DE102007025578A1 - Device under test analyzing i.e. network analyzing, method for measuring wave frequency i.e. S-parameter, involves forming reference signal by repetition of arbitrary time limitation signals e.g. chirp signal, in specified range - Google Patents

Abstract

The method involves producing reference signals for a predetermined frequency break point, where the reference signals exhibit a period T. A device under test (DUT) (130) is supplied with the reference signal, and a measurement signal is received from the device under test supplied with the reference signal. Wave frequencies i.e. S-parameters, of the device under test is determined based on the measurement signal. The reference signal is formed by repetition of arbitrary time limitation signals e.g. chirp signal, in a specified range. Independent claims are also included for the following: (1) a computer readable medium has a set of instructions to perform a method for network analyzing a device under test (2) a measuring device i.e. network analyzer, comprising a reference signals producing unit.

Description

The The present invention relates to a method and an apparatus for analyzing a measurement object, hereinafter referred to as DUT (Device Under Test). The invention particularly relates to a method and a device for fast network analysis, i. H. to surveying the S-parameter of active and passive modules. The S parameters describe the reflection and transmission of the measurement object in forward and backward direction. These Parameters of the measurement object are provided by vector network analyzers determined in amount and phase.

In the previous concepts of network analysis, a DUT is measured by means of a single sine tone whose frequency is tuned piecewise over the frequency range to be measured. In this regard, z. B. on the DE 102 46 700 A1 directed.

The general block diagram of a typical network analyzer is in 1 shown, wherein only the measurement of the transmission S _{21 is} shown here by way of example. The measurement of the other S parameters takes place in the same way, with evaluation of the assigned wave quantities.

Referring to 1 creates a source generator 110 a sine wave at the frequency to be measured. With the sinusoidal oscillation becomes a measuring object (DUT) 130 applied. Over a bridge 120 the trailing and the returning wave are separated and analyzed separately according to magnitude and phase. This analysis is based on an IF and baseband mix ( 160 . 170 . 180 respectively. 160a . 170a . 180a ) in a digital signal processing 190 instead of.

There In this procedure, only with a single sine wave Consequently, the course of the S-parameters must be measured over the frequency by a pointwise measurement at each discrete frequency points respectively. This scanning of the frequency response is typically referred to as 'sweep', where there is no continuous detuning the test frequency takes place, but the frequency response to each Fixed frequency points must be measured separately.

advantage This method is a very high measurement accuracy, as by very small measuring bandwidths and a very fine grid of frequency points a high resolution of the frequency response can be achieved.

disadvantage However, this procedure is that the scanning of the frequency response leads to high measurement times, since the measurement is essentially is determined by the switching and settling time of the oscillators. Furthermore, the measured object is always only with a single sine wave acted on what just active assemblies such. B. transmit amplifiers, does not correspond to the typical operating conditions of the test object.

Just for measurements on active assemblies these are limitations hardly acceptable. Here measuring methods are needed, the Measure the test object under typical conditions of use, d. H. The measured object should be as close as possible to those in real use applied signals are applied. Still are straight desirable for production test shortest measuring times, by the turnaround time of the products not by the final one Stop testing.

task The present invention is therefore a measuring method and a Specify measuring device, in particular for network analysis, which allows the shortest possible measuring times while at the same time the measured object subjected to real signals.

These Task is according to the invention by a method and a device for network analysis according to achieved independent claims. advantageous Embodiments are in the subclaims Are defined.

These and other aspects and advantages of the present invention more clearly reading the following detailed DESCRIPTION OF THE INVENTION, in conjunction with the attached drawing, in which

1 shows a general block diagram of a network analyzer;

2 shows a block diagram of a network analyzer with modulated measuring signals;

3 Measuring points of the DUT frequency response shows;

4 shows the spectrum of a Fourier series;

5 shows a simplified measuring arrangement;

6 shows the signals of the measuring arrangement;

7 showing the digital signal processing;

8th shows the structure of the polyphase filter for the interpolation of the measured values;

9 shows the interpolation of the measured values with only one measuring section and

10 shows the interpolation of the measured values for several adjacent measuring sections.

2 shows the block diagram of a network analyzer according to an embodiment of the invention. In order to perform a network analysis with modulated measurement signals, the general concept is out 1 extended by a waveform generator and a mixer.

It It will be seen that this modification is merely an extension the scope of the device is reached, the previous Methods of network analysis are with such a device still applicable without restrictions. In a simple Embodiment may be a measurement with modulated signals also performed with a conventional network analyzer However, the generation of the measurement signal by an external Signal generator must be replaced.

