Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
A COMBINATION wEIGHrN~ SYSTEM lZ20497
.
BACRGROUND OF THE INV~NTION
_
The present invention relates to weighing ~ystems
and is concerned in particular with weighing systems util-
izing a plurality of scales to achieve a minimum qualified
weight from a selec~ed combination of the scales.
Many products such as fruits, vegetables, candies
and other small items are produced or manufactured with
varying sizes and weights, and are handled in bulk quanti-
ties prior to being separated in groups and packaged.
Combination weighing systems have been developed for se-
lecting from a plurality of individ~al scales containing
the product a particular combination of scales which cum-
ulatively provides a total weight closely approximating or
equaling the target or ~tated contents weight~ Such
weighing ~ystems are described, for example, in ~S. Pat-
ent Nos~ 3,939,928 and 4,267,89~.
Combination weighing systems have become more
common through the advent of the microprocessor which i-
~capable of sampling multiple combinations of scales in a
very short period of time and determining which combina-
tion most satisfactorily provides a target weight. When
the combination has been identified, those scales belong-
ing to the combination are dumped into a common chute
which discharges the collected product into a film wrapper
or other container in a packaging machine7 The process
may be carried out repeatedly by a microprocessor with the
scales reloaded or with the dumped scales eliminated from
the search processes until they are reloaded.
~zz~
The flexibility of microprocessors allows a mult-
itude of scales to be examined during the search process
and permits weight parameters to be readily adjusted in
accordance with varying product and production demands~
However, it is important that the weight information from
each scale be accurate ~hro~ghout extended p~riods of use
and not be affected by drift in the components which pro-
cess the weight information. For this reason, calibration
systems are general}y employed in the scale, and the sys-
tems are periodically activated to upda~e the weighing
parameters used by the processor.
Sampling of a weight in a given scale is fre-
quently complicated by the environment in which the scales
operate. Scales are commonly loaded from a vibrating
feeder, and in order to isolate the scales and the weight
mea~urement from the effects of the vibrator, resilient
mounts support the critical measuring sensors and the
scales. Nevertheless, spurious errors are introauced into
the weight signals and produce inaccurate results in the
final weight.
In spite of ~he speed wi~h which microprocessors
operate in comparison to the mechanical weighing deYices,
cycle times for performing the microprocessor functions
are important because they are added to the sampling and
reading times, and one microprocessor may service a number
of scales which are loaded in staggered sets.
It is accordingly an object of the pre~ent inven-
tion to provide soluti~ns to the problem~ mentioned above.
~ . .
3L~2Z~)4~7
SUMMARY OF TH E I NVENT I ON
The present invention resides in a co~bi~ation
weighing system that is designed to s~arch for and obtain
a minimum gualified weight of product from among multiple
quantities of the product. The system includes a plural-
ity of scales, each of which receives and weighs a
quantity of the product an~ provides a weight signal rep-
resentative of the weight in the scale~
In one aspect of ~he invention, the scale has a
weighing tray for holding the produ~t and a member such as
a strain gau~e that is strained by the weight of product
during a weighing operation. Calibration means ~re pro-
vided in the scale to determine the parameters of tare and
s~ope in a weighing calculation, and included in the cali-
bration means i~ a weight of known amount that is lowered
and raised from the scale in a calibration process. The
weight is joined with an actuator means, such as an air
cylinder, by a coupling having first and second inter-
locking ~embers ~hat disengage automatically when the
weight is resting on the s~ale~
In another aspect of the invention, the output
signals from the scale are sampled and processed to im-
prove accuracy and eliminate spurious errors caused by
vibrations and other disturbances. Signal averaging means
are connected with the scale to receive the output signal
and include sampling means for sampling the signal multi-
ple times to establish the average value of the signal
from the various samples.
