An Add-on to the Minsk-1 Electronic Digital Computer for Fault Detection in the Control Unit
Abstract Generated abstract
The paper addresses fault localization in the control unit of the Minsk-1 electronic digital computer, a component responsible for a substantial share of computation-mode failures and requiring significant diagnostic time. It proposes an attachment used with test programs to monitor selected basic elements by checking the timing sequence of control-unit pulses, using two channels, delay elements, counters, flip-flops, and sequence comparison to register missed or extra pulses and identify the likely faulty element group. The authors report that the circuit was built from Minsk-1 elements and tested on the machine, and they estimate its economic effect with reliability-loss formulas. Under an assumed 20% reduction in control-unit fault-location time, the calculated average unreliability losses fall by about 8 rubles per hour, or 10%, with an annual reduction of about 15,600 rubles.
Full Text
Attachment to the “Minsk-1” Electronic Digital Computer for Fault Finding in the Control Unit
M. N. Strogonov, V. M. Razin
(Presented by the scientific seminar of the Faculty of Automation and Computer Engineering)
The control unit (CU) of the “Minsk-1” electronic digital computer is the most complex part of the machine. Processing of statistical data over a one-and-a-half-year period shows that approximately 30% of all machine failures occurring in computation mode are attributable to the CU of the “Minsk-1” computer. Finding one fault in the CU of the “Minsk-1” computer requires, on average, about 3 hours. Since the CU is structurally implemented according to the block principle, it may be considered that the principal recovery time is the time spent searching for the failed element.
Checking the correctness of the machine’s operation and locating faults are carried out by means of test programs and special measuring-and-monitoring equipment. Test monitoring has a low resolving power. The low resolving power of the tests requires active intervention by the engineering and technical maintenance personnel during the search process. This leads to the fact that the time spent locating faults depends on the qualifications of the maintenance personnel.
In the present work it is proposed, for fault finding in the CU of the “Minsk-1” computer, to use partial circuit monitoring together with test programs; that is, not all elements of the CU are checked, but only a portion—basic elements. A similar method of fault finding, using the basic elements when checking a faulty machine by means of another serviceable machine, is described in [1]. However, another machine that can be used to check the faulty machine is not always available; therefore, it is proposed to use, for checking the basic elements, an attachment that checks the timing diagram of the execution of operations in the computer CU according to the sequence of operation of the elements. If pulses taken from the machine’s test points overlap in time, they are shifted in time relative to one another by delay elements (mercury relays).
The circuit of the attachment consists of two identical channels. One of the circuit channels is shown in Fig. 1. Each channel is divided into two subchannels. Outputs from the CU test points are fed to the inputs of the summing circuits of the subchannels. Positive pulses from the computer test points are fed to the first channel, and negative pulses from the computer test points are fed to the second channel. Signals taken from the input summing circuits or from the input mercury relays are fed to the inputs of the gates (K-2, K-3) of the corresponding channels. A level from the “failure” flip-flop—T-3—is applied to the potential input of the gates. From the gates, the pulses are fed to the pulse-count counter and to the
Fig. 1. Functional diagram of a channel of the attachment for fault finding in the control unit of the “Minsk-1” electronic digital computer.
a sequential circuit [2], which monitors the order in which the pulses applied to the channel pass through.
Before the start of operation, all flip-flops of the “reset” circuit are first set to the initial “zero” state. T-1 is set to the “one” state.
Thus, the sequence-control flip-flops T-1 and T-2 are in different states. These flip-flops will remain in different states while the circuit is operating until some pulse is missed or an extra pulse appears in the timing diagram of the operation being performed. At the moment when the correspondence of the states of the sequence-control flip-flops is disturbed, the comparison circuit for these states operates, and a signal appears at its output which throws T-3 into the “one” state. T-3 blocks the inputs of the subchannels and the “reset” circuit. From the state of T-1, T-2, T-3 and the state of the counter, we find, using previously compiled charts, the group of elements containing the faulty element. The channel circuit remains in the position in which the fault was registered until the operator, by pressing the “reset” button, restores the circuit to its initial position.
