US3028958A - Sorting machine - Google Patents

Description

April 10, 1962 w. H. P. PoULlART ET Ax. 3,028,958

SORTING MACHINE 4 Sheets-Sheet 1 Filed April 13, 1959 G. VAN MEC N By Attorney By l o Alorney SORTING MACHINE 4 Sheets-Sheet 2 W. H. P. POULIART ET AL April 10, 1962 Filed April 13, 1959 W.H.P. POULIART G. VAN MECHE April 10, 1962 W. H. P. POULIART ET Ax. 3,028,958

SORTING MACHINE Filed April 13, 1959 4 Sheets-Sheet 4 @1C PBC 00,2 t *i 4' :\"SV 535 CMP @S4 4 swg Inventor W.H.P. POULIART- G. VANMECI-IE By Attorney April 10, 1962 w. H. P.. PouLlART ETAL 3,028,958

SORTING MACHINE 4 Sheets-Sheet 3 Filed April 13, 1959 CMP W.H.P. POULIART- G. VAN MEC A ttorne y United States Patent @ddee 3,928,958 Patented' Apr. 1o, 1962 3,U28,958 SORTING MACHINE Willy Hortense Prosper Pouiart and Guillaume Van Mechelen, Antwerp, Belgium, assiguors to International Standard Electric Corporation, New York, N.Y., a corporation of Delaware Filed Apr. 13, 1959, Ser. No. 805,860 Claims priority, application Netherlands Apr. 16, 1958 12 Claims. (Cl. 209-72) The invention relates to a sorting machine. More particularly it concerns a sorting machine of the general type disclosed in the commonly assigned U.S. application of l. Young, Serial No. 636,309, tiled January 25, 1957, and which is adapted to the sorting or classification of articles constituted by substantially flat documents such as cheques, punched cards, or any substantially flat individual carrier able to enclose or support such types of documents.

The sorting machine disclosed in the above patent application is provided with a plurality of sorting stages which each have two inputs and two outputs, the two inputs ofv the iirst stage of the machine constituting the two inputs of the complete machine, while the two outputs of the last sorting stage constitute the two outputs of the complete sorting machine, the two outputs of any stage except the last feeding the corresponding inputs of the next stage except the iirst. Input and output stores are provided for each stage to hold the documents, and they are constituted by continuously advancing conveyors carrying the documents 4stacked against one another. At each stage, a sorting mechanism is provided so that the first document at one input may be compared with the first document at the other input and one of these may then be made to advance towards one or the other output of the .stage concerned.

ln this manner, sorting by merging can be accomplished. Alternatively, sorting by digits could also be achieved with such a machine and more precisely, sorting by binary digits since the machine and each of its stages have only two outputs. lt will be recalled that sorting by merging consists in merging sequences of like rank with one another, the merged output sequences being alternately distributed to one and the other output. By sequence, one should understand a succession of documents of which the characteristic numbers which determine the sorting, are already ordered in the desired manner. In this way, the sequences will gradually grow in size and the total number of sequences will gradually diminish until one is left with only two sequences which will then be merged into the single desired final sequence. On the other hand, sorting by digits is performed also by a series of sorting passes but during the first, the documents are distributed to one or the other of the two outputs in accordance with the value of the binary digit of lowest rank part of the sorting number. The documents so sorted will then go through a second sorting pass and this time they will be `distributed to one or the other output in accordance with the value of the binary digits having the next to last lowest rank, and so on.

In the rst case of sorting by merging, it is the total number of documents which determines the number of sorting passes required and hence the number of sorting stages in a multiple-stage sorting machine. The number of sorting passes required is at most equal to logz N, or rather the next higher integral value of this expression, where N represents the total number of documents. O11 the other hand, the number of sorting passes required when sorting by digit and more particularly when sorting by binary digits, is simply equal to the number of binary digits used to characterise each sorting number. Thus,

whenever the number of binary bits required to characterise the sorting numbers is rather high, e.g. for sorting numbers having ten decimal digits, sorting by merging will prove to be advantageous. one thousand documents to `be sorted, only ten lsorting passes are at most necessary.

ln a multi-stage sorting machine operating on the principle of sorting by merging, such as disclosed in the above mentioned patent application, there are nevertheless some disadvantages. indeed, the number of stages depends on the total number of documents to be sorted at one time and if this is not always substantially the same, the sort will sometimes already be completed after the documents have passed through only `some-of the stages of the machine. Then, means must be provided to extract the sorted documents at some arbitrary intermediate stage of the machine, or alternatively, the sorted documents have then to go through the remaining stages of the machine without being actually sorted, the remaining stages merely passing the documents from an input to an output.

Moreover, `such an inefficient and awkward use of the multi-stage sorting machine may also occur even if the number of documents constituting a lot to be sorted is substantially constant. passes depends on the number of sequences initially present in the Unsorted lot and this number of initial sequences cannot be controlled. One may in fact reason that the average length of an initial sequence is of two documents. indeed, without making au accurate study of the probabilities, the sorting number of a document may oe said to have an equal chance of either being higher or lower than the sorting number of the immediately preceding document. Then, on the average, one might sa'y that the sequences will be of two documents, since documents having alternately a sorting number higher or lower than that or the immediately preceding one follow one another.

Thus, on the average one may count on the number of initial sequences being half the number of documents whereby one sorting pass will be saved. But, one cannot use this reasoning to omit the last stage of the sorting machine, since this corresponds merely to average conditions, and in some cases the number ot' sequences will be higher and the last sorting stage will be needed.

Moreover, during the sort, what is known as accidental merging may occur and this will also reduce the number of sorting passes required since when accidental merging takes place, a sorting pass reduces the number vof output sequences to less than half the number of input sequences. lt will be recalled that, in the case of lower sorting number in front, accidental merging occurs when the highest or last number of a pair of sequences of like rank about to be merged are both lower than the lowest number of the pairs of sequences having the next but one rank.

rthe general object of the invention is to realize a sorting machine which is more llexible than the multi-stage sorting machine hitherto known, and which does not possess its disadvantages while offering many advantages, particularly for sorting by merging.

In accordance with a iirst characteristic of the invention, a sorting machine in which each sorting stage has two inputs from -any of which an article to be sorted may be directed to one out of two outputs, is characterised by the fact that said machine comprises two stages with the outputs of each stage feeding respectively the inputs ot' the other stage.

A machine of the type characterised above has many advantages over the earlier multi-.stage machine. Its outstanding advantage is the flexibility of the arrangement since the machine may be designed for a maximum load of documents to be sorted but then, smaller numbers of indeed, with batches of indeed, the number of sorting documents may be sorted at any time without the drawbacks of the multi-stage sorting machine. Moreover, in a multi-stage sorting machine the document stores beween the successive stages will need to be designed so that they can accommodate an ever increasing number of documents, whereby the stages are not all identical with one another. Such a problem does not occur in a twostage sorting machine as characterised above and the two stages may be entirely identical. Also, due to the limited number of stages, it is clear that faults will be much easier to cure. Admittedly, in a two-stage sorting machine, only two stages may be working at the same time, whereas in a ten-stage sorting machine for instance ten stages may be working at the same time. But, before each stage in the ten-stage sorting machine is effectively engaged in sorting documents, some time will be lost. This may be called the loading time for each stage. lt is clear that in a two-stage sorting machine these time losses will be less. For a given amount of sorting work, it will be necessary to provide more two-stage sorting machines than ten-stage sorting machines. But again, this arrangement is advantageous since a faulty two-stage sorting machine will not be such a great inconvenience as a faulty ten-stage sorting machine which normally has to do more work.

Another object of the invention is to provide a sorting machine of small size.

Indeed, with a multi-stage sorting machine involving for instance ten consecutive stages, these have to be cascaded one behind the other and this will mean an exceedingly long machine.