In still a further aspect of the invention, the
combination of scales which provides the minimum qualified
~;~20~97
weight is identified through a search operation based upon
an ordered search seq~ence of all combinations of the
scal~s. The search is conducted by a search con~rol means
which has means for omitting from the search, combinations
of scales hav}n~ subcombinations previously searched and
found to be qualified at or above a target weight. Elim-
ination of certain combinations frsm the search sequence
reduces cycle timeO The ~earch sequence is also esta-
blished by adding one new or different scale to the com-
binations or subcombinations previously searched. In this
manner~ the volume of data manipulated during each step of
a search sequence is minimized with corresponding improve-
ments in cycle time.
~RIEF DESCRIPTION OF THE DRAWINGS
Fig. 1 is a block diagram schematically lllU5-
trating a combination weighing system embodying the pre-
sent invention.
Fig. 2 is a horizontal elevation view partially
in section of a weighing scale including calibrating and
weight-~ensing mechanisms.
Fig. 3 is a fragmentary view of the calibrating
mechanism in Fig. 2 and shows the coupling engaged.
Fig. 4 is a fragmentary view of the calibrating
mechanism and shows the coupling disengaged.
Fig. 5 is an alternative em~odiment of the coup-
lin~ in FigsO 3 and 4.
Fig. 6 is an electrical diagram of the weight
signal acquisition components of the combination weighing
system.
';~ 4
l~Z()4~
Fig. 7 is a diagram of the signal averaging and
calibration elements in a microprocessor of ~he combina-
tion weighing system.
Fig. 8 is a chart showing all of the combinations
of scales in the search sequence of a four-scale system.
Fig. 9 i~ a diagram ~howin~ the details of the
combination searching elements of the microprocessor in
the combination weighing system.
Figs. l0~ & B are a flow chart of the combination
search routine.
Fig. ll is a schematic diagram of the sequencer
utilized in the search ro~tine.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Fig. l ill~strates the principal components of a
combination weighing system which searches for and obtains
the best combination of scales which collectively provide
a measured weight of product not les~ than a predefined
target weight. The product~ may be fruit, nuts or other
items which have random weights and which are loaded in
:
groups into the plurality of:scales from which the combin-
ation is selected. When a "best" combination has ~een
estabIished, the~corresponding scales are dumped into a
collector device~ or funnel for wrapping or deposit in a
sn~1e~ package at the measured weight. The system may
have any number Nn" of~:scales l0, which are respectively
~designated 10-l, lO-Z . . ~ lO-~.
Each of th:e scales lO has the same basic con-
struction, which is described:in greater detail below in
connection with Figs. 2-5, ~nd produces an electrical
~ 5~
..",~. ,,
0~9~
output signal indieative of the weight of product loaded
in the scale for sampling and calibration circuits 12.
After suitable sampling and other processing, the weight
data acquired from ~he ele~trical signals is loaded into a
weight memory table 1~ to more easily faci~itate the loca-
ti.on of weight data auring a search for th~ best combina-
ti.on~ ~n a preferred embodiment of the invention, the
memory table and all of the components in Fig. 1 apart
from the scales and the dump controls 18 are comprised by
a microprocessor having the capability of formulating the
illustrated elements when programmed. One co~mercially
available microprocessor sui~able for this function is a
model 6809E manufactured by Motorola, Inc. of Austin,
Texas.
Within the microprocessor, the weight memory
table 14 is formed from part of a random access memory
that is shared with a sequence memory table 20 . Th2
sequence memory table is operated in conjunction with an
index register 22 and a search controller 24 through a
data bus 26 to maintain a current record of the unex-
hausted sub~combinations and corresponding weights during
a search operation. A more detailed description of the
function and operation of the tables 14 and 20 is provided
below in connection with Fig. g.
