JPS6119286A - Correction circuit of semiconductor linear image sensor - Google Patents

Abstract

PURPOSE:To correct a drift and to detect a subject with high accuracy by fetching odd or even pixel signals whichever is larger, from the 1st memory, then fetching a signal corresponding to the smaller pixel from the 2nd memory. CONSTITUTION:A detection signal V from a linear image sensor part 6 is A/D converted 11 and stored in a medmory 14, while an average Voffset of adjacent pixel is calculated in a CPU15 as shown in an expression I. The Voffset is obtained and added 16 at every line of (N+1) pixels, and stored in a memory 14A. Data B0-BN on (N+1) number of pixels are stored in the memory 14A in the same period as shown in an expression II. Assuming that the level difference between odd and even pixels among several frames is approximately constant delta and levels V0-V2 of even pixels are larger by delta than odd pixels V1, V3..., the even pixels V0, V1... from the memory 14 and B1, B2...=V1+delta, V2+delta... corresponding to the odd pixels from the memory 14A are transmitted to the CPU15, and their positions are detected. The higher levels of the detection signals are approximately equal, and the adverse influence caused by the drift can be eliminated.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a correction circuit for a semiconductor linear image sensor, and in particular, a correction circuit for a semiconductor linear image sensor suitable for use in highly accurate positioning of semiconductor exposure equipment, etc. Regarding.

[Background of the invention]

FIG. 1 shows the semiconductor linear image sensor section 6 and its peripheral circuits. Here, the linear image sensor 1 is arranged so that every other pixel is divided into odd and even pixels, so that the outputs can be taken out separately.

The clock pulse generator 2 alternately applies a set of clock pulses ψ1° to the odd picture element and the even picture element, whereby the even picture element Vt and the odd picture element ■ are transmitted from the sensor 1 to the even picture element Vt.
0 are output alternately to the buffers 51, 52 . Synthesized in buffer 5 via analog switch l, 42,
Output as one video signal ■. In addition, the clock pulse generator 2 outputs a clock 7 and a start signal 8, which are supplied to the signal V processing system. In such a conventional configuration, not only the buffers 51 and 52 but also the analog switches 41 and 42 are actually composed of semiconductor elements, so if fluctuations or offsets in their gains occur, the even system and odd system Imbalance may occur in the pixel outputs, resulting in an output difference δ as shown in FIG. 2 (α).

The detected waveform at this time appears to be a double signal as shown in FIG. 2Cb). Conventionally, such a signal V is processed by a clock 7 and a clock 7 delayed by a delay circuit 12, as shown in FIG.
D, the sample and hold circuit 10 samples and holds the data, the A/D converter 11 digitizes it and stores it in the memory 14, and this stored data is sent to the processing device (CPV).
) I imported it into 15 and filled it in IA. Naori Unta 1
5 is initialized by the start pulse 8 from the converter generator 2, counts the clock signal 7A of the A/D converter, and uses the counted value to determine the address of the memory 14 where the A/D conversion output is to be stored. is specified. By the way, when such a linear image sensor is used, for example, to detect the position of a circuit pattern in an aligner for semiconductor manufacturing equipment used to manufacture fine circuits, an error of one pixel will occur due to the level difference between the odd and even systems. this·
is a large deviation for precise positioning as described above,
This can be a fatal flaw in semiconductor performance.

[Purpose of the invention]

An object of the present invention is to provide a correction circuit for a semiconductor linear image sensor that eliminates pattern position detection errors due to drift within the linear image sensor circuit and enables highly accurate position detection.

[Summary of the invention]

The characteristics of the elements constituting the peripheral circuit of a linear image sensor fluctuate mainly due to the influence of temperature, etc., causing a drift in the circuit output signal, and this fluctuation is slow in terms of time. Therefore, in the present invention, the signals of adjacent odd picture elements and even picture elements are converted to 1/l/f), and the average of the differences is set to a certain 1.
It is calculated from the data between frames (7 cycles/1 scan period), and in the next frame, this average value is added to the odd and even low level signals, and the result is combined with the high level signal. This is characterized in that the combinations are extracted as detected waveforms.

