Logic State Analyzer With Sequential Triggering and Restart

Logic state analyzers are used to monitor and record sequences of states that occur in a collection of digital signals in a system under test. A "state" is simply any one of the 2.sup.n logical patterns that n-many digital signals may experience. A sequence of addresses or a sequence of fetched instructions are examples of electrical activity describable as states in a microprocessing environment and that can be monitored by a logic state analyzer to record their "state flow."

To monitor the ongoing sequence of states in a system under test a logic state analyzer samples the electrical values of the signals of interest at times determined by one or more clock signals associated with the system under test. The sampled electrical values obtained are compared to thresholds of selected value and polarity to determine their logical values, each of which will be either true or false, one or zero. Each resulting collection of ones and zeros for a sample is a state in the ongoing sequence of states. It is also simply a binary value that may be stored in a memory. A series of such stored values is a record of the activity occurring in the system under test. Such a record may be termed a trace.

To store an indefinitely long record or trace of such activity would require a corresponding indefinitely large memory. This is impractical for at least two reasons. First, the cost of huge memories is prohibitive. Second, and more to the point, no human being using a logic state analyzer has time to sort through several hundred thousand stored states, or even a few tens of thousands, to find the data that is meaningful for the problem under investigation. Yet this could easily be required if it were necessary to trace an entire high level transaction such as finding the tangent of an angle, or reading a file from a mass storage peripheral. What are needed are ways to reduce the amount of trace data stored to manageable proportions while ensuring that the information in the trace is most likely to be pertinent to the investigation at hand.

Accordingly, logic state analyzers are typically equipped with sufficient memory to store traces that range from a few hundred to a few thousand states. The memories are used circularly. That is, the newest data to be stored is written in the memory location currently containing the oldest data. The size of the memory is then the size of the trace, which then always reflects the last "memory's worth" of activity occurring to that time.

The storage of state data into the memory halts subsequent to the detection of some specified condition in the incoming state data. That condition is called the trigger condition, or simply the trigger. If the storage of state data halts immediately upon the detection of the trigger the trace then represents the activity that preceded the trigger. This can be called an "end-on" trigger. If the storage of state data were instead to continue for one additional "memory's worth" of state data the trace would then represent the activity that occurred after the trigger. This can be called a "start-on" trigger. Typically, the user specifies some number of additional storage operations that is less than one memory's worth, e.g., half that number of locations. This produces a "center-on" trigger. A "center-on" or any other "middle of trace" trigger produces a trace that records both what states led up to the trigger, as well as those that followed it.

One of the ways to help ensure that the trace contains principally the data that is most likely to be pertinent to the investigation at hand is to allow the trigger to represent a condition more sophisticated than simply any occurrence of a specified state. A sequential trigger, for instance, produces a trigger only when a predefined sequence of states has already occurred previous to the trigger state. In terms of a flow chart for the state flow of the system under test, the user can use a sequential trigger to say "trigger when it gets there by going that way." That is, simply getting "there" is not sufficient to create a trigger; getting "there" must also be preceded, in order, by certain other events (states) that represent "going that way." For example, the trigger may be desired upon fetching an instruction from the first address in a complement routine, but only if that routine is called in the context of floating point division, not for floating point subtraction, etc. Thus, sequential triggering allows the user to specify a trigger condition that is not simply a mere place on a flow chart, but is one that has historical criteria associated with it as well.

Sequential triggering is accomplished by equipping a logic state analyzer with a mechanism to allow the user to define a sequence of states to precede a designated trigger state and a sequence detection mechanism to issue an appropriate internal trigger signal upon detection of the trigger state subsequent to the satisfaction of the sequence.

It may also be convenient to equip a logic state analyzer having a sequential trigger with a restart capability. This useful feature allows the detection of a specified state to cancel the progress made to that time toward satisfaction of the sequence, requiring satisfaction to begin afresh.

FIG. 1 illustrates the interactive format specification display.

FIG. 2 illustrates the interactive trace specification display.

FIG. 3 illustrates a trace list display of the stored data states.

FIG. 4 illustrates a trace graph display of the stored data states.

FIG. 5 illustrates a trace compare output display list.

FIG. 6 illustrates the input keyboard.

FIG. 7 illustrates a block diagram of the present invention.

FIG. 8 illustrates the distributed memory addressing of the present invention.

FIG. 9 illustrates the relationship between physical and logical addresses of the distributed memory of FIG. 8.

