Introduction ' This invention relates generally to systems for providing digital representations of analog values and, more particularly, to such systems as used with analog sensor elements for producing digital output values representative thereof with relatively high resolution and accuracy.

Background of the Invention

A system for providing digital representations of the analog values from an analog sensor element is described in United States patent application, serial number 147,220, filed May 6, 1980 by Neil Brown. The system described therein utilizes a successive approximation conversion technique using a transformer D/A converter and a dual slope integration technique, which techniques provide more useful digital representations of the analog values, particularly when such values may be changing relatively rapidly during the digital sampling interval, and also provide a more rapid operation than systems theretofore known. In accordance with the system described in such application, the most significant bits of the digital output are determined using a successive approximation register and a precision transformer D/A converter, the latter providing an AC output voltage having an amplitude which is proportional to the digital value currently provided by the successive approximation register.

The transformer D/A output is compared with the AC analog input signal from the sensor to provide a comparator output voltage representing in a suitably designed comparator an error voltage which is supplied to the successive approximation register to determine the value of each bit. In this manner, both the "sign" bit and the nine most significant bits, of an 18-bit digital output signal, for example, can be determined.

The residual error voltage which is present following the determination of the most significant bits is then demodulated and integrated for a fixed period of time, the resulting value of such integration process being digitized to provide the additional nine least significant bits in accordance with a dual-slope integration processs, as described in the aforesaid application.

When using such system to provide a digital output signal, particularly one having more than 18 bits, the resolution and accuracy tends to deteriorate because the comparator circuit thereof tends to respond not only to the fundamental frequency of the basic AC oscillator signal but also to the harmonics thereof when such signal is an imperfect sine wave. Hence, it is desirable to modify the circuitry as disclosed in the above referenced application so that the comparator circuit responds only to the fundamental frequency of the basic oscillator signal so that the desired resolution and accuracy can be achieved, with digital outputs particularly of greater than 18 bits, e.g., as high as 26 bits.

Brief Summary of the Invention

In the circuitry utilized in the aforesaid application it has been found that the generation of the most significant bits seems significantly less affected by the problem of harmonic presence in the AC reference signal than does the generation of the least significant bits. Accordingly, in the invention described herein the comparator circuit is arranged to operate in two modes, i.e., a "most significant bit" (MSB) mode and a "least significant bit" (LSB) mode. In the LSB mode, the input signals which are compared are supplied via a preamplifier to a band-pass filter/amplifier circuit which has a very precise gain and a substantially zero phase shift. Such filter then has a high harmonic rejection capability. The output of the band-pass filter is supplied to a phase detector the output of which then passes through a first conventional, low pass filter and a second Bessel low pass filter

OMP

^S UB ^ST .

'to a dual-slope integrator circuit. The integrator output is thence supplied to a limiter to provide the comparator output for generating the least significant bits.

During the MSB mode, the output of the preamplifier bypasses the band-pass filter and is supplied to the phase detector and thence through the first low pass filter to a fixed-time integrator circuit and thence to the limiter for producing the comparator outuput used in generating the most significant bits. Such comparator circuit configuration can be contrasted with the previously suggested approach of the aforesaid application wherein the output of the preamplifier circuit during the LSB mode is supplied directly to the phase detector and thence to a dual-slope integrator and limiter, while during the MSB mode the ±AC reference signals are supplied directly to the limiter to produce the comparator output.

Description of the Invention

The invention can be described more specifically with the help of the accompanying drawings wherein Fig. 1 shows a broad block diagram of the system of the invention;

Fig. 2 shows a block diagram of the comparator circuit of the system depicted in Fig. 1;

Fig. 3 shows a diagrammatic representation of the band pass filter/amplifier of Fig. 2;

Fig. 4 shows a circuit diagram of one embodiment of the band-pass filter/amplifier of Fig. 3;

Fig. 5 shows the amplitude and phase characteristics of the band-pass filter/amplifier of Fig. 2; Fig. 6 shows a circuit diagram of an exemplary bi-quad amplifier used in the circuit of Fig. 4;

Fig. 7 shows a circuit diagram of the phase detector/filter of Fig. 2;

Fig. 8 depicts the waveforms of the reference voltages used in

SUBSTITU

^" the phase detector of Fig. 7;

Fig. 9 shows a circuit diagram which is functionally equivalent to that depicted in Fig. 7;

Fig. 10 shows waveforms (A) through (E) useful in explaining the operation of the circuit of Fig. 9;

Fig. 11 shows a more detailed circuit diagram of a particular embodiment of the preamplifier of Fig. 2;

Fig. 12 shows a circuit diagram of the low pass Bessel filter and dual-slope integrator of Fig. 2; Fig. 13 shows a circuit diagram of the finite time integrator of Fig. 2;

Fig. 14 shows a circuit diagram of the limiter of Fig. 2; and Fig. 15 is a diagrammatic representation of the timing sequence of operation of the system of the invention.

