminals. It also has a high enough intrinsic resistance that (again in theory) the V+ rail should survive intact until the protection operates even across a dead short, assuming decent battery power in the first place (that is, the V+ rail cannot be in alarm due to low battery level already - if it is, you wouldn't fire the solenoid anyway.) It automatically resets, so having it trip is ok - it's turn itself back on once the short is removed. They're the size of a midsized ceramic cap.

The efficacy of this approach is yet to be proven. So far it looks reasonable, but I'm not completely happy with the temperature rise on the output stage during a sustained, direct short, and I managed to provoke an unsolicited reset this afternoon during torture testing - albiet right at the ragged edge of the regulator's dropout voltage. Interestingly enough the "crash log" caught the event and successfully saved it to EEPROM before the processor went down!

I may have to step down on the value of the protective resettable fuse; I used one with a decent margin over the expected draw, but may have given it a bit more room than I really should have.

In any event this is the one remaining serious design verification challenge that I have left "unproven"; all other I/O on the boards are safe against shorts to either voltage rail. The V+ regulator will fold back if shorted as well - the problem with the solenoid drive is that it is intentionally connected to unregulated V+ because a foldback on the regulator kills the CPU, and that's not acceptable.

A potential solution might be to use a second regulator only for the solenoid and trust its foldback and internal protection to solve the problem. If I have to go there then the head board needs to be re-fabbed to accomodate that IC.

LS_DK

Hey,

I have also made a set of CCR electronics (for my inspiration). At the moment the electronics does not control the solenoid, but merely monitors the 3 sensors and a depth sensor. It does a VPM/BuhlMann hybrid deco algorithm.
(8 bits MCU and 4 MHz clock, was not enough for my VPM-B algorithm.. ,, 45 seconds calculation time when Ascend time exceeds 30 minutes is not good enough)
I have prepared my board for driving the solinoid as well.
Had the same sort of considerations as you regarding the protection of the gateing transisitor and the battery..
My solution was to use a current mirror. I.e. figure out how much current the solenoid requires for opening, add a safety margin, and then tune your mirror to that value. The mirror wil deliver the current into any load unless the load voltage exceeds the supply voltage. I appreciate current mirrors are a bit "old school", but in this situation, where you have an unknown load resistance it does the job for you. The current mirror can be controlled by a GPIO on the MCU. The MCU would then be "behind" two Collector/Emmiter-Base transisistions which should be enough protection for most of todays MCU's.


Cheers,
Lasse