An arithmetic unit ~0 receives the weight data
for the scales of each combination and adds the weights
together to obtain a subtotal for the combination. The
target weight comparator 32 compares the subtotal with a
predefined target weiqht that represents, for example, the
desired weight in each package p~oduced by the system., If
.
g7
a subtotal is equal to or exceeds the target weight, ~hen
that combination of scales will p~ovide a qualified weight
for paekaging. In comparator ~4 all qualified weights are
compared with the best qualified weight previously located
during the search operation. If the previously located
weight is larger r then the currently searched combination
and weight replace the previous best weight and combina-
tion stored in the search memory 36. ~s the search pro-
cess continues, the best weight stored in the search
memory may be periodically replaced by lower qualified
weights, and when the search process has been completed,
the minimum qualified weight and corresponding combination
may be read from the memory through a decoder 38. The
decoder supplies dump information to the dump controls 18,
and those scales comprising the best combination are
dumped and then refilled for another search. The process
continues in cyclic fashion as long as there is product
and packages to be filled.
The combination weighing ~ystem in a preferred
embodiment employs a programmed microprocessor to conduct
the search and ccmparison operations because microproces-
sors provide the speed and accuracy for performing the
arithmetic and comparison operation~ in cyclic fashion.
The processors also permit system parameters, such as the
number of scales and the magnitude of the target weight,
to be varied by simple changes in programmed data.
Pig. 2 illustrates the structure of one scale lO
with proYisions for calibrating the output signal automat-
ically between weighing operations. The scale includes a
tray structure comprised by a we;ghing tray 4U suspended
7 _
34~7
from a balance beam ~2 within the scale housing 44. Th2
tray ~0 is suspended by chains 41 from the balance beam
and may be dumped by the controls lB in Fig, 1. The
balance beam is supported by flex hinges 46, 48 from the
housing and by a range spring 50 which is placed in ~en-
sion by the ~ray 40 and weight thereon. ~rhe range spring
is secured at its lower end to the balance beam and at its
upper end to a cantilevered arm 52 of the housing by an
adjusting screw 54. The screw can be adiusted in the tare
condition to approximately center the balance beam 42
within the housing. A dash pot 56 extends between the
balance beam and the cantilevered arm 52 to damp oscilla-
tions of the balance beam brought about by the mass and
,spring components of the system.
A load sensor which in Lhe preferred embodiment
i5 a ~train gauge 60, i`R mounted at the projecting end of
a support block 62 secured in the housing. A straining
member or wire 64 is connected between the ~auqe and the
balance beam 4~.
When product is loaded into the weighing tray ~,
both the ran~e spring 50 and the strain gauge 60 are
strained; however, due to its greater stiffness, the gauge
absorbs the principal portion of the load and produces an
output signal which is:representative of the weight of the
product loaded plus any tare weight, that is the weight of
the balance beam ~2 and tray ~ which is not supported by
the range spring 50 in the unloaded condition. Prod~ct
which adheres to the weighing tray 42 after a dump opera-
tion also becomes a part of the tare weight.
In order to calibrate the scale 10 and eliminate
~;~2~ 7
tare weight from the weighing operations, a calibration
weight 70 is suspended from the housing 44 by a pneum~tic-
ally operated spring and cylinder assembly 72. Normally
the cylinder assembly i5 deactivated and the spring 74
surrounded the piston rod 76 at the upper end of the
assembly lifts the calibration weight clear of the balance
beam 72 and other ~ray struc~ure. However, during a cali-
bration operation, the assembly 72 is actuated and ~he
piston lowers the calibration weiyht 70 onto the balance
beam as shown. Through a unique coupling 78 the cylinder
assembly 72 is totally disengaged from ~he calibration
weight, and thus only the addition of the calibration
weight 70, which is of known amount, is felt by the strain
gauge 60.
Figs. 3 and 4 illustrate the unique coupling 78
and its meth4d of operation in greater detail. In Fig~ 3,
the coupling is comprised by a ~irst link 80 depending
from the piston rod 76 and a second link 82 projesting
rigidly upward from the body of the calibration weight 70.
