Suggestions for linear displacement sensors, durable, underwater, low rez.

Our frame is all aluminum because it has to be water resistant, but I'll make sure to keep iron out of the frame around it.

For the force sensor and spring idea, I was thinking I could employ one of these:
https://www.amazon.com/Electronic-B...1?s=aps&ie=UTF8&qid=1331088641&sr=1-1-catcorr
It's probably a very low precision load cell for that price, but like I said, I don't need precision, and it's probably more linear and reliable than those force sensing resistors. For $6 it seems worth a try.

That looks good, has absolute output, bolts right on, needs an amp but still seems fairly simple to read, can probably waterproof it with a shot of spray-on something. It's little light, though - if you apply more than 2 kg force (4.4 lb) you will permanently deform or break it.

I noticed that, I was going to go with the 5kg or 10 but they ran out. I will probably go with 5 if I can find it elsewhere.

This french place has two of the 5kg load cells in stock -

**broken link removed**

There's this version, too -
https://www.sparkfun.com/products/10245

$10, 50kg, plenty in stock.

My sensors came in today! I'm getting to experimenting immediately, I'll post about how it goes.

The magnetic sensor seems steady and reliable, and it doesn't drift too much from other magnetic fields until they get very near, within 2 or 3 inches of the range the magnet you intend to sense is in. But the swing at 5v of excitation isn't 120 mV, I'm getting more like 80 mV (+40 to -40) before it hits the top of the sin wave and curves back down. Which is fine because I can build the amp to whatever I get, it would need to be adjustable anyway, though I do wonder if these sensors will all be the same, unfortunately for now I only have one. I've gotten some neodymium magnets and I'm going to see if a stronger magnet gives a wider swing, but I don't think it will as the distance from the magnet to the sensor seems to effect the accuracy at the edges of the linear range, but not the span of the readings.

The accelerometer and the bar load cell are in too, I'm just not mentioning them so much because they are normal and well known. Which is good, that its a regular load cell for $9. I think I'll include the accelerometer in the final design for it's ability to pick up high frequency movement.

Could you you explain a little about the science this robot is assisting with? I read the article, see that it's getting temperature, pressure and tilt from the surf zone, but I'm baffled as to what the upshot of all this is.

LMAR is a underwater science platform. All of it's capabilities are to allow it to navigate and operate in water while supporting extra equipment. For example last October we mounted a sonar device which measures flow speed and direction every 1.5 cm, from the platform to the surface up to 6 meters above. In one case this data was compared with concentration of invasive muscles to determine which flow and temperature conditions support them. In previous missions LMAR carried equipment to sample the bottom of the lake or test for levels of various contaminants and bacteria in the water. Soon we will be working with an organization that wants to search for unexploded ordinance underwater. Basically LMAR is like a truck, you can do a lot of things with it, that's why it's under the school of freshwater science and not engineering, they use it for all sorts of things.

Fascinating! I'm working with a guy doing something with those damn zebra mussels at Northern Products Development Group (I'm a consultant there). He's had some dealings with the DNR over it concerning bacterial counts from decomposing soft bodies. It would be pretty weird if these two projects crossed paths at some point.

The unexploded ordinance angle is scary. There have been some bad outcomes with this in Germany in recent years, and they often seem to find them on shorelines at low tide. They are even still finding unexploded ordinance from the Civil War here in the US - there was a bomb found in the Pamlico River during a drought recently, and the explosive was still as full of ka-BOOM! as it was 150 years ago.

There are magnetic axial position sensors such as these: **broken link removed**

A good use for them is to do "drive by wire" for accelerator pedals. You would need to convert your travel to angular position. Some are 10 bits or so.

Drive by wire is a good thing to consider. Since the magnetic sensors work I'll probably keep using the Honeywell ones, I've found that they with linear displacement they give a linear output between +/-45 degrees from center, that is, with the magnet pointed at the sensor and in line with its center you get 0v, and if you do not rotate the magnet, but slide it side to side, you get a linear change until the magnet passes the point where the line between it and the sensor is 45 degrees from center line of the sensor. After that it waves up and down sinusoidally and falls off as you get further away. If you rotate the magnet you get 4 sine waves per rotation, which I think would be harder to track.