FIG. 10 is a block diagram of the acquisition system.

FIG. 11 illustrates a multiple pattern recognition unit.

FIG. 12 illustrates a simplified sequential triggering circuit.

FIG. 13 illustrates the measurement and control module.

FIG. 14 illustrates the data format of the data memory.

FIG. 15 illustrates the format of the label format file.

FIG. 16 illustrates the logic flow of the display formatting logic.

Data formatting permits the partitioning of 32 input data channels into parameters of interest. Contiguous data channels which behave as a single parameter may be assigned to one of six labels (A-F). For example, in FIG. 1, illustrating the interactive format specification display, 16 bits of an address bus have been assigned to label "A", 8 bits of a data bus have been assigned to label "D", 1 bit of data on pod 1 has been assigned to label "F", and 7 bits have been left unassigned (labeled "X"). Further specifications and data manipulations are made by referencing these labels. Each assigned label may be independently declared to have a positive or negative "logic polarity" and converted to an independently selected radix which can be binary, octal, decimal or hexedecimal. Further, the slope of the positive or negative clock transition at which time the input data channels are sampled can be selected ("clock slope").

Keyboard entries to the microprocessor 800, as shown in FIG. 16, permit the construction of the lebel format file, shown in more detail in FIG. 15 which, contains the format specification parameters. This is used to process the stored data states in the construction of the alphabetically concatenated ASCII display data file and the graphic display data file. Either of the display data files is subsequently selected and used for display purposes by the display control module 700 and the CRT display 1000.

The assigned input data channels are sampled at the specified clock transitions and are treated as one sampled state. The trace specification defines which of the sampled states are to be stored for display and which sampled states are to be counted for count measurements. The trace specification comprises a definition of state conditions specifying the trace position, the selective trace, and the count measurement. Each state condition defines a state of the assigned input data channels in any combination of 1's, 0's, and/or X's (don't care). In octal, decimal or hexedecimal bases the definition is defined in terms of the appropriate alphamumerics and X's.

A trace position may be selected to a start, center or end the selective trace in response to the input data satisfying a predefined state sequence. In this description it will be assumed that the trace position starts the selective trace. A state sequence of up to seven state conditions must be satisfied in a specified order, ignoring intermediate states which do not satisfy the state sequence. The simplest state sequence is a single state condition. Specific segments of branched, looped or nested forms of state flow may be directly located by properly defined state sequences. In addition, each state condition in a state sequence may be required to occur from 1 to 65536 times before the state condition is satisfied. This form of positioning will locate the nth pass of a loop beginning at a given state condition. Clock delay may be incorporated by defining the nth occurrence of any state (an all don't care state specification). The trace logic may also be specified to restart the satisfaction of the predefined state sequence if it is not satisfied before or concurrently with the location of a predefined restart state condition. A restart on "any state" requires that the state sequence be satisfied without any unspecified intermediate states. For example, FIG. 2 illustrates the interactive trace specification display for a trace position starting upon the satisfaction of 4 state conditions in sequence. A restart state condition is also defined.

The selective trace is a qualification to determine which sampled states will be stored for display. One to seven state conditions may be "OR" specified for collection. Selectively tracing only sampled states of interest eliminates the clutter of unnecessary states and magnifies the apparent size of the trace beyond its 64 terms. Also, an occurrence term may be specified so as to store only every nth satisfaction of an "OR" specified state condition. FIG. 2 illustrates the selective trace of every occurrence of a single state condition.

The count measurement performs a "time" or a "state" count associated with each of the (64) states stored and can be displayed in one of two formats:

absolute--the count from the trace position

relative--the count from the previous trace state

The time count is performed by counting the occurrences of an internal clock between sequentially stored states and the display is in the units of seconds. A state count similarly counts the number of occurrences of a specified state condition ("count") between sequentially stored states. For example, specifying "any state" would result in a count of the selected clock transitions of the input data. In FIG. 2, a state count is performed on the occurrences of a specified state condition intermediate to each sampled state stored.

One complete measurement of 64 sampled states, which includes the sampled states satisfying the state conditions defining the state sequence and specifications of the format, trace, and display, may be internally stored. This "current measurement" may be stored or exchanged with a "stored measurement" for later analysis. A "trace compare" mode of operation (described more fully below) compares results of a previously stored trace with the current measurement and may be utilized as a further qualifier on data storage.

The output display format of the current measurement may be selected from a trace list, a trace graph, or a trace compare.

