Replies to This Discussion

ModeratorPermalink Reply by Chris Anderson on July 10, 2008 at 5:48pm

In general, it's a good idea to keep as much data as you can as long as you can so you can use the maximum amount of information in your filters. Analog has more information than digital, so it's better from that perspective. Also, the ATmega168 has loads of ADC pins, so doing the ADC conversion at the chip level is no computational hardship. In short, I don't see the advantage of converting to digital at the sensor level.

Am I missing something?

• ▶ Reply

Permalink Reply by paul hubner on July 11, 2008 at 9:06am

I think I understand now. The only thing saved would be the number of wires between each sensor and main board, which is not a big deal.

I do want to take a look at your code when ready to release to keep learning!

Thanks,
Paul

• ▶ Reply

ModeratorPermalink Reply by Giles Barton-Owen on July 11, 2008 at 12:14pm

If you wanted to have your themopiles a long way away from the autopilot then you might want to use an adc to reduce noise but you would have code a lot for it!!!

• ▶ Reply

Permalink Reply by Xander on July 11, 2008 at 11:18pm

I may be mistaken, but I believe on the ATmega168 it takes more time to do an analog read than a digital one. So if you're really pressed for processing power and want to perform analog reads often, offloading the ADC to an external chip (very cheap!) might help.

Also, as Giles mentioned, the readings may be more resistant to noise if you convert it to digital closer to the source. But...that really shouldn't be a lot of code, they make chips designed to do exactly that.

• ▶ Reply

Permalink Reply by paul hubner on July 13, 2008 at 12:53pm

Yeah - I was toying with the idea of a 5-way thermopile - Pitch x2, Roll x2, and one for vertical orientation. The 6th shouldn' be needed because if you place #5 on top, it should always read colder that the average of the other 4.

Then I could mount it on top of the craft and not worry about seeing through the plane.

Digitizing at the source would get rid of noise, but I was just thinking of it as a digital sensor pack.

Paul

• ▶ Reply

ModeratorPermalink Reply by Chris Anderson on July 13, 2008 at 1:27pm

Actually, you need six so that you can ensure that the plane isn't inverted in flight (it happens more often that one might think!). The other ones can't tell. The Z-axis pair need to be on the side of the plane.

• ▶ Reply

Permalink Reply by paul hubner on July 14, 2008 at 6:57pm

I agree the needing a Z for inversion is needed, but here is a better description of what I mean.

For both the X and Y pairs, the temperature will always be 1/2 ground+1/2 sky. This is regardless of their orientation because they work in pairs: if one is up the other is down and average is maintained. So we have a near constant average temperature between the four X and Y sensors in all positions

If we point the single Z sensor up, it should always read colder than the average of the all the Xs and Ys. It we are ever inverted it would read warmer.

Now I understand that the range is now limited to 1/2 the full-earth-to-full-sky reading, especially you pull a 90deg turn and are perpendicular to the ground. but in that case a 2xZ would not be useful either. But your X+Y sensors would still know what to do. I think this would work despite the dynamic range limit.

So is the 6th really required?

This will continue to be a thought experiment until I get off my keister and get some boards etched!! ;-)

Paul

• ▶ Reply

ModeratorPermalink Reply by Chris Anderson on July 14, 2008 at 7:45pm

Are you sure it's true that the sky is always "hotter" than the ground? Even on an overcast day? At dusk? At night?

• ▶ Reply

Permalink Reply by paul hubner on July 15, 2008 at 11:05am

I wanted to say "yes" (though the sky is colder) but I can understand your point - weather inversions, warm days with snowy fields, etc. Unfortunately, any thermopile pair system would work backwards in this situation unless there was a way to tell it "up from down".

Left to its own devices, a leveler component would operate in reverse under these conditions. Leveling sensors would see an elevated wing as hot (earth). But because the sky is colder, a warm reading from that direction would make the aileron lower and exacerbate the problem by raising the wing more.

All I am proposing is to use the 'lower half' of the thermopile to sense the ground temperature. I know I cant really use half a sensor, but the aggregate average of the X and Y will always read (sky+ground temp) / 2, which I contend will be warmer than the sky alone.

I see what you are getting at though - relying on the temperature delta, regardless of weather conditions, is probably the best indicator, and a strong differential is a better than a weaker one

Thanks for the discourse! I really should get some empirical data before I let my thought experiment go any further...

Do you have anymore thermopiles available? I'll take six: (5+1 spare LOL).

Paul

• ▶ Reply

Permalink Reply by Xander on July 15, 2008 at 1:00pm

I realize you're trying to go for something simpler than a gyro + accelerometer setup, but past a certain point is it really still simpler?

• ▶ Reply

Permalink Reply by paul hubner on July 15, 2008 at 2:29pm

I know I know... the rule of diminishing returns. I still like the ideas of a small FMA-like sensor with a 3 wire output with all the data there. Somehow it seems like an elegant solution.
Paul

• ▶ Reply

ModeratorPermalink Reply by Giles Barton-Owen on August 1, 2008 at 1:35pm

Adding to Chris Anderson's point what about water?
This has been particularly bugging me as I would like to be able to work on water.

• ▶ Reply

• ‹ Previous
• 1
• 2
• Next ›
• Page