In accordance with a further characteristic of the invention, a sorting machine of the type already defined is further characterised in that four parallel transverse conveyors are provided to pass the articles to be sorted between the outputs and the inputs of the two sorting stages of the machine, the actual sorting stages using edgewise conveyors to interlink the transverse conveyors and the edgewise conveyors of one stage being parallel to those of the other and perpendicular to said transverse conveyors linking the two stages.

Thus, such a construction as characterised above is substantially rectangular and uses a minimum volume, an awkwardly long sorting machine being avoided.

lt may be remarked that in the commonly assigned U.S. application of G. Van Mechelen, Serial No. 635,884, filed January 23, 1957, and now Patent No. 2,987,705, a two-input two-output sorting machine consisting of only one sorting stage was already envisaged.

In such a machine however, the documents coming from the two stores which would act simultaneously as joint input and output stores would have to be returned from one end of the store to the other after having passed through the distributing mechanism deciding to which of the two stores the documents have to be sent and in what order. In this manner, however, very little mechanisms will be saved and little place will be gained in View of the necessity for the return journey without the documents undergoing an actual sorting operation. Thus, the slight economy which might eventually be made would, in comparison with the two-stage sorting machine now proposed, be substantially offset by the loss of efficiency due to there being only one sorting stage working at the sante time, instead of two as in the sorting machine now proposed. Moreover, the two-stage sorting machine has the advantage of symmetry which facilitates its construction and maintenance.

The above and other objects and characteristics of the invention and the best manner of attaining them will be better understood from the following description of the invention to be read in conjunction with the accompanying drawings in which:

FlG. l represents diagrammatically a two-stage sorting machine in accordance with the invention;

FIG. 2 `represents synchronising circuits used in the control of each stage of the sorting machine;

FIG. 3 represents circuits adapted to the control of the rst sorting stage of the machine used as input stage;

FIG. 4 represents circuits additional to those of FIG. 3 and used for the control of the other sorting stage of the machine.

Referring to FIG. l, the latter diagrammatically represents a two-stage sorting machine in which each stage is provided with two inputs and two outputs, the outputs of each stage respectively feeding the inputs of the other stage.

The two sorting stages are essentially identical and may be termed normal sorting stage NS and reverse sorting stage RS. This terminology, which will be used in the remaining part of the description, may be justilied by the fact that the documents or document carriers bearing magnetically `encoded signals and which may be read as the document goes through a sorting stage, are preferably only subjected to a translation movement. Hence, if the necessary relative displacement between the document or document carrier bearing the magnetic signals and the reading magnetic heads is obtained by using fixed magnetic heads reading while the document or document carrier is in motion, reading of the serially inscribed magnetic signals will necessarily take place in reverse sense for the two stages. But, the physical arrangements of the two stages are quite identical, the differences may only occur in the controlling circuits.

As shown, each stage includes two input positions such as A4 and B3 for the normal sorting stage NS, the corresponding input positions for the reverse sorting stage RS being respectively labelled A4 and B3. In these input positions, the documents or document carriers are supr ported in a vertical plane by resting on their bottom long edges and they will be displaced in parallel paths by means of edgwise conveyors until they reach the so-called merging positions A1, B1 and A1, B1. In so doing, the document inserted in input position B3, for example, will pass through the reading position B2 which is provided with a reading magnetic head arrangement. Likewise, during its edgewise travel to reach the corresponding merging position A1, the document inserted in the input position A4 will pass through a similar reading position A2, but prior to this it will go through an intermediate position A3. 'Ihe latter position is located behind the input position B3 as shown on FIG. l and is necessary in view of the input positions A4 and B3 being shifted with respect to one another in order to permit the documents to be fed to these two input positions from two transverse parallel conveyors A and B which are perpendicular to the two edgewise conveyors.

The reading positions A2 and B2 need not necessarily be provided as special positions having a length corresponding more or less to that of the document. They would only be required if the documents were read while they were at rest with the help of displaceable magnetic head arrangements. Hence, from the input positions such as B3, the documents might immediately reach the merging positions such as B1 although the additional intermediate position A3 would still be necessary to take care of the fact that the documents going through input position A4 must travel a longer distance through the edgewise conveyor than those admitted through input position B3.

With the help of the two merging positions A1 and B1, the documents are afterwards made to follow a single path through a single channel edgewise conveyor which will successively dispatch a document to the output position C, the output position D and eventually the ejecting position E. From the output position C, a document may be prevented from travelling farther into output position D and instead be displaced on the output transverse conveyor B' which corresponds to the input traverse conveyor B (which respect to the normal sorting stage NS) and which leads to the input position B3 of the reverse sorting stage RS. Likewise, a document having reached the output position D may be prevented from further travelling into the ejecting position E, and instead 4be placed on the output transverse conveyor A' which constitutes the input transverse conveyor for the input position A4 of the reverse sorting stage RS. It a document is not diverted either towards the transverse conveyor B or towards the transverse conveyor A', it will be dispatched to the ejecting position E. from which it will fall into a reiect bin.

When a document is allowed to leave the reading position such as A2 to go through the merging position A1 and then reach the output position C, as it is allowed to leave, the document eventually located in the reading position B2 stays put and it is only as the first document is allowed to leave the merging position A1 that this second document will -be able to enter merging position B1. In this way, the documents in the output positions C and D will follow one another, either coming from one or the other input, just as the documents follow one another in the input path such as A4, A3, A2 and A1.

The principle ot' sorting used in such a two-stage sorting machine is preferably to continuously merge sequences of documents of like ranks and to distribute them'alternately to one and the other output, eg. C or D. By sequence of documents it will be recalled that one should understand a series of successive documents which happen to be already ordered in the required order. For example, in the case of bank cheques, the sorting may be based on the account numbers, it being desired that at the end of the complete sorting operation, the cheques initially fed to the sorting machine in any arbitrary order for their account numbers, should finally be found with all the account numbers in numerical order and for instance with the lowest account number in front.

Initially, the twin continuously advancing transverse conveyors A and B constituting respectively the first and the second input of the normal sorting stage NS may be loaded with the documents to be sorted, approximately half the total number of documents being loaded on each of these two transverse conveyors. The documents will be made to advance continuously along both paths so that the lirst of each half lot reaches the correspond-l ing reading position, i.e. A2 and B2. Then, in the case of bank cheques :tor example, the account numbers of the two lirst cheques in each path may be compared and the cheque with the lowest account number will be allowed to travel to the merging position, its place in the reading position being immediately taken by the next following cheque which is in turn icompared lwith the cheque still located in the other reading position. The cheque having an account number higher than that of the one which has already left and which account nurnber is moreover lower than that of the other cheque in the other reading position will be allowed to reach the corresponding merging position to travel behind the tirst cheque having already left. Both cheques will be transferred to the same transverse conveyor, eg. the two cheques will be successively admitted through output position C and successively dispatched on the transverse conveyor B. The operations will continue in this manner, a significant feature being that as each cheque leaves a reading position, it is automatically followed by the next cheque having come through the same input path. When neither of the two cheques in the two reading positions has an account number which is higher than that of the cheque which left immediately before, that cheque with the lowest account number will be sent forward, but it will be made to follow a different output path than the previous cheque. Hence, for instance, if that previous cheque was sent along transverse conveyor B', the next cheque will be sent on the transverse conveyor A.

Thus, it A is the account number of the cheque in.

6 one reading position and B is the account number of the cheque in the other reading position, while C is the account number of the cheque which left one of these two reading positions immediately before, the A cheque will be sent forward provided:

or its two cyclic derivations are satisied, whereas the B cheque will be sent forward provided:

or its two cyclic derivations are. satisfied. Moreover, a change of output wil take place ir" both A and B are smaller than C;

In other words, the first sequence to be found in each input lot will be merged into one sequence which will be dispatched to one output. The next sequence in each respective half lot will be similarly merged into a single sequence which will be sent to the other output. The following sequences of like ranks will be similarly merged and dispatched to alternate outputs.