If the control unit is functioning correctly, the last pulse of the timing diagram is applied to the channel input and to the “reset” circuit. The “reset” circuit, with a delay for the time required to set the counter, sets all flip-flops to their initial position, i.e., prepares the circuit for operation in a new cycle.
This circuit is assembled from elements of the Minsk-1 electronic digital computer and has been tested on the machine.
The average economic losses due to unreliability of an electronic digital computer may be estimated by the formula [3]:
\[ \overline{W}_{R} = \sum_{i=1}^{m} \left[ \frac{1}{T_{\mathrm{sr}i}K_{\mathrm{epn}i}+\overline{t}_{\mathrm{un}i}} \left(A\overline{t}_{\mathrm{un}i}+B_iK_{\mathrm{epn}i}\right) \right] + W, \tag{1} \]
where \(\overline{W}_{R}\) is the average economic loss due to unreliability of the electronic digital computer,
\(T_{\mathrm{sr}i}\) is the mean time between failures of the \(i\)-th device of the machine,
\(i=1,2,\ldots,m\) is the number of separate devices of the machine,
\(K_{\mathrm{epn}i}\) is the efficiency coefficient of the fault-finding system for the \(i\)-th device of the machine,
\(\overline{t}_{\mathrm{un}i}\) is the average time required to eliminate a fault in the \(i\)-th device of the machine,
\(A\) is the specific loss (loss per unit time) during downtime of the electronic digital computer, equal to the average income that the electronic digital computer would yield if it were operating properly,
\(B_i\) is the average loss for repair and elimination of one failure in the \(i\)-th device of the machine,
\(W\) is the cost per unit time of general maintenance not directly associated with eliminating failures.
The average economic losses due to unreliability of the Minsk-1 electronic digital computer, \(\overline{W}_{R}\) and \(\overline{W}'_{R}\), before and after introducing automatic fault finding in the control unit, can be obtained from formula (1):
\[ \overline{W}_{R} = \frac{1}{T_{\mathrm{sr}1}K_{\mathrm{epn}1}+\overline{t}_{\mathrm{un}1}} \left(A\overline{t}_{\mathrm{un}1}+B K_{\mathrm{epn}1}\right) + \]
\[ + \frac{1}{T_{\mathrm{sr}2}K_{\mathrm{epn}2}+\overline{t}_{\mathrm{un}2}} \left(A\overline{t}_{\mathrm{un}2}+B K_{\mathrm{epn}2}\right) + W; \tag{2} \]
\[ \overline{W}'_{R} = -\frac{1}{T_{\mathrm{sr}1}K'_{\mathrm{epn}1}+\overline{t}_{\mathrm{un}1}} \left(A\overline{t}_{\mathrm{un}1}+BK'_{\mathrm{epn}1}\right) + \]
\[ + \frac{1}{T_{\mathrm{sr}2}K_{\mathrm{epn}2}+\overline{t}_{\mathrm{un}2}} \left(A\overline{t}_{\mathrm{un}2}+BK_{\mathrm{epn}2}\right) +W, \tag{3} \]
where the parameters without an index refer to the entire machine, the parameters with index 1 refer to the control unit, and the parameters with index 2 to the remaining units. The parameters with a prime index refer to the control unit when automatic fault finding is introduced in the control unit of the electronic digital computer.
Taking the difference between formulas (2) and (3), we obtain the value of the reduction in the average economic losses due to unreliability of the electronic digital computer when automatic fault finding is introduced in the control unit:
\[ \Delta \overline{W}_{R} = \overline{W}_{R}-\overline{W}'_{R} = \frac{1}{T_{\mathrm{sr}1}K_{\mathrm{epn}1}+\overline{t}_{\mathrm{un}1}} \left(A\overline{t}_{\mathrm{un}1}+BK_{\mathrm{epn}1}\right) - \]
\[ - \frac{1}{T_{\mathrm{sr}1}K'_{\mathrm{epn}1}+\overline{t}_{\mathrm{un}1}} \left(A\overline{t}_{\mathrm{un}1}+BK'_{\mathrm{epn}1}\right). \tag{4} \]
The total annual economic losses due to unreliability of the electronic digital computer, \(W_R\) before and \(W'_R\) after the introduction of automatic fault finding in the control unit, can be expressed as follows:
\[ W_R=\overline{W}_R\left(T_v+t_{p1}+t_{p2}\right); \tag{5} \]
\[ W'_R=\overline{W}'_R\left(T_v+t'_{p1}+t_{p2}\right), \tag{6} \]
where \(T_v\) is the total annual operating time of the machine in computing mode;
\(t_{p1}\) and \(t'_{p1}\) are the total annual time spent finding and eliminating faults in the control unit of the electronic digital computer that occurred in computing mode, respectively before and after the introduction of automatic fault finding in the control unit;
\(t_{p2}\) is the total annual time spent finding and eliminating faults in the remaining units of the machine.