[Embodiments of the invention]

The present invention will be explained below using the embodiment shown in FIG. In the figure, the difference from the conventional configuration in Figure 1 is the mesori 14.
In addition, a memory 14A and an adder 16 are added, and the processing content of the processing device 15 is different from the conventional one. That is, the detection signal V from the linear image sensor section 6 is sampled and held, then digitalized by the A/D converter 11, and stored in the address of the memory 14 designated by the counter 16. Up to this point, the process is the same as the conventional one, but in the present invention, instead of immediately processing the data Vo to ■N stored in the memory 14 as a detection signal, the data stored in the memory 14 is first processed by the processing device 15. The average value of the difference between adjacent picture elements Voffsgt
is calculated according to the following formula.

However, N+1 is the number of picture elements for one line, and one line (
frame) every 11CVoffJ? gt is calculated, the result is set in adder 16, and in the next frame, adder 1
6 adds the set 'Jofjxtt to each picture element of the output V and stores it in the memo 114A. During this time, the processing device 15 simultaneously executes the calculation of equation (1) for the long sword at that time. Therefore, the data of N-) 1 pixel stored in the memory 14 in one frame period is the output V(+% VN) from the image sensor unit 6 as shown in Figure 4.
However, the data of N+1 pixels stored in the memory 141 during the same frame period (3o - (3N is B for voffεet calculated by formula (11) one frame before)
'=A) + Voffztt,)'=o ++
N... f21. Temperature fluctuations that cause drift change much more slowly than this one-frame period, so the level difference δ between odd picture elements and even picture elements over several frames is
(Fig. 2) can be regarded as almost constant, so Vo of the formula fll
ffzet can be considered to be approximately δ. That is, Voffget
N = -Σ (δ, 1st edition δ21+...). Therefore, now N t'L = 1 Level of even picture element "' m etc. is the level of odd picture element, ~'3....If we assume δ Dogedogu, then the memory is as shown in Figure 4. From 14 onwards, Even Emoto Atsushi, Hato,...
..., from memory 14A to the Odo-picture compatible one,...
... = +δ, v3+δ, ... processing device (CPV
) 15 and performs processing such as position detection. Conversely, if the level 'I IVs+... of the odd picture element is higher, the remaining *Vl *Bll y■8 s to the processing device 15
... are taken in sequentially from menus 14A and 14. In this case, the detection signal becomes almost equal to the higher level shown in FIG. 2 (α), and does not have a double waveform as shown in FIG. 2 (h), so that the shape position of the target object can be detected with high accuracy.

〔Effect of the invention〕

As is clear from the above explanation, according to the present invention, pattern detection position errors caused by detection signals can be eliminated, and when applied to semiconductor manufacturing equipment aligners, the alignment accuracy can be improved and the yield of products can be improved. There is.

[Brief explanation of drawings]

Figure 1 is a diagram showing a conventional linear image sensor and its peripheral circuit, Figure 2 (α) is a diagram showing a state in which a difference occurs between odd and even video signals due to analog circuit drift, and Figure 2 (z ) is a diagram showing the pattern detection signal in the state of FIG. 2 (α), FIG. 5 is a circuit diagram showing an embodiment of the present invention, and FIG. 4 is an example of the order in which data in the memory is taken into the processing device. FIG. 6...Linear image sensor section +4, 14A...Memory 15...Processing device 16...Adder

Claims (1)

[Claims] an average value calculation means for calculating the average value of the difference between adjacent odd pixel signals and even pixel signals of the linear image sensor over one frame of the linear image sensor; 1st to store
and a second memory that stores the result of adding the average value calculated by the average value calculating means one frame before the current frame to each of the output pixel signals of one frame of the linear image sensor. At the same time, the larger one of the odd picture element signal and the even picture element signal is taken out from the first memory, and the signal corresponding to the smaller one is taken out from the second memory. A correction circuit for a linear image sensor, characterized by having a function of correcting a drift occurring in an output path of a pixel signal.