As can be seen in the block diagram of Fig. 1, which is a modified version of the general configuration discussed in the aforesaid Brown application, a sensor circuit 10 is supplied with an input reference signal from AC signal generator (oscillator) 11 and provides an output analog voltage Eg which represents a sensed parameter value, such as a pressure or a temperature, for example, depending on the sensor transducer which is being utilized. The AC signal from generator 11 is supplied to a digital-to-analog (D/A) converter circuit 12 which provides an analog signal EJJ which is an analog representation of the output of a successive approximation register (S.A.R.) 13. Eg and Erj are compared at AC comparator circuit 14, to provide an error signal utilizing suitable finite time integration and dual-slope integration techniques using integrators 16 and 17, respectively, for correcting the successive approximation register 13 via control logic 15. Such integration techniques are described in the aforesaid Brown application and are further discussed

SUBSTITUT

^" below.

During a first mode of operation (when switch 21 is in the position shown), the output of the finite time integrator 16 is supplied to an up/down counter 19 via limiter 18 to successive approximation register 13 via suitable control logic 15 as discussed in the aforesaid Brown application. The output of S.A.R. 13 is supplied to a suitable microprocessor 20 and effectively represents the most significant bits (MSB of a number which in turn represents the sensor output. During a second mode of operation (when switch 21 is in its alternate position), the output of the dual-slope integrator 17 is supplied to limiter 18 and up/down counter 19 and thence to the microprocessor 20. Such output efferctively represents the least significant bits (LSB) of the number which represents the sensor output. The microprocessor 20 operates on such MSB and LSB values so as to provide a suitable indication of the value of the sensed parameter which can be suitably used by and/or displayed for the user.

While general circuitry shown in Fig. 1 is substantially similar to that shown in the above-referenced application, the AC comparator and integration circuits and their associated circuitry, in accordance with the invention, have a different configuration from that shown in the prior application, which new configuration is depicted in more detail in Fig. 2. As can be seen therein, a comparator preamplifier circuit 22 is supplied with the input difference signal (Eg-Ep _j ) representing the difference between the signal from sensor circuit 10 and the signal from D/A converter circuit 12. The output of preamplifier 22 is supplied either to a band-pass filter/amplifier circuit 23 or directly to a phase detector/filter circuit 241, depending on the positions of switches 25 and 26, as discussed in more detail below. The output of phase detector/filter 24 is supplied to a low pass Bessel filter circuit 27 via switch 29 and to the input of a fixed, or finite, time integrator circuit 16.

SUBSTITUTE

The output of Bessel filter 27 is supplied to a dual-slope integrator circuit 17 via switch 28. The output of either dual-slope integrator 17 or fixed time integrator 16 is supplied to limiter 18 in accordance with the position of a switch 21, the limiter output thereby providing the desired comparator output signal for supply to u /down counter 19.

In Fig. 2 switches 25, 26, 29 and 21 are shown in their selected positions for operation in the LSB mode (the solid lines used with such switches) while in their alternate positions (the dashed lines) the switches are arranged for operation in the MSB mode. Switch 28 operates in a slightly different manner, as discussed below.

The design approach for the band-pass amplifier 21, as used in combination with the preamplifier circuit 20, is shown in Fig. 3 in diagrammatic form. As seen therein, the output of preamplifier 20 is supplied to a circuit which acts effectively as a cascaded pair of tuned feedback amplifiers 35 and 36 each effectively represented diagrammatically in the figure as comprising an AC current source 37 and 39, respectively, and parallel resonant inductance-capacitance (LC) circuits 38 and 40, respectively.

Such circuits are arranged to have, in effect, an ideal open loop gain of infinity at the selected operating frequency (in a particular embodiment as described herein the operating frequency is selected to be 384 Hz) and to be completely stable under closed loop conditions. Acceptable performance can be achieved with a feedback factor of 2.5x10?. The diagrammatic approach of Fig. 3 depicts two LC circuits each having high Q factors, e.g., up to 2000 at low audio frequencies.