When the weight is supported free and clear of the balance
beam 42 as shownt the links 80, 82 are contacting and in
load transmitting relationship due to gravitational force~
on the weight. However, when the calibrativn weight 70 is
lowered onto the beam 42 as shown in Fig. 4, the links 80,
82 become automatically disengaged although they are still
interlocked, ~nd the full mass of the calibration weight
including the li~k 82 is supported on the beam. The coup-
ling 78 formed ~y the links is simple in structure and
precise in nperation. The coupling totally uncouples the
air cylinder 72 whi~h lifts the calibration weight: ~rom
.~ _g_
~20~7
the beam and insures that it is only the weight and not
the cylinder or any portion thereof which influences the
measurements taken while calibrating.
It is important to have an acc~rate weight signal
from the scale especially when the contents of seYeral
scales are being measured to o~tain a desired target
weight. If each scale is inaccurate, the inaccuracies are
earried forward cumulatively into the combination weight.
Additionally; drift in the electrical or measuring por-
tions of the system may cause further error. All of these
errors can be circumvented by periodically calibrating the
output of the scale.
A calibration operation is performed by first
reading the output of the strain gauge 60 while the scale
is empty and by then lowering the calibration weight onto
the tray str~cture and taking a second reading~ The first
reading constitutes the tare weight and the second reading
represents the magnitude of the tare and calibration
weight which is known. By subtracting the two readings, a
scale factor or slope of the output signal is obtained for
use in accurate}y measuring subsequent loads above and
below the calibration weight.
Fig. 5 shows an alternate embodiment of the ~oup-
ling between the cylinder 72 and the calibration weight
70. In this em~odiment, a cup 86 secured to th2 weight 70
loosely e~velopes a ball 88 suspended from the piston rod
76 by a cord 90. The cup is swaged or otherwise closed to
orm at its upper end an aperture 92 smaller than the dia-
meter of the ball 88. The weight 70 may be raised by the
pi~ton rod 76 and be lowered and disengaged from the rod
-10-
~2;~
when the ball reposes at a central position within the.
cavity of the cup. The cup 86 and the ball 38 are func-
tionally equivalent to the links 80, 82.
Fig. 6 ill~s~rates the electrical circuitry whichacquires weight data ~rom ~he strain gauges 60 in the
scales of the combination weighing system of ~ig~ l. The
strain gauge i8 'cypically a bridge structure, and the out-
put of the bridge is fed to a high-gain instrumentation
ampliier lO0. In one embodiment of the invention, the
amplifier converts the differential ~iqnal of the strain
ga~ge to a single ended signal and amplifies it by a fac-
tor of 600.
The output of the instrumentation amplifier is
applied to a low-pass RC filter circuit 102 which has, ~or
example, a 30 millisecond time constant to suppress high
order siqnal oscillations due to vibrations of feeders and
other environmental factors surrounding the scales. The
filtering circuit operates in combination with mechanical
isolators in which the scales are generally mounted.
The sampling and calibration circuits l~ of Fig.
l include a scale selecting and sampling circuitry 106 ln
Fig. 6. ~his circuitry receives the strain gauge signals
from each scale in a multiplexer 108 controlled by a chan-
nel decoder llO. The decoder causes the multiplexer to
sample the output signals from each scale in order, and
the sampled values axe sequentially loaded into a sample
and hold circuit 112. Timing and control of the decoder
llO and sample circuit 112 is controlled by the circuitry
114, and the sampled ~ignals are transferred serially from
the circuit 112 to an analog-to-di~ital converter llfin
- ~2~
In o~der to improve the rel;ability of the weight
signals from the scales~ the signals are sampled several
times and then averaged. For example, in one embodiment
of the inven~ion having ten scales, the conversion-tO-
digital format is postponed until approximately 130 to 140
milliseconds prior to dumping of the scales. Each output
signal is then converted to a digital value onee every
five milliseconds, and the digital signals are relayed to
a data bus 122 through a buffer amplifier 118 and data bus
driver 120.
