WO1995011460A1

WO1995011460A1 - Method and apparatus for detecting and removing corrupted data

Info

Publication number: WO1995011460A1
Application number: PCT/US1994/011151
Authority: WO
Inventors: John D. Schlesselmann; Kevin L. Pettijohn; William H. Frye; Mary J. Hewitt
Original assignee: Santa Barbara Research Center
Priority date: 1993-10-22
Filing date: 1994-09-30
Publication date: 1995-04-27

Abstract

In accordance with a method of the invention, and apparatus for accomplishing the method, the following steps are executed: (a) receiving data samples from a data stream; (b) storing a first received data sample; (c) adding a predetermined offset value to a next received data sample to produce an offset data sample; (d) comparing an amplitude of the stored first data sample to an amplitude of the offset data sample; and (e) outputting the next received data sample during a predetermined output period only if the next received data sample differs in amplitude from the first received data sample by an amount that is substantially equal to or less than the predetermined offset value.

Description

METHOD AND APPARATUS FOR DETECTING AND REMOVING CORRUPTED DATA

STATEMENT OF GOVERNMENT INTEREST

This invention was made with Government support under Contract No. DASG60-90-C-0128 awarded by the United States Army. The Government has certain rights in this invention.

BACKGROUND OF THE INVENTION

The present invention pertains to a method and apparatus for detecting and removing corrupted data from a data stream. More particularly, the present invention pertains to an adaptive lowball technigue for detecting and removing corrupted data from a data stream.

A gamma detection circuit capable of detecting and removing corrupted data from data sets of a data stream is useful in strategic and theater defense sensors which may be subject to adverse nuclear environments. For example, gamma radiation is known to cause "gamma spikes" on the data output of a focal plane array of radiation detectors. Conventional techniques for detecting and removing gamma corrupted data include two methods known as a "lowball" method and an "adaptive threshold" method. However, the lowball method suffers from low sensitivity, while the adaptive threshold method is characterized by high complexity.

More specifically, the first conventional method for detecting and removing gamma corrupted data is known as the lowball (or low pick) method. In this method, two adjacent samples of data are compared, and the sample having the lowest value is accepted while the other is rejected. As a result, one of every two samples is discarded, and half of the information content of the original data stream is lost. This method can thus cause unacceptable degradation of the system sensitivity. However, the analog circuitry reguired to implement this method is relatively simple, and can be formed in a small area of an integrated circuit.

In the second conventional method, known as the adaptive threshold method, an average of some number of samples is taken and compared to subsequent incoming samples. If an incoming data sample differs from the average by some defined quantity, then that data sample is rejected. This technique offers the advantage of little or no data loss in a non-corrupted environment. However, the circuitry required to implement this method is relatively complex, and thus requires a relatively large circuit area.

In that the conventional lowball method retains only the lowest of N consecutive samples, the result is a reduced "clear" performance (i.e. 1/N¹² impact on the sensitivity of the method) . Furthermore, the lowball method does not provide an adjustable threshold for optimizing sensitivity according to the magnitude and distribution of corrupted samples. The adaptive threshold method, on the other hand, requires that an un-degraded running sample average be maintained and used for identifying and rejecting corrupt samples. This requires a complex circuit and, therefore, high cost and large circuit area.

Therefore, there is a need for a simple gamma detection method and apparatus which is applicable for detecting any type of single event upset in a data stream, with little or no data loss in an un-corrupt environment. SUMMARY OF THE INVENTION

The present invention is intended to alleviate the drawbacks of the prior art. It is an object of the present invention to provide a method and apparatus for detecting and removing corrupted data from data sets of a data stream. It is another object of the present invention to provide such a method and apparatus for detecting and removing corrupted data, capable of eliminating gamma spikes on the data output of a focal plane array of radiation detectors. It is still another object of the present invention to provide a method and apparatus for removing corrupted data, which can be implemented using analog circuit construction to reduce power consumption and circuit area as compared with conventional techniques. It is still another object of the present invention to provide a method and apparatus for detecting and removing corrupted data having a user definable offset which can be optimized according to the magnitude and distribution of corrupted samples.

In accordance with the present invention, a method for detecting and removing corrupted data from a data stream is provided. The teaching of the present invention is not limited only to detecting gamma effects, but is also applicable in detecting any type of single event upset in a data stream. Also, the present invention can be implemented for detecting differences above an adjustable threshold between two adjacent samples of data. The present invention can further be used to provide thresholding between consecutive frames of data, which is useful for thresholded target detection (i.e., a target moving against a stationary or slow moving background) .