Suitable mechanisms may be used to advance the documents or document carriers in the required manner, and suitable control circuits may be used to control these mechanisms.

In particular, as far as the mechanisms are concerned, reference may be made to the following patents and applications: Belgian Patent No. 577,758, issued Oct. 16,`

1959, in the name ot' G. X. Lens covering an arrangement for feeding lots ol loaded document carriers on the two transverse conveyors such as A and B, as well as for recovering the sorted lots at the end of the sorting operations; commonly assigned U.S. application No. 806,286, tiled April 14, 1959, in the name of G. X. Lens et al. covering a guillotine arrangement permitting the separation of the first document carrier from an input load in order to drop it into the input position such as A4; Belgian Patent No. 577,760, issued October 16, 1959, in the name of G. X. Lens et al. covering an edge- Wise conveyor adapted to move the document carrier through the various positions of a sorting stage; Belgian Patent No. 577,765, issued October 16, 1959, in the name of G. X. Lens et al. covering the arrangement of an output position such as C permitting a document carrier to be lifted from such a position to the transverse conveyor such as B'.

Document carriers provided with lateral extensions enabling them to be supported on transverse conveyors constituted by two separate parallel endless chains have been described in the commonly assigned, now abandoned U.S. application of J. Young, tiled January 25, 1957, Serial No. 636,310.

Electrical and electronic control circuits for a sorting stage provided with two inputs and two outputs as those shown in FIG. l, have been disclosed in the U.S. application of G. Van Mechelen tiled Ian. 23, 1957, Serial No. 635,884.

Although a two-stage sorting machine of the type illustrated by FIG. 1 can be realized in various ways, the diagram of FIG. 1 shows it is not essential to provide a points position which will direct a document carrier entering said points position to one out of at least two output positions. Such Was vthe case in the multi-stage sorting machine envisaged in the U.S. application of J. Young, filed January 25, 1957, Serial No. 636,309. If it has appreciable inertia, a points position may be deemed undesirable in that it tends to slow down the operation of the machine. y

The merging of the two edgewise vconveyors into one, as shown by the merging positions such as A1 and B1 in FIG. 1, offers the advantage that the same pair of output positions can be used for the two input paths. Otherwise, the two input paths need not be merged, but then the output positions such as C and D would have to be duplicated and serious difficulties might be encountered in associating these two pairs of output positions per sorting stage to their respective transverse conveyors.

Though the alignment of the output positions such as C and D one after the other and the merging positions such as A1 and B1 permit dispensing with a points position, care should be taken to avoid jamming along the edgewise conveyors and particularly the common output edgewise conveyor which receives document carriers from two sources. It is in this respect that the concept of more or less defined positions along the edgewise conveyors path, together with a step-by-step or cyclic operation of the machine may be proved useful.

A possible circuit to obtain such a step-by-step or cyclic operation is shown in FIG. 2 which represents the equipment necessary for the cyclic operation of one stage, e.g. sorting stage RS.

The synchronizing circuit SY comprises a series of electro-magnets which are used to directly control the mechanisms permitting the edgewise conveyors to advance the document carriers to the positions extending between AC1 or B3 and E. It also comprises a series of relays. A master ground is provided at the armature of changeover contact e1 of error detecting relay Er, and this master ground is used for the selective operation of the desired electromagnets and relays. This master ground will only be interrupted upon relay Er being operated due to the closure of one of the fourteen parallel contacts such as kx and upon the interrupter contact IN1 moving from the position shown. Then the changeover contact e1 moves over to the operated position while relay Er remains operated and thereafter locked until the closed contact(s) kx is again open. A pair of such contacts kx is provided in each of the seven positions A1, A'2, A'g, B'1, B2, C and D', thus in all positions of a stage except the input positions A'4 and B'3 and also the ejecting position E. In each of the positions where they are provided, there is one contact near the entrance of the position and one near the exit so that they may be used to control the correct centering of a document carrier in the corresponding position. If it is insufficiently advanced or if there is a partial overshoot of the position when the interrupter contact IN1 moves from the position shown, then relay Er will be operated to suppress the master ground. In case of such a faulty condition, the stage concerned is therefore effectively stopped until the document carrier is properly positioned.

Interrupter contact IN1 is part of a series of cyclically operated contacts which are provided by cams operating in synchronism with the mechanical means provided for the advancement of the document carriers.

Neither these mechanisms nor these cams are shown here, but a timing chart of the closure of these various cam contacts is represented in FIG. 2.

At the beginning of a cycle, one of which is represented, the three make contacts PU1-PU3 will be closed to produce the ground pulses PU at terminals 5, 4 and 3 respectively, which terminals are connected to the control circuit (not shown). At a later moment in the cycle the ground pulses DR will be produced by the closure of the make contacts DR1 and DRZ which are used to control the dropping of a document carrier from the transverse conveyors into the input positions B3 and A21 respectively. An example of mechanism suitable for dropping the document from the transverse conveyor into the input position is shown in the previously mentioned U.S. application Serial No. 806,286. Shortly after this, a longer ground pulse SH is produced by each of the four make contacts SH1-SH.1, which pulses will be used to shift the document'carriers from the position in which they stand to the next position. Slightly before this shift pulse, a so-called positioning pulse PT is produced by the closure of the malte contacts PT1-PT4 which will cause the application of the master ground in such a way as to authorize the departure of a document carrier from a position by opening a mechanical gate, normally closed and locked, at the end of various positions.

Of the ten positions used in each stage, all except the ejecting position are provided with a shift electromagnet such as SA.1 for the input position AC1 and which shift electromagnets may be operated by the shift pulse SH, i.e. by the closure of the corresponding contacts such as SH3. All the positions except the ejecting position E' and the two input positions AC1 and B'3 are further provided with an individual positioning electromagnet such as PA3 for the intermediate position A'3, and which positioning electromagnets may be operated by a positioning pulse PT, i.e. by the closure of the Contact such at PT.1. Instead, the two input positions AC1 and B3, in addition to their shift electromagnets, are provided with dropping electromagnets such as DA for the input position A4, and which dropping electromagnets may be actuated by a drop pulse DR, i.e. by the closure of the contacts such as 13H2. Finally, the two output positions C and D' are each provided with a lifting electromagnet such as LC for position C and these lifting electromagnets may be operated by a so-called lifting pulse LI, occurring in the latter part of each cycle, i.e. by the closure of malte contact LI1 transmitting the master ground at break contact e1 to LC or LD.

At terminals 9 and 1t) signals may respectively appear to Cause the energisation of either relay Ar or Br depending on whether the A or the B input is to be advanced. These signals will be provided from an electronic and electromechanical comparator arrangement which may be of the general type disclosed in the U.S. application Serial No. 635,884 and which may readily be adapted to the control of a synchronized two-stage sorting machine of the type envisaged here. Alternatively, electromechanical and electronic control circuits particularly advantageous in the case of a two-stage sorting machine envisaged here and operating in a step-by-step fashion, are detailed in the commonly assigned U.S. patent applications of W. Pouliart et al. tiled April 13, 1959, Serial Nos. 806.025 and 805,841.

Upon the control circuits having applied a signal, preferably by gating a PU pulse for instance at terminal 9, relay Ar will be operated. This will lead to an advancement of all the document carriers present in the positions A2, A3 and A3 together with the advancement of the first of the documents which may occupy the transverse conveyor leading to input position A'4, while the documents in the positions B2 and B3 and in the transverse conveyor leading to this last input position will stay put.