One can achieve a very high gain at a selected frequency, e.g., 384 Hz, by the use of a single high-Q tuned circuit. The frequency at unity gain will be given in that case approximatley as 4.8M Hz. Since the tuned circuit has a phase shift approaching 90° at frequencies well above or well below resonance, the untuned stages would have to have a combined phase shift of less than 90°

O OMPPII

"for all frequencies up to 4.8M Hz if the amplifier is to be unconditionally stable under closed loop conditions, a situation which is clearly impractical to achieve. The difficulty increases when two tuned stages are used. If a small resistance Ri is added in series with the first tuned circuit, as shown in Fig. 3, the desired stability can be achieved. At resonance, R is negligible but at frequencies remote from resonance Ri predominates and the phase shift of the first stage approaches zero at low gain. If this stage is followed by an unmodified second stage, as in Fig. 3, it is found that the overall amplifier meets the required very high open loop gain at resonance while maintaining excellent closed loop stability.

At the low audio resonant frequency of 384 Hz the circuit of Fig. 3 becomes impractical since the physical size of the inductances required are too large for most applications. An alternative technique for achieving the same result is the circuit shown in Fig. 4. Preamplifier 20 supplies an input signal to a pair of cascaded amplifiers 41 and 42 each of which can be embodied as a conventional "bi-quad" amplifier. Such circuits are Well known to the art and can be found, for example, in the text "Rapid Practical Designs of Active Filters," Johnson and Hilburn, John Wiley and Sons, 1975. The resistors R and R2 (where Rl = R2) effectively produce the same effect as does the resistance R in Fig. 3. The gains of each of the bi-quad amplifiers can be arranged to be 1000 and the Q's of each to be 2000. The overall open loop gain, for a preamplifier gain of 37.5, is greater than 25x10^ at 384 Hz and falls below unity below 20 Hz and above 7500 Hz with phase shifts ±91° at these frequencies.

If input and output transformers are used, their phase shifts over the 20-7500 Hz range must be relatively small (±30° or less) to maintain good closed loop stability. Such requirements are relatively easy to satisfy and the use of such transformers

^• assists in providing optimum signal to noise ratios and impedance matching and provides the further advantage of permitting the use of floating input or output windings if desired in any application. A specific embodiment of a bi-quad amplifier which can be used for each of the amplifiers 41 and 42 in Fig. 4 is depicted in Fig. 6 as comprising three operational amplifier (op-amp) stages and having a closed loop feedback resistor R _a connected from the output of the final op-amp to the input of the initial op-amp. The overall circuit operates in effect as a parallel resonant circuit the amplitude characteristics of which rise from a zero level to a peak at the resonant frequency and then fall to the same zero level. The phase goes from +90° at low frequencies through 0° at the resonant frequency to -90° at high frequencies. Accordingly, when used in the band-pass amplifier configuration of Fig. 4, the overall operation thereof provides the desired closed loop amplitude and phase characteristics shown in Fig. 5. The amplitude vs. frequency characteristic aprovides a substantially flat response (a constant gain) in the relatively narrow pass band region about the center frequency, f _r , reduces rapidly to low levels near the resonance pass band, and approaches zero at extreme off-resonant frequencies. The phase shift is zero over the pass band and is effecitvely +90° or -90° elsewhere, as shown. The output from the band-pass filter/amplifier 23 is then supplied to phase detector/filter 24, the phase detector portion of which, as shown in Fig. 7, comprises synchronous detector circuits 45, 46, 47 and 48 which are supplied with synchronous signals M, M, N and N, respectively, (where the "barred" terms are the inverse waveforms of the non-barred terms). The waveforms of such signals are shown in Fig. 8. Over the time period T representing the period of the resonant frequence f _r the wave shapes are of the form shown wherein each is at maximum at one or the other polarity (i.e., ±1) during

^" the middle third of its half cycle, and is zero otherwise.

The detector circuit of Fig. 7 is functionally equivalent to the simplified circuitry shown in Fig. 9 and can be used to illustrate the operation thereof. In Fig. 9 an input A-C signal Eg (as from a generator 50) is supplied directly to one terminal Si of switch 51 and the inverted A-C signal (-Eg) is supplied to another terminal S2 of switch 51. The center terminal is grounded. If Eg has the waveform (A) shown (for one cycle) in Fig. 10 and the switch is operated as shown by waveform (B), the detected output EQ is of the form shown by waveform (C). The third harmonic of the input signal, E 3, is shown by waveform (D) and the third harmonic of the output signal Eo3 _> is shown by waveform (E). As can be seen, in the latter waveform the areas of the positive output third harmonic portions are essentially equal to those of the negative portions thereof so that the third harmonic output essentially averages out to zero and is effectively eliminated. Such operation can be contrasted with a conventional square wave detector in which the presence of third harmonics adversely affects detector operation in many applications. Such operation also substantially reduces the fifth harmonics by a factor of four.