The present invention is less complex than the conventional adaptive threshold method, and can be readily implemented on a focal plane array of radiation detectors, or on an associated integrated circuit that is coupled to the array of radiation detectors. The present invention offers the advantage of preserving all data in a non-corrupted environment. The invention can be implemented in analog circuitry, which allows for the real time elimination of delayed gamma spikes (or any anomalous spikes) on each pixel in a focal plane array of radiation detectors. An analog implementation also consumes significantly less power than a digital circuit.

In accordance with both method and apparatus of the invention for detecting and removing a data sample from a data stream, the removed data sample differing in amplitude from an immediately preceding data sample by a predetermined amount, the following steps are executed: (a) receiving data samples from a data stream; (b) storing a first received data sample; (c) adding a predetermined offset value to a next received data sample to produce an offset data sample; (d) comparing an amplitude of the stored first data sample to an amplitude of the offset data sample; and (e) outputting the next received data sample during a predetermined output period only if the next received data sample differs in amplitude from the first received data sample by an amount that is substantially equal to or less than the predetermined offset value.

The predetermined offset value can be user defined and selected to optimize the performance of the inventive method. If the offset value is selected to be zero, the inventive method collapses to a lowball method.

The present invention improves on the conventional lowball method, which outputs only one data sample from each set of two data samples of the data stream. In accordance with the inventive method, only the second data samples of each set which are greater in magnitude than the first data sample, plus the offset, are removed from the data stream. Therefore, a relatively uncorrupt data stream will not have the quantity of its information substantially reduced. The inventive method also improves on the other conventional method, known as the adaptive threshold technique, which requires that an un-degraded running sample average be maintained and used for identifying and rejecting corrupt samples. The adaptive threshold technique requires more power and a more complex circuit design (and, therefore, a larger circuit area) than does the circuit utilizing the inventive method.

In accordance with the present invention, an apparatus for detecting and removing corrupted data from data sets of a data stream is provided. A first data sample signal and a second data sample signal of a data set from the data stream are received by a receiver circuit. An adder circuit adds an offset value to the first data sample signal to produce an offset data signal. A result is determined by comparing a value dependent on the offset sample data signal and the second data sample signal. An output circuit selectively outputs the second data sample signal depending on the result, so as to detect and remove corrupted data from the data set of the data stream. The determining circuit includes a subtracter circuit for subtracting the offset data sample signal from the second data sample signal to determine a difference. In accordance with the analog circuit construction of the present invention, the subtracter circuit comprises a coupling capacitor that is connected to an input of an amplifier. The result is a function of the difference.

The circuitry of the invention also includes a comparator circuit for comparing the difference with a preset value. The comparator produces the result as an affirmative output result only if the offset data sample is at least equal to the second data sample. Thus, the output circuit outputs the second data sample only if the result is the affirmative output result. In accordance with the analog circuit construction of the present invention, the comparator circuit comprises an amplifier, such as a high gain differential amplifier or an open loop amplifier.

An adder circuit adds a user definable offset value to a received data sample signal to produce the offset data sample signal. Therefore, in accordance with the present invention, if the second data sample signal of a received data set, plus the offset, has a value greater than the first data sample signal, (in other words, if the second data sample signal is corrupted) , it will not be output and is effectively removed from the data stream.

Therefore, in accordance with the present invention, corrupted data is detected and removed from a data stream, using a relatively uncomplicated analog circuit which requires a comparatively small surface area of an integrated circuit. The use of the circuitry of this invention does not result in the 1/N^1/2 deficiency associated with the conventional low ball method, and furthermore requires less power and less complex circuitry, as compared with the conventional adaptive threshold method.

BRIEF DESCRIPTION OF THE DRAWING

Fig. 1 is a schematic drawing of an analog circuit constructed in accordance with the present invention;

Fig. 2 is a flow chart describing the functioning of the circuit shown in Figure 1; and

Figs. 3 and 4 are each a timing diagram that illustrate the operation of the circuit of Fig. 1. DETAILED DESCRIPTION OF THE INVENTION

Referring to Fig. l, in conjunction with the timing diagrams of Figs. 3 and 4, an analog circuit 1 constructed in accordance with the inventive apparatus for detecting and removing corrupted data from data sets of a data stream will be described. In Fig. 3 and 4 the signal names refer to those shown in Figure 1 for activating the various switches, it being realized that in a preferred embodiment the switches are implemented with transistors, such as MOSFETs. Furthermore, the duration of each timing interval in Fig. 3 (e.g., Tl to T2, T2 to T3) is on the order of microseconds, while the duration of each timing interval shown in Fig. 4 is on the order of tens of nanoseconds. As shown in Fig. 3, the total time duration illustrated in Fig. 4 (TO'-TS¹, or TA) occurs within a short period at the initial portion of the timing sequence T0-T5. As employed herein, the total time shown in Fig. 3 (T0-T5) is referred to as a timing frame. Timing frames repeat in a consecutive manner, such that the beginning of TO of a second timing frame follows the end of T5 of a first timing frame.