The sampl~d signals are averaged in the elements
of the data processor shown in Fig. 7. The conseoutively
sampled values from a given scale are combined in the
adder 124. The added signals are stored during the samp-
lin~ period in an accumulating register 126-1, 126-2,
126-3 . . . or 126-N corresponding to the particular scale
from which the si~nal originatedO For ~xample, the digi-
tal value of the signal from scale No. 1 is sampled six-
teen times at 5 millisecond intervals for a total sampling
period of 80 milliseconds. The sixteen samples are se-
quentially added and stored in the accumulating register
126-1. After the sampling periodr the accumulated sum is
divided by 16 in divider 128 before the signal is used in
calibration and tare circuitry 130. The process of aver-
aging the sampled signals multiple times prior to utiliza-
tion provides a more reliabl~ signal less affected by
disturbances in and around the scale. In one embodiment
of th~ invention, it has been found that the averaging
technique improves the accuracy of the system by a factor
o~ 2 to 4 time~ in comparison to a single-sample system.
-12-
z~
The calibration ~nd tare circuitry 130 obtains
the tare weight and slope during calibration to provide
during weighing a signal represen~ative of the net weight
of product in the scale. The ne~ weigh~ signal ;s then
transmitted ~o the sc31e memory table 14 for s~orage and
use during a search operation.
COMBINATION ~EP:RCH
In other combination weighing systems, the ~ech-
nique of locating the best combination of scales i~ com-
prised of examining eYery combination possible and
comparing the various combinations with one another until
the minimum combination providing a weight equal to or
greater than a given target weiqht is found~ This tech-
nique requires that the microprocessor examine 2n _ 1
combinations where "n~ i~ the number of operative scales
being examined. However, by establishing a special search
sequence havin~ consecutive steps i~ which the preqiously
examined combinations are added to one new scale not exam-
ined in the previous step; it is possible to omit or s~ip
certain steps of the seguence and thereb~ reduce the cycle
time for each search operation, For example, if a partic-
ular combination of scales has been previously searched
~nd that combination yields a total weight equal to or in
excess o the target weight, there i~ no further need to
examine other combinations in which the previously
searched and qualified combination is included. In other
words t another combination of lesser weight cannot be
found by adding other scales to the previously searched
and qualified combination~
-13-
4~
This concept is more clearly understood by exam-
ining a search sequence established in accordance with the
present invention and by discussing an example whi~h il-
lustrates the point. ~ig~ 8 shows a chart listing all lS
possible combinations in columns a-o that can be yenerated
with four different scales. Furthermore, ~he combinations
have been arranged in acoordance with the seq~ence of the
present invention which establishes a prio:rity order among
the scales. For purposes of illustratlon, it will be
assumed that the priority corresponds to the n~merical
designation on the scale, the No 1 scale having highest
priority. The sequence thus established adds one new
scale in each step of the sequence as illustrated in
combinations a-d and then replaces the lowest priority
scale in the exhausted combinations in order of prior;ty.
In other words, when combination e has been reached, all
possible combinations, including the subcombination of
scales 1, 2 and 3 have been examined and thus the com-
bination of scales 1, 2 and 3 has been exhausted. The
lowest priority scale, scale No. 3, is replaced by the
next scale, scale No. 4, in the series~ The same comments
apply to combination f since the subcombination of scales
1 and 2 is exhausted and when combination h has been
searched, the combination consisting of scale 1 itself has
been exhausted. Thus in combination i, scale No. 1 has
been replaced by the next scale, scale No 2, in priority
order.
The search sequence begins with combination a
consisting only o~ scale No. 1. Assuming that scale No. 1
does not reach the minimum combination or t~rget weight,
-14-
~;~Z~7
the next combination b is examined. Assume also that com-
~ination b does not reach the target weight and therefore
combination c is examined. If combination c exceeds th
target weight and is thus a ~ualified combination, there
is no purpose in examining combination d because it is
impossible for that combination to yield a lesser quali-
fied weight than its subcombination consisting of co~bin-
ation c. Accordingly, combination d is omitted or skipped
i n the search sequence.
In order to establish the number of steps that
can be skipped in the sequence, an ADDE~D equal to the
number of steps to advance is assigned to each combina
tion. The addends are illustrated in Fig. 8 with ~heir
associated combinations, and a brief analysis of the
addends indicates that they are equal to 2n ~, where "n"
is the number oE scales being searched and "N" is the num-
ber of the lowest priority scale in the combination.