Other specific circuits used in a particular embodiment of the invention are shown in Figs. 11-14. Fig. 11 depicts the preamplifier stage 22, the elements used therein having selected values as shown for a particular operating embodiment of the invention. A particular operating embodiment of the band-pass filter/amplifier 23 has already been shown in Figs. 4 and 6, while the phase detector/filter configuration 24 has been shown in Fig. 7. The low pass Bessel filter 27 is depicted in Fig. 12, the filtered output of Bessel filter 27 being then supplied to dual-slope integrator 17, also shown in Fig. 12.

The Bessell filter 27 is a well-known two-pole filter, the characteristic of which is that its output becomes essentially completely stabilized in response to a step signal input more

TITUTE SHEET f ^OMPI

^" rapidly than other filters having equivalent attenuation characteristics. The filter also provides effective attenuation of the beat frequency signals which result between the center frequency (e.g., 384 Hz as selected for use in the particular embodiment discussed) and the seventh harmonic of the line frequency (i.e., either 60 Hz or 50 Hz, for example) which line signal harmonic is closest to the center frequency. In some applications it may be desirable to achieve more effective operation by providing a conventional four-pole Bessel filter instead of the two-pole version depicted in Fig. 12.

The output from phase detector/filter 24 is also supplied to finite time integrator 16 which is depicted in Fig. 13. The outputs from iontegrators 16 and 17 are supplied through appropriate switching means 21 to limiter 18 as depicted in Fig. 14 to provide the output signal which is thereupon, supplied to up/down counter 19 as shown in the configuration of Fig. 1. Thus, during the most significant bit operating mode, the switches shown in Fig. 1 are in their dashed-line positions wherein the output from preamplifier 22 by-passes the band-pass filter circuit 23 and is supplied directly to phase detector/filter 24 and thence to finite time integrator 16 for supply to limiter 18. During the least significant bit operating mode the switches are in their alternate positions wherein the output of preamplifier 22 is supplied to band-pass filter/amplifier 23, to phase detector/filter 24 and, thence, to Bessel filter 27 and dual-slope integrator 17 before being supplied to limiter 18.

Fig. 15 shows in diagrammatic form the timing sequence of operation of the system of Fig. 1. Thus, for an overall computation time of 1.25 seconds, divided into fifteen 83 1/3 millisecond (msec.) segments, the most significant bit computation occurs in segment 0, the first 83 1/3 msec. Once the filters become stabilized in segment 11, the least significant bit measurement begins. In such operation the dual-slope integrator first integrates up, or down, depending on the polarity of the

error voltage of the input to preamplifier 22. Such first upslope (downslope) integration occurs over a time period from segment 2 through segment 9 at which point the integrator output reaches a steady state value. Segment 10 is used to stabilize the filters (the integrator maintains its steady state value) and the downslope (or upslope) integration occurs from segment 10 through segment 14. The direction of such second integration depends on the polari of the A-C reference signal supplied via switch 32 which is selected in accordance with the direction of the first integration.

During such operation, switches 23, 241, 29 and 21 are in their dashed line positions during the MSB operating mode (segment 0) and in their solid line positions during the stabilizing segment (segments 1 and 10) and during the LSB operating mode (segments 2-9 and 11-14). Switch 28 is in its dashed line position during the MSB operating mode and during the stabilizing time segments (segments 1 and 10) so as to assure that dual-slope integrator 17 begins its integration with a zero input during both integration time periods. Switch 28 is in its solid lines position during the LSB dual-slope integrations (segments 2-9 and 11-14).

Because the error signal and the selected A-C reference signal both utilize the same components during such dual-slope integration process, any changes in the characteristics of such components affect the two integrations in the same manner and no error arises in the integration process due to any such component changes.

Thus, the operation of the circuitry depicted and discussed with reference to Figs. 1-14 provides a more effective and improved operation over that which is described in the above-referenced U.S. patent application, Serial No. 147,220, and permits the generation of the least significant bits in a manner which is not adversely affected by the problem of harmonic presence, particularly the presence of third harmonics, in the A-C reference oscillator. In accordance with the circuitry described

SUBSTITUTE SHEET ' ° ^MPI

"above, the circuitry responds essentially only to the fundamental frequency of the basic oscillator circuit and the desired improved resolution and accuracy of the digital output signal can be better achieved. '

eUSSTVTUTE SHEET