EDDY CURRENT INSPECTION INSTRUMENT WITH NOISE SHAPING FILTER

Abstract

Quantization noise in an oversampled eddy current digital drive circuit is reduced using a noise shaping filter.

Claims

1. A noise reduction circuit for non-destructive testing (NDT) of a test object with a NDT probe, the circuit comprising: a direct digital synthesizer (DDS) producing a digital waveform having (N+M) bits, wherein the digital waveform has a periodic excitation frequency; a noise shaping filter having the digital waveform as its input, and wherein the noise shaping filter produces a digital shaped output having N bits; a digital-to-analog converter (DAC) having the digital shaped output as its input, the DAC configured to sample the digital shaped output at a sampling frequency and to produce an analog shaped output having a quantization noise power distributed according to a noise frequency distribution; an analog filter having the analog shaped output as its input, the analog filter configured to remove from the analog shaped output all signals having a frequency greater than a filter bandwidth, thereby producing an analog filtered output; and, at least one amplifier for amplifying the analog filtered output to produce an amplified analog filtered output; and, wherein the noise shaping filter is configured to shape the noise frequency distribution so that the noise power corresponding to the excitation frequency is smaller than the noise power corresponding to frequencies greater than the excitation frequency.

2. The circuit of claim 1 wherein the NDT probe is an eddy current probe.

3. The circuit of claim 1 wherein the amplified analog filtered output is a drive signal for the NDT probe

4. The circuit of claim 1 wherein the at least one amplifier is a pre-amplifier and a power amplifier.

5. The circuit of claim 1 wherein the digital waveform is a sine waveform.

6. The circuit of claim 1 wherein N=12, M=8, the DAC is a 12-bit converter and the DDS is a 20-bit synthesizer.

7. The circuit of claim 1 wherein the sampling frequency is greater than twice the excitation frequency.

8. The circuit of claim 1 wherein the sampling frequency is greater than one hundred times the excitation frequency.

9. A method of reducing noise in a circuit for non-destructive testing (NDT) of a test object with a NDT probe, the method comprising the steps of: producing a digital waveform by means of a direct digital synthesizer (DDS) having (N+M) bits, wherein the digital waveform has a periodic excitation frequency; shaping a noise frequency distribution of the digital waveform with a noise shaping filter, wherein the noise shaping filter produces a digital shaped output having N bits, converting the digital shaped output to an analog shaped output by means of a digital-to-analog converter (DAC) sampling the digital shaped output at a sampling frequency and wherein the analog shaped output has a quantization noise power distributed according to a noise frequency distribution; filtering the analog shaped output to produce an analog filtered output, wherein the filtering comprises removing from the analog shaped output all signals having a frequency greater than a filter bandwidth; and, amplifying the analog filtered output with at least one amplifier to produce an amplified analog filtered output; and, wherein the step of shaping comprises shaping the noise frequency distribution so that the noise power corresponding to the excitation frequency is smaller than the noise power corresponding to frequencies greater than the excitation frequency.

10. The method of claim 9 wherein the NDT probe is an eddy current probe.

11. The method of claim 9 wherein the amplified analog filtered output is a drive signal for the NDT probe

12. The method of claim 9 wherein the at least one amplifier is a pre-amplifier and a power amplifier.

13. The method of claim 9 wherein the digital output waveform is a sine waveform.

14. The method of claim 9 wherein N=12, M=8, the DAC is a 12-bit converter and the DDS is a 20-bit synthesizer.

15. The method of claim 9 wherein the sampling frequency is greater than twice the excitation frequency.

16. The method of claim 9 wherein the sampling frequency is greater than one hundred times the excitation frequency.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0012] FIG. 1 is a schematic of an EC probe drive circuit according to the present invention.

[0013] FIG. 2A is a graph of spectral density vs frequency showing quantization noise of a digital-to-analog converter.

[0014] FIG. 2B is a graph of spectral density vs frequency showing filtered quantization noise of a digital-to-analog converter.

