Cheap proximity sensor

[danst83]

I'm looking for a proximity sensor that can detect metal about 7mm away. A capacitive or inductive solution. The sensor should have a relatively low cost (under $1). I've used the tda0161 chip from ST in the past but they are phasing the chip out so I can't use it any more.

 

[robismyname]

Im thinking a hall effect ic can work for you. i found one for under .80 cents.

**broken link removed**

 

[danst83]

I plan on using this sensor in a container that contains magnetic particles so a hall effect sensor won't work cause it detect magnetic fields so I'll have to use a magnet instead of metal which is not feasible in my case.

 

[shabazmondal]

Have u tried searching for IR sensors,there are a wide variety of IR sensors,though im not sure about their price

 

[ard]

Hi,

Cant get much cheaper/simpler than this!

NEW:: picoDetector : How to detect metal with a PIC

Regards,

Anand Dhuru

---------- Post added at 06:14 ---------- Previous post was at 06:14 ----------

Hi,

Cant get much cheaper/simpler than this!

NEW:: picoDetector : How to detect metal with a PIC

Regards,

Anand Dhuru

 

[danst83]

Ard, Thank you! I think your solution might work. I just need to figure out if a 7mm range is too much for that picoDetector. I'll build it something this week. I'll keep you posted.

 

[danst83]

I built that picoDetector a couple of days ago and it is not very stable. Sometimes it signals metal presence when there is none. The detection range for the sensor is more like 4mm. It wasn't able to get 7mm out of it. Another requirement that I have that I forgot to mention is the detection speed needs to be less 1ms. I'd really appreciate any input on the subject. IR sensors won't work there's a 3mm piece of plastic that separates the sensor from the target. Thanks!

 

[ard]

Replacing the inductor shown with a larger diameter hand-wound coil might help. But you'll need to experiment with the number of turns to get approximately the same vale as the one shown.

Regards,

Anand

 

[danst83]

I played with different inductor sizes but 7mm was very hard to achieve. But my main concern is not the distance. It's the stability of the sensor and the response time. (I believe the stability problem comes from the fact that I use a bigger inductor that requires more samples) If I keep increasing the size of the inductor then the OSC frequency will drop to the point when I'll have to increase the number of samples the PIC takes before it can accurately detect metal presence, which will increase the response time. But even now, the sensor is too slow for my needs. The response time of the sensor is 18ms (watchdog timer) * 2 (pre-scale factor they use) = 36 ms, which is way above 1ms that I'm aiming for.

 

[ard]

How about using a variation of this: Any sensor that can detect mettalic objects and magnetic objects : 8051 Microcontroller Projects AVR PIC Projects Tutorials Ebooks Libraries codes

Looks sensitive, and fast enough.

Once you have calibrated for beat frequency, any presence of metal would produce a signal on pin 11, which your CPU can measure.

Anand

 

[danst83]

Anand,

Thank you for your help. It's a feasible option. I like the idea of that metal detector. It's a basic colpitts OSC that it's constantly being compared to a stable frequency reference. The only problem that I see with it is that it might go out of calibration over time. It's not temperature-compensated and i don't think it's linearized either. It will also require a separate uC to process the output. But still a good option. I'll have to see how much the frequency of the colpitts OSC changes with the presence of metal.

 

[ard]

Anand,
The only problem that I see with it is that it might go out of calibration over time. It's not temperature-compensated and i don't think it's linearized either. It will also require a separate uC to process the output.

You know, the calibration might not be a problem; you could use another small cpu to sample the beat frequency before every use, and make it output the same frequency thereon. Perhaps add a calibrate button that saves the value so you dont have to do it till you suspect the calibration is off, in which case you just hit the button.

For the linearity I guess you'll need a look-up table or a curve fitting algorithm. But if you just need a go, no-go reading, the linearity wouldnt matter, would it?