Taking the difference between expressions (5) and (6), we obtain the average annual gain \(\Delta W_R\) from the introduction of automatic fault finding in the control unit of the electronic digital computer:
\[ \Delta W_R = W_R-W'_R = \Delta \overline{W}_R\left(T_v+t_{p2}\right) + \overline{W}_R t_{p1} - \overline{W}'_R t'_{p1}. \tag{7} \]
The effectiveness coefficient of the fault-finding system is determined by the formula:
\[ K_{\mathrm{epn}} = \frac{\displaystyle\sum_{j=1}^{k} t_{\mathrm{un}j}} {\displaystyle\sum_{j=1}^{k}\left(t_{\mathrm{ot}j}+t_{\mathrm{un}j}\right)} = \frac{\overline{t}_{\mathrm{un}}}{\overline{t}_{\mathrm{ot}}+\overline{t}_{\mathrm{un}}}, \tag{8} \]
where \(t_{\mathrm{un}j}\) is the time spent eliminating the \(j\)-th failure;
\(t_{\mathrm{ot}j}\) is the time spent locating the faulty element in the \(j\)-th failure.
In processing statistical data for the “Minsk-1” machine for 1964, the following data were obtained, presented in Table 1.
The values \(t'_{\mathrm{p}1}\), \(t'_{\mathrm{ot}1}\) (Table 1) are taken on the assumption that the introduction of automatic fault finding in the control unit of an electronic digital computer, on the basis of checking the basic elements, gives a 20% reduction in the time required to locate the failed point.
All calculations carried out using formulas (1)—(8) are presented in Table 2.
Table 1
| \(T_{\mathrm{cp}1}\), h | \(T_{\mathrm{cp}2}\), h | \(\bar t_{\mathrm{un}}\), h | \(\bar t_{\mathrm{ot}1}\), h | \(\bar t'_{\mathrm{ot}1}\), h | \(\bar t_{\mathrm{ot}2}\), h | \(t_{\mathrm{p}1}\), h | \(t'_{\mathrm{p}1}\), h | \(t_{\mathrm{p}2}\), h | \(T_{\mathrm{v}}\), h | \(W\), rub./h | \(B\), rub. | \(A\), rub./h |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 70 | 32 | 0,05 | 3 | 3,2 | 1,1 | 104 | 83 | 63 | 1600 | 15 | 5 | 739 |
Table 2
| \(K_{\mathrm{epn}1}\) | \(K'_{\mathrm{epn}1}\) | \(K_{\mathrm{epn}2}\) | \(\bar W\), rub./h | \(\bar W'_R\), rub./h | \(\Delta \bar W_R\), rub./h | \(\Delta W_R\), rub./h |
|---|---|---|---|---|---|---|
| 0,0123 | 0,0154 | 0,0435 | 82,0 | 74,0 | 8,0 | 15690 |
From the data obtained in Table 2 it is evident that, when the time for locating the failed element in the control unit of the “Minsk-1” electronic digital computer is reduced by only 20%, the average economic losses due to the unreliability of the machine decrease by approximately \(8 \frac{\text{rub.}}{\text{h}}\), i.e., by 10%, while the total annual economic losses due to unreliability are reduced by 15,600 rubles.
REFERENCES
-
A. K. Olefir. On the search for faulty elements in a system of two electronic digital computers. Computing Systems, issue 6, Novosibirsk, 1963.
-
Lee Fred. An automatic Self-checking an fault-locating method. “IRE Trans. Electronic Comput.,” 11, No. 5, 1962.
-
V. M. Razin. An efficiency criterion for a fault-finding system in devices of long-term use. The present collection.