A trace list, illustrated in FIG. 3, displays a listing of the stored states in their order of occurrence. Twenty trace states, (one per line) are simultaneously presented on the CRT display. The "ROLL" keys allow scanning of the 64 stored states. Each line comprises a line number, the stored state alphabetically sorted into assigned labels in their numerical base, and the time or state count if selected.

A trace graph, as shown in FIG. 4, presents a graph of the data magnitude of a specified label versus the storage location for all 64 stored states. Each state is given a vertical displacement corresponding to its binary magnitude and an increasing horizontal displacement for successive states in order of their occurrence. The result is a waveform analogous to oscilloscope displays of voltage magnitude. The label to be graphed is selected by specifying the "graphed label". Scaling of state magnitude is controlled by specifying the "upper limit" and "lower limit" on the vertical axis. Limits can be specified directly or dynamically varied with logrithmic autoranging controls. These facilities allow any portion of a graph to be magnified to a full scale presentation. The 20 points corresponding to the lines viewed in the trace list are intensified. The intensified portion also responds to the "ROLL" controls, and their corresponding absolute value may be read in the trace list.

A trace compare as illustrated in FIG. 5 presents a tabular listing of the difference between results in the "current measurement" and the data in the "stored measurement". The listing is formatted and rolled as in the trace list. The results of the two measurements are exclusive OR'ed such that identical corresponding bits are displayed as zeros and unequal bits are displayed as ones. In an octal base a "03" is equivalent to a binary "000 011" and indicates that the right two bits are different in the two measurements. Trace compare also offers a "compared trace" mode which reruns a measurement until the current and stored measurement are either equal or not equal. (STOP=, or STOP.noteq.) For example, in FIG. 5 of the instrument has rerun trace measurements until the "current measurement" equaled the "stored measurement", as indicated by the "STOP=" specification and revealed by the array of "0"'s in the comparison.

Three trace mode options are provided. "Trace" executes a single current measurement. "Continuous trace" repeats the execution of a current measurement continuously. "Compared trace" repeats the execution of a current measurement until the desired comparison with the stored measurement is obtained.

A trigger output provides a triggering pulse for external instrumentation such as oscilloscopes. A 50 ns pulse is generated each time the trace position is found. The clock enable output is useful for gating clocks or interrupting the device under test. A high signal level indicates that the instrument is actively searching for the trace position. It remains at the high signal level until the trace position has been found or the halt key is depressed. Both outputs are suspended when the format specification is displayed to allow measurement of channel activity.

Referring to FIG. 6, an illustration of the keyboard, the keys are functionally segregated into four blocks, the "current measurement display", "entry", "edit", and "execute". A power up sequence initially defines a default set of specifications, displays the default format specification, then automatically selects a hexadecimal trace list display. Activation of the "ROLL DISPLAY" keys permits the presentation of any portion of the 64 states stored. To change the format specification, the "FORMAT SPECIFICATION" key is pressed. The cursor keys in the edit block are used to move the cursor, designating a selectable entry field by a blinking inverse video field on the interactive display.

The trace specification can be edited by selecting the trace specification interactive display by activating the "trace specification" key. Editing is accomplished in the same manner as the format specification is edited. A general description of the functions of the individual keys is given in Appendix A. A detailed description of the interactive display entry fields is given in Appendix B.

Input states are sensed through 32 high impedance variable threshold data probes at rates up to 10 MHz. The data probes 100, illustrated in FIG. 7, are segmented into four 8 bit data pods and a fifth pod for clock sensing. Each pod may be preset to TTL logic threshold or variably adjusted in the range of +10 to -10 volts to interpret input logic levels.

The 32 input data channels and the clock signal from the data probes 100 are input to the state recognition module 200. An internal sampling clock is generated in response to the selected clock slope, the input data signals are compared to the selected threshold voltages and interpreted, and the data signals are latched in response to occurrences of the internal sampling clock. The state recognition module 200 outputs the sampled state to the high speed acquisition system bus 500. The index module 300 accesses the sampled state on the acquisition system bus 500, compares the sampled state to the selected state conditions and determines the trace position, selective storage events and state count events. The measurement control module 400 also accesses the acquisition system bus 500 and stores state or time counts and sampled data states in response to the events detected by the index module 300.