On the other hand, a signal at terminal 10 will have inverse effects.

Assuming that relay Ar operates, it locks to the master ground at break contact e1 through its make contact a1 in series with the interrupter contact IN1, in the break position. As shown on the pulse diagram, this interrupter contact IN1 will only move to the make position, and thus interrupt the holding circuit, at the end of each cycle. Relay Ar being operated and locked, the make contacts A1-A7 are also closed. The contacts ka and kb are closed whenever a document is being pressed by the continuously advancing transverse conveyor at the entrance of the corresponding input position such as AC1 for contact ka.

Therefore, during the drop pulse, i.e. the closure of make contact DR2, the drop electromagnets DA will be energised to deliver a document in the input position AC1. The corresponding shift pulse, i.e. the closure of make contact SH3, will operate the shift electromagnet SA.; to immediately transfer this document to the next position A3. At the same time, the document which might have been located in that position is moved out of it due to the operation of shift electromagnets SA3 and the simultaneous operation of the positioning electromagnet PA3 during the positioning pulse PT, i.e. the closure of 'make contact P'l`4. Similar operations take place in the reading position A'z and the document eventually located in that position is sent forward to the merging position A'l, its place being taken by the document eventually coming from the intermediate position A3.

As soon as a document reaches the merging position, the further advancement along the edgewise conveyor will be automatic since the contacts of relays Ar and B1' no longer intervene. But, it will be noted that since either the A or the B input side only is advanced at the same time, only one of the two merging positions A'l, or Bl may be tilled by a document at any time.

When a document is transferred from one of these two merging positions to the output position C', from there it may have to be diverted to the transverse conveyor associated with that output position. This occurs under the control of output relays Cr and Dr, the former being used to divert a document from output position C' and the latter being used to divert a document from output position D' if it is allowed to reach that position after first passing through output position C', and not being diverted from the latter. These output relays Cr and Dr are not directly operated from the control circuits. The latter provide a signal (a gated PU pulse) either at terminal '7 or at terminal 8 to cause the energisation of an electromagnet CC or DC respectively, upon these control circuits having ascertained that the docuent contained in one of the two reading positions A'2 or B'2 must advance and must eventually be diverted from output position C or output position D' respectively. This discrimination by the control circuits will thus take place two or three cycles before the document actually reaches the output position C or D' respectively, and from which it must be diverted to the associated conveyor. Thus, a memory arrangement is required and although this can be done by a relay arrangement, an attractive simple mechanical memory exists, and the detailed realization of which in conjunction with the present application is fully disclosed in Belgian Patent No. 577,759, issued Oct. i6, 1959, in the name of G. X. Lens et al.

Briefly, the operation of the electromagnet CC or DC will trigger a constantly rotating mechanism which will memorize the operation of the electromagnet which caused the triggering action. The rotating mechanical memory is provided with the required number of such mechanisms, i.e. two sets of three, and the rotation is synchronized with the various mechanisms and in particular the cams providing the pulses of which the waveforms are shown in FIG. 2. In the same way as for the input relays Ar and Er, once an electromagnet such as CC is energised, it is locked to the master ground at break contact e1 through the special interrupter contact Sil. Before SI1 opens to release the operated electromagnet, the next mechanism to pass the armature arrangement of this operated electro-magnet will have been triggered as required. By the time the document which caused this triggering action has reached the ouput position C' or D', the rotating mechanical memory will deliver a pulse PC or PD one cycle later, corresponding respectively to the application of the master ground through maire contact PCl or PDI. Thus, in the case of the operation of the electromagnet CC, as the document is about to he shifted from the merging position such as A'1 to the output position C', relay Cr will be operated and will continue to memorize the required output by locking to the master ground through its make contact c1 in series with the interruptor contact IN1. Therefore, during the lifting ground pulse corresponding to the closure of make contact L11, the master ground will energize the lifting electromagnet LC through make contact c2 thereby causing the dispatch of the document to the corresponding transverse conveyor. Of course, at that time relay Dr may also be energised so that a document already located in the output position D will be simultaneously lifted on the corresponding transverse conveyor by the same ground pulse Ll. if due to some fault, for instance if the required output has not been registered by the operation of one of the electromagnets CC or DC, the document will then be allowed to go through the output positions C' and D' to finally reach the ejecting position E. In such a case it is therefore taken out of the closed loop circuit to be placed in a reject bin.

Terminals 1 and 2 are shown connected to the armature leads of contacts kb and ka respectively for the purpose of issuing a drop pulse at these terminals when the corresponding tranverse conveyor is lled by at least one document next to the corresponding input position. Such signals may be used by the control circuits (not shown). Finally, a further pulse coinciding with a ground shift pulse (Sill) may be provided at terminal 6 to the control circuits, whenever make contact a8 or bg is closed.

Some considerations will now be given on the principles governing the special control measures which have to be taken in order to ensure a proper control of a two-stage sorting machine of the type described.

It is clear that if both stages of the machine work on the principle that they stop the processing of documents whenever one of their two inputs is empty, a completely continuous and fully automatic sorting operation involving a plurality of sorting passes will not be possible without additional special controls. Indeed, the machine may be blocked without the sorting being completed and special measures would then have to be taken for starting again at that time. Such a situation may readily be visualized by a simple example. Assuming that at some time, e.g. at the beginning of the sort, there are only two sequences at one input of stage NS and one sequence at the other input of this stage, while both inputs of the other stage RS are empty, the rst sequence at the first input of NS will be merged with the single sequence at the other input of NS to deliver a single sequence at one input of RS and leaving another single sequence at the second input of NS. The sorting machine would therefore stop although the sort is not completed.

Thus, it is useful to adopt the rule, during the whole of the sorting operations, that at least one of the two sorting stages continues to work when one of its two inputs is empty. In such a case, the sorting stage concerned would pass the remaining sequences at the other input alternately to one and the other output. It will be clear however, that while such an arrangement is desirable in order to avoid blocking of the machine, indiscriminate use of this feature should not be made since it does not constitute an actual sort, there being no merging of sequences but merely a distribution of the latter betweenv the two outputs. Therefore, it is preferable to adopt this rule of splitting the sequences following one another at one input while the other input is empty, only for one stage. In any event, if the stage for which this splitting feature is not adopted may eventually remain inactive during short periods due to one of its two inputs being empty, since the other stage will always provide sequences at its two outputs, irrespective of whether the two inputs of that other stage are both filled by documents o-r only one input is iilled by documents, the one stage which cannot work in the absence of documents at its two inputs will neve-r be in such a state for a long period.

Moreover, once the principle of splitting for one stage only is adopted, it is advantageous to choose the sorting stage NS which is the first sorting stage, or input stage, through which the documents are passed at the beginning of a sort. indeed, while the documents to be sorted will initially be divided into two half lots containing approximately the same number of documents, and one half lot being fed to one input and the other half lot to the other, this cannot ensure a perfect equality of the number of sequences initially present at the respective inputs of NS before the sort begins. Even if care was taken to make an exact division of the total number of documents obvispaanse ously this would still not ensure equality of the number of sequences on the respective sides. Thus, after that input which contains the smaller number of sequences has been emptied, there will generally remain a certain number of sequences at the other input of sorting stage NS, abstraction being made of the fact that some documents have already come back through the other sorting stage of the machine. Even though, without special measures, documents will in fact come back through the sorting stage RS to fill both inputs of sorting stage NS, there will remain an imbalance of the number of sequences on the two sides and this unbalance will be substantially equal to the difference between the numbers of the sequences initially fed to the two inputs of NS since both sorting stages distribute their sequences alternately to both outputs. To prolong this state of unbalance is undesirable. Indeed, the unbalance will eventually disappear as sequences are continuously merged but the original surplus sequences may only be merged with other sequences after the sorting operation has been in progress for some time. In general, there will come a time when a stage delivers an output sequence and is then left with an empty input and with the surplus sequences at its other input. If there are empty inputs for the other stage, but if that stage is allowed to pass a sequence at one input while the other input is empty, the sequence delivered by the first stage will be able to go round and eventually be merged with the first residual sequence. Then, the newly merged sequence will again go round before it can be merged with the next residual sequence. It is clear that such a process is quite inefiicient since a number of double passes will be required equal to the number of residual sequences, and each time, merely for the purpose of merging only two sequences. Moreover, these double passes will each time involve the whole series of documents, except those forming the residual sequences still to be merged. If these residual sequences have not been able to reach a size such that the sum of the documents forming these residual sequences is nearly equal to the total number of documents, it is clear that each of these double passes merging only two sequences, will be even more inefiicient since a large number of documents will have to go round for each of these double passes.