In order to obtain a great flexibility of the concept becomes an Arbitrary Waveform Generator 210 (Generic Waveform Generator), which allows the input of any measurement signal. This measurement signal is generated in the baseband and subsequently with a mixer 230 to the RF oscillation of the source generator (source generator) 110 impressed. Followed by the bridge 120 the selection of the outgoing wave as reference signal u _{ref, HF (t)} as well as of the wave which generates the DUT 130 has passed as measurement signal u _{mess, HF} (t). Subsequently, the two wave quantities are separated by an IF and baseband mixture ( 160 . 170 . 180 respectively. 160a . 170a . 180a ) of digital signal processing 190 fed. With this concept, large areas of the frequency response of the DUT can be measured with a single measured value recording. 130 be measured.

3 shows the desired measurement points of H _DUT (f). The measuring points have the distance 1 / T. In conventional network analysis, the RF frequency (high frequency) is set to each measurement point. As a result, the RF oscillator must settle at each measurement point. Likewise, the resolution filter must settle separately for each measurement point. This method leads to long measuring times, especially at low bandwidth of the resolution filter (in the Hz range). In the proposed method, the RF frequency is not set to each frequency point but only to the frequencies f _{RF, ν} shown in the figure. The distance between the RF breakpoints is BW. Due to the smaller number of breakpoints results in a correspondingly shorter overall duration for the settling of the RF oscillator. The measurement points within the bandwidth BW occur simultaneously by a corresponding broadband signal. This means that the resolution filter only has to settle once, which reduces the total duration of the measurement accordingly.

Consequently becomes in the presented concept a combination of piecewise Tuning of the RF oscillator and broadband measurement with bandwidth BW performed. In the case that the bandwidth BW for the considered analysis is sufficient, even on the tuning of the RF oscillator can be dispensed with.

wobei u_ref,T(t) ein zeitbegrenztes Signal im Bereich 0 ≤ t ≤ T ist. Das zu wiederholende Signal u_ref,T(t) kann beliebig gewählt werden und kann beispielsweise ein CDMA-Signal, ein OFDM-Signal oder ein Chirpsignal sein. In einigen Anwendungen ist es sinnvoll, das DUT mit einem ähnlichen Signal wie im Betriebsfall zu stimulieren, weil dann gleiche Bedingungen vorliegen und sich das DUT ähnlich wie im Betriebsfall verhalt.For measuring the frequency response within the bandwidth BW, a periodic reference signal u _ref (t) is used, which for each frequency stop f _{RF, ν} by definition starts at time t = 0 and has the period T. This results in the reference signal through

where u _{ref, T} (t) is a time-limited signal in the range 0 ≤ t ≤ T. The signal u _{ref, T} (t) to be repeated can be chosen arbitrarily and can be, for example, a CDMA signal, an OFDM signal or a chirp signal. In some applications, it makes sense to stimulate the DUT with a similar signal as in the case of operation, because then the same conditions exist and the DUT behaves similar to the operating case.

erzeugt wird, wobei k der Zeitindex, T_α die Abtastperiode und N_FFT die FFT-Länge ist. Die Periodendauer ergibt sich aus der Abtastperiode und der FFT-Länge gemäß T = N_FFT·T_α.From a metrological point of view, it makes sense that the reference signal is band limited exactly to the analysis bandwidth BW. This condition is achieved by taking the digital sequence u _{ref, T} (kT _α ) from a Fourier series according to

where k is the time index, T _{α is} the sampling period and N _{FFT is} the FFT length. The period duration results from the sampling period and the FFT length according to T = N _FFT · T _α .

FFT of the sequence u _{ref, T} (kT _α ) results in transforms U ref (μ) = α μ (3) with μ = [-N _FFT / 2 + 1, N _FFT / 2] at f = μ · 1 / T. By appropriately selecting α _μ ≠ 0 within the bandwidth BW and otherwise α _μ = 0, an exactly band-limited signal can thus be generated. In 4 an example of the FFT length N _FFT = 16 is shown. The example is for illustrative purposes only. In practice, larger FFT lengths are usually used.

The periodic sequence u _ref (kT _α ) is applied to the D / A converter and the subsequent analog anti-aliasing filter suppresses the repetition spectra. Thus, the analog reference signal u _ref (t) has only the in 4 shown discrete spectral components at the frequencies f = μ · 1 / T within the bandwidth BW. In order to reduce the implementation effort of the anti-aliasing filter, in practice it is often "upsampled" to a higher sampling rate before the D / A conversion in the digital domain.

