TABLE OF CONTENTS 

In particular an optimization of:
Digital ultrasonic instruments have been in use for some time as components of NPP inspection. So far, guidelines and standards have not contained characterizations of signal recognition or evaluation, as long as the Digitization has been applied for gate (monitor) techniques only. That means there is no documentation concerning the Digitization of ultrasonic echo signals in DIN 54119 (08/1981) "Ultrasonic Testing  Terms (Ultraschallprüfung  Begriffe)" and DIN 25450 (09/1990) "Ultrasonic Test Systems for Manual Testing (Ultraschallprüfsysteme für die manuelle Prüfung)".
Only in DIN 254351 is it mentioned that the resolution must be equal to or greater than 1 dB if analog to digital conversion is applied (DIN 254351 [11/1987] "Wiederkehrende Prüfungen der Komponenten des Primärkreises von Leichtwasserreaktoren  Mechanisierte Ultraschallprüfung"). For our interpretation a limit of the quantifying error is determined and in particular the applicable echo dynamic range is limited (e.g., for a 8 Bit A/D converter approx. 20 dB).
In ultrasonic testing procedures for defect description certain methods are used which cannot be detailed within the framework of a guideline stipulation. Nevertheless, standards of inspection procedures and instruments must be applied for defect detection, to ensure repeatability and comparison with previous inspections. This must also be a consideration when digital ultrasonic instruments are used.
Therefore, in new or modified ultrasonic testing standards it is necessary to include and define applicable terms, as well to define the characterization of ultrasonic instruments, such as A/D converters, sampling error, sampling frequency, data recognition and reduction, evaluation procedure, etc.
Fig 1: Sampling of an analog signal at the times nT_{a} Fig 2:The Shannonsches' sampling theorem: The signal can be reconstructed if fa > 2fm is applied. (fm upper frequency limit) Fig 3: a). Sample and hold, b). Amplitude curve of the hold device c.) Sampling error Fig 4: Sampling rate f_{a} as function of the sampling error  F and the upper frequency limit f_{m} Fig 5: Sampling error for an 2 MHz shear wave probe Fig 6: ALOK  Algorithm: Determination of the paramenters i and k: Probapility of detection, registration level, Wave mode, Probe frequency and bandwidth, Material Fig 7: Digitization and pixel generated Ascan display Fig 8: Digitization and pixel generated RF signal 
The evaluation of large amounts of data during the search process results in a considerable slowdown of the inspection process, without an improvement in information quality. The Shannonsche sampling theorem (also called Nyquist criteria) delivers an excellent solution /1/. The Shannonsche sampling theorem states that the curve of an analog signal of bandwidth of a certain frequency can be described if the sampling rate is twice that of the signal's upper frequency limit f_{m} (Fig 2) (ideally the description can be complete).
Figure 2 shows that the spectrum X(f) of the sampled signal is periodic. The spectral period duration is equal to the sample period T_{a} = 1/f_{a}. The overlap of the partial spectrums (higher degree folding products) can be prevented if the Nyquist criteria f_{a} ( 2f_{m} is applied. The complete reconstruction of the measured signal would be possible with an ideal square wave low path filter, which delivers the value 1 for the sampling point nT_{a} and delivers the value 0 for all other sampling points. A Sample and Hold is in practice an applied unit which confirms the Shannonsche sampling theorem (Fig 3a).
The curve of such a Sample and Hold is depicted in Fig 3b. The amplitude deviations of the recorded analog signal from the original signal are as high or as small as the sampling rate chosen in relation to the upper frequency limit. Those deviations are in practice efficiently described with the term relative sampling error ( F) (further called sampling error) (Fig 3c).
Figure 4 shows a graphic presentation of the sample frequency f_{a} as a function of the sampling error (  F) and the upper frequency limit f_{m} shows that a sampling frequency 4 f_{m} > f_{a} > 2 f_{m} describes an analog signal incorrectly, however, based on the Shannonsche sampling theorem the frequency was determined as being sufficient. The reconstruction shows amplitude errors of 10 % to 33 % of the actual signal height. For achieving sampling errors less than 10% it is necessary to apply a sample frequency higher than 4 times the upper signal frequency limit.
The determination of sampling frequency for sampling errors of 10% and 5% is shown in Figure 5, an example of the Siemens probe TTK2 MHz. Considering a 3,35 MHz upper frequency limit f_{m} and a sampling error of 5% ( f_{a} approx.6 f_{m} ) a 20 MHz sampling rate of the ultrasonic signal can be determined.