In Fig. 1 a data stream is inputted to the circuit 1 through a receiver circuit 10 that includes a first XI amplifier 11. By example, the data stream is sourced from a focal plane array (FPA) of radiation detectors (not shown) . Further by example, the FPA may comprise 128x128 individual detector elements, or pixels. The output of each the 16,384 pixels is multiplexed into a single bit-serial data stream that is output from the FPA to an associated integrated circuit that includes a 128x128 array of the circuits 1. The VIN terminal of each of the circuits 1 is coupled in common to the bit-serial data stream. The PEN_DMX signal (Fig. 4) is activated (pulsed low) during TO' in order to cause a particular one of the circuits 1 to input and store, on capacitor 9, a data signal from the bit-serial data stream. As shown in Fig. 4, the PEN_DMX signal for the circuit 1 of Fig. 1 is activated at TO' for causing the associated switch to close and store the voltage appearing at the VIN terminal on the input storage capacitor 9. The second activation of a PEN_DMX signal that occurs at T4' illustrates the activation of another PEN_DMX signal for another one of the circuits 1. As such, the outputs from a plurality N of data sources are multiplexed onto a conductor, and are then demultiplexed from the conductor into a plurality N of the circuits 1 in response to assertions of N associated PEN_DMX signals.

In accordance with the foregoing, a first data sample signal of a data set from the data stream is received at the input of the receiver circuit 10 through the switch activated by the timing signal PEN_DMX. By example, the data sample represents the output of one pixel of the FPA. Previously, a comparator 20 of a comparator circuit 12, and an associated node of an auto-zero capacitor (Caz) 16, has been reset by timing signal PRST, as has a storage capacitor 27 of a latch circuit 26.

After PEN_DMX is deasserted, timing signals PFS and PSH are asserted (Fig. 4). PFS, not shown in Fig. 1, is employed internally by an input amplifier 11, which is constructed as a clocked MOSFET type buffer. PSH is employed to couple the buffered VIN signal to an offset adder circuit 14 and to the input of a second XI buffer 13.

The adder circuit 14 adds an offset voltage VOFF to the first data sample signal by switching, with timing signals PCMP1 and PREF, a reference voltage of the adder circuit 14 between VOFF and ground. The upper node of a sample and hold capacitor 15 is thus switched to the potential appearing at VOFF during the time that PCMPl is asserted, and is switched to ground potential during that time that PREF is asserted (Fig. 3). In accordance with an aspect of the invention, the value of VOFF may be adaptively programmed in real-time so as to accommodate changes in operating conditions that influence the integrity of the input data stream, such as the presence of gamma spikes on the output of the FPA. If VOFF is set to zero volts (no offset) , the circuit 1 will operate in a manner similar to a conventional lowball circuit.

PCMPl being asserted, in conjunction with PEN_DMX being asserted, impresses a voltage across capacitor Caz 16 of a subtracter circuit 18. The impressed voltage is representative of the voltage magnitude of incoming first data sample signal, gated by timing signal PSH from the output of amplifier 11, plus VOFF.

That is, the adder circuit 14 adds the user definable offset value VOFF to the first data sample signal to produce an offset data sample signal.

It is assumed for this discussion that the data sample signal just received is the first data sample signal. As such, the effect of timing signals PCMP2 and PSND will not be discussed at this point. Next, PRST is asserted during T4 to reset the comparator circuit 20 and the sample and hold capacitor 27 of the latch circuit 26. The assertion of PRST, during the time that PCMPl is deasserted and PREF is asserted, causes Caz 16 to store a voltage equal to the value of the first data sample minus VOFF. That is, the upper node of the sample and hold capacitor 15 has been re- referenced to ground potential by the assertion of PREF. As such, the potential stored on Caz 16 is equal to the previously sampled value of VIN minus VOFF.

The timing illustrated in Figs. 3 and 4 then repeats for processing a next data sample. That is, PEN_DMX, PFS, PSH and PCMPl are asserted, and VOFF is added to the second data sample. If the magnitude of the second data sample, plus VOFF, is greater than the potential stored on Caz 16 from the first data sample, then the output of comparator 20 will assume a first state. If the magnitude of the second data sample, plus VOFF, is equal to or less than the potential stored on Caz 16 from the first data sample, then the output of comparator 20 will assume a second state. The first state may indicate that the second data sample has been corrupted by a gamma spike having a magnitude that exceeds the value of VOFF. The second state indicates that the second data sample has not been corrupted, and thus represents a "good" data sample.