So far we haven't gotten to actually look for ordinance, but it certainly is an issue that needs addressing. I think it will be a big project here soon because it's the most likely funding source for this lab in the near future.

The sensors all seem to work well. Now there's the matter of amplification. I'm not very experienced with analog circuits, but I've at least got an instrumentation amplifier to work. Since this lab uses so many sensors with differential outputs in the 20-500 mV range, what would be really great is if I could make a general design for them to use that could be customized with just a few resistors so we could print a few dozen PCBs and make an amplifier quickly whenever we need one. It would need to be able to be customized to take that range of inputs and scale it to either 0-5 or 0-10v, then everyone here could use it.

Here's what I did, if that's any help.


Plugs into an A/D port, span and offset adjusts were done in software.

That's a huge amount of help. I'm looking and I keep finding amp designs that really sweat the filtering and precision, and all I need is a reading that lets me know if I've crashed into something or not, and I need it fast so layers and layers of analog conditioning isn't the best thing.

I've finished fabricating the mechanical part of the bumper.

This is the bumper itself. The flat plates allow easy mounting of sensors, they have holes in several positions.

**broken link removed**

This is me holding where it will fit on LMAR.

**broken link removed**

Optionally it can be moved lower to the ground.

**broken link removed**

Thanks for the pix, looks pretty good! Seems to be a lot less than 3" - 5" of travel.

For testing I installed a lighter compression spring than it will probably use in the water so it's sitting further back then it can. You're right though, it will probably only move +/- 1 inch with this spring. Even with the heavy spring it's only about +/-1.5.

The magnetic displacement sensors work great, they are very reliable and easy to read! Now the sensor aspect I'm looking at is how to collect useful information from the accelerometers. For example, in the materials lab, on a full PC workstation, I've seen vibration data broken down into duration (before vibration dies off to a set level), power during that duration, number of beats (times the signal breaks a given amplitude), it even allowed these to be broken down into the same stats for the most prominent frequencies. Obviously an 8 bit PIC microcontroller running at 20 MHz cannot do all of this (just the computer for the sensor), and it wouldn't be practical to try to stream all the vibration data to the CompactRio which serves as the main computer of the robot. I think the most useful thing would be to get the magnitude of an impact, followed by the frequency and duration. But those aren't easy things to read without collecting a lot of high sample rate data. I'm thinking of analog conditioning schemes that might help, but I'm not very experienced at that, for example maybe a threshold could be set to trigger a counter, so that a slower digital pulse which is proportional in frequency to the vibration could be read. I'm sure this is something someone has studied a lot. Any ideas?

Shift the data from the time domain to the frequency domain, possibly with an FFT: https://en.wikipedia.org/wiki/Fast_Fourier_transform
Some micros (Renasys?) have instructions that make it easier ton perform.

Another possible way to look at the problem is through a "multi-channe (MCA)l" analyzer. Where you might have an acceleration(x) and a magnitude. Not really sure how to handle that. An MCA is used to detect X-Ray events. Your essentially looking at the number and the energy of the event: https://en.wikipedia.org/wiki/Energy-dispersive_X-ray_spectroscopy. Analysis software basically strips off the bottom "junk" and then tries to find a peak using the FWHM (Full Width Half Maximum) method.

Or even spectral analysys: https://www.electro-tech-online.com/custompdfs/2012/04/Spectrum-Analysis.pdf

The thing that might be wierd with your situation is that the the accelerometer data seems n-dimensional, so maybe collect it as a phasor.

Hey, you just asked for "ideas". I'm not sure what might work. In my head, I'm seeing a 3-D surface. # if events, direction, magnatude.

So, I'm seeing binning the levels. e.g discrete levels. (window dector) so to speak.
Dividing the direction into binnable levels 1 degree apart?)
And counting the events,

So you have a "surface". The surface can be displayed in "intensity modulated", "color modulated" or "y-modulated" form. The "Y-modulated" data could also be shown in color.

Hey, you asked for ideas. I did what I could.

For now I just have it measuring low frequencies using a cycling buffer. I'm thinking I could possibly use a frequency to voltage converter connected to the ADC of the controller to capture higher frequencies.