A
Anonymous
Guest
In hand-held pulse induction metal detectors, demodulation is normally done with CMOS transmission gates, which are either on or off, no in-between. In order to facilitate the use of nonsaturating amplifiers, some receiver topologies also use a transmission gate between the receiver coil and the preamp. Prior to the use of CMOS, bipolar and/or JFET transistors were probably used as switches. I've used bipolar differential amplifier preamps as switches. Go back farther and you'd probably find electromechanical relays and motor-driven commutators.
All these methods do sampling with a step function. The rising and falling edges of a step function demodulate extraneous high frequency frequency signals outside the range of interest down into the baseband, resulting in electrical interference. Heretofore it has been one of the Achilles' heels of pulse induction.
A partial solution to this problem is for the demodulators to use soft turn-on and turn-off. Coil/preamp gates can make good use of soft turn-on but will not ordinarily require soft turn-off.
There are several methods of achieving this.
1. Use discrete JFETs or MOSFETS as switching elements, applying a ramp to the gate(s) to accomplish the desired soft-switching function. The most obvious method is to use an RC circuit, but an RLC circuit may provide sufficient improvement in the windowing function to make the extra complexity worthwhile. Ordinarily a DC bias adjustment will be necessary, but the use of matched pairs may make possible a fixed DC network needing no adjustment.
2. Use a controlled-transconductance amplifier, for instance a bipolar differential pair, as the preamp and/or as the demodulator(s). Ramp the control current to provide the desired windowing function.
3. Rather than using a "linear" (continuous-time) soft switching system, use sequenced multiple transmission gates to approximate the desired window function. This has the advantage of wide dynamic range, but the disadvantages of complexity and ineffectiveness above a certain frequency.
4. Electronically controlled potentiometers could, in principle, be used, but charge injection and slow speed probably make them an unsatisfactory solution.
5. Integrated circuit multipliers, most of which are based on a modified Gilbert Cell, can be used; however, an ordinary long-tailed differential pair would usually be superior from the standpoint of signal-to-noise ratio.
DIGITIZED SYSTEMS
Until recently, the A/D resolution and speed required to do demodulation in software has been incompatible with the cost and power consumption constraints of hand-held consumer metal metal detectors. The power consumption of microcomputers needed to crunch the numbers has also been an impediment. However, the communications industry is driving rapid changes in this arena, and I expect we'll be seeing within the next several years metal detectors which do demodulation in software.
This doesn't make the problems go away. High resolution sampling A/D's (the most likely candidates) are fast switched devices capable of demodulating (unwanted) signals several orders of magnitude higher in frequency than their speed rating. The problem of electrical interference can be reduced by taking many samples and processing the data in a (presumably FIR) filter, but this approach runs into the same kind of upper frequency limit that the multiple-gate system (#3 above) does. Introducing pseudo-random phase noise into the A/D sample timing will reduce potential problems with synchronous interfering signals, but won't help with broadband interference.
Of course, we can cheat and use some really good anti-aliasing filtering ahead of the A/D, and then guess what? We've just spread our lovely pulse out like jam on toast. Well, maybe we can store the flyback smear as a constant in software.... but the dynamic ranges have to be juggled, and now signals demodulated by the protection diodes are being tossed into the data....
I'm not saying the problems have no solution. Just pointing out that the problem with free lunches, is how darn expensive they are.
A FIRST PREFERRED EMBODIMENT OF A DIGITALLY DEMODULATED PI SYSTEM
Use continuous-time soft turn-on gating of the preamp. This can be done with transistors used as variable resistors, between the coil and the preamp; or, by modulating the preamp itself by varying its transconductance. Another possibility, but probably not a desirable one, is to control gain by modulating the impedance of shunt elements, as is done in some analog multipliers.
The output signal from the preamp can be anti-alias filtered and then digitized.
A SECOND PREFERRED EMBODIMENT OF A DIGITALLY DEMODULATED PI SYSTEM
Use a fast saturating amplifer, as is frequently done in this industry. Soft-gate its output and feed the resulting signal into an anti-aliasing filter. Then digitize the filtered signal.