That is, Caz 16 serves to subtract the first data sample signal, plus VOFF, from the second data sample signal, and presents the difference D to the input of the comparator 20. The comparator 20 compares the difference with a preset reference value to produce a result R. The comparator 20 may be implemented with, by example, a high gain differential amplifier or an open loop amplifier. In the case of an analog circuit, the preset reference value may be the reference voltage of the comparator 20. As a function of the difference, the comparator 20 produces the result R. The result R is an affirmative output result only if the magnitude of the second data sample signal, bus VOFF, is equal to or less than the magnitude of first data sample signal.

The latch circuit 26 is responsive to the assertion of the timing signal PCMP2 (T1-T2) to store the result R, supplied through amplifier 25, on the sample and hold capacitor 27.

If the stored value of R is the affirmative output result, the magnitude of R activates (closes) the switch 29, thereby enabling the first assertion (during T2) of the signal PSND to pass through to the switch 30 in an output circuit 22. The first assertion of PSEND will thus close switch 30, thereby coupling the second data sample to a sample and hold capacitor 31 and the input of an output buffer 23. The value of the second data sample is then applied to the output terminal VOUT.

It should be noted that, at this point in the timing cycle, PCMPl is deasserted and PREF is asserted. As a result, the magnitude of VOFF is no longer present on the data sample appearing at the output of buffer amplifier 13. Thus, only the potential of the received data sample is applied to the sample and hold capacitor, and not also the potential of VOFF.

Downstream electronics (not shown) , specifically a sample and hold circuit, is activated during the first assertion of PSND to store the value of the signal appearing at VOUT.

If the affirmative output result is not obtained from the comparator 20, indicating that the second data sample, plus VOFF, is greater than the first data sample, then the voltage stored on capacitor 27 during PCMP2 is not sufficient to close switch 29. As a result, switch 30 is not closed by the first assertion of PSND, and the second data sample is not coupled through to the external sample and hold circuit. In other words, a possibly gamma- corrupted second data sample is not passed through to VOUT to be latched by the external sample and hold during the first assertion of PSND.

The second assertion of PSND, between T4 and T5, is provided to refresh the sample and hold capacitor 31 for those cases where switch 30 is not closed by the first assertion of PSND. This mode of operation is advantageous when the input data samples are increasing in amplitude over time, with the increase between two successive data samples being greater than the magnitude of VOFF. For this condition, the circuit 1 will determine that each successive data sample is greater than the preceding data sample, and switch 30 may not be closed, during the first assertion of PSND, for a considerable number of successive timing frames (T0-T5 of Fig. 3) .

The second assertion of PSND occurs after the assertion of PRST. The assertion of PRST places the potential VSS on the sample and hold capacitor 27. The magnitude of VSS corresponds to that of an affirmative output result R from comparator 20, and therefore causes switch 29 to close. As such, the second assertion of PSND during T4 is always coupled through to switch 30, causing switch 30 to close and couple the magnitude of the second data sample to the sample and hold capacitor 31. As a result, during the first assertion of PSND during a time frame where an affirmative output result is not obtained, the downstream sample and hold circuit will sample a value of VOUT which will reflect the magnitude of the data sample of the preceding timing frame. This mode of operation enables a gradually increasing series of data signals to be tracked by the downstream electronics.

Fig. 2 is a flow chart showing the functioning of the circuit 1 of Fig. 1, and will now be used to describe a method for detecting and removing corrupted data from a data stream. As shown, a first data sample Sn of a data set from the data stream is input to the receiver circuit 10 (Step 1) . The user definable offset voltage VOFF is added to the first data sample Sn to produce an offset data sample (Step 2) . In accordance with the analog circuit shown in Fig. 1, at Step 2, the user defined offset value is added to the first data sample Sn by switching the upper node of capacitor 15 to VOFF with PCMPl. This impresses a voltage across the coupling capacitor 16 at the subtracter circuit 18 (which is representative of the offset data sample which is the incoming first data sample Sn plus the offset VOFF) . When the reset pulse (PRST) is applied, VOFF is removed, and Caz 16 a potential equal to that received at VIN minus VOFF. A second data sample Sn+1 of the data set is subsequently received by the circuit (Step 3).

At Step 4, using the comparator circuit 12, the difference (D) is determined, and then the result (R) by the comparator 20 (Step 5) . In the analog circuit shown in Fig. 1, Caz 16 functions to subtract the first data sample Sn from the second data sample Sn+1, plus VOFF, and presents the difference D to the input of the comparator 20.

