| 摘要 |
<p>952,804. Automatic character reading: selective signalling. STANDARD TELEPHONES & CABLES Ltd. March 30,1962 [April 1, 1961], No. 12286/62. Headings G4H and G4R. An error correcting arrangement for the output of a code conversion matrix e.g. of the kind described in Specification 939,134 is adapted correctly to determine the output lead which should have the highest signal, even though this lead is defective and gives no signal at all. In the form of Fig. 2 the input consists of three parallel bits each applied on true and complementary leads X1, #X1; X2, #X2; X3, #X3. Couplings in the matrix B are arranged, in normal operation, to give outputs on each of the eight leads YO-Y7 the value of which represents the match between the input pattern and the pattern set up on the lead. The input pattern 0, 1, 0 should give a highest output on the "2" lead Y2, the value of the output being 3 i.e. all three digits agree. Other leads have values of 2, if two digits agree, 1 if one digit agrees and 0 if no digits agree. If the lead Y2 or its couplings fail in some way its output may be zero (as shown in brackets). Each lead is applied to a threshold circuit having a level of between 1 and 2 which converts the signal into two-state values for application either directly, or on true and complementary leads to an error correcting matrix P. This generates a new set of eight values which, if the B matrix is working correctly, distinguish the "2" lead with a value of 8 while all the other leads have values of 4 or zero. The defective operation of lead Y2 in matrix B produces the values shown in brackets, the "2" lead still having the highest value but with a smaller margin. The threshold device S2, set say at 6, isolates the lead b2 whether the first matrix is working correctly or not. For larger input combinations it is more economical to split the second matrix into groups. For example with four input bits there are sixteen output leads. These may be divided into two groups of eight and applied to P matrices as in Fig. 2, only one of which will give an output. In the form of Fig. 5 the intermediate threshold devices S1 are omitted, the couplings of the Pa matrix being weighed according to the output patterns expected. Thus in normal operation the 0, 1, 0 combination gives the values shown, the highest value 3 being on the "2" lead. The "2" column lead is therefore coupled by proportional couplings to the eight row leads, thereby to magnify the response to the correct combination of values on the row leads. For correct operation the response on the "2" lead is (adding from upper row lead) 2<SP>2</SP> + 1<SP>2</SP> + 3<SP>2</SP> + 2<SP>2</SP> + 1<SP>2</SP> +<SP> </SP>0<SP>2</SP> + 2<SP>2</SP> + 1<SP>2</SP> = 24. This is higher than the response on any other lead by at least 4. For defective operation the "2" lead response is 15 which is also higher than any other lead but now only by a margin of 1. The weightings of the couplings may be modified and simplified by eliminating all the "1" values so that only couplings of "2" and "3" remain.</p> |