Applying any one of the addends to the combinations shown
in the chart illustrates their utilityD For exampler if
the combination c produces a qualified weight, then there
is no need to examine combination d and thus the search
sequen~e should skip from combination c to e. The addend
value of 2 indicates that the search sequence should ad-
vance or be increased by two steps rather than one which
omits combinati~n d ~rom the search sequence. The ability
to define the addends by the expression 2n ~ is directly
related to th manner or formula by whirh the search ~e-
quence is established as d scribed above. Furthermore,
the expxession for the addend is valid regardless of the
number of scales being searched as long as the search
15-
~ ;22~
sequence is established as described.
Fig~ 9 illustrates in det~il the weight memory
table 14, the sequence memory table 20, the index register
22, the search controller 2~ and the search me~ory 36.
These elements of ~he microprocessor are the primary com~
ponents involved in the search operation apart from the
arithmetic operations performed by the arithmetic unit 30
and comparators 32, 34.
The weight memory table 14 includes a number of
memory locations or addresses which as illustrated store
the weight information and associated addend for each
scale. For eKample, the weight measured in scale 1 and
the associated addend are stored at address Ql~ Access to
either the addend or the weight data is accomplished by
the index register 22 which has a dual pointer 136 that
moves between the various addresses to read the data. The .-
addresses of the memory table are read in a predetermined
order to formulate the various combinations in the search
sequence. 5pecifically~ the index register 22 starts with
the data in the addre~s Q1 and builds the combinations
downwardly from that point. Therefore, the order in which
the addresses are scanned establishes the priority order
of the scales in the search sequence as descri~ed in con-
nection with Fig~ 8, and by loading the weight data of a
particular scale at a given address, a ~iven priority is
assigned to that scale. It will b~ understood that be-
cause a gi~en address represents a ~iven priority, the
addend data at a given address only changes with the total
number of scales being searched.
The ability ~co establish priority among 'che
--16--
~Ls~
plurality o~ scales may be used advantageously ~o ensure
that all scales in the system receive approximately the
same activity over ~ number of loading and dumping cycles.
For example~ if one scale is not dumped over a period of
cyeles, the product may gather moisture or otherwise
deteriorate, or n~ay not flow satisfactor:ily through the
packaging machine in the same manner as the fresher pro-
d~ct from other scales. ~his si~uation is remedied by
ensuring that the oldest scales, that i5 the scales that
were not included in the best combination dumped in the
previous cycle, are given priority in subsequent search~
The assignment of priority is readily accom-
plished in the weight me~ory table 14 by pushing the
weight data from the lower priority addresses toward the
top of the table after dumping and clearing the weight
data of the d~mped scales from the table. Pushing or
moving data upwardly from the bottom toward the top of a
memory stack is an elementary data proce~sing function.
The memory table 1~, in addition ~o having one
address for each scale, has one additional address
Q(n + 1) in which a lar~e negative weigh~ is stored with a
zero addend. This data is used in the search routine as
described in further detail below to indicate that the
last scale in the priority order has been examined and
that either the search is done or other subcombinations
should be examined.
Additionally, the memory table 14 may have fur-
ther addresses Q(S ~ 1) to handle spare scales that are
used to achieve best combinations from a greater plurality
of scales.
-17-
~2~ 7
The search of various combinations of scales in a
search sequence is regulated primarily by the search con~
troller 24, and the controller causes the pointer 136 of
the index register 22 ~o move back and forth ~e~ween the
various addresses of the weight memory table 14 to obtain
the weight data for proces~ing through the arithmetic unit
30 and the comparators 329 34.
