JPS63246681A - Apparatus for simultaneously measuring impedance of numerous electronic components - Google Patents

Abstract

PURPOSE:To enable simultaneous measurement of numerous elements in a space sharing manner with one unit, by driving numerous elements to be measured with a narrow band function waveform individually perpendicular thereto to determine a correlation value with respective drive waveforms. CONSTITUTION:A signal source 10 has (n) signal sources S1-Sn, output amplitude of which is all constant. Output waveforms of the signal sources are of a narrow band to a projected extent while perpendicular to each other to drive DUTsi 11. Other ends of the DUTs are all grouped to be connected to a current- voltage converter 12 for detection of current, an output of which is inputted into a signal processor 14 through an A/D converter 13 to determine correlation values of signals to be analyzed with individual drive function waves by frequency analysis or the like. Thus, a complex amplitude measurement of a corresponding component in a signal being analyzed is performed for each frequency pin thereby enabling separate measurement of impedance values of the DUTs without interference.

Description

Detailed Description of the Invention (Field to which the invention relates) This invention relates to inductance (L), capacitance (C
) or resistor (R), etc., and is used to measure the impedance (or admittance) of passive electronic components such as resistors (R), particularly for measuring chip components in large quantities, at high speed, with desirable accuracy, and with good throughput. Regarding.

(Overview of technical background and problems) In conventional LCR meters, basically one
The number of devices to be measured in one measurement system is one, and speeding up the measurement requires shortening the measurement operation time of the measurement electronic circuit and switching between supply/discharge or contact/disconnection (scanning) of the electronic component to be measured to the measurement terminal. Only the faster speed was the most impressive. That is, Fig. 2 shows such a conventional example, in which the impedance to be measured DUTn is "scanned" by a supply/discharge mechanism that also serves as a mechanical switch, and captures each DUT. In a short period of time, the measurement circuit side generates the drive waveform dw from the oscillation source S, learns the response current waveform 1dut of the DUT via the current-voltage conversion circuit ivc, and generates the drive waveform dw as necessary or just in case. Waveform small ■ or V
dut, and finally obtained the ratio of 1dut and vdutO using an amplitude phase measurement circuit VRD called a vector ratio detector circuit to obtain values representing various quantities such as Z, θ, R, and jx. The disadvantage of this configuration is that even if the switching is performed at high speed at the expense of measurement accuracy and dynamic range, the measurement circuit requires convergence time or time required to achieve equilibrium with pseudo-phase bridges, etc. However, by using an ultra-high-speed operational amplifier as the null amplifier A in the iVC, the range from direct current to high frequencies exceeding the measurement frequency was treated as the baseband, and equilibrium was achieved in one period, thereby eliminating the problem. However, this time, transient phenomena resulting from the entry and exit of the DUT or the rise of dw become a problem, leading to a situation where it is difficult to further increase the speed. Furthermore, since such a high-speed multiplex system generally has to be broadband, it has poor resistance to external noise, and its own thermal noise limit cannot be lowered.

There are also quite fundamental problems with this type of approach. That is, if you want to find the impulse response or step response of the DUT, impedance Z = R + j
If x or admittance Y=G-1-jB is to be determined according to the convention of complex AC expression, the observed frequency must be stable. In other words, CW
(continuous wave) CW under drive or in as narrow a band as possible
Measurements must be made under conditions that are compatible with the above. However, high-speed switching is contrary to this idea, especially since the impedance to be measured is simple and is not C, R or similar, in other words, it is not a function of energy storage or redistribution such as co-shooting characteristics or delay response. This becomes a significant constraint in cases where the In other words, it is impossible to measure the long transient response of such an element until it has subsided.

(Summary of the problem to be solved by the present invention or the device which solves the problem to be provided) The present invention aims to solve the above-mentioned difficulty in increasing the speed or throughput of the conventional method. It is an object of the present invention to provide a device devised so that one device can simultaneously measure a large number of DUTs (n + where n is a natural number greater than 1) in a space-divided manner. In this case, the device is essentially a single, highly integrated and intelligent device, unlike conventional devices operating in parallel. Yet another aspect of the present invention is that the present invention essentially allows the entry and exit sequence of each DUT to be performed at almost any time, even while the device is measuring other DUTs. The aim is to provide the equipment.

Yet another aspect of the present invention is that, despite its overall high speed or high throughput,
The drive waveform applied to and the response to it is sufficiently narrow band, has strong cW properties and almost no pulse properties, yet is optimized for high throughput performance. The present invention aims to provide such a device.

Yet another aspect of the present invention is that, despite its overall high speed or high throughput, the overall bandwidth is at most limited, compared to conventional CW drive and VV drives.
It is an object of the present invention to provide such an apparatus that constitutes a measurement system having only a reception bandwidth equivalent to that of the individual single method using RD.

Still another aspect of the present invention is to provide such an apparatus that once adds all the individual responses of a large number of I)UTs to be measured, and then analyzes and separates and recognizes the responses. It is also an object of the present invention to provide the apparatus, characterized in that almost all of the processing of the addition results including the separation and analysis process is performed digitally, and the digital signal processing is performed on the DUT and The part of the device that requires a response to that drive is essentially remote.
Furthermore, we aim to provide a device that can be used at high speed and efficiently by adding a commercially available function expansion board to a commercially available personal computer without using any specially designed software. From a practical point of view, the present invention has a throughput of tens to more than 100 capacitors per second when used, for example, in selecting medium-value ceramic chip capacitors for consumer equipment, and moreover, with a throughput every time the loft changes. It is an object of the present invention to provide such a device that can automatically adapt and continue measuring even if the nominal value changes without the need for reprogramming.

(Essential Techniques Adopted by the Present Invention to Solve the Problems) The basic and essential techniques adopted by the present invention in order to realize a high-throughput LCR meter that solves the above-mentioned problems are as follows. This can be accomplished by following the three steps below.

fl) Driving a number of DUTs with predetermined narrowband functional waveforms that are all individually orthogonal.

