The PI projects presented here are very simple in concept, primarily to prevent the basic principles of PI from getting lost in extraneous functions. There are a num-ber of ways to enhance the basic PI design for better performance, so let’s look at some of those techniques.
Power Supply
We’ve already mentioned the possibility of using a charge pump voltage inverter instead of a second battery or a rail splitter. Most commercial designs now use this approach which, of course, only requires a single battery (or battery pack). A charge pump voltage inverter is not difficult to design, but there are chips available specifi-cally for this task. One is the ICL7660; for a positive input, the output is negative at about 80-90% of the input under moderate load. For example, a +12V input might yield a -10V output; these levels could then be regulated down to 5 volts for opamp power.
One disadvantage of a charge pump is that they are not 100% efficient, some energy is lost in the process. Another potential problem, and one that can be serious in pulse induction, is that the switching action in a charge pump generates a lot of noise, and that noise can feed through the power rails and seriously degrade performance.
For the same reason as mentioned for the audio, PI detector charge pumps are often synchronized to the overall pulse rate. Figure 12-17 shows a simple discrete charge pump that can run at most PI pulse rates.
Preamp
PI-1 through PI-4 use a preamp with a gain of 100; PI-5 has a gain of 1000. Many commercial PI designs have a preamp gain of 1000 as well. A high gain requires an opamp with a high gain-bandwidth product, and even then the slower transient response limits how early the target sample can be taken, usually around 15s. Earlier target sampling almost always yields better sensitivity, but so does a higher preamp
Fig. 12-17: Charge Pump Circuitry
VCC
2 x VCC 10k
10k
100u
100u 1u
1u CLK
gain. A good solution is to split the preamp into 2 stages, each with a lower gain. This provides an overall high gain while maintaining a fast transient response.
Sampling
In the sampling integrator PI designs (PI-1 through PI-3) a JFET was used as the sampling switch. While this makes for an extremely simple switch, it is not the best solution because it has a rather high “on” resistance with a lot of variation, so the gain of the second stage is not very predictable. Also, the JFET is a depletion device, so it has a very limited analog voltage range. A better analog switch can be found in the CD406611 logic chip. Actually, this chip contains 4 switches. Each switch is a CMOS transmission gate12 made up of both a PMOS and NMOS device in parallel, so it works very well over a wide range of analog voltages. It also has a lower and more consistent “on” resistance of about 50.
Also in the sampling designs, a single sample was taken early in the decay curve.
This early sample point corresponds to expected variations in the decay due to metal targets, where effects on the decay curve tend to occur in the first 5-50s after the pulse. After that, the effects from metal targets tend to die out so they have little effect at longer sample delays. Sometimes there are anomalies which persist for much lon-ger times, and affect the decay not only in the region of the sampling but also farther out. One such effect is the induced signal from the Earth’s magnetic field as you swing the coil. It appears as an overall offset in the decay and can result in a response that comes and goes as you swing the coil (a “breathing” effect).
A common method of eliminating these unwanted responses is to take a later sample of the decay voltage, and subtract its value from the normal earlier sampled voltage. For long response times (such as Earth field effect) the two samples will largely cancel and there will be little to no differential voltage passed through the 11. Or the CD4016.
12.Also called a bilateral switch.
Fig. 12-18: Differential Sampling Integrator
R C
N.O.
N.O.
R
Vin Clk1 Vo
Clk2 C
integrator. Metal targets with short decay time constants will not be affected. Figure 12-18 shows a differential sampling integrator.
Ground Balance
This concept can be taken a bit further to provide a method of ground balance.
Most ground responses to pulsed signals are due to magnetic viscosity effects and have a fairly predictable 1/t response. By taking two samples fairly close to each other and subtracting a properly amplified second sample from the first sample (Figure 12-19), the ground response can be nullified while still passing signal responses. Because the samples are fairly close together (generally 10-20s apart) the subtraction tech-nique will reduce target sensitivity somewhat. However, some soils are nearly impos-sible to hunt without the ground balance feature, so a little target loss is better than no detecting.