Furthermore, it is desirable for the spectrum to be constant within the bandwidth BW so that the noise floor is the same in all estimated measurement points. This is achieved simply by taking the Fourier coefficients within the bandwidth BW according to | α μ | = const (4) be chosen with equal amounts. It should also be noted that in some applications, the reference signal should have the lowest possible crest factor. This condition can be approximately achieved by obtaining the coefficients of equal size according to equation (4), a quadratic rising phase. By this further measure approximately a chirp signal is generated with a constant envelope.

berechnet. Allerdings sind in der Regel die berechneten Koeffizienten α_μ außerhalb der Bandbreite BW ungleich Null. Im nächsten Schritt wird die gewünschte Bandbegrenzung durch Multiplikation mit einer diskreten Tiefpass-Übertragungsfunktion αμ' = HTP(μ)·αμ erzwungen. Im einfachsten Fall ist das TP-Filter ein idealer Tiefpass mit der Bandbreite BW, alternativ kann ein Transitionband mit endlicher Steilheit verwendet werden. Im letzten Schritt erzeugt man die gewünschte bandbegrenzte Referenzfolge u_ref,T'(kT_α) durch Einsetzen von α_μ' in Gleichung (2). Durch diese Vorgehensweise wird zwar das ursprüngliche Signal etwas verändert, allerdings bleiben die grundsätzlichen Eigenschaften des Signals (z. B. Crestfaktor) im Wesentlichen erhalten. Es sei angemerkt, dass bei einem beliebig gewählten Signal auch innerhalb der Bandbreite BW kleine Fourierkoeffizienten α_μ auftreten können, was die Messung an diesen Stellen μ ungenau macht.In order to force an exact band limitation for a generally selected signal u _{ref, T} (t), the following procedure is recommended: First, the Fourier coefficients are determined according to

calculated. However, as a rule, the calculated coefficients α _μ outside the bandwidth BW are not equal to zero. In the next step, the desired band limitation is multiplied by a discrete low-pass transfer function α μ '= H TP (Μ) · α μ enforced. In the simplest case, the TP filter is an ideal low-pass filter with BW bandwidth; alternatively, a finite-slope transition band can be used. In the last step, the desired band-limited reference sequence u _{ref, T} '(kT _α ) is generated by substituting α _μ ' in equation (2). Although the original signal is somewhat changed by this procedure, the basic properties of the signal (eg crest factor) are essentially retained. It should be noted that, given an arbitrarily selected signal, small Fourier coefficients α _μ can also occur within the bandwidth BW, which makes the measurement inaccurate at these points μ.

5 shows a simplified measuring arrangement in the equivalent baseband. The reference signal u _ref (t) is passed to the DUT and there is the measurement signal u _mess (t). The DUT has the impulse response duration T _DUT . In the right part of the figure some examples are shown.

6 explains the time sequence of the measurement (within the bandwidth BW) using the reference and measurement signal. In the upper part of the image the reference signal u _{ref, T} (t) is shown. The measurement starts at time t = 0. For the sake of illustration, only the time profile is represented by the two Fourier coefficients α _-2 and α ₂ . It can be seen that with the reference signal no signal jumps occur due to the periodic repetition at the transition points.

Below this, the measured value recording is explained. Note that the measurement signal in the range 0 ≤ t ≤ T _{DUT is} in the transient phase (see 5 ). In this area no measured value recording is carried out, because the linear convolution with h _DUT (t) does not coincide with the cyclic convolution and therefore the frequency response of interest can not simply be determined by FFT calculations. Thus, the measured value recording is only started from t = T _DUT . When taking the measured value, the measured values of the α-th period T are stored in the vector x _α .

führt. Im vorliegenden Beispiel wird eine Mittelung über NofAvg = 3 Vektoren durchgeführt.In principle, the first vector x _{1 is} already sufficient for the estimation of the transfer function. Nevertheless it makes sense to record several vectors. Coherent averaging of multiple vectors provides noise immunity, comparable to reducing the resolution bandwidth of conventional network analysis. The analysis time is thus T mess = NofAvg · T (5) resulting in an equivalent resolution bandwidth of

leads. In the present example, averaging over NofAvg = 3 vectors is performed.