(2) The responses of each DUT to the above driving are finally added together to obtain at most one channel of analog signals to be analyzed.

(3) By determining the correlation value between the analog signal to be analyzed and each of the drive function waveforms, the impedance value (the same applies to the admittance value) of the corresponding DUT is finally determined.

In addition to the above-mentioned basic method, various practical devices are possible by taking advantage of the characteristics of the method, and all of them can produce beneficial implementation effects under each applicable condition. The invention will be better understood with reference to some embodiments described below.

(Preferred Embodiment of the Present Invention) FIG. 1 is a block diagram showing the method principle for introducing the basic method of the present invention, in which the signal source 10 is n
Number of signal sources S! -, whose output amplitudes are all constant, whose output waveforms are all narrowband to a predetermined extent, and are mutually orthogonal. A typical example of such n orthogonal fixed-width narrowband waveforms is a group of sine waves each having a slightly different frequency. Of course, there are many others as described below, but this is the most familiar and quickest. To express it more precisely, let m be a natural number with a relatively large value (such as z 1024 to 4096), and [
S (t) = A cos (mOt) within a limited range of n (for example, a width of 32 to 128 before and after)...
・・・・・・・・・・・・・・・M or mo is the frequency one below the lowest frequency, S (i, t)=Acos ((mo+i) ωt
)−・・・・・・(1') The coming harmony (
...-Monisox) group car. Here, A is the number of threads used to represent the drive level, and has no other meaning. As is obvious, when this group is observed with a period of 5, all components are orthogonal to each other, and as a whole, mO
is itself sufficiently large, and is sufficiently larger than the width n of the range of change of i, so that both the individual components and the whole preserve the narrowband property well.

Such Si drives each DU't"ill, and the other ends of each DUT are all introduced together into the input terminal under virtual ground conditions of a current sensing current voltage converter (iVC) 12. As an example of a conformal iVC, a high-speed wideband operational amplifier A fed back with a high-quality reference impedance Zref can be used.According to the well-known principle of this type of circuit, the output of this iVC has V(1)= −Zref * n1dut =−
−・−・==(';'r signal waveform is obtained.1d
ut is a response current corresponding to each drive waveform of each DUT. Expanding the above formula a little further, it appears as follows (21).

Such V(t) is converted into a digital data flow as a signal to be analyzed by the A/D converter 13 with necessary and sufficient bit precision and sampling rate, and then subjected to correlation analysis or frequency analysis (
It is also subjected to nothing but a type of correlation analysis). The signal processor 14 is conformally a carrier of the Fl''T algorithm for executing the DFT and is a general-purpose signal processing microprocessor, such as the T1 320
It can be made by adding necessary peripheral elements around one or more signal processor LSIs selected from the series. Here, we are not looking for characteristics or equipment specific to this signal processor itself. As will be described later, in addition to PFT/DFT, the calculation to be performed may be a textbook correlation with a group of replicas or templates, that is, convolution integration. Therefore, the signal processor 14 may be a dedicated hardware as in the above example, or may be constructed from a general-purpose microprocessor and software if the processing speed meets the requirements. Here, the communication path 15 is a path for obtaining timing or phase information from the signal source 10 in order to inform the signal processor of the time period from where to what time period should be processed as one unit.

For each unit of time, the output of the signal processor 4 is, in the above example (i.e. performing FFT/DFT), a complex amplitude spectrum, i.e. for each frequency bin in the signal to be analyzed V(t). Complex imaging width measurement of the corresponding component will be performed. If the analysis interval is correctly synchronized and the overall level is within the dynamic range of the system, the orthogonality of the drive signal group is maintained, and there is no interference between bins, so the impedance values of each DUT can be separated in one period without interference. can be measured. If we ignore the fixed number of threads, the net part is nothing but the equation (2') decomposed for each i. In other words, the output of the first bin of the frequency analysis result of V(t) is l11i1=F
x(il+ j
), and ignoring the sign, each of the impedances to be measured can be found as RlI 1 . Actually, in order to actually obtain the inductance or capacitance from the above Xi, conversion including individual correction based on the frequency ωi of the corresponding bin is required.

Since the above process can be performed every two synchronizations mentioned above, all n DUTs can be replaced in total in this period, and this enables the realization of the high throughput that is the gist of the present invention. Become the source. For example, if ω is set to about 5 to 10 m5ec, the frequency difference between each frequency bin (that is, the notch on the frequency axis) can be set to about 100 Hz to 200 Hz. On the other hand, FF with 64 points to 256 points
Since T can be completed within this amount of time using the signal processing microprocessor, the processing side does not become a bottleneck for throughput. Therefore, as an example, if the nominal measurement frequency is IMHz, 100 to 2
To achieve a total of 64 to 256 bins in increments of 00 Hz. This means that it is possible to measure the number of DUTs in the above bins.

In this case, it is necessary to consider the dynamic range of the system. In other words, since the above frequency step or observation time (integration time) gives the same equivalent noise bandwidth,
The measurement environment is set to room temperature (ζ300°IO, and the DUT value is also Z
Considering an example where the value of ref is approximately 50 to 100Ω,
The background noise (also called noise floor) of the entire system is approximately: 1; = $ = -147 dBV - (4). This is approximately the same value as in the conventional method. On the other hand, n
When the sine wave signals of approximately the same level are summed, the level becomes only y times higher on average due to the power addition side than the level when each signal exists individually, but even instantaneous distortion occurs. If we are to treat it as if it were a gradient, we must allow an instantaneous level that is n times higher. In practice, it is often possible to compromise between the two, but in this example, an additional dynamic range of is required. As an impedance meter, the remaining margin in the dynamic range can be used to improve measurement accuracy and expand the impedance range that can be handled, so for example, accuracy 1
If you try to measure it in percentage terms, at least for the reception analysis system, (1:100) 0 dB all at once...
...(6) dynamic range is required. If we adopt a value that is neither excessive nor insufficient as the allowable limit level for the drive level or ■tl, approximately IVrms, this is OdB, so the remaining margin is 1=
17-48-40 toma 60 (dB), which is the above 50
Based on the standardization of ~100Ω, impedances that are easy to handle can be handled in three digits each in the higher and lower directions, that is, from 50 to 100 mΩ to 50 to 100 MΩ, which can be handled with the above-mentioned throughput. Although in reality it is much narrower than this due to stray impedances and necessary corrections, the above estimates nevertheless give a good outline of the practical capabilities of the system.