[0015] FIG. 2C is a graph of spectral density vs frequency showing the effect on quantization noise of noise shaping according to the present disclosure.

[0016] FIG. 2D is a graph of spectral density vs frequency showing the effect on quantization noise of noise shaping and filtering according to the present disclosure.

[0017] FIG. 3 is a schematic illustration of a noise reduction method according to the present disclosure.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENT

[0018] FIG. 1 shows a schematic of a drive circuit according to the present disclosure for driving an EC probe 26. EC probe 26 may be any type of EC probe including a single-coil probe, a differential probe, a remote-field probe or a probe array. The drive circuit comprises a digital section 12 and an analog section 14, separated by a digital-to-analog converter (DAC) 20. In an embodiment, DAC 20 is a 12 bit converter. Digital section 12 comprises a direct digital synthesizer (DDS) 16 and a noise shaping filter 18. Analog section 14, comprising an analog filter 21, a pre-amplifier 22 and a power amplifier 24, is the generator circuit that provides an excitation signal S-24 having an excitation frequency to drive EC probe 26.

[0019] DDS 16 is a digital circuit that generates the EC drive signal, which is typically a sine wave, but may be any combination of multiple sine waves, a chirp, or any other desired form of excitation signal. In order for noise shaping filter 18 to operate as designed, DDS 16 should have a higher number of bits than DAC 20. In an embodiment, DDS 16 is a 20 bit synthesizer.

[0020] Noise shaping filter 18 is a key part of the present invention. Noise shaping filter 18 is implemented in digital electronics such as an FPGA, an ASIC, a microprocessor, or a microcontroller, and its function is to take an (N+M) bit input from DDS 16 and convert it to an N bit output, where N is the number of bits of DAC 20 and M is the number of additional bits of DDS 16.

[0021] Noise-shaping is a technique generally used to increase the SNR in a specific frequency band of a signal. In the present invention, noise shaping is used to increase the SNR of excitation signal S-24. The SNR of excitation signal S-24 has an important effect on the productivity of EC NDT/NDI because with lower SNR it is necessary to average the EC measurement over more excitation cycles in order to filter out the noise. This reduces productivity by forcing a reduction in the probe scan speed to compensate for the added averaging cycles. In the worst case, the noise in excitation signal S-24 could altogether prevent a meaningful EC measurement. It is thus important to reduce the noise, thereby increasing the SNR in excitation signal S-24 as much as possible.

[0022] An important part of the noise in excitation signal S-24 is the quantization noise in DAC 20. For an ideal DAC generating a sine wave, the SNR is given by (see for example W. R. Bennett, “Spectra of Quantized Signals,” Bell System Technical Journal, Vol. 27, July 1948, p 446-471):

SNR (dB)=6.02*N+1.76 dB   (1)

where N is the number of bits. For an ideal DAC, the only noise is the quantization noise, hence the SNR is the ratio between the amplitudes of the sine wave and the quantization noise. For DAC 20 in an embodiment having 12 bits, the SNR according to equation (8) is 74 dB.

[0023] DAC 20 samples the excitation signal at a sampling frequency f.sub.s which must respect the following equation (see, for example, Stremler, F. G. (1992), Introduction to Communication Systems, Addison Wesley Longman):

f.sub.s>2×excitation frequency   (2)

According to equation (2), the sampling frequency, f.sub.s, of DAC 20 should be at least twice the excitation frequency of the drive signal of probe 26. If f.sub.s is much higher than the excitation frequency, then the signal is said to be oversampled. For example, to generate a 100 kHz excitation signal the minimum sampling frequency is 200 kHz. Using a 100-MHz sampling frequency would be oversampling by a factor of 500.