In principle, there is no restriction on the choice of T for a given impulse response duration T _{DUT of} the DUT. In principle, T should also be chosen smaller than T _DUT . However, this makes little sense because the measuring points are calculated in the frequency spacing of 1 / T and thus the frequency response between the measuring points would hardly be correlated. In addition, the sampling theorem in the frequency domain would be violated, resulting in a subsequent interpolation in the frequency domain (see below 7 ) would not allow. For these two reasons, it makes sense if the period T is the condition T ≥ 2 × T DUT (7) Fulfills. By overreaching the sampling theorem in the frequency domain by a factor of 2, accurate interpolation can be achieved with relatively short FIR filters.

In comparison to conventional network analysis, the method according to the invention advantageously has a considerably higher measuring speed: on the one hand, the analog LO only has to settle once per RF frequency stop f _{RF, v} (see FIG 3 ). In conventional network analysis, the analog LO must settle on every measured frequency point (distance 1 / T). This reduces the necessary settling time of the method described by the number of frequency points in the bandwidth BW. If, for example, the FFT length N _FFT = 1024 is used and the inner 800 points span the bandwidth BW, the necessary LO settling times would be reduced by a factor of 800. Furthermore, the necessary measurement time T _mess according to equation (5) is required only once per FFT, while this measurement duration is required for each individual measured frequency point (distance 1 / T) in conventional network analysis. It is assumed that the transient period T _DUT can be neglected compared to T _mess . This reduces the measurement duration of the presented measurement concept by a factor of 800 compared to the conventional network analysis. It can thus be concluded that the measurement duration is reduced according to the number of measurement points within the bandwidth BW. Thus, this method is particularly interesting for small resolution bandwidths, because the measuring time is considerably shortened compared to the conventional case.

aus allen NofAvg Vektoren berechnet ist, erfolgt im nächsten Block 730 die FFT-Berechnung. Als Ergebnis der FFT-Berechnung erhält man im oberen Zweig das geschätzte Spektrum des Messsignals Û_mess(μ).In 7 the simplified block diagram of the digital signal processing is shown. The measurement signal u _mess (t) is sampled by the A / D converter. In the window block 710 the samples are taken from the time t = T _DUT . The N _FFT measured values of the α-th period T are transformed into the vector x _α and to the following averaging block 720 given. In the averaging block 720 the previously received vectors are averaged. This coherent averaging is permissible because the reference signal is periodically repeated at a distance T and thus the vectors are identical in the no-fault case. Thus, by the coherent averaging, a noise reduction is achieved. Once the mean according to

calculated from all NofAvg vectors is done in the next block 730 the FFT calculation. As a result of the FFT calculation, in the upper branch, the estimated spectrum of the measurement signal Û _mess (μ) is obtained.

• – des Referenzsignals im Synthesezweig und
• – des Referenz- und Messsignals im Empfangszweig (angenommene Voraussetzung: die beiden Frequenzgänge sind identisch, was zumindest bei den Digitalfiltern ideal zutrifft)

bei der Berechnung des DUT-Frequenzgangs durch die Quotientenbildung gemäß

automatisch kompensiert werden. Man beachte, dass der DUT-Frequenzgang nach Betrag und Phase berechnet wird.Due to the identical measuring arrangement ( 710a . 720a . 730a ) in the lower branch, the spectrum of the reference signal Û _ref (μ) can be estimated. The received reference signal is designated u _{ref, Rx} (t) for avoidance of misunderstanding. Alternatively, it is possible to dispense with the measurement of Û _ref (μ) and to use the spectrum U _ref (μ) analytically calculated a priori and analytically calculated previously according to equation (3), by appropriate multiplexing. However, the measurement of the reference spectrum is recommended, because thus the meter's frequency response

• - The reference signal in the synthesis branch and
• - the reference and measuring signal in the receiving branch (assumed assumption: the two frequency responses are identical, which is ideally true, at least for the digital filters)

in the calculation of the DUT frequency response by the quotient formation according to

be automatically compensated. Note that the DUT response is calculated by magnitude and phase.

If the user wishes to zoom between the calculated frequency points, the desired frequency points can be interpolated by an interpolator block 750 calculated (see below). If the condition in equation (7) is satisfied, a short filter of the order NofTaps = 10 is sufficient to achieve a very accurate interpolation. To also at the edge subcarriers (see 4 ) to interpolate quality without loss of quality, it is advisable to broaden the reference spectrum on the left and right by NofTaps / 2 subcarriers.