A critical requirement of the method in Figure 12-19 is that the integrator time constants match well, usually to about 1%. Another approach avoids this by using only one integrator. If the S2 switch is fed with an inverted version of the preamp out-put, then it can directly feed the same integrator as S1. The gain of this signal (and thereby the ground balance) can be controlled either by varying the gain of the pre-amp inversion stage, or by varying the pulse width of S2. The latter has the advantage of integrating more noise and usually improves noise performance.
Coils
So far, all of our PI circuits have used a simple mono coil, which serves for both transmit and receive. In terms of do-it-yourself construction, this is much easier than trying to build carefully balanced coils for induction balance. However, the mono coil approach has a disadvantage: the optimal design of a transmit coil might not be opti-mal for the receive coil. It turns out that PI can also use separate transmit and receive coils and, unlike VLF detectors, they do not have to be inductively balanced.
All of the designs in this chapter have used a basic MOSFET coil switching cir-cuitry, and very few changes need to be made to accommodate separate coils. Figure 12-20 shows that just splitting the circuit at the coils and adding a second damping resistor is sufficient. We can continue to use a fast (low inductance, low capacitance) transmit coil, but the receive coil need not be as fast, especially if it is inductively
bal-Fig. 12-19: Ground Balance
anced. This gives us the freedom to increase the windings which improves sensitiv-ity13. Another advantage is the RX coil can be referenced to a different point than the TX coil, making the preamp’s power supply design easier.
VLF coils are almost always wound from enameled magnet wire. PI coils can also be wound using magnet wire but the thin insulation results in a large self-capaci-tance which slows the flyback response and limits how early the decay can be sam-pled. A faster sample delay is often desirable (especially for hunting small gold nuggets) so many PI designs focus on getting a fast flyback response. One way is to decrease the coil self-capacitance by using wire that has a thicker insulation or insula-tion with a higher dielectric material. Teflon-insulated wire is often chosen for this.
Furthermore, stranded wire has shown to have a slight improvement in self-eddy gen-eration, with Litz wire being the best.
Really Advanced Topics
We have not discussed discrimination because traditional PI designs have not had that ability, at least not like VLF. The subtractive ground balance method can be expanded to provide some level of iron discrimination. A number of experimenters have proposed sampling the decay curve at multiple points and identifying targets by their characteristic decay. This has so far not been realized, probably because the decay signal is primarily resistive and is missing the reactive component.
Another approach is to look at the target response during the transmit pulse, which requires the use of an induction balanced coil. During the transmit pulse both reactive and resistive signal components are available so target ID is easier, much akin to a VLF. However, di/dt is much lower so depth will suffer. A compromise is to combine decay sampling for raw depth and ground balance with transmit sampling for identifying shallower targets.
Another technique is to transmit two or more different pulse widths and process them through different channels. The channels can be compared for both ground and 13.More turns on the RX coil increases the signal strength of targets, but also increases the
sig-nal strength of the ground sigsig-nal as well as extersig-nal noise sources (interferers). Another adjunct to non-existent complementary banquets.
Fig. 12-20: Separate TX & RX Coils
RD2 CLK1
TX +VB
RX
-VB RD1
target information. The idea is that ground viscosity effects are generally independent of transmit pulse width, whereas target responses are not.
Finally, pulse induction is only one type of time-domain detection. There are many other transmit waveforms that can be used for time-domain, including square wave, triangle, sawtooth, half-sine, and multiple slope. Furthermore, any style wave-form can be generated as a complementary bipolar signal which inherently cancels Earth field signals (and some noise) and can allow the use of higher pulse rates.
Even though pulse induction has been around for 60 years or so, it has been rele-gated to specialty detectors and has not been explored as deeply as induction balance.
There is tremendous room for technological advances, especially in the realm of tar-get identification and discrimination. PI has the ability to detect deeper than IB, so with better ground balance techniques and the addition of discrimination it could eas-ily displace IB as the mainstream technology.