FIG. 3 is a timing chart showing a working procedure when all n DUTs are to be measured at intervals close to the period of ω by the method described above.

As you can see, it is not data collection or calculation that can actually become a bottleneck in work speed, but nn's DUT
It is easy to see that the work involves completely replacing the

For example, in this type of impedance measurement, DU
Since the temporary phase grounding end of T must be handled with great care in terms of accuracy, it cannot be carried around for a long time via cables or switches, so for example 64 to 25
This is because the six DUT chip components must be physically in front of the virtual phase ground terminal, ie, the inverting input terminal, of the operational amplifier A in the iVC to enter and exit. In other words, a UT transport mechanism (DUT-handler) that can achieve such a state is required, which is semi-assembled and semi-batch processing.

Therefore, Fig. 4 is a schematic diagram showing an example of the configuration of such a UT handler, in which a chip component that is a DUT is dropped onto a rail 40 made of an insulating material, and is transported or removed. They come in from left to right. In the center of the rail 40, there is a required length of iv.
C-side contact 41 A large number of drive source-side contacts 42i are arranged. The pitch at which these drive source side contacts are aligned is made to be equal to the pitch at which the DUTs are lined up inside when being transported. Further, a pressing mechanism including a spring 43 and the like pushes up each DUT to ensure upper and lower contact. It is clear from this figure that the above-mentioned FIG. 1 is realized electrically, so further explanation will be omitted, but the operation of this device can also be easily understood from this figure. In other words, when n DUTs fit between each electrode as specified, the conveyance is stopped using sprocket B4 wheels 4 = 1a, b, etc., and the signal of V(i l is collected, and then the collected data is analyzed. During this period, transportation will be resumed and all DUTs will be replaced for several years.

Although this method can be called semi-batch processing and is not a very efficient method, it is still an important stepping stone for the various improvements shown below in response to the most basic method of the present invention introduced in Figure 1 above. It is something.

On the other hand, in the first embodiment (Fig. 1.3.4), the measurement nominal frequency is 1 MHz, and the surrounding area is ±12.8 KH.
z is received and converted by iVC and A/D to form digital data, which is processed and analyzed directly by a signal processor in a narrow band manner. However, it is well known that it is a great loss to directly sample, A/D convert, or harmonic analyze such narrowband waveform data while observing its own Nyquist conditions.

This is because the actual amount of information included is only commensurate with the bandwidth. The usual improvement in such cases is to move away from an RF system that directly handles RF data and replace an IP system via heterodyne conversion with a baseband system via coherent detection. FIG. 5 shows an example in which the present invention is implemented using an IP system based on such a policy, and FIG. 6 shows an example in which the invention is similarly implemented using a baseband system, both of which show only the main parts.

In FIG. 5, the output V(tl) of the iVC is heterodyne-converted using a sine wave of another frequency fLo generated by the local LO, and the frequency is shifted.For example, in the above case, the measurement frequency is IMHz±12.8 KHz. Since it is distributed within the range of The signal is passed through the anti-aliasing filter AAF and subjected to A/D conversion, and then treated in the same manner as in Fig. 1. In this method, care must be taken regarding phase rotation in such heterodyne conversion and filter processing.

In this case, phase rotation occurs due to a phase shift of ω between the basic timing shared by the local LO and each drive signal (with a cycle of 15), and when it occurs during the filtering or A/D conversion work described above. There is. The former applies commonly to all frequencies (i.e., to measurements of all DUTs),
The latter is determined by the frequency characteristics (phase characteristics) of the amount of phase shift provided by the filter and the aperture effect of A/D conversion, so its effectiveness is not common to all frequencies. These solutions include:
First, since W should also be phase-coherent with a period of 'lx', it is set as one of the outputs of the harmonics generator to obtain the signal source for driving the DUT.
One good idea is to get it as one. Regarding the filter and A/D stage, this filter is incorporated into the digital side, and a so-called phaseless filter or phase linear filter is created and implemented digitally, and in order to make this possible, A/D conversion is required. One ultimate solution is to perform oversampling by 2 to 8 times. This type of technology has already been well established and widely used as a digital audio technology. Note that if the pit accuracy of A/D conversion is directly performed at a necessary and sufficient sampling rate, the necessary (or naturally occurring) measurement accuracy cannot be achieved unless it is 16 bits or more, but if oversampling or other If you measure the period and average it, you can also use it with 12ω to 14 pits. In addition, all the DUTs mentioned above
It would be even easier to divide the IP frequencies into several subbands and handle them separately instead of handling them on a single line. In any case, even if all of the above phase rotation factors cannot be eliminated, if they always appear with good reproducibility, then F
It can be made harmless by correcting the complex amplitude output of the pin after F'T using software. Amplitude (Gain) Frequency Characteristic Board Even if there is one, it is harmless if the observed value of the standard impedance is finally saved and calibrated (or normalized) and output as described above. can be converted into -On the other hand, in the example of Fig. 6, L
O is located approximately at the center of the band, preferably any one of Si
It is advantageous to use this as the LO without using it for driving the DUT (although it may be used, it is preferable to avoid the DC component in the demodulated output), and use it as a reference carrier that differs by 90' depending on the phase shifter PSN. With these, the output (t) of the iVC is subjected to coherent detection (homodyne detection) by a 21-unit balanced demodulator BDWt, 2. Both the i and q baseband signals obtained are filtered as necessary and then A/D converted, as in the case of the previous IP system. In this case, the highest frequency appearing in each 1yq channel is half that of the previous IP system. i and q
By arranging the sign of the result obtained by applying complex FFT and performing a linear combination, the component USB above fLo and the component L below fLo can be obtained.
It is well known that analysis and evaluation can be performed without confusion with SB. After all, the amount of processing required for the hardware for FFT/DFT is the same for both the IF system and the baseband system in this case, requiring or paying attention to phase rotation and phase coherency. matters,
The countermeasures or correction policies are essentially the same.

