multi-point sensing

[_Raz_]

Hi Brian,

Thanks for the tip on the Tanjil Bren area. I’ve never been there, but I have spent some time nearby, around Erica and Walhalla, as my dad was involved in the construction of the Thompson River dam.

Unfortunately, I’ve never actually detected this particular area, but hope to in the future. My friends and I did spend a weekend detecting further east, in the mountains just north of Briagolong about a year ago. This was our first serious trip to look for gold. It was hard going (those mountains are steep!) and we only had one good detector (a Minelab SD2100) between the 3 of us, but we had a lot of fun.

In-fact, it was this trip that convinced us that we all needed a detector of our own, and since we weren’t too confident in our ability to convince our wives that $3000-$5000 for Minelab detectors was a good investment, the only other option was to design/build something ourselves, and maybe learn a thing or two along the way.

I’m a programmer by trade, so I am much more comfortable in the digital world than the analog one, but we reasoned that it was probably a good idea to keep things as simple as possible, at least until we had something that worked, then we could build on that. The only digital component that made it into the first design was the Atmel AVR, which handles the TX pulse / RX sample timing.

The Atmel AVR was chosen primarily because I already had some experience with the device, it has a nice straightforward architecture, is easy to code for and we had software that could simulate the device in near real-time, making development easier. It isn’t the most powerful device around, being an 8-bit device with a max clock speed of 16Mhz (~16MIPS), but it was more than adequate for our purposes.

I have since acquired some Microchip DSPic devices, being 16-bit with a max clock of 120Mhz (~30MIPS), these could be a useful upgrade from the AVRs, should our design ever need a bit more processing capability. I haven’t had the time to actually play with these devices yet, but ‘on paper’ they look promising.

To answer your question; our design does not use a differential front-end. I did consider it, and no doubt it would have reduced the ground problem significantly, but we already had a collection of Minelab / Coiltek mono and DD coils which we wanted to use with our detectors.

I was very interested in your description of what you are doing. It definitely sounds like it is an order of magnitude beyond my experiments here. I think I understand most of what you wrote, but looking at the flow charts would probably make it clearer. I would have thought that retrieving the positional information, with enough resolution to be useful, would be a nightmare.

If you don’t mind, I have a question regarding your front-end / ADC; How much analog gain are you using before the ADC, and how many bits of resolution did you find was required for the ADC itself?

Anyway, I think I’ve waffled enough for one post.


Mark (_Raz_)

[Muntari]

Hi Mark,
I am currently involved with an ultrasonics sensor that uses the DsPIC although only on the hardware side. Excellent for controllers PWM for motors etc. ADCs are ok, not very fast but more than adequate for what they were designed for. At around $20 they are excellent value though.
The front end of my design uses 2 stages first a gain of 10 then a variable gain of between 1 and 100 both differential. The ADC I have two paths to choose from. One 16 channel within the DSP or DSC (TMS320F2812) @10msps 12 bits and another higher speed at 80 - 125msps 14 bits also a TI unit (ADS5500 series).
The DSC is a flash device with some great features and well suited to my application. They operate at around 150mips max.
All devices are 3volt logic. I have an option to clock the converter at different speeds to suit the dynamics of the system. Noise can be a problem but not as much as one might think. The fact that high resolution just gives more converted noise is true but as 12 bits is enough for my application, a 14 bit converter is used for the main sampling however the number of bits actually used in calculations is dependent on what I am doing. I use 8 bits, 12 bits or 13 bits. The external adc reference is also dynamically set depending on the average signal strength. The DSC internal ADC is used for ground and position data.
Basically as this is an experimental board, all options that could be made programmable, were. Analog Devices have a number of accelerometers that can be used for tilt sensing in tandem with Murata Gyros for position data.
As this is on a search coil swing by swing basis it is not all that demanding. The data from the position signals also modifies the gain of the front end. When the signal from a target is at it's peak, the gain is at it's lowest. Don't need anything more than a peak detector here (analog or digital). I use digital. However, when the peak is detected, the position data is logged and continues to track it's relative path as the coil is moved over and away from the peak signal area. The gain is increased until (a) the coil's vertical position moves outside a preset tolerance or (b) the original signal peak is reached. In this way I can determin a rough signal shape and width of the object. It does not mean simple translation to depth of the object but it could with little more work. If you can imagine an oscilliscope displaying video horizontal line data, then that's basically the lcd graphic output of this experiment albiet somewhat slower. The display output is not updated on every swing , rather as required by the user. It is meant to aid in target shape and boundary measurements but with higher speed circuits, could be realtime.
Hope I have explained things well enough. There are a lot of details left out but I have been thinking lately of pushing the design onto it's own web page where I can just upload documents as I do them....hmmm time...and I have so much to do..oh what fun!

Cheers for now and good luck with your design.

Regards

Brian K