Therefore it will be found advantageous to have the input stage NS work on the principle of splitting the remaining sequences at one input while the other input is empty, and moreover load balancing means may be envisaged in order to ensure that all the documents will have an opportunity to be passed through the input stage NS before documents having gone through a double pass,

i.e. through NS and RS, are again allowed to go through the input stage NS. In this manner, during the first sorting pass one of the two inputs of NS will eventually become empty and the remaining sequences on the other side will be distributed equally between the two outputs, and from that moment the machine will work with an equal number of sequences on both sides, abstraction being made of the slight unbalance of one sequence eventually produced when the total number of sequences attacking any stage is not even.

This initial load splitting eature may readily be secured for instance by providing stop members behind the two ha'f loads of documents initially feeding the input stage NS so that the documents coming out of RS will be arrested by these stops until the whole of the documents has gone through the input stage NS, at which time these stops may then be removed once and for all.

The detection of an empty input at the normal input sorting stage NS should therefore cause the remaining documents at the other input to be dispatched through the sorting stage NS with the output sequences being distributed alternately between the two outputs of that stage. The comparator part of the control circuits disclosed in the U.S. application Serial No. 635,884 includes three bistable devices one of which memorizes the comparison between the value of a document on the one side and that of the document on t.e other side, while the other two bistable devices respectively memorize the relation between the value of the document on one side and that of the document on the same side which went through the stage immediately before it. Thus, if A and B are the values of the two documents respectively present at the two inputs while C is the value of the document which went through the stage before A and B (C being used indiscriminately both for the A and the B side), the three bistable devices respectively indicate whether A is greater or smaller than B, whether A is greater or smaller than C and whether B is greater or smaller than C. In accordance with the rules already discussed, the smaller of A or B goes forward through the stage unless this smaller number is also smaller than C while the higher number is greater than C. In such a case it is that higher number which is sent through. Also, if both A and B are smaller than C, there is a reversal of output since this indicates that the two input sequences have just been merged and the following two sequences of like rank must now be merged.

lf upon an input of NS becoming empty, one produces a signal which continuously forces the bistable for that side in the condition indicating an end of sequence for that side, and irrespective of the signals given by the comparator, this can be used as already discussed to some extent in the U.S. application Serial No. 635,884 already referred to, to cause the continuous advancement of the documents still present at the other input.

Assuming that the A side becomes empty as the last A document goes through the stage and that its place is therefore not taken by a following document on the A side, the corresponding bistable will thus be forced to register C /i. Since A became empty at the last A document went through, this can only mean that the comparator indicated B A C or A C B. In fact, the condition C B A would also have led to the departure of the last A document, but in accordance with a particular feature of the comparator disclosed in the Bclgian patent mentioned above, such a state of affairs as expressed by the last inequality is never allowed to persist. Indeed, the comparator is adapted to exploit this particular condition (or C A B) to produce the required reversal of output, but immediately thereafter the bistables respectively recording C A and C B, i.e. end of sequences on both sides, will be automatically restored to the opposite conditions, i.e. A C and B C respectively, and the document which must be sent through the stage as the first document of the new merged sequence will be determined merely by the relation between A and B i.e. the smaller goes through.

ln the case ofthe first possible situation, i.e. B A C, forcing a registration of C A will transform this inequality into B C A. This means that the residue of the sequence on the B side which was being merged with the now finished sequence on the A side will also be dispatched through the stage, as required to complete the output sequence. When the next sequence comes on the B side, the comparator will register the end of sequence on both sidcs which will produce the required switching of the output. Therefore, the comparator will be immediately reset to the conditions A C and B C, but the first will be ineffective in view of the empty A input forcing at the end or" sequence condition C A. Of course, the first document of this next sequence on the B side has now been read and the previous relation B A established by the comparator may no longer hold. ln fact, the corresponding bistable comparator might now record A B which is in contradiction with B C coupled with C A. However, the comparator is designed so that the bistable recording the relation between A and B is ineffective in case there is an end of sequence on one side and no such condition on the other side. Hence, the probe pulse examining the state of the comparator will act on the indication B C A and will therefore continue to advance the documents on the B side as required, and this as long as the A side remains empty.

In the case of the second condition, i.e. A C B, this means that when the last document on the A side goes out, the sequence on the B side which was being merged had already gone through the stage since the condition indicates an end of sequence on the B side. Therefore, upon an end of sequence being forcedL on the A side, switching of output will be produced and after the automatic reset, the comparator will record B C A again leading to all the documents on the B side being dispatched through the stage as long as the A side remains empty.

Preferably, the signal forcing the end of sequence condition on the side which has become empty should be produced with a certain delay to take care of normal discontinuity in the flow of the documents through the transverse conveyors, but as soon as an input contact for instance, has detected that one input side has become empty for `a certain time interval, splitting the sequences on the other side wil-l be started.

Eventually however, the input side which became empty may again be filled by documents and it is then desirable that the input sorting stage NS should resume its normal operation by merging corresponding sequences appearing at its two inputs. Provided the appearance of a document on the previously empty input side cannot only be detected, but also that this document can be read, normal sorting will be automatically resumed.

Indeed, when the iirst document is read, it will produce a normal end of sequence signal which may now supersede the artificial end of signal which had been given as long as the corresponding input was empty. Therefore, the document going out on the other side will continue to do so since there is no change in the conditions of the comparator but as soon as the outgoing sequence is ended, there will be a switching of outputs immediately followed by a reset olf the bistables so that they will indicate A C and B C respectively. There is no longer an end of sequence which is being forced and accordingly, provided the new document having come on the previously empty input side has been read the bistable comparing A and B will be able to decide which of the documents of a new merged output sequence is smaller and must go hrough rst.

The only possible problem is to ensure that the document coming on the previously empty A side will be read. In this way the corresponding value can be stored and upon a new sequence coming on the B side, the value of the iirst B document of that sequence can be compared with the value of the first A document.

FIG. 3 shows an adaptation of the circuits disclosed in the U.S. application Serial No. 635,884 mentioned above and will permit the operation of the normal input sorting stage Ns in the desired manner according to which splitting of one input is produced when the other is empty. The relays Adr and Bdr are those used for the A and the B side respectively to produce a forced end of sequence for that side. This may occur in the manner previously disclosed in U.S. application Serial No. 635,- 884. The contacts such as kal for the A side are normally operated as long as there are documents ready to be 4taken from the corresponding input through the stage NS. Whenever there is absence of documents on the A side for a certain time covering short absences, contact kal will assume the break position shown in FIG. 3 and from that moment, as soon as relay Aar operates in a manner disclosed in said U.S. application No. 635,884 to dispatch the last document at the A input through the stage, an operating circuit will be closed for relay Adr through make contact m17. Relay Adr operates and locks through a circuit which includes its make contact adl in series with the paralleled break contacts acm and aeg of time. Relay Acr is a relay which is used to dispatch a document on the A side through the corresponding read- Y therefore, as long as it is operated, force the end of sequence condition on the A side and accordingly the splitting of the documents on the B side.