In this respect, there is almost no significant difference between the IP system and the baseband system in practice. However, the work equivalent to such heterodyne conversion and coherent detection work is R.
It is well known that this can be done by digitally editing or processing the results of sampling the F area Mt) according to appropriate rules. Also, if the sampling aperture is narrow enough, S
/N or if there is room for the total throughput, it is also possible to substitute the heterodyne conversion or coherent detection using a high-order monic sampling technique where the power gain of the conversion is poor.

In either case, careful attention and management are required for the overall phase relationship and its constancy (in other words, coherent), including the stage where it is finally subjected to FFT/DFT processing. be.

By the way, in the explanations given above, we have assumed that various operations are always carried out in units of time, and that each signal source Si is a constant amplitude CW, although it has a harmonic relationship with respect to lrr. . The reason for doing this is that Si or Mt)
As the length of the cutout section! This is because, strictly speaking, if a value other than Tomoe is used for ω, the orthogonality between each harmonic component will not be perfect, causing interference between pins and eventually interference between measured values of the DUT. However, if Si or V(t) is amplitude-modulated (weighted) on the time axis in advance, even if the cutout section is out of order, it will not cause trouble to other pins. FIG. 7a shows a conformal example of this, which is called a Gaussian pulse, and its frequency spectrum (FIG. 7b) also quickly converges to the next pin (arrow). Of course, this is not the only option; in practice, various weighting functions (or window functions), such as Hanning and squared cosine, can be usefully used for the same purpose. It is very beneficial to weight the pins in such a gentle mountain shape so that there is no (or sufficiently little) interference with other pins on the frequency axis. In this way, even if you have to perform a certain task in parallel time ω with the tapping period roughly as planned, there is no need for strictness that includes the instantaneous phase of each RF multiplied signal. However, the main requirements are only the constancy of the steady-state deviation and the subsequent software correction. A constant amplitude burst wave that is simply extracted without weighting, as shown in Figure 7C, exhibits the side lobes of the well-known sink-type spectrum, as shown in Figure 7D, and the observation period perfectly matches that of 1. Otherwise, other pins will fall at locations other than the zero-crossing points (arrows) in the diagram, causing interference.

Obviating the requirement for strict coherency by time-axis weighting each 81 or Vtt) as described above also brings about the following essentially important advantages. In other words, no matter how limited the bandwidth is,
If the extracted burst waves shown in Fig. 7C are coherently aligned so as to include instants in which their instantaneous phases are in phase with each other, their sum will be much larger than the average value l5 times as mentioned earlier. It can even exceed n times. As mentioned above, including such moments is extremely disadvantageous in designing the dynamic range of the system. However, each Si
If the phase (or delay time) is distributed little by little to the corresponding i-th component in V(t), which is the sum of the responses of n DUTs with the same nominal value, then this Such situations can be avoided. Alternatively, the same effect can be produced by arranging the entire process so that the moment when all components harmonize comes at the end or start point of the weighted function, at a time when it is definitely zero. Neither of these modifications can be performed without the presence of a weighting function. However, if saturation due to instantaneous simultaneous phase as described above does not occur frequently, and the total time is small enough to be ignored compared to the data collection time, then this can be regarded as mere noise or measurement error, and its existence is assumed. It is also possible to design a system by taking this into account, which is within the freedom of design.

By the way, the gist of the present invention is to simultaneously measure a large number of DUTs using an orthogonal function system, but the orthogonal function system can manage the energy (power) and the range in which the amplitude and frequency appear. Even within the scope of what can be done (that is, the necessary condition that it can be made into a real electrical signal that is easy to handle), there are many other things besides harmonic sine waves. Some of them can be applied to implement the present invention as described below. First, for understanding,
If we discuss the duration T and the frequency spectrum width B of a group of constant frequency sine/sinusoidal bursts or time-amplitude weighted bursts as shown in Figure 7 above, this is the so-called Heisenberg uncertainty. It is well known that by using the principle of gender, T B ≦1. However, as has been experienced in the areas of radar, ν-ner, and wireless communications, it is possible to reduce the TB product to a value much larger than 1 using a sublends vector system (distributed spectrum transmitting/receiving system). Are known. The advantage of applying a drive signal with TB > 1 to the present invention is that each DU
The point is that all measurement frequencies or drive waveforms for T can be unified, which eliminates the trouble of correcting the frequency based on the observed impedance value by handling it differently for each DUT and converting it to the value of L or C. The point is that it can be omitted. In addition, the orthogonality of the driving waveforms of each DUT from moment to moment can only be obtained by time-shifting the waveforms, so the time at which one DUT enters and leaves the measurement system, or in other words, the time at which measurement starts and ends, is , as long as it is sufficiently different from other DUTs as planned. Therefore, n due to the sine wave harmonic group
An approach that follows timed patching of individual DUTs is rather unsuitable in this case, as new DUTs and drive sources enter the market, and as the oldest pairs are separated, data acquisition is slowed to a signal. A system in which processing and correlation analysis are carried out in a continuous manner is more suitable for this case.

A spectrally distributed drive signal is conformally defined as PN.
There are sine waves that are phase-modulated with a code (m-sequence, Barker code, etc.), and sine waves that are FM-modulated using a linear or predetermined monotone function. It is well known that the former is commonly called a phase code signal (or method), and the latter is commonly called a PM chirp signal (or method). The autocorrelation of these drive signals is essentially zero in every bin of time shift except for unity at the origin. In practice, side lobes of the order of the reciprocal of the bit length of the code occur, but if the code is long enough, in other words, if it is sufficiently large as evaluated by the TB product, it can be ignored. Therefore, in order to implement the present invention, each DUT is driven with such a distributed spectrum drive signal that has the same waveform but different start times, and the obtained V(t) is determined using the waveform as a template. All that is required is a correlation analysis, and the impedance value of each DUT will be obtained as a correlation value at each time-shifted pin.

