Sensor Array

Discussion in 'The Projects Forum' started by bconnaghan, May 9, 2012.

  1. bconnaghan

    Thread Starter New Member

    May 9, 2012
    I have a hobby project I am about to start and need some general guidance since this my first attempt in embedded systems.

    I am planning on having an array of 80 sensors that will be outputting signals in the 0-15 millivolt range.

    I want to use these sensors to determine the location of an event. When this event happens there will be a very brief voltage spike on a few of the sensors. I need my system to determine which of the signals has the highest amplitude and to output a digital identifier of that sensor. I will then record this information (which sensor was closest to the event). I do not need to record any information about the signals themselves, I only need to know which sensor output the strongest signal.

    Since I am such a beginner I was hoping someone could simply point me in the right direction of the general architecture of this system so I can know what I need to start reading about. Right now I am a little lost since I don't know how to go about accomplishing this and there is an abundance of information to read through. Any guidance, no matter how general, is much appreciated.
  2. Brownout

    Well-Known Member

    Jan 10, 2012
    That makes two of us lost.
    We don't know what event you want to sense
    We don't know what sensors you want to use
    We don't know the objective of the project
    We don't know what analysis you want to perform
    We don't know....
  3. bconnaghan

    Thread Starter New Member

    May 9, 2012
    Sorry about that, I am actually in the process of trying to patent this idea and this was my initial attempt at making a working prototype. Being the paranoid type I would prefer not to disclose exactly what I am doing yet, as it is a semi simple solution to a problem lots of people have thought about. I understand this is probably poor forum etiquette and I apologize. Perhaps I should repost after submitting a patent claim.

    Still I am just looking for general guidance and I know any answer will be stunted by my lack of information, but the concept is just an array of sensors, and when an event happens a few of these will pick it up and output a millivolt signal. I just need to record which sensor picked up the strongest reading. These events will be happening every few minutes.

    Again I apologize for being vague.
  4. mcgyvr

    AAC Fanatic!

    Oct 15, 2009
    You would just use a microprocessor to always be reading these sensors and do whatever logic/functions you want when something is detected.

    i2c/1 wire might work here in that you can have a bunch of sensors on that single line and each one is assigned a 7 bit address and blah blah blah. :)
    of course I suspect these sensors aren't all close to each other (within 300ft) and maybe spread out around the world or state or whatever.. In that case I'd use ethernet where each one has a mac address and can be located via that or any of the other thousands of ways to accomplish that.

    A hobby where a patent and an engineering team are required no longer becomes a hobby though. I hope you have plenty of thousand dollar bills laying around.
  5. mcgyvr

    AAC Fanatic!

    Oct 15, 2009
    Not to mention that sensor nodes/clusters,etc.. are already done/patented/in process,etc.. for just about anything you can think of.

    MIT's got flying robot bee sensors in process. Bridge/earthquake,etc.. sensors have been done for a long harvesting sensor nodes/clusters,etc.. all done.

    Research TONS before you even try to patent something as that process is FAR from cheap and I'd bet you'll find its already been done/prior art,etc..
  6. bconnaghan

    Thread Starter New Member

    May 9, 2012
    Thanks for the input guys, I guess the question i have now is how do I got from analog signals to something a microcontroller can process? Doesn't this require some type of additional equipment/ interfacing with the micro controller? Is a mixed FPGA something that might work? Really im just looking for a high level overview of how a system like this might work? What converts analog to digital. How does the controller store and manipulate the data once its digital? How does someone actually go about programming a system like this?
    Last edited: May 9, 2012
  7. wayneh


    Sep 9, 2010
    A lot depends on the timing element, whether you're looking at seconds or microseconds or less. If the process is "slow", you can multiplex, sampling all 80 sensors in series, one at a time, maybe all within a second or less. Or use four channels to each multiplex 20 sensors, you get the idea. A cheap ($100) data acquisition device such as a LabJack, DATAQ, or Arduino would be able to do that. Each of those will have an upper limit on how fast they can sample. Faster generally equals more expensive. If you need big iron, look into National Instruments line. They're also well known for their Labview software.

    You could consider putting a peak detector circuit on each sensor, so that it would lock the highest value seen between resets. With 80 sensors, that's at least a possibility. This would allow you to sample and hold until the analog-to-digital sampling can come along and catch up. You could even have a corresponding LED light up for sensors that have produce a voltage above x volts.
    Last edited: May 9, 2012
    bconnaghan likes this.
  8. GetDeviceInfo

    Senior Member

    Jun 7, 2009
    I would think that you'd have to establish what constitutes an 'event'. Would it be a rate of change, a change in level, or maybe even no change in an otherwise continually changing environment. Would an 'event' be a brief spike on many sensors? Define your event, then work towards capturing it.

    A project I worked on years ago had a small array of infrared detectors with programmable opamp gains. Each sensor was supervised by a small footprint 8051 micro. Sensor output was smoothed and loop'd back to control gain and setpoint to aquire a high sensitivity to 'events' within a changing background. The sensor(s) capturing an event notified the master with thier ID and a packet of data, over a token ring serial scheme.
    Last edited: May 9, 2012
  9. milesguidon

    New Member

    Dec 4, 2010
    I agree with wayneh on this one--I'm doing this with some microphones at the moment, and a PIC18F4550. The client wants to know from which direction a noise event came from. You might consider amplifying your signal (nothing fancy, just put it through a simple op-amp stage with filtering if you need it). Your "missing link" is just finding a controller with an on-board A/D (analog to digital) converter. Your A/D converter will read the analog signal and give you a proportional digital value. You just need to do some quick math on the digital result, and then you've got the analog value.

    I don't know about 80 sensors--that's a lot--but you'll want to go more in depth than just choosing "the largest" signal. Simply choosing the largest signal leaves you vulnerable to noise and unintended/unimportant events. Depending on the sensitivity of your sensors and how much you trust your setup, you should look into characterizing a pulse based on the shaping of the data that multiple sensors feed back. Instead of just focusing on each sensor individually, consider them as feeding you back a "shape" of the signal. If you're getting a parabolic response, where the second and third largest values are from the sensors directly adjacent to the sensor reporting the largest value, and the fourth and fifth largest values are from the sensors adjacent to the second/third largest value sensors, etc., then you know you're likely experiencing an actual interesting event.
  10. MrChips


    Oct 2, 2009
    There is a solution for everything (well almost everything) but the devil is in the details.

    As others have indicated, we need to know the details such as:

    What is an event?
    What is the sensor?
    How far apart are the sensors?
    What is the timing information from each sensor, waveform, repetition rate, resolution needed?

    (I am doing this with nine sensors and my timing resolution is 1μs. I am detecting gamma radiation).
    (I have another project on the go with 160 sensors).

    Without knowing the details I doubt if we can be of much help.

    I suggest if you would like to learn more about electronics, scrap the patent idea and tell us what you are trying to do. Hobbyists don't get rich by spending money on patent applications.
    Last edited: May 14, 2012