Offbeat antenna design questions - RF Cafe Forums

 Post subject: Offbeat antenna design questions

Posted: Thu Jan 22, 2009 11:45 am 

Most of the design software and classes etc are geared towards the microwave communication/RADAR industry. I've got a unique situation where I work in the 2-5KHz range. Yes Kilo, it's not a fat finger on the keyboard. I work for an industrial metal detecor manufacturing company that uses Pulsed Eddy current for detection of metal in a product. I am trying to move our Analog systems completely over to Digital and already have a pretty good Digital starting point. That being said, our digital system isn't very sensitive because it was designed for our workhorse system. We have analog systems that are able to detec a piece of tramp metal that is 1.5% of the size of the apperture between the transmit and receive antennas. I would like to get some ideas/thoughts/ramblings on how I can improve the sensitivity of the receive antenna, to be able to work at slightly higher frequencies(i.e. 10KHz, and 40KHz). The truly unique part is we wind our antennas by hand. The transmit coil follows the basic rules for the inductance of a coil and can be modled fairly easily. The receive coil isn't as easy. Here is a link to a question i posted to eng-tips.com ( http://www.eng-tips.com/viewthread.cfm? ... 049&page=1 ). It has a graphical representation of the receive antenna towards the bottom of the page. There are more details to the receive antenna available, such as buck winding for higher turn ratio and use with higher frequencies.

I look forward to any thoughts.

 Post subject:

Posted: Fri Jan 23, 2009 12:37 pm 

Hi mgburr
Golly that question takes me back a way, to when I was running a tech department serving our surveyors who were surveying city streets in the UK to map buried pipes and cables.
We developed a combined tone and metal detector based on a 90 degree crossed transmit and receive coil pair, at a frequency of 2500 hertz.
initially we used a linear bipolar cascode amplifier front end followed by a diode detector. This was fairly sensitive in the absence of mains electrical noise, but quickly saturated when confronted with high levels of same. We used a Q mulitplier single resonator filter with current fed back into a parallel tuned L with a tapped winding.
Eventually we developed a FET bomb-proof front end amp and a cross-coupled quad of igfets as the synchronous detector.
With this we were able to locate steel pipe 24 inch diameter at 20 times coil separation.
How is that for a start? Maybe I have some circuits back home. I will see what I can find, but we are talking 35 years ago here so the circuit technology is from the caves era.

_________________
At bottom, life is all about
Sucking in and blowing out.

 Post subject:

Posted: Fri Jan 23, 2009 1:13 pm 

nubbage, thanks for the thoughts and ideas. Our antennas are anywhere from a coil length of 18 inches up to a custom 144 inches. We don't have the antennas crossed, but inline. The ability to measure the desired material comes from looking for the eddy curent decay after the pulse. If you had a chance to see the receive antenna in the link to the eng-tips website you might get an idea of some of the problem. It is essentially wrapped as a figure 8, but then is bent in the middle and placed side by side. Although I wonder if it might work a little better for the recieve to use a flat wind instead of a bundle wind? I've got an instrumentation amplifier as the differential amp to get a fairly large amount of sensitivity, and using a programable gain amplifier to bring the signal down to a manageable level. Most of what this was designed from(i.e. the digital crossover from analog) was designed 40 years ago, so it might be usefull and an idea on how I can proceed. Hoping to get as much sensitivity and signal strength as possible without having to "Fix" the previously designed hardware.

