**LOW** Power application

Not true. Consider a force that partially countered gravity like a landing aircraft. There will be more than one third force that if added to the existing two will result in the same magnitude of acceleration. I can't think of any cases that would be sustainable but it should be possible to come up with short term cases. Whether they are realistic.....

;)

In that case so will accel of course, so there is no need to distinguish. However if you are that sensitive to tilt maybe it's worth looking at packing differently. I have noticed packaging tilt sensors though so it appears that there exits a fair number of shipments that have that as a constraint.

Robert

Insolent anarchist.

Yes, you are right. It does not require three forces, it just requires a force in the right direction of the right magnitude. But it would be the odd case that would put you there without a transition through a value of different magnitude.

Here the difference is duration. It would be nearly impossible to accelerate a package so that it appeared to be upside down for more than a moment. This sort of acceleration would not likely spill the contents of various items either. But if they are left upside down for some minutes, hours or days, it might be a different matter with the contents spilling out.

Of course in physical terms you are right, there is no difference. But in practical terms it would be hard for an acceleration to mimic a tilt.

In this case, the guy acknowledged that he was new to the newsgroup so it was appropriate to point out what some consider to be etiquette. But other times people must simply be full of themselves and make a big deal about this top/bottom posting thing. I have been posting for years and sometimes I just think a top post is warranted. A couple of weeks ago someone actually sent me an email about newsgroup etiquette and how top posting is bad! WOW!!

All I can say is some people need to get a life...

So we are looking at frequency content and relying on the momentum of the liquid to keep it from moving to much. Practically speaking more likely both need to be considered.

Although in those cases (particularly the battery case) the obvious solution is to ship them empty and fill on receipt.

Robert

I just figured three forces would be easier to explain without resort to diagrams :)

I'm not so sure it would be that hard to have a smooth transition though. I thinks it's unlikely to occur in a commercial environment though, you'd probably have to deliberately set out to create it. Reminds me of a science fiction short story I read some time ago.

I can think of two cases where acceleration would mimic tilt that would affect this one in particular. First is straight acceleration, the other is 'rounding a corner'. The second acceleration in particular could easily mimic fairly large tilts. Yes the total accel is moderately larger than G but it really doesn't matter if the source is tilt or acceleration, the result is the same.

Robert

In both cases you describe the acceleration may be smooth, but the magnitude will not be constant. A tilt of any angle does not change the magnitude of the force felt, just the direction. It would be an unique case where the magnitude of the force felt is constant, but the direction changes smoothly. Bascially it requires dropping support of the package while pushing it from another direction in a smooth motion. Hard to create on purpose, much less accidentally.

I thought that was what I just said? I do believe we are in agreement.

Which only leaves the question why would anyone care? Whether the total magnitude of the off axis accel is 1G or 1.25 G, it's still off axis which strikes me as far more significant. Anyway that's a question for the OP.

I went through some of this looking at a tilt interlock on a scissor lift. I quickly came to the conclusion that whether the off-axis accel was because it was parked on a slope or because it was turning a sharp corner really didn't matter. Unstable is unstable.

Robert

But it does make a difference if it is because you hit a bump or touched the brake. For the shipping container it will not experience accelerations that are sustained unless it is in free fall. So frequency would be a good discriminator between shock and tilt.

--snip--

--snip--

Use an analogue peak detector for each sensor (you should be able to get the leakage current low enough with low bias op amps), then periodically wake the micro and read each peak detector with it's A/D. Have a reset on the detectors so you can then look for the 'next' set of highest readings.

Cornering more likely a concern than braking or bumps in that case. Braking happened on a more stable axis. Bumps could be deadly though, there was a set of mechanical interlocks to deal with those.

Really? It's never going to stop, start or turn a corner(1)? Those all approach tilt magnitudes and time scales.

Absolutely, less so between longer accels and tilt. Although any accel that lasts for hours is probably tilt :) There's a range in between that's harder to distinguish though and since the effects are the same (barring something actually is sensitive to fractional G differences) why bother?

Robert

(1) With work you probably can distinguish the corner turning case from the simple fact that the unit vector keeps changing but it would have to be a special case to matter.

Yep - use alkalines. If there's space say 6 x D cells - 15000 or so mAhr depending on manuf.

or if not enough space 2 lithium D's , 20000 or so mAhr

The OP required 12 weeks or about 2000 hours, the average current must be below 7 mA. Considering that current electronics operate well below

Subtract power supply losses and common static current consumption. Divide the remaining current by 10 (number of sensors) to get the current available for measuring and processing the signal from one sensor.

Subtract the single sensor and channel specific preamplifier consumption. The remaining current is multiplied by the CPU voltage to get the power available to do all calculations for the sensor.

Find out the amount of energy the processor needs to execute a single instruction (on average). Divide the available power by the energy for a single instruction to get the number of instructions available each second to process a single sensor.

Divide the number of instructions by the required sample rate and you know how many instructions are available to process a single sample. Then it is up to you to squeeze your program into the number of instructions available for a single sample on a single channel :-).

This is a simplistic approach, but it can be used as a reality check before continuing any further. If you get a negative result in some of the calculations, there is not point of continuing any further with this set of parameters.

Paul

We estimate active current of 30mA and idle current of 1mA, so power management is critical.

Depends on the clock, for an AVR (PIC and others are similar), less than 1mA at 32KHz and over 10mA at 8MHz. A2D and sensors add another

Instructions do not matter much. Active subsystem affect power consumptions more than number of instructions. For example, to ramp up the A2D, you need to enable the analog subsystem and wait for it to settle. You could be running thousands of instructions to check the status register.

If the AVR is running 10 mA at 8 MHz, you might be better off with an ARM processor such as the AT91SAM7S from Atmel. We did a comprehensive comparison and found that when running, the ARM is lower power, clock for clock, and the ARM does more in a clock cycle than does the AVR. We did not compare the ultra low power modes where the processor or clock is stopped, so I can't say which is better there.