The comparator 20 switches appropriately as a function of whether the offset second data sample Sn+1 is greater than or less than the first data sample Sn. In either case, R is stored by the latch circuit 26 (Step 6) . The second data sample Sn+1 is selectively output during the first assertion of PSND (Step 7) , depending on the result R, so as to detect and remove corrupted data from the data set Sn, Sn+1 of the data stream.

In accordance with an aspect of the invention, the magnitude of VOFF can be controlled during operation to optimize the response of the circuit 1 to external operating conditions, thereby providing an adaptive gamma correction capability. That is, the magnitude of VOFF can be controlled so as to select a magnitude of a data disturbance that is to be detected.

As was previously stated, the teaching of this invention is not limited only to detecting gamma effects, but is also applicable in detecting any type of single event upset in a data stream. Also, the teaching of this invention can be used for detecting differences above an adjustable threshold (VOFF) between two adjacent samples of data. The teaching of this invention can further be used to provide thresholding between consecutive frames of data, which is useful for thresholded target detection (i.e., a target moving against a stationary or slow moving background) . It should also be realized that the teaching of this invention is not limited to a discrete or an integrated circuit construction, as depicted in Fig. 1, but is also amenable to being implemented with programming instructions executed by, for example, a digital signal processor (DSP) .

Furthermore, for an embodiment where demultiplexing of a data stream is not required, the timing signal PEN_DMX could be eliminated, as well as XI buffer 11 and capacitor 9.

Thus, for the purposes of promoting an understanding of the principles of this invention reference has been made to the embodiment illustrated in the drawings, and specific language has been used to describe these embodiments. It will nevertheless be understood that no limitation of the scope of the invention is thereby intended, there being contemplated such alterations and modifications of the illustrated method and apparatus, and such further applications of the principles of the invention as disclosed herein, as would normally occur to one skilled in the art to which the invention pertains.

Therefore, the foregoing is considered as illustrative only of the principles of the invention. Further, since numerous modifications and changes may occur to those skilled in the art, it is not intended to limit the invention to the exact construction and operation shown and described. Accordingly, all suitable modifications and equivalents may be resorted to, falling within the scope of the invention.

Claims

1. A method for detecting and removing a data sample from a data stream, the removed data sample differing in amplitude from an immediately preceding data sample by a predetermined amount, comprising the steps of:

receiving data samples from the data stream;

storing a first received data sample;

adding a predetermined offset value to a next received data sample to produce an offset data sample;

comparing an amplitude of the stored first data sample to an amplitude of the offset data sample; and

outputting the next received data sample during a predetermined output period only if the next received data sample differs in amplitude from the first received data sample by an amount that is substantially equal to or less than the predetermined offset value.

2. A method according to claim 1, wherein the step of comparing includes a step of subtracting the offset data sample from the first data sample to determine a difference.

3. A method according to claim 2, wherein,the step of comparing further comprises a step of comparing the difference with a preset value, and producing a result as an affirmative output result only if the offset data sample is equal to or less than the first data sample; and wherein the step of outputting outputs the second data sample only if the result is the affirmative output result.

4. A method according to claim 3, and further comprising a step of storing the second data sample without regard for the result.

5. A method as set forth in claim 1, and further including a step of varying an amplitude of the predetermined offset value.

6. Apparatus for detecting and removing a data sample from a data stream, the removed data sample differing in amplitude from an immediately preceding data sample by a predetermined amount, comprising:

means for receiving data samples from the data stream;

means for storing a first received data sample;

means, responsive to a next received data sample, for adding a predetermined offset value to the next received data sample to produce an offset data sample;

means for comparing an amplitude of the stored first data sample to an amplitude of the offset data sample; and

means for outputting the next received data sample during a predetermined output period only if the next received data sample differs in amplitude from the first received data sample by an amount that is substantially equal to or less than the predetermined offset value.

7. Apparatus as set forth in claim 6, wherein the comparing means includes means for subtracting the offset data sample from the first data sample to determine a difference.

8. Apparatus as set forth in claim 7, wherein the comparing means further comprises means for comparing the difference with a preset value, and producing a result as an affirmative output result only if the offset data sample is equal to or less than the first data sample; and wherein the outputting means outputs the second data sample only if the result is the affirmative output result.

9. Apparatus as set forth in claim 8, and further including means for storing the second data sample without regard for the result.

10. Apparatus as set forth in claim 6, and further including means for varying an amplitude of the predetermined offset value.

11. Apparatus as set forth in claim 6, wherein the data stream is output from an array of radiation detectors.