Alternatively, the spectrum at only one frequency stop f _{RF, ν} (see 3 ) and the measurement is repeated cyclically. This operating mode is useful if the DUT is modified during the measurement and the change of the measuring parameters is to be observed virtually in real time. In this case the vector averaging is reset after NofAvg averaging and the calculation is repeated cyclically.

Farther can with large averaging length NofAvg (eg. NofAvg = 1000) the total measurement time will take a long time. In this case could already use intermediate results of vector averaging (eg, at intervals of 10 newly added vectors) the DUT frequency response be calculated. This procedure is already after a short time Time a first rough measurement is displayed and the measurement accuracy the subsequent interim results decrease with increasing averaging to.

In As a rule, fewer measuring points are calculated when calculating the spectrum determines when pixels are present on the screen. To the accuracy To increase the display, optionally an interpolation of Measurement results are performed. It must be ensured that the frequency response of the DUT sufficiently high in the frequency domain was scanned. To reduce the effort should be the interpolation be carried out by means of a polyphase filter.

The structure of a polyphase filter is in 8th shown. The original samples are in a sample memory 810 stored. The filter coefficients are in a coefficient memory 820 deposited.

The interpolation is realized by means of FIR filters which each calculate a polyphase, ie intermediate values with a specific distance to the original value. Thus, the actual interpolation is achieved by switching between the individual polyphases. The number of polyphases corresponds to the maximum Interpolation. If smaller interpolation factors are to be realized, this can be achieved by a jump width greater than 1 in the polyphase selection. The number of filter coefficients per polyphase determines at the same time how many samples are additionally required for the correct tuning in and out of the filter. These samples must be calculated additionally in the calculation of the frequency response if only a single section of the frequency response with the bandwidth BW is to be determined, see 9 ,

If the network analysis is carried out over a larger frequency range, which consists of several adjacent analysis sections with the bandwidth BW, then the original measured values of the neighboring sections can be used for the swing-in phase. This procedure is in 10 shown.

The Invention is not limited to the illustrated embodiments limited. All described and / or drawn features can be combined as desired within the scope of the invention become.

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• - DE 10246700 A1 [0002]

Claims (13)

A method for analyzing, in particular for network analysis, a DUT, comprising the steps of: generating a reference signal for a predetermined frequency stop point, the reference signal having the period T; Applying the measurement signal (DUT) to the reference signal; Receiving a measurement signal from the measurement object (DUT) loaded with the reference signal; Determining wave quantities, in particular S parameters, of the measurement object (DUT) based on the measurement signal, characterized in that the reference signal is formed by repeating any signal limited in the range 0 ≦ t ≦ T. The method of claim 1, wherein the reference signal is band limited. The method of claim 1, wherein the time-limited Signal a time-limited CDMA signal, OFDM signal or chirp signal is. The method of claim 2 or 3, wherein the reference signal additionally filtered to the spectrum of the reference signal to restrict to a given frequency band. The method of claim 1, wherein for determining Wavelengths of the DUT an average Measurement signal is determined within a section whose length the period T of the reference signal corresponds. The method of claim 5, wherein the average Measuring signal by coherent averaging over several Sections of the measuring signal is determined whose length the Period T of the reference signal corresponds. The method of claim 1, wherein the period T of the reference signal is at least twice as large as the pulse response time (T _DUT ) of the measurement object (DUT). Method for analysis, in particular for network analysis, of a test object (DUT), the steps of the method according to Claim 1 cyclically for the particular frequency breakpoint be repeated. Method for analysis, in particular for network analysis, of a test object (DUT), the steps of the method according to Claim 1 for a plurality of frequency breakpoints performed become. The method of claim 1, wherein determining wave quantities of the measurement object (DUT) comprises determining the frequency response (Ĥ _DUT ) of the measurement object (DUT). The method of claim 10, wherein the frequency response (Ĥ _DUT ) of the measurement object (DUT) is determined based on a frequency spectrum of the reference signal. Computer-readable medium on which instructions stored on a computer Process according to the preceding claims realize. Measuring device, in particular network analyzer, comprising: means ( 210 . 220 . 230 ) for generating a reference signal of the period T for a predetermined frequency stop point; Medium ( 120 ) for applying the reference signal to the measurement object; Medium ( 140 . 150 . 160 . 170 . 180 ) for receiving a measurement signal from the measurement subject to the reference signal; Medium ( 190 ) for determining wave sizes of the measurement object based on the measurement signal, characterized in that the means ( 210 . 220 . 230 ) for generating the reference signal by repetition of any signal limited in the range 0 ≤ t ≤ T.