FIG. 8 is a block diagram showing an embodiment of the present invention using the above phase code method, in which the I MH
While the Z carrier generator 8o gives a common carrier to the n balanced modulators 81, the output of the PN sequence generator 82 is sent through a tapped shift register (or delay line) 83 so that the phase is slightly different. It is distributed to each balanced moshulator 81 as shown in FIG. An m-sequence generator with m28 (255 clock periods) is used as an example of the PN sequence generator. 2, which is obtained by dividing the frequency of IMI-1z of the carrier generator 80 by 40.
5 K1-1z is used as its step clock and also as the step clock of the shift register 83. The total length of the shift register 83 is 256 stages, which accommodates one period of the In series that is constantly generated.

Each output of the balanced modulator 81 is power amplified by the power amplifier 8.1, made low impedance, and then used to drive one end of each DUT as before, while the current response of the DUT is also summarized. i The signal to be analyzed is converted to V (tl) via VC.The A/D conversion and signal processing are the same as the example in Figure 1, and the IF system and baseband system (preferably,
(FIG. 8 shows an example of a baseband system) is the same as above.

In this case, the periodic length of the sequence is 25 KHz (40 psec
) 0255 times, it becomes about 10 m5 ec or more, which gives the minimum unit of data collection time. Further, as is well known, the sine wave carrier binary phase modulated with the PN code has a frequency of repetition of the above series (in this case, 100I (100I)).
Arrange equal-sized spectrum groups around the center frequency in increments of , weak).

In this case, digital signal processing can be carried out by cross-correlation or convolution integration with the PN signal used, and for this purpose, as mentioned above, a signal processing microprocessor such as the Tl320 series and a suitable group of peripheral elements are used. can be used. However, such processing generally does not have as efficient and sophisticated algorithms as in the case of FFT/DFT, so a replica of the code signal (either in an IP system, in a baseband system, or in an RF
(The style differs depending on whether it is direct processing) is prepared in ROM, and if a processor called Stray is used with appropriate peripheral elements while shifting the relative phase, the instruction cycle can be executed in steps of 100 nS (10 MHz), and the product-sum Since the accumulation work can be done with one instruction, 255-point convolution can be done in a time not much longer than the 2545μ formula. On the other hand, as mentioned above, the step clock when executing the PN code is 40μ9[! c(25Kl-1z), so it is much slower than this, and in the end it takes a while for this PN code to return to its original state after being damaged. ) It is possible with enough time to finish calculating all the correlation values in the 256 time-shifted bins.

On the other hand, FIG. 9 shows the FM when using the signal with TB≦1.
This shows the case of chirp method, in which VC
O90 continues to periodically generate a linear chirp signal according to instructions from ramp function generator 91. This period is separately generated by the local oscillator (LO) 95. In this case, the signal of 975 KI (z) is divided by 39, and the width of the 2M chirp is ±25 KHz.
It is weak. This is input to the tapped delay line 93 to obtain a signal shifted by 40 μm from each tap, which is amplified by the power amplifier 94 to drive the DUT in the same manner as described above. The steps below to obtain V(t) are exactly the same as the example above, and it may be preferable to handle V(t) using the IP method (Figure 9 shows that case) or the baseband method. The same is true.

In this case, the tapped delay line is the required number of crystal delay lines with a propagation time of 40μ (256 if n=256).
A preferred example is to use the amplifiers in a cascade, or insert an amplifier in the middle as necessary to compensate for attenuation.

Since the result of hetero gain conversion by the LO of 975 KHz falls within the range of dc to 50 KHz, it can be treated roughly in the same way as an audio signal, as in the previous example. In the end, the waveform that should be the partner (template) for cross-correlation can be obtained by hetero gain converting the output of the VCO 90 itself using LO, so prepare it in a ROM etc. and use the same method as in the previous example. Perform signal processing. It is self-evident that the signal processing procedure constitutes a software matched inverse dispersion filter.

In addition, in all of Figure 1, Figure 5.6.8.9, etc., i
The VC does not necessarily have to be constructed in such a way that the response currents of all DUTs are added to one point and virtual grounded using only one operational amplifier.
have unit circuits 101 and their output Vi(t)
It may also be configured such that V(t) is obtained by adding them using the summing amplifier 102. Further, although not shown in the figure, it is not necessary to disperse the DUTs to this extent, but it is also possible to group several DUTs into units, create a unit iVC for each group, and combine and add the outputs. The advantage of the method shown in FIG. 10 is that the degree of freedom regarding the distribution of the physical locations of DUTs is greatly increased compared to the example shown in FIG. In other words, the routing of Vi(t) in the intermediate section can be routed over a considerable distance as long as attention is paid to matching (preventing reflections and compensating for attenuation and delay) at the beginning and end. In this case, it is obvious that the signal to be analyzed V(t) itself can be carried around a considerable distance with the same care taken.

Incidentally, in the examples of FIGS. 8 and 9, it is clear from the system principle that the sequence of entry and exit of the DUT can be started at almost any time. If anything,
If you wait an amount of time equivalent to one pin chirp or code shift (this time is stated as approximately 40μ in all of the above examples), then any other DU
This is because it is possible to start its own drive signal without interfering with the response signal that T is giving. This property enables so-called flow-through work. This is extremely useful for designing DUT handlers and improving overall throughput. That is, FIG. 11 is a schematic diagram showing an example of a DUT handler based on this idea, in which a continuously rotating turntable 110 has a large number of tweezers-like DUT gripping limb means 111 around its periphery. Chip components are continuously supplied from the left through a duct 112a, fitted into the chip at a point A, removed at a point B, and carried away to the right through a duct 112b. At point B, if necessary, a means for excluding parts whose measurement result is defective is also used. However, what I would like to discuss here is not the minute details of such a mechanical mechanism, but the electrical aspects of its design.