The determination of the upper frequency limit f_{m} is not commonly used in actual practice. It is more accurate and easier to determine the probe center frequency /2/. Also if the spectrum is unknown the manufacturer delivers the value of the probe center frequency.
The reconstruction of an analog signal can be performed by using linear interpolation, as long as the sample rate is 10 times higher than the nominal or center frequency of the probe /3/. For a sinus signal and a sampling error of 5 % a sample frequency of 20 MHz can be also determined for the above mentioned example. The example shows the same result for sample frequency whether upper frequency limit or the nominal or center frequency of the probe is used.
The sampling error is the most important value of digital instruments; the necessary sample rate is based on that. Considering the state of the art of inservice inspection techniques of NPP, a sampling error of 5 % of the ultrasonic echo signal can be achieved.
The data reduction techniques must be verified on reference bodies containing artificial reflectors. A high data reduction rate can be achieved with the proven ALOK  Algorithm. This method is based on the determination of the half wave maximums of the RF signal. The precise determination of the half wave maximum is achieved by using a high sampling rate (usually 50 MHz).
The ALOK  Algorithm recognizes only those amplitude and time of flight values, which lie within a time gate, and are greater or equal to the "i" preceding half wave value maximums and are higher than "k" successive half wave maximums. The measurement resolution of this technique can be adjusted with the "reduction parameter" i and k.
Figure 6 shows when using high i and k values neighboring echo maximums can not be detected. To evaluate a defect indication (e.g. difference between one large and many small indications) it is usually necessary that the axial and lateral sound resolution be in the range of a wave length . In other words, for a sound area of 2 at least one measurement point must be taken. The parameters i and k can be calculated:
This result is confirmed by most of ALOK applied to inspections and correlated with the results of the guidelines for determination of parameters i an k published by Fraunhofer Institute for NDT (IZFP) (i and k with the amplitude interval: Amplitude maximum  6 dB) /4/. The ALOK method loses the signal phase information (e.g., phase shift at crack tips, not for i = 0, k = 0) and based on the amplitude and time of flight values (raw data) it is not possible to generate a correct Ascan display later. Since the relevant maximums and time of flight values are memorized together as pair values, a characterization of indications on hand of the ALOK method is clear and reproducible.
Another method to reduce the number of measurement values is possible with the "Pixel method" of the ultrasonic signal. The pixel method memorizes or displays the amplitude value, usually the maximum value, within a time gate (pixel width). Pixel method values are:
The pixel method of the Ascan is an pragmatic solution to storage of UT signals in the field of automated testing /5/. Since for this kind of data reduction no high sample rate is necessary (e.g., . 10 MHz for nominal probe frequency of 14 MHz), the amount of data is small. This method has been applied with the transducer array technique for many inservice NPP inspections and has been proven as a search technique. The pixel width in the sound beam axis was in the range of a wave length. This resolution should also be kept for future inspections. By means of rectification and filtering of the RF signals the amplitude data (without the phase information) can be precisely determined. But the related time of flight values depend on the adjusted pixel width. For failure analysis, the transducer array technique uses beam swiveling (analysis of multiple angle of incidences).
A higher resolution can be achieved by pixelizing the digital RF signal. With a high sample rate (f_{a} > 10 f_{n} ; f_{n} = nominal frequency) and a pixel width < /2 the phase of the ultrasonic signal is determined. If not otherwise stated, (see above), the pixel resolution of the RF signal in the beam axis should also have a width of < .
Under item 8.7 "Group 1 tests  Digital Ultrasonic Instrument" additional performances are mentioned. The draft standard points out that "conventions are still developing".
In light of the recently modified standard DIN 25435 1
(Wiederkehrende Prüfungen der Komponenten des Primärkreises von
Leichtwasserreaktoren. Mechanisierte Ultraschallprüfung), and with
DIN 25450 2 (Ultraschallprüfsysteme.
Mechanisierte
Prüfung) still in development, this is the ideal time to determine
characterizations and
performance standards of Digital Ultrasonic Instruments.
We recommend the use of the following items:
DIN 254351
Pkt. 5.2.3 Ultrasonic Instrument
DIN 254502
Item 3 Definitions and Symbols
This article is seen as the first step for a clarification of definitions and performances of digital Ultrasonic Instruments. The aim is to ensure the quality of inspection and the reproducibility of inspection for inservice inspection of NNP as well to achieve a reduction in inspection and evaluation time.
NDTnet 