 Post subject:

Posted: Mon Jan 26, 2009 10:20 am 

Hi mgburr
I have now had time to look at the link. I still think the best antenna configuration is 2 cylindrical coils wound on ferrite and placed so they are at right angles, say transmitting across the conveyor, and receiving through the conveyor (hopeully made of non-metal such as rubber). With only the transmit horizontal field vector present the pulse will generate a horizontal B vector. Using a vertical receive coil on ferrite there will be no energy coupled into it, since it responds only to a vertical B field.
In the presence of even a small piece of metal, the eddy current at 2kHz running around the conductor generates a complex randomly oriented B field vector. When resolved into its horizontal and vertical components, there will be some energy in the vertical vector direction, that will be picked up as a pulse in the receive coil.
On the topic of filtering covered on the link message board, we transformed the impedance of the receive coil to 1000 ohms using a small signal transformer, and got rid of most of the L component of impedance by a series capacitor. The same way was used to shape the drive current waveform into the transmit inductor/antenna. At a 1000 ohms nominal resistive load impedance, classical filter design can be used. However in the early models we just used a Q-multiplier circuit by feeding the signal into a transistor base of a common emitter current amplifier, and connecting part of the emitter current back through the near-ground few turns of the filter inductor. We easily obtained a loaded Q of 200 this way.
Using a pulsed source should enable you to use a synchronous detector clocked by the transmitter pulse, thereby getting rid of mains supply harmonic noise.

_________________
At bottom, life is all about
Sucking in and blowing out.

 Post subject:

Posted: Mon Jan 26, 2009 1:50 pm 

Hi nubbage,
The antennas were designed quite few years ago and we don't want to drastically change the concept. We want to be able to replace an antenna and have it operatie identically to what was replaced. That being said, the hand winding does tend to negate that because of inconsistencies caused by human error. As far as the operation over the conveyor. We do have some conveyors in which we have to "ignore" steel cords that are used for ripstop. so it can be an issue where there is metal in the conveyor that isn't tramp. If you were to look at the transmit and receive antennas ligned up you would almost see the 90 degree of seperation. The transmit is a single coil approximately 5&3/4 inch across and is centered over the receive section. The gap between the coils on the receive section is usually about 1/4 inch, and they are wound with 5&1/2 inch gaps in the middle. So we have a way to see the polarity as it passes over each receive section, i.e. creates a positive transition and negative transition as the material travles across the receive field of view. both antennas are setup on the horizontal plane and but the eddy current decay happens in all 3 classical directions. So this is what we are picking up. The design of being under the conveyer limits our ability to build up, yet we can build out forward and backwards. This can sometimes improve reception, however it also becomes sensitive to hands running within the field.

 Post subject: Re: Offbeat antenna design questions

Posted: Thu Jan 29, 2009 8:51 am 

Hi nubbage,

I did start looking at Grover's equation for polygon antennas, but I'm not sure how the crossover in the center between the 2 lobes will affect the formula. It works out fairly accurate for antennas with a 42 inch or longer side. When I go below that however, it tends to run the inductance a little low. I know the turn ratio for the transmit antennas is correct in that I have wound them and tried them. Could the shorter antenna coil length, with the low frequency be affecting the inductive reactance?

 Post subject: Re: Offbeat antenna design questions

Posted: Fri Jan 30, 2009 9:46 am 

Hi mgburr
You've reminded me there is a third antenna config, which places the wires of a flat plane transmit coil over the mid-point of the received coil circle. This also causes a null, but I have not tried this method myself. It has the merit it is virtually the same coil geometry as what you are using. I think most commercial amateur metal detectors for archeology:treasure hunting use this config.
The coil impedance at lower frequencies may well be influenced by mutual capacitance due to having to wind a large number of turns on the coil. That was why we used ferrite: to reduce the number of turns needed at 2kHz for the inductance required. The stray inter-winding capacitance was therefore lower.
If you do not wish to change the coil geometry, then (assuming you are achieving a good null in the absence of metal on the conveyor, and a good received signal in the presence of metal), the rest comes down to filtering of noise (particularly supply noise) which we achieved by a simple Q multiplier. The next is the detection process so you just detect the size of metal you are interested in, and it is repeatable reliably. As I said before, because you are using a pulsed transmit signal, the best approach to the detector would be a synchronous type. There are many designs available for these.

_________________
At bottom, life is all about
Sucking in and blowing out.