That is, each Vincera) 111 mutually insulated upper and lower metal fittings, all of the upper metal fittings are individually connected to the signal source Si.
The lower ones are connected to the iVC all together, in groups, or individually. The main parts of all of these electronic circuits such as Si and iVC are located on the turntable 110, and signals and signals are connected to the outside world by means such as slip rings, rotating transformers, rotating photocouplers, and telemeters (not shown). Transfers and receives electricity. The essential point of this configuration is that each tweezers has an individual Si, and the D
At the point where the DUT completes driving for at least one cycle while traveling along the same path as the UT from point A to point B, the journey experienced by such an individual DUT is relatively slow. Even so, this turntable as a whole can achieve a very high throughput. This can be interpreted as allowing the sequence of each DUT to be slow compared to the amount of throughput required; for example, a large capacity chip electrolytic capacitor of 1 μF or more can be Frequency (e.g. 10
This is useful when trying to continuously measure a large amount of frequency (0 Hz to IKHz, etc.).

The signal processing device to be coupled to the turntable-like DUT handler has a dynamic file of the signal to be analyzed in a stream, and has a phase corresponding to the drive waveform of the DUT that has just completed one cycle of drive. If you control the timing of the calculation by using the template as the target for correlation at that time, you can always know the impedance value of the DOT that is about to leave, and judge whether or not to pass the test. It can be performed. During this time, as mentioned above, the convolution calculation for one pin can be performed using at most several tens of microns, so in practice, this time does not cause a decrease in throughput. Therefore, most of the time spent rotating around the turntable in a set of bins can be spent on driving with Si and collecting data on V(tl). DUTs can also be measured continuously at high speed and in large quantities with high precision.In the example shown in Fig. 11, a DUT-handler for rotating circumferential flow work was described as an example.
This is just one example, and based on this idea,
The driver has infinite degrees of freedom when considering robotic movements. The entry and exit of the DUT does not necessarily have to be performed sequentially; for example, although not shown in the drawings, several fixed hole-like actuators such as arm-like actuators such as robot-like arm-like actuators may be sequentially or sequentially arranged. A mechanism for loading and unloading the DUT is also well suited to the spirit of this embodiment. In either case, one DUT handler has one system of iVC and a signal processing device (plus a summing amplifier for peace of mind), and has many DUs.
The signal to be analyzed V(
t), or in other words, in the V(t) thus obtained, essentially a large number of D
Any configuration in which mutually orthogonal response signals of the UTs are superimposed and awaiting analysis can be said to be an example of the implementation of the present invention. The decision as to whether the DUT's discrete signal processing is to be performed in patches or in a continuous manner is determined by the first step.
5.6.8.9, it can be easily understood that this is within the degree of freedom in implementation.

Incidentally, all of the various ideas and precautions described below bring about beneficial effects in implementing the present invention. That is, first of all, when using the above-mentioned dispersion spectrum drive signal, if the fractional bandwidth as a result of the dispersion is too wide (although such implementation is not desirable for the idea of complex AC expression of impedance, it will reduce the throughput). (There may be cases where it is necessary to adopt it according to the requirements) If the DUT is essentially C or R, the Zref will be set accordingly.If C and R are not used, frequency characteristics will occur in the iVC operation and correlation will occur. The calculation will not be completed correctly. If the f-characteristic gain difference at the upper and lower ends of the band is 1 dB or more due to this reason, the detailed warping will not satisfy the required accuracy of ±1π according to the trial calculation, and the orthogonality between each Si will be disturbed, and other Disturbs the measurements of other DUTs in the bin. Therefore, even if a distributed spectrum system is adopted, it is preferable to design it so that its fractional bandwidth is no more than 109g.

Furthermore, even if adaptive or empirical correction is performed when inter-bin interference is observed, essentially D
Therefore, it is important to simultaneously and accurately measure items with different C and R values, as well as groups that include items that are the same but whose values differ by more than twice or less than half the average value. That's a mistake. However, as long as these constraints are observed, Zref can be satisfied in all cases by using high-grade pure resistors.

On the other hand, as an LCR meter or impedance analyzer, components that exhibit a resonant response or delayed response, such as 2
There may be cases where it is necessary to measure a ceramic filter (resonator), etc. There may also be cases where it is necessary to measure the transfer impedance of a 3- or 4-terminal circuit network. In such a case, in the distributed spectrum method, the delay time, that is, Si
It should be noted that the time difference between the pins is an essentially important point, and when measuring a DUT with a response delay that is large enough to disturb it, it is necessary to
It is essential to devise ways to limit the use of DUTs only to tsuoki and others. Furthermore, since the measurement frequency of the measurement using the distributed spectrum signal is not uniquely determined, it is natural that the result will be different from the result of CW measurement performed slowly over time. High throughput by driving with distributed spectrum signals is beneficial only if the difference is significantly less than the required measurement accuracy.

On the other hand, as long as the method is mainly used to find fine differences between a group of DUTs whose properties are fairly predictable, it may be reduced to relative measurement. In other words, taking Figure 1 as an example, the DUT
1 to DUTn-1 as the object under test, DUT
If you include a standard whose value is known in detail, you don't have to worry about its accuracy, as it only needs to be constant even if the Zref value itself is not detailed. Become.

The purpose can be sufficiently achieved by back-calibrating or normalizing other values using the measurement data of the DUT and outputting the results. In addition, even if Zref is unreliable in the worst case, statistical data of n DUTs can be obtained relatively, for example, ±2 to
It is possible to perform work such as discarding items with a distance of 3 sigma or more.

However, in practice, good DU is limited to a certain range.
It is difficult to deny that incorrect or broken items suddenly enter the T' group.