As disclosed inv the U.S. application No. 635,884 this is done not only by forcing the corresponding bistable to register C A but also by producing an artificial probe pulse to extract the state of the comparator, the normal probe pulse being produced by a document itself after it has been read.

iter the A side has been empty for some time, a new sequence of documents may then arrive and it is desirable that as soon as the outgoing sequence on the B side has gone through the stage Ns, the next sequence on the B side should now be merged with the first of the new sequences having reappeared on the A side. Due to the reappearance of at least one sequence on the A side, contact kul will again be operated and ground will be applied to make contact ham. This contact pertains to a relay Har (not shown) while the following break contacts in the chain, hcw and hdm, pertain respectively to relays Hcr and Hdr (not shown). These three relays are however disclosed in said U.S. applicationvSerial No. 635,884 and are operated to dispatch a document out of the stage at a time no new documents are actually being read. Relay Har operates first followed by the subsequent operation of either relay Hcr or Hdr according to the direction taken by the document going out of the stage. the circuit permits the reappearance of documents on the A side to be rendered effective at a suitable time without hindering the operations which are taking place in connection with the splitting of the documents on the B side. ln the circuit just disclosed, relay Aer will thus receive a ground pulse during the time Contact ham remains closed before the opening of either break contact 11cm or hdm, and through make contact adm. Relay Aer locks to ground through its make contact nel, in series with make contact adm, the connection of the winding of relay Aer extending to battery through break contact bea of relay Ber.

The function of this additional relay Aer is to ensure that the iirst document arriving on the A side can now be read. A rst result of the operation of relay Aer is to open break contact neg thereby preventing the operation of relays Am' and Bar which serve to advance documents on the A and B sides respectively through the sorting stage. Their respectively helping relays Abr and Acr are also prevented from operating. Hence, as the B document is sent out t-hrough the stage subsequent to the operation of relay Har causing among other things the closure of make contact halo, the normal circuits are prevented from further advancing documents.

A second result of the operation of relay Aer is to force the master bistable circuit B87 exploiting the results of the comparator CMP, containing in particular the three bistable devices previously mentioned in order to finally determine whether the document on the A or on the B side should be advanced. As shown, in addition to the signals normally provided to the two inputs of the bistable device B57 from the comparator CMP, a suitable fixed potential at terminal P1 is interconnected to the A and the B inputsof BSq through relay contacts. The arrangement is obviously symmetrical, together with the rest of the circuits of FiG. 3, and for the A input the connection involves 'break contact aeg in parallel with make contact acm. Normally, the connection is therefore present, but it is interrupted as soon as relay Aer opens its break contact aar. This opening of the connection to terminal P1 will have the result of placingV B87 in the A condition, whereas it was previously in the Hence,

l B condition indicating the advancement of the documents on the B side.

This may for example take place in practice by causing the grid of one of the tubes forming the ip-fiop BS, to have its potential increased due to the opening of the connection to the fixed potential at terminal P1. This might cause an increase of the resistance between this grid and the negative HT supply, this grid being on the other hand connected to the positive HT supply for the tube through a resistance leading to the anode of the other tube forming the flip-flop, and thus in conventional manner. The plate-cathode space of the tube whose grid potential has been modified in this manner will now be ionized while the other tube will become nonconductive.

Further, make contact nel, is located in parallel across make contact abz of the helping relay Abr in a circuit normally provided for the operation of relay Aer upon relay Abr being operated and relay Aar having released as a consequence. At the present time, both these two relays, and particularly relay Aar, are released due to break contact ae3 opening the battery supply; and by the closure of make contact ae will permit the operation of relay Acr which, as stated previously, is the relay causing the dispatch of the first document on the A side through the reading position.

Hence, the first of the new A documents will be read not only due to it being advanced through the corresponding reading position but also due to the electronic comparator circuit allowing the information to be read in view of the master bistable BS, being in the A condition. As soon as relay Acr operates to cause the dispatch of the first A document through the reading position, make contact acm closes and rc-establishes the connection between terminal P1 and the A input of B57. The latter bistable circuit remains in the A condition, but from now on it is no longer forced in that condition and if the comparator probe pulse should now detect that a document on the B side should be advanced, the comparator CMP will be able to trigger BSI; in the B condition.

Due to the operation of relay Acr in the special circuit disclosed above, break contact new opens and since break contact lez is already opened, the locking circuit for relay Adr is interrupted and this last relay releases. lndeed, at that time it is no longer necessary to force an end of sequence of the A side and while the bistable device (not shown) registering that condition in the comparator, CMP will remain in the end of sequence condition, it is now free to assume the contrary condition. In fact, however, as the first document coming from the A side is read due to the operation of relay Aer, a normal end of sequence condition will be registered by that bistable device.

Due to the release of relay Adr, make contact adm opens the locking circuit for relay Aer which also releases. Through break contact ac3 battery is again applied to the normal circuits involving the relays such as Aar and Bar.

Due to the opening of make contact ae5, relay Aer will also release and upon the comparator probe puise having determined which document should be advanced through the stage, either relay Ar or relay Br will be operated and correspondingly either make contact al or make contact b1 will be closed depending on whether BS, is in its A or in its B condition respectively.

Two cases may arise upon the first of the new A documents having been read. The first possibility is that the first document on the B side now ready to pass through the stage is not the last document of a sequence. Therefore the comparator records C A and'B C whereby BS? will be placed in its B condition by the probe pulse, resulting in the closure of make contact b1 and in the dispatch of the remaining documents making the last part of the unfinished sequence on the B side.

Later on, upon there being an end of sequence on the B side also, the bistable device (not shown) which records the comparison between A and B will determine whether the first document on the A side or the first document of the next sequence on the B side should be sent forward, and sorting will again proceed in the normal manner.

lf there is already an end of sequence on the B side as the rst document on the A side is read, a switching of output will be given as required and again the bistable device (not shown) recording the comparison between A and B will determine which of the first two documents on the A and the B side respectively should first be advanced through the stage.

While it is desirable as previously explained, that the second sorting stage RS should not be provided with the splitting feature, there is nevertheless the following difficulty in relation with that stage. Assuming that, at the end of the sorting operation, stage NS delivers the last two sequences one at each output, as the second and last sequence reaches the second input of RS, that stage will start to merge it with the other sequence already at its input. As sorting stage RS produces the final sequence, there will come a time when there are no longer any documents at one of the inputs of RS while there remains at least one document at the other input. If the sorting stage RS does not possess the splitting feature used for NS, the RS stage will therefore be blocked. Yet, the final sequence produced by RS will travel to one of the inputs of NS and will also pass through that stage which can work with one input empty, but there is no insurance that the final sequence passed through NS will be directed to the empty input of RS. Thus it might be placed behind the residue of the unfinished sequence at the input of RS and while the sorting machine would be stopped, the sort would not actually be completed.

lf it is desired to remedy this situation, it is therefore necessary that the reverse sorting stage RS should also be provided with'means to detect an empty input in order to advance the residue of the longer sequence at the other input. But, one cannot adopt exactly the same circuits as those for the normal sorting stage NS, since if an empty input condition arises at RS in any other circumstances than those envisaged above, RS would split the sequences on the other side exactly in the same way as NS. and as stated before, this is undesirable.