[0024] FIG. 2A is a graph of spectral density vs frequency showing a quantization noise 40 of an ideal DAC, which is an approximation of real DAC 20. Quantization noise 40 has a total noise power P.sub.q and a constant spectral density N.sub.q spread out over all frequencies from 0 Hz up to a cut-off frequency of f.sub.s/2, where

[00001]$\begin{array}{cc}{P}_q=\frac{q}{\sqrt{12}}& \left(3\right)\\ N_q=\frac{P_q}{\sqrt{f_s/2}}& \left(4\right)\end{array}$

where q is the least significant bit of DAC 20. Note that with oversampling, f.sub.s has a higher value, thereby spreading the noise over a larger frequency range. FIG. 2A also shows the filtering bandwidth, BW, of analog filter 21 and the corresponding bandwidth cutoff frequency 44. Also shown is an excitation signal 42, which is shown as a line corresponding to the excitation frequency of the generated sine wave.

[0025] FIG. 2B shows a filtered quantization noise 46, which is the result of filtering quantization noise 40 with analog filter 21, wherein analog filter 21 removes frequencies higher than bandwidth cutoff frequency 44. The SNR of excitation signal 42 in FIG. 2B is higher than the SNR of excitation signal 42 in FIG. 2A by an amount:

10 log.sub.10 (f.sub.s/2.BW).   (5)

This SNR improvement is known as the process gain. For example, when generating a 100-kHz excitation sine wave with a sampling frequency of 100 MHz with an ideal 12-bit DAC, the SNR is 74 dB (equation (1)). By filtering frequencies higher than 1 MHz, the process gain given by equation (5) is 10 log.sub.10 (100/2)=17 dB. Therefore the SNR of excitation signal 42 in FIG. 2B is improved to 91 dB.

[0026] FIG. 2C illustrates the effect of noise shaping on quantization noise 40. Noise shaping filter 18 is used to shape the quantization noise of DAC 20, displacing the noise to higher frequencies and minimizing the noise at lower frequencies. Implementation of noise shaping such as in noise shaping filter 18 is a technique known in the art. In general, to implement noise shaping filter 18, the (N+M) bit signal at its input is compared with the N bit signal at its output, and an error signal is fed back to the input via a digital filter. The higher the order of the filter, the more the quantization noise is displaced to higher frequencies and the more the quantization noise is minimized in the frequencies of interest.

[0027] FIG. 2C shows a shaped quantization noise 48, prior to filtering by analog filter 21. The area under the graph of shaped quantization noise 48 is the same as the area under the graph of quantization noise 40, meaning that noise shaping filter 18 has not changed the total noise energy but has only displaced some of that energy to higher frequency.

[0028] FIG. 2D shows the result of filtering the noise of FIG. 2C with analog filter 21 after DAC 20, wherein analog filter 21 removes frequencies higher than bandwidth cutoff frequency 44. Shaped quantization noise 48 is filtered to a shaped filtered quantization noise 50. It is seen in FIG. 2D that the area under the graph of shaped filtered quantization noise 50 is much less that the area under the graph of filtered quantization noise 46, meaning that the noise energy has been significantly reduced by application of noise shaping filter 18. The reduction of noise energy is associated with an increase of SNR by approximately 10 to 20 dB, depending on the order of noise shaping filter 18. Therefore, in the example cited above of a 100-kHz excitation sine wave, a sampling frequency of 100 MHz, an ideal 12-bit DAC and a 1 MHz filter cut-off, application of noise shaping filter 18 will increase the SNR from 91 dB to at least 101 dB. This increase of SNR due to noise shaping is a key aspect of the present invention and represents a significant improvement in NDT/NDI using EC or other types of NDT/NDI probes.

[0029] FIG. 3 is a schematic illustration of a noise reduction method according to the present disclosure. In step 30 a digital waveform is produced, having (N+M) bits and a periodic excitation frequency. In step 32 the digital waveform is shaped with a noise shaping filter having N bits. In step 34 the shaped digital signal is converted to an analog signal by sampling with a DAC at sampling frequency L. In step 36 an analog filter is used to filter out signals having frequency greater than the filter bandwidth. In step 38 the filtered analog signal is amplified and in step 39 the amplified signal is used to drive the NDT probe.

[0030] Although the present invention has been described in relation to particular embodiments thereof, it can be appreciated that various designs can be conceived based on the teachings of the present disclosure, and all are within the scope of the present disclosure.