As shown in Figure 1, and as everything explained below, in a circuit that uses constant voltage drive to determine the response current, the lower the impedance of the DUT, the larger the current flows, the more power is required, and in the worst case, a short circuit occurs. If the bar gets stuck, it will cause a problem. In other words, in this case, the obstacle is D
Even if the UT, Si, or PA does not burn out, it saturates the operation of the iVC and makes it impossible to collect and evaluate the response data of all other DUTs that are being tested at the same time. Hail to you. The only way to prevent this is to provide each Si with a current limiting function. In this case, a so-called clipping type limiter will generate a large amount of distortion due to non-linearity when the limit is reached, which may cause trouble in analyzing data in other bins, so it is preferable to use a simple series resistor. type IJ
Mitta is good. FIG. 12 shows an example. In this case, since the final impedance calculation value includes this series resistance Rs, each value must be subtracted and output as a measurement output. When measuring n≧100 DUTs, the dynamic range margin of the system is 3d.
If B or more is secured, the minimum impedance between each Si and DOTO is 1/100 to 115 of the average impedance of the DUT.
If there is a series resistance of 0, at least one or two
Even if a few short (punk) DUTs or similar low impedance devices are mixed in, other D'J
The measurement of T is seamless. In order to be a current limiting means that is effective not only for C but also for R,
The current limiting resistor needs to be a pure low value R.

In the case of voltage drive and current response, an open DUT (cracked, chipped electrode) accident is harmless for group measurement, but only causes the data in the bin corresponding to that DOT to be filled with residual noise and interfering components. be.

Reducing the accuracy required for Zref by the above method is very advantageous for making it variable. Figure 13 shows group A taking advantage of this property at the iVC stage.
GC (or ALC) is implemented, and in this, a well-known photoconductor element Cd5131 is used as Zref while being controlled by a light emitting diode LED"2. The current of the LED is controlled by software from the signal processing device 134. Based on the analysis results of V(tl), it is controlled to a moderately appropriate level.Such control can also be performed by monitoring the Mtl level.

The concept of ALC or AGC can be extended to automate range switching. FIG. 14 shows such an example, in which Zref is realized using a multiplier type D/A converter as a variable transfer resistance (three-terminal resistance network). As is obvious, the operation of this circuit is I)U
The range is automatically selected digitally for the average value of T1 to DUTo-1, and the signal processing device DSP
is set to an appropriate operating level by taking into account either the level of the input V(t) itself or the average value of the size of each bin of the output of the correlation operation or FFT/DF71''. In this case, Zref is a multiplication type D/A called LOG-DAC that can be electrically controlled.
One example of a suitable converter is a converter such as the AD7118.

In each of the above embodiments, the description has been made on the premise that all of the many DUTs to be measured are individual components. However, it will be understood from the following examples that the practice of the present invention need not be under such conditions. Figure 15 deals with standard samples of substrates such as organic materials, plastics, or ceramics as treasures to be measured.
The purpose is to simultaneously measure the distribution of resistivity, permittivity, dielectric loss, etc. within that plane. Although the figure schematically shows the cross section of such a UT-ndler, the actual product has a two-dimensional structure. That is, one flat plate electrode 151 is applied to the entire surface of the measured sub-stray 150 from below, while from above there are numerous spring-loaded bottles 1 whose heads have a constant shape.
52 descends and is brought into close contact. The shape of the head of each of these bottles may be round or square if corrections based on the area occupancy are corrected, but since adhesion is extremely important, they are all finished smoothly. The importance of adhesion is the same in the case of the lower flat plate electrode. Each of the upper bins is connected to the same Si as on each side, and the lower common electrode is connected to iVC. The configuration and operating principle of the following parts are exactly the same as those of the previous parts.

Here, the pedestals 153 and 154 made of an insulator hold the lower and upper electrodes, respectively, and are used to extend or move up and down by a mechanism (not shown) as necessary.

With this configuration, it is possible to know the distribution of electrical properties of the substrate to be measured, as well as the bias and distribution of its baking or composition, at one time.

The measurement results are 1D as an image of the distribution of ε, ρ, and tanδ.
It is very easy to understand if the information is transferred from the IP to an external personal computer 155 or the like and displayed in color on the screen of the color display 156.

Such a system for directly evaluating the planar distribution of electrical properties of a sample substrate can be widely applied to evaluation of prototyping manufacturing processes for various organic and inorganic artificial composites, not just ceramics and plastics intended for electronic components. Be expected. Of course, it can be applied not only to artificial objects but also to the evaluation of natural rock samples. It can also be applied to the evaluation of membranous substances derived from living organisms, living biological membranes, monolayer cell layers in culture, active enzyme retaining layers, etc. Returning to electronic components, a group of ceramics and plastics whose piezoelectric or pyroelectric properties are used have a very close relationship with the dielectric constant after firing or drawing. It can also be useful for indirect evaluation of the objective nature of Alternatively, the DUT handler shown in Fig. 15 may be reduced to a one-dimensional structure, and the upper and lower electrodes may be made movable, or the DUT 150 may be moved continuously as a long plate or sheet using a structure such as a roll. It is also possible to transform it so that it can be measured.

(Effects of the Invention) As described above, according to the present invention, a large number of DUTs can be measured simultaneously with high precision and high throughput, so it is greatly beneficial when applied to industries such as electronic components and electronic materials. It is.

(Definitions of Abbreviations) Among the abbreviations and abbreviations used in this specification, those that are not commonly known and commonly used abbreviations are as follows.

DUT cod・・・・・・Device Under
Abbreviation of Te5t. Generally refers to an electronic device or its parts to be measured.

1°V RD n--Vector Ratio
Abbreviation for Detector. It is an electronic circuit that calculates the ratio of two complex alternating current vectors of sine waves that are different in amplitude and phase.

”dw,・・・・・・・・・・・・driving w
Abbreviation for aveform. Drive waveform.

i dutn・・・・・・DUT current.

Vdut-・・・・・・・・・Terminal voltage of DUT.

``iVC, ・・・・・・・・・Current voltage converter.

"V(t) -... Signal to be analyzed.

″DSPs----DIgital Signal
Abbreviation for Processor.

”Z refn・・・・・・Reference impedance.

Z s tcly・・・・・・Standard impedance.

"S+n......i-th signal source.