A possible solution for the control of the reverse sorting stage RS is shown in FlG. 4 representing supplemen tary circuitry to those used in the above mentioned U.S. application Serial No. 635,884 and in FIG. 3. The main principle of the control achieved by the circuits of FIG. 4 is to have a switching of output signal used to suppress the splitting feature for RS. Hence, if an input of RS becomes empty while the sequence which was being merged at the other input is not yet completed, the residue will be allowed to go through. But, the immediately following sequence appearing at the input which is still filled by the documents, will cause a switching of output signal which will cancel the empty input signal for the other side. As a consequence, the sorting stage RS will therefore be stopped as required until documents again appear at the empty input and can be merged in the normal manner. if on the other hand, no documents come to fill the empty input, this means the end of the sort and the nal sequence will remain at one input of RS.

In case for instance the A input of the reverse sorting stage RS becomes empty when the last two sequences are being merged to produce the final sequence, the residue of the last sequence on the B side will therefore be advanced through RS, and on the other hand the final sequence will be passed by the normal sorting stage NS to reappear at one of the two inputs of RS.

It will be observed that at the end of the sorting operation, if the A input becomes empty, the final sequence must necessarily come back on the B input, since othersignal will suppress the splitting feature. empty A input side will be made to advance with the result that the inal sequence remains at the B input of u on the A side.

sequence ends on the B side.

wise the A input would not have become empty, the

, final sequence following the shorter penultimate sequence on the A side out of whichv it is partially built. As the final sequence reaches theB input of the reverse sorting stage RS, it will produce a switching of output signal, which, as mentioned above, is adapted to cancel the empty input signal coming from the A side. Therefore, the final sequence on the B side will not be forced to advance through the reverse sorting stage and the sorting operation will be ended.

Alternatively, at the end of the sorting operation the A input of the reverse sorting stage RS may become empty as the final sequence has already appeared at the B input of the reverse sorting stage RSV, thereby causing an end of sequence for the B side. Hence, upon the A side becoming empty and forcing an end of sequence conl dition for the A side, the two end of sequence conditions will cause a reversal of output andin accordance with the rules for sorting stage RS, the switching of output Hence, the

the reverse sorting stage R8 as required.

Referring now to the detailed embodiment of FIG. 4

l for the control of the reverse sorting stage RS, wherey the circuits must be read in conjunction with those of FIG. 3 already described in connection with the operation of the normal sorting stage NS, it will again be assumed that at some moment during the sort, the A input becomes empty.

As previously described, relay Adr` (FIG. 3) operates and locks, and by opening its break contact ada (FIG. 4) it interrupts the connection between terminal PAC and an input of bistable device B85 which is part of the comparator CMP and used to register an end of sequence In a manner previously explained above, this interruption may be used to trigger the bistable to the end of sequence condition C A.

At the same time, as the last document on the A side goes through R8, relay Car (FIG. 4) will be operated v or alternatively relay Dar depending on the direction taken by this last document coming from the A input.

As this last document goes out, the bistable device BS5 indicating thatY it should be directed to the C or to the D output will permit the operation of either relay Cr or relay Dr respectively through break contact d or C10. Assuming that relay Cr is operated and temporarily locked `through its make contact c1 through undisclosed circuits already detailed in the U.S. application Serial No. 635,884, make contact Cnwill therefore be closed and upon the Voperation of relay Adr, make contact adn also closes thereby establishing an operating circuit for relay- Car through break contact daz. Relay Car operates and locks to ground through a circuit including its make contact cal in series with break contacts ae., and beq. Make contact ca5 is also closed but this is yet ineffective since .BS5l is still in its C condition. l

As before, the operation of relay Adr (closure of make contact ada) produces an artificial probe pulse to investigate the state of the comparator and replacing the probe pulse normally initiated by the document itself after it has been read. If there is no end of sequence on the B side when the A side becomes empty, the forced end of sequence .on the A side thereby compels the documents on the B side to continue to pass through RS until the It may also happen that the A side becomes empty as the first document of a new sequence on the B side was standing ready to be passed through the stage. Then there isr already an end of sequence on the B side when the forced end of sequence appears on the A side and consequently it leads to a `switching of output.

tion, it will now pass to the D condition with the result that relay Swr becomes energised since make contact cas is closed. Also, in the comparator, the switchmg of output results in a pulse trying to perform the normal reset of BS5 and B84 to the respective condition A C and B C.

The operation of this relay Swr to record a switching of output, plays an essential part in the circuits since otherwise, break contact udg being still open, the comparator probe pulse would be unable to perform the electronic reset of BS5 to the condition A C and only B84 would be normally reset to the condition B C. This would indicate the advancement of the next sequence on the B side which-is undesirable for the reverse sorting stage RS which should not possess the splitting feature afforded to the normal stage NS. The operation of relay Swr will however remedy this situation since although when it operates its change over contact SW1, it tends to force BS5 into the condition A C, but without success since break contact adz is still open, as soon as sw, has moved to the make condition, it by-passes the open break contact adz and accordingly BS5 iscorrectly triggered to the condition A C. Actually, this circuit arrangement is symmetrical for the two sides with the result that B84 will also be forced to the condition B C due to changeover contact swg having moved to the make condition thereby interrupting the connection between terminal PBC and the corresponding input of B84. This merely duplicates the previous electronic reset but it has the advantage of preventing undesired operations at a later stage.

However, despite the operation of relay Swr having caused the bistab'e devices BS5 and B84 to record respectively A C and B C, there is yet no insurance that the empty A input will now be giving the signal to advance, or more precisely that the B input at which a newv sequence of documents is standing, will not be given the order to advance while there are no documents at the A input.

This will be produced by the opening of break'contacts swg and sw., in the connections between terminal P1 (FIG. 3) and the two inputs of the bistable B87 determining which of the A or of the B input is to be advanced. Soleiy for the A input, the opening of break'contact swg will be effective since break contact adm is already open. Therefore, B87 will be triggered to the A condition thereby indicating advancement of the A input which is empty. The `result is that the reverse sorting stage RS is now stopped in the desired manner until a new sequence of documents again appears at the previously empty A input.

As already described in connection with FIG. 3, this will cause the operation of relay Aer and a subsequent series of operations designed to permit the reading of the first document of the new sequence on the A side. After this new document `has been read, the comparator will record a normal end of sequences for the A side and if this was allowed to be effective, and bearing in mind that B84 indicates B C, the joint state of the bistable devices filled, the documents being advanced from the other input` or allowed to continue through the stage until the sequence is ended, at which time thetwo sequences at thel ytwo inputs are being merged in the conventional manner. In the case of the reverse sorting stage RS however, conditions are different since there is at least one sequence standing at the B input as the sequence arrives atithe A input. Then, it is undesirable, particularly at the end of the sort to allow this sequence at the B input to be passed through the reverse sorting stage RS without being merged with the new sequence at the A input.

However, in view of the contacts SW1 and SW2 being in their make conditions, B85 and B54 are forced to remain in their conditions A C and B C respectively i whereby as the first A document is read, it cannot modify the state of BS5.

' observed that this is permitted by make contact sw3 shortcircuiting the break contacts ae4 and acm whereby the operation of relay Aer is fully effective to prevent B87 frombeing forced in the A condition. Consequently, the comparator' probe pulse will be abie to set B57 in the condition corresponding to the advancement of the first document at the A or at the B side which has the lowest number. This will be the case since the bistable device determining whether A is greater than B or not will also determine the signal to be sent to BS7 in view of B55 and B34 recording respectively A C and B C.

It will be noted that since the reverse sorting Stage remains idle when, the A side being empty, end of sequence is reached at the B side, means must be provided to operate relay Aer upon a new sequence reaching the A side. This is afforded by the dotted line connection of FIG. 3 including make contact sw5.

Also, as shown by PIG. 4 and contrary to the circuit of FIG. 3, the operation of relay Aer is no longer required to set the master bistable B57 to the A condition since it is already in that condition permitting the first A document to be effectively read by the comparator circuit.

After relay Aer has operated relay Acr, both break contacts ae7 and acm are temporarily and simultaneously opened, resulting in the release of relay Car and hence that of relay Sm' due to the opening of make contact caa.

At the end of the sorting operation, relay Adr (FIG. 3) being assumed to be energised to indicate that the A input side has become empty while the longer penultimate sequence on the B side was still passing through the stage RS, relay Car for example is also energised and locked.

`Hence, upon the first document of the final sequence reaching the B input, the switching relay Swr, operating as a result of a change of output, will be energised with the results already explained above and consequently the final sequence will be stopped at the B input and will not be able to go through the reverse sorting stage RS, as required. If the A input becomes empty as the first document of the final sequence has already been read to indicate an end of sequence for that side, the operations will be similar.

Thus, the control circuits for the closed loop machine described above may be readily designed to ensure a satisfactory and ethcient automatic operation, and are no more complicated than the control circuits which must be associated with a multi-stage machine of the type disrclosed in the U.S. application Serial No. 636,309.

On the other hand, a closed loop sorting machine of the type described above has numerous advantages, some of which may be recapitulated below:

(l) The machine is of overall smaller size.

(2) An eventual fault is easier to trace and it immobilizes only a less costly machine of smaller size.

(3) The mechanical construction of the two stages is identical, and maintenance is facilitated.

(4) The machine is more flexible, and in particular, further documents may be added at the inputs of the `normal sorting stage while the sorting operation is al- (6) The machine is readily adapted to the eliicient sort ing of lots of documents substantially smaller than the i maximum size allowed by the capacity of the document favourable conditions may be a smaller number of initial sequences than the average, which can be reckoned as equal to half the number of documents; the so-called accidental merging during the sorting operation, which also tends to reduce the number of sequences; and a subing stage, whereas in a multi-stage machine this is impossible to foresee, since the final sequence might be obtained at any stage of the machine.

(10) There is much less loading time lost, since at the beginning of the sorting operation, as soon as the normal sorting stage delivers the second output sequence, the reverse sorting stage will start to function.

(l1) The efiiciency of the two-stage endless sorting machine is substantially higher than that of a multi-stage sorting machine.

To illustrate this last point, some comparative figures relative to the eiliciency of the two-stage sorting machine versus a ten-stage sorting machine are given below. When using sorting by merging, a ten-stage sorting machine may deal with lots of 1000 documents since 210:1024. The efficiency, for which the relative figures are given below, may be defined as the ratio between the theoretical minimum number of steps to perform the 'whole sorting operation in accordance with the principle of binary sorting, divided by the aggregate of the number of steps actually performed by any stage of the machine plus the number of steps during which this stage was stopped, this sum being made for all the stages of the machine. The number of steps actually performed by any stage and the number of steps during which that stage was inactive should be reckoned from the beginning until the end of the sorting operation. The number of steps during which the sorting stage is inactive corresponds for instance with the time during which the first sequence is delivered by the first stage; then, the second stage is yet inactive.

Initial loads at the 5l2(1) 25(5(2) 261(2) 266(2) 27(5(2) 20(5(2) 12(1(2) two inputs 5l2(1) 25(5(2) 251.(2) 24(5(2) 2360) 216)(2) 100(2) Ten-stage -.pereent. 48 29 20 29 30 30 20 Two-stage .do. 84 90 78 66 68 59 60 Two-stage .-do 84 90 76 77 78 79 63 (with load balancing) not only splits the remaining sequences at one input, after the other is empty, but also prevents the sequences having already undergone a double pass from being able to pass through the normal sorting stage a second time before all these residual sequences have gone through that stage.

In the calculation of these efiiciencies, it should be noted that no account has been taken of accidental merging, and it has been assumed that only end delivery is used for lthe ten-stage sorting machine. 7 The various columns refer to the different initial possibilities which have been considered. The number bcfore the brackets corresponds to the number of initial sequences which is initially fed at an input of the input stage While the number between the brackets indicates the size of the sequence which has been assumed. Thus, the first column identities the case of 1024 documents of which 512 are initially fed at each input, and each set of 512 is ordered in exactly the reverse order than the desired one (since each sequence is assumed to contain only one document). The next ve columns also envisage an initial lot of 1024 documents, but with each sequence containing two documents and with gradual unbalance in the initial loading at the two sides. Finally, the last column refers to an unbalanced case of sorting only 440 documents, again with initial sequences of two documents.

While there is not much difference between the particular yfigures given for the two-stage sorting machine with or without load balancing, the improvement in the eiiiciency with respect to the ten-stage sorting machine is striking.

While the principles of the invention have been described above in connection with specic apparatus, it is to be clearly understood that this description is made only by way of example and not as a limitation on the scope of the invention.

We claim:

l. A sorting machine comprising two sorting stages, each stage having two inputs and two outputs, the outputs of each stage feeding the inputs of the other stage, the articles to be sorted being initially applied to the two inputs of one of said stages in lots of approximately equal size, the lots containing sequences of articles, each of said sorting stages merging from the two lots sequences of articles of equal rank according to their sorting characteristic and applying the resulting sequences alternately to one and then the other of the two outputs of said sorting stage.

2. A sorting machine as claimed in claim l, further comprising two pairs of transverse conveyors, each pair passing articles to be sorted from the outputs of one stage to the inputs of the other stage of the sorting machine.

3. A sorting machine as claimed in claim 2, wherein each of said sorting stages further includes two input edgewise conveyors -which merge into a single output edgewise conveyor, the input ends of said two input edgewise conveyors being adapted to be fed from the corresponding pair of said transverse conveyors, while said output edgewise conveyor is adapted to feed successively the other pair of transverse conveyors.

4. A sorting machine as claimed in claim 3, further comprising a synchronizing control circuit for each of the two sorting stages to provide a stepwise advancement of the articles through said edgewise conveyors.

5. A sorting machine as claimed in claim 3, wherein said input edgewise conveyors include at least an input position followed by a merging :position leading to the corresponding output edgewise conveyor, and an intermediate position located between said input and merging position in one of the two input edgewisc conveyors for each stage, thereby staggering the input positions of the two associated input edgewise` conveyors topermit feeding of said two input positions from the associated pair of transverse conveyors without interference from one another.

6. A sorting machine as claimed in claim 3, wherein K.

said output conveyors further include three positions following one another, the first two positions being output positions leading to the corresponding transverse Yconveyors, while the third position is a reject position for removing any article from the sorting machine, e.g., in case of a fault.

7. A sorting machine as claimed in claim 6, wherein the stepwise advancement of the article through said edgewise conveyors is such that all positions in said edgewise conveyors may simultaneously be occupied by a docul ment, the advancement of one liberating the position for the following article, except for the two output positions only one of which may be occupied at one time by an article since they are fed sequences of articles alternately.

8. A sorting machine as claimed in claim 7', wherein the stepwise advancement of articles along one of said input edgewise conveyors takes place simultaneously with the advancement of articles located on said output edgewise conveyor, the articles located on the other input edgewise conveyor remain at rest. l

9. A sorting machine as claimed in claim l, wherein a predetermined one of the two sorting stages is utilized as the input stage and is adapted to alternately distribute sequences of articles, appearing at one of the two inputs of the stage while the other input is empty, towards one and then the other output of said stage, the other sorting stage is adapted to prevent the low of articles whenever one of its two inputs becomes empty.

l0. A sorting machine as claimed in claim 9, wherein the articles to be sorted are fed to the two inputs of said'.

predetermined sorting stage.

1l. A sorting machine as claimed in claim 10, further of the non-predetermined sorting stage, when the other input is empty, is allowed to pass through the stage but upon the arrival at this input of a new sequence, passage through the stage is prevented until the empty input again receives articles.

No references cited.

Images