A NOTE ON FAST SATURATING AMPLIFIERS
It's been the custom in this industry to saturate the amplifier itself. Another approach is to saturate the feedback network (for instance using diode clamps), maintaining the amplifier itself biased in its linear region. This allows a wider choice of amplifiers and may provide faster recovery.
--Dave J.
All these methods do sampling with a step function. The rising and falling edges of a step function demodulate extraneous high frequency frequency signals outside the range of interest down into the baseband, resulting in electrical interference. Heretofore it has been one of the Achilles' heels of pulse induction.
A partial solution to this problem is for the demodulators to use soft turn-on and turn-off. Coil/preamp gates can make good use of soft turn-on but will not ordinarily require soft turn-off.
There are several methods of achieving this.
1. Use discrete JFETs or MOSFETS as switching elements, applying a ramp to the gate(s) to accomplish the desired soft-switching function. The most obvious method is to use an RC circuit, but an RLC circuit may provide sufficient improvement in the windowing function to make the extra complexity worthwhile. Ordinarily a DC bias adjustment will be necessary, but the use of matched pairs may make possible a fixed DC network needing no adjustment.
2. Use a controlled-transconductance amplifier, for instance a bipolar differential pair, as the preamp and/or as the demodulator(s). Ramp the control current to provide the desired windowing function.
3. Rather than using a "linear" (continuous-time) soft switching system, use sequenced multiple transmission gates to approximate the desired window function. This has the advantage of wide dynamic range, but the disadvantages of complexity and ineffectiveness above a certain frequency.
4. Electronically controlled potentiometers could, in principle, be used, but charge injection and slow speed probably make them an unsatisfactory solution.
5. Integrated circuit multipliers, most of which are based on a modified Gilbert Cell, can be used; however, an ordinary long-tailed differential pair would usually be superior from the standpoint of signal-to-noise ratio.
DIGITIZED SYSTEMS
Until recently, the A/D resolution and speed required to do demodulation in software has been incompatible with the cost and power consumption constraints of hand-held consumer metal metal detectors. The power consumption of microcomputers needed to crunch the numbers has also been an impediment. However, the communications industry is driving rapid changes in this arena, and I expect we'll be seeing within the next several years metal detectors which do demodulation in software.
This doesn't make the problems go away. High resolution sampling A/D's (the most likely candidates) are fast switched devices capable of demodulating (unwanted) signals several orders of magnitude higher in frequency than their speed rating. The problem of electrical interference can be reduced by taking many samples and processing the data in a (presumably FIR) filter, but this approach runs into the same kind of upper frequency limit that the multiple-gate system (#3 above) does. Introducing pseudo-random phase noise into the A/D sample timing will reduce potential problems with synchronous interfering signals, but won't help with broadband interference.
Of course, we can cheat and use some really good anti-aliasing filtering ahead of the A/D, and then guess what? We've just spread our lovely pulse out like jam on toast. Well, maybe we can store the flyback smear as a constant in software.... but the dynamic ranges have to be juggled, and now signals demodulated by the protection diodes are being tossed into the data....
I'm not saying the problems have no solution. Just pointing out that the problem with free lunches, is how darn expensive they are.
A FIRST PREFERRED EMBODIMENT OF A DIGITALLY DEMODULATED PI SYSTEM
Use continuous-time soft turn-on gating of the preamp. This can be done with transistors used as variable resistors, between the coil and the preamp; or, by modulating the preamp itself by varying its transconductance. Another possibility, but probably not a desirable one, is to control gain by modulating the impedance of shunt elements, as is done in some analog multipliers.
The output signal from the preamp can be anti-alias filtered and then digitized.
A SECOND PREFERRED EMBODIMENT OF A DIGITALLY DEMODULATED PI SYSTEM
Use a fast saturating amplifer, as is frequently done in this industry. Soft-gate its output and feed the resulting signal into an anti-aliasing filter. Then digitize the filtered signal.
A NOTE ON FAST SATURATING AMPLIFIERS
It's been the custom in this industry to saturate the amplifier itself. Another approach is to saturate the feedback network (for instance using diode clamps), maintaining the amplifier itself biased in its linear region. This allows a wider choice of amplifiers and may provide faster recovery.
--Dave J.