[Brief explanation of the drawing]

FIG. 2 shows a conventional example, and FIG. 1 is a block diagram showing the most basic embodiment of the present invention. 3 to 15 are block diagrams or schematic diagrams, waveforms, spectra, etc. showing different embodiments of the present invention. Figure 2 (Prior art) Figure 12 Submitted on April 1, 1988 (Patent application dated April 2, 1986 (application number notification not received) (Measuring device (3) Person who makes corrections
(Relationship with the case Patent applicant address: 2-16-11 Takaido Higashi, Suginami-ku, Tokyo, Japan Detailed description of the invention column (5) Contents of amendment
(As shown in the attached document, the error in the correction details will be corrected accordingly.) 2) On the 7th line from the top of page 37, it says "If the PN code is -, it will return to the original state..." Insert omissions such as "The PN code goes through one cycle and returns to the original...". 3) On the 12th line of the same page, “TB≦1” means TB≧
4) In the 5th or 3rd line from the bottom of page 46, there should be a “One-sided DU” error.
``T... may be degenerate'' is changed from ``DOT'' is used mainly to find fine differences within a group of objects whose properties can be predicted fairly well. , the measurement may be reduced to a relative measurement.''

Claims (1)

[Scope of Claims] 1. n drive sources that output functional waveforms whose output amplitudes are controlled and are all orthogonal to each other, and these n drive sources are connected to one side of each of n electronic components to be measured. and the other terminal of each of these n electronic components to be measured, individually or in groups, or all collectively, while operating to virtually ground the other terminal of each of the n electronic components to be measured. current detection means (current-voltage converter) for detecting the total value of ground currents derived from the drive outputs of the drive sources at a virtual ground point; an output waveform of the means; and a drive waveform of the n drive sources; A method for detecting a large number of electronic components, comprising means for obtaining each correlation value and integrating or smoothing them as necessary, and configured to obtain each impedance to be measured from the group of correlation values. A device that measures impedance simultaneously. 2. The device, wherein each of the sine waves is a harmonic of a different order in a harmonic relationship with respect to the integral interval length of the correlation value. 3. In the above device, each driving source is a sine wave burst of each of the above-mentioned different frequencies multiplied by a weighting function on the time axis, and the product of T (duration) and B (bandwidth) (TB product) )
is approximately 1. 4. The device as described above, wherein each driving source is a phase-modulated or frequency-modulated burst, and the TB product thereof is significantly larger than 1. 5. In the device, a group of drive waveforms output from each drive source falls within at least one frequency band whose relative bandwidth is not large, and the outputs of the virtual grounding means and the grounding current detection means are integrated into one station. The device described above is characterized in that it is configured to include a step of performing heterodyne conversion or homodyne detection (coherent detection) using radiation. 6. In the apparatus, the values of the large number of impedances to be measured are determined based on complex frequency spectrum data obtained by subjecting the IF signal as a result of heterodyne conversion or the baseband signal as a result of coherent detection to fast complex Fourier transform (FFT). The device, characterized in that: 7. In the above-mentioned FFT type device, the length of the data section to be subjected to FFT is set to a time length such that a harmonic relationship is obtained for all the driving sources, and the relative phase of all the driving sources with respect to the data section is managed. The device is characterized in that: 8. The apparatus is characterized in that it is configured to perform batch processing, including a procedure for replacing all of a group of electronic components to be measured, using the data collection period as one time unit. 9. The data acquisition section is not particularly divided, and the data processing device is configured to perform the corresponding measurement only when a significant response corresponding to each drive source has been inputted,
The apparatus is characterized in that the part to be measured is configured so that it can be attached to and detached from, and supplied to and discharged from, a connection terminal for measurement according to a random or predetermined procedure. 10. The above-mentioned parts to be measured are loaded and unloaded by a serial transport mechanism, and one drive source is arranged corresponding to the part to which each part of the transport mechanism is to be fitted, and the parts and the drive sources are connected to one another. The apparatus is characterized in that it is configured to travel as a set in the conveyance mechanism in one-to-one correspondence. 11. In the device according to claims 1 and 3, each drive source is obtained by binary phase modulating one common carrier with a PN code signal,
The apparatus is characterized in that the code signal has a plurality of modulation results having different phases and is used as a group of driving sources. 12. In the device according to claim 1, each driving source is composed of a constant amplitude sine wave of the same length and FM chirped in the same manner, and only the start and end times of the chirps are different. The device is characterized in that it is a group of driving sources at a location. 13. In the device according to claim 11 or 12, the degree of spectral dispersion of the modulation result is sufficiently small to the extent that it does not interfere with the treatment of the impedance to be measured using a complex AC representation, and This device is characterized in that it is configured so that a pure resistance is used in common as the reference impedance of the current-voltage converter, regardless of whether the impedance to be measured is L, C, or R. 14. In the device according to claim 1, n
The apparatus is characterized in that it is configured to perform an auto-ranging operation with respect to the average of the absolute values of impedances of the parts to be measured. 15. In the device according to claim 1, n
- A known standard impedance is connected to the remaining one for measuring the impedance of one component under test, thus reducing the accuracy requirements for the reference impedance of the current-voltage converter. The device is characterized in that: 16. A variable impedance that can be controlled by voltage, current, light intensity, etc. is used as the reference impedance, and by separately controlling this variable impedance, automatic level adjustment (ALC) or Automatic gain adjustment (
1. The device is configured to include and execute operations of AGC. 17. In the apparatus according to claim 1, each drive source is configured to drive each DUT through a current-limiting series impedance of a predetermined value, so that it is possible to inadvertently The device is characterized in that it is configured so that even if a short-circuited DUT is mixed in, impedance measurement of other DUTs can be performed without interference. 18. In the device according to claim 1, the large number of impedances to be measured are the large number of impedances formed by electrodes placed at many points on the front surface of one member and common electrodes uniformly provided on the back surface. impedance, and is characterized in that it is configured so that the impedance density distribution of the member can be determined by measuring these impedances as a group.
The device.