The modules of the acquisition system 250 communicate with other system modules via the communications bus 600, which provides a means for addressing selected modules and for transferring selected data. The entire system functions as a distributed memory, as illustrated in FIG. 8. For instance, addresses between 1800 and 1FFF on the communications bus 600 access the state count measurements and the sampled data states stored in the measurement control module 400 memories. FIG. 9 shows another representation of the system architecture, illustrating the relationship between the physical couplings of FIG. 7 and the logical addresses of FIG. 8.

Referring to FIG. 10, the index module 300 detects the trace position by first comparing the sampled state on the acquisition system bus 500 with a qualifier state condition stored in the multiple pattern recognition unit 315. The multiple pattern recognition unit 315 comprises a digital pattern triggering circuit as described in the copending patent application entitled "DIGITAL PATTERN TRIGGERING CIRCUIT", U.S. patent application Ser. No. 743,188 filed Nov. 19, 1976, by WILLIAM A. FARNBACH now U.S. Pat. No. 4,100,532. As illustrated in FIG. 11, the multiple pattern recognition unit 315 comprises 2 pairs of 8 sixteen by four bit memories providing for the detection of up to eight qualifier state conditions, where each qualifier state condition is identified by a 1, 0, X input, format (in binary). Pattern selector 325 of FIG. 10 selects one of the eight lines output from the multiple pattern recognition unit and passes the selected output to the occurrence counter 345. The occurrence counter 345 counts the occurrences of the selected qualifier state conditions and provides an output in response to counting a specified number of occurrences of the selected qualifier state condition. This output is termed a "break event" and the sequencer logic 350 in response requests the pattern selector 325 to select the next sequential qualifier state condition and requests the occurrence counter 345 to select the corresponding count. The sequencer logic 350 also outputs a "N-1" event flag in response to detection of the occurrence of the "NEXT TO LAST BREAK EVENT". A simplified sequential triggering circuit is illustrated in FIG. 12 where the multiple pattern recognition unit 316 incorporates the functions of the multiple pattern recognition unit 315 and of the pattern selector 325. The sequence logic 351 incorporates the functions of the sequence logic 350 except that the final trigger is output in response to the completion of the state sequence. Another method of implementing the multiple pattern recognition unit 316 would be to have 3 selector bits be the most significant bits in the address, allowing the comparator to sequence through various segments of memory when comparing sequential state conditions of the state sequence.

Referring again to FIG. 10, the selective trace is incorporated in a similar manner except that the trace selector 320 of FIG. 10 can "OR" any combination of the AME lines. A trace occurrence counter 340 outputs a trace event flag upon counting each "nth" "ORED" AME event.

The restart unit 310 causes the sequence logic 350 to restart the satisfaction of the state sequence subsequent to the detection of a selected restart state condition. The restart unit is disabled for the data state corresponding to the detection of a break event by sequencer logic 350 which permits the state sequence to be satisfied without any unspecified intermediate state by setting the restart state condition to "any state".

The state count unit 305 strobes a counter in the measurement control module 400 each time the selected state condition to be counted is detected.

The measurement and control module 400 is illustrated in FIGS. 10 and 13. The event flags from index module 300 are input to the high speed control 460 and determine which sampled states on the acquisition system bus 500 are to be stored. The high speed control 460 addresses the data memory 410 and the count memory 420 accordingly. FIG. 14 illustrates the data format of the data memory 410. The sampled state conditions resulting in break events are sequentially stored in locations 1-(N-1). Upon detection of the "N-1" event flag, sampled state conditions are sequentially written into the remaining memory locations, writing over the oldest data when the memory is filled. The trace position address of the memory location containing the state condition resulting in the final trigger is stored in a register and sampled states are written into the appropriate number of remaining storage locations. For example, if the trace was defined to end on the detection of the trace portion, no sampled states would be written subsequent to the detection of the trace position. The order of occurrence of the stored data is easily reconstructed by recovery of the trace position address appearing on the communications bus 600 as illustrated in FIG. 8. Count selector and synchronizer 450 controls the measurement counter 430, whose contents are stored in count memory 420 upon update of the memory address. The low speed control 480 provides a low speed interface for programming the high speed control 460 and for selecting and latching data for the communications bus 600 interface.

The strobe generator 400, illustrated in FIGS. 10 and 13, generates a sequence of strobes which, when coupled with a series of data latches (not shown) and timing logic (not shown) effectuate the orderly performance of machine tasks. In effect, a number of sampled states are simultaneously in various stages of processing at any one time and are "pipelined" through the required logic blocks.