The search memory 36 maintains a current record
of important parameters during a searrh operation. At
address 140, the memory stores the surrent subtotal for
the particular combination of scales being examined at a
given step in the search sequence. The step of the se~
quence corresponding to that combination is stored in
coded form at address 142. The best weight revealed in
previous steps of the search sequence and the correspond-
ing best combination identified by the seq~ence number
which produced that weisht are stored in memory addresses
14~ and 146, respectively. The best weight data i9, of
course, utilized by the comparator 34 in each comparison
step. ~he combination of scales which produces that
weight is uniquely associated wi~h the sequence number at
address 146 and the actual combination can be derived from
the sequence number through the decoder 3B. The target
weight is a fixed parameter and is stored at the beginning
of th~ search operation at address 148~ The last addend
at address 150 is stored at the end of a search operation
and is utili2ed to decode the best combination sequence
number and determine the combination. S-flag information
at address 152 is utilized as an instructional command to
permit the search sequence to start.
18-
~;~2~ 7
The structure of the sequence memory table 20 is
essentially similar to the structure of the memory table
14, ~nd data is entered, read and expunged from the table
20 at addresses identified by the index register through
the dual pointer lS~. It must be unde:rstood that the
index register 22 is a single register w~ith the pointers
136 and 156 ganged and operated jointly.
The function of the sequence table 20 during a
search i5 to maintain a record of the exhausted and unex-
hausted combinations ~ogether with the assoc~ated subto-
tals of the weights in those combinations. Additionally~
the table 20 utilizes the sequence number or alternatively
a negative punctuation mark or number to assist the index
register in advancing through the search sequence with
steps omitted as defined above. Specific operations of
the sequence table 20 are discussed in greater detail
below in conjunction with Figs. lOA and lOB~
Figs, 10~ and 108 comprise the flow chart defin-
ing the programmed search routine performed by the search
controller 24. The search routine is entered at 160 and
initially examines the S fla~ at branch 162 to determine
if a search routine should be carried ~ut. The S flag is
stored in memory 36 after each of the scales has been
loaded and a weight on the scale is confirmed~ If lo~ding
is not complete, no S flag signal is given and the ~earch
routine is exited at 163. The search controller may then
wait and reenter the search routine or take other remedial
steps until the S 1ag indicates that a search may proceed
with selected scales identified in the weight memory 14.
At the first instruction 164 of the s~earch
--19--
:~2~317
operation, initial values o the search sequence number
and weight subtotals are set in the sequence table 20.
~he ini~ial value set in the table at address S0 in ~ig. 9
represents step 0 of ~he sequence and tlle corresponding
weight is al50 0; however, in a microærocesssr, the digi-
tal codes entered for these numbers may not be numerically
equivalent.
As indicated at in~truction 166, the current se-
quence n~mber and subtotal are stored in the sequence
table and the memory 36 without consequence at this step
of the sequence, and the pointers of the index register
are stepped downwardly to addresses S1 and Ql. At in-
struction 168, ~he weight of scale 1 is added to the
current subtotal, then being 0, at address 140 in the
search memory 36. The new subtotal is then examined at
branch 170 to determine if the l~st and lvwest priority
scale in the table 14 had been examined, and thus a sub-
combination was exhausted. A large negative value is
always obtained under these circumstances by storing at
address Q ~n + 1) of the memory table 1~ a negative weight
value well in excess of any expected weight produced by
the combinations of the scales. If the negative val~e is
obtained, then the program ~ranches to a subroutine Pl at
172 as described further below~
If the subtotal is positive at branch 1709 a
second test is ordered a'c branch 174 to determine if 'che
subtotal is less than the target weight stored in the
search memory 36. ~his test is performed by the compara- ;
tor 32. If the subtotal is less than the target weight~
the corresponding combination is not qualified to form a
--~0
~a~2~
package since it is a general rule of the search process
that an.y qualifiea combination must equal or exceed th~
target weight. Unqualified weights ollow the program
branch P0 from the branch 174 to instruc~ion 176 which
causes the current sequence number, being 0, in memory 36
to be increased by lo
The instruction 176 is pursued with the aid o~ a
digital sequencer within the search controller 24, and
Yig. 11 illustrates one embodiment of the sequencer in
~reater detail. The se~uencer includes a CURRE~T SEQUENCE
register 180 which is loaded through a clocked control
gate 182 from one of several sources~ A preset sequence
number is loaded into the register 180 from preset data
184 during the initialization operation detailed at in-
struction 16~ in Fig. lOA. During the course of a search
operation, the instruction 176 of Fig. lO~ enables the +l
adder 186 so that ~he current seq~ence number is increased
by one and the increased value is then loaded into the
register through the gate 182. The output of the register
is also stored in the search memory 36. An ADDEN~ adder
1~8, which may be enabled at a later phase of the search
operation~ adds to the current sequence number the addend
associated with the new scale in a searched combination
and loads the sum in the register 180 through the gate
182~ Thus, the sequencer of Fig. 11 can be advanced in
sinyle or multiple steps d~pending upon the enabled adder
and ~he value of the addend.
In ~ccordance with instruction 166 in Fig~ lOA,
the new, current sequence number, now step 1, together
with the associated subtotal t which is the weight of the
~Z~
first combination, that is, scale 1, are now stored at
address Sl in the sequence register 20 a~d in ~he search
memory 36. Since the subtotal has not qualified at or
above the target weight, ~he subtotal is needed to esta-
blish the weight of other combinations in subsequent steps
of the search sequence. Storage of the subtotal in this
manner minimizes the number of arithmetic operations that
must be made ~o determine subtotals developed from a plur-
ality of ~cales.
Instruction 1~6 also indexes both pointers 156
and 136 of the index register downwardly to the next
address spaces in the tables 1~ and 20, and the cycle
through the branch PG is repeated with additional weights
from lower priority scales being added until eventually
the test 174 identifies a weight in ex~ess of the target
weight. The associated co~bination is now qualified and
the search process advances tn subroutine P2 at 190~ ¦
Subroutine P2 is illustrated in greater detail in
Fig. lOB. At test 19~ conducted by the best-weight com-
parator 34, the subtotal for a qualified combination is
compared with the best previous weight to determine if a
iesser qualified weight i5 pYovided by the subtotal. If
the subtotal i~ less, then instruction 194 causes the new
best subtotal to be stored in the search memory 36 togeth-
er with the sequence number. If the subtotal is equal to
or greater than the previous best sum, no chan~e in the
search memory occurs. Rejection of cornbinations equal to
the previous best weight combinations gives priority to
the older ccales at this point~
-22-
~ egardless of the results of test 192, instruc-
tion 196 causes the addend circuitry 188 in Fig~ o add
to the current seq~ence the ~ddend associated with the
lowest priority ~cale in the qualified combina~ion. It i8
this feature of the ~earch operation which allows one or
more steps of the search sequence to be omitted as ex-
plained above in connection with ~ig. 8. The omission of
unnecessary steps in the search sequence which is an
iterative process reduces the se~rch time and allows the
system to be shared among several group~ of scales or to
handle larger numbers of scales more effi~iently. Erom
instruction 196, the search operation proceed~ to subrou-
tine Rl at 198.
~ he Rl sub~outine is illustrated in Fig. lOA ~nd
includes instruction 200 that causes the previous subto-
tal, that is the subtotal stored ln the previous address
of the the sequence table, to be entered in the next
addres~ of the ~equence table. This ~tep of the operation
enables unexhausted subcombinations of the qualified com
bination~ to be further considered in the search process.
For example, if the combination c in Fi~. 8 proved to be a
qualified combination, then t~e su~otal of combination b
would be ~tored at the next address in the sequence table
f~r use in determining the weight of combination e.
In instruction 202, the address bearing the pre-
vious subtotal i~ flagged by entering a negative punctua-
t~on ~ark or number with the pr~viou~ ~ub~otal to identify
that subtotal a~ being exhausted the next time that that
addres~ of the table i~ examined through the pointer~ 156
of the index regist~r a3 explained below ln ~onnection
~ 23-
with ~ubroutine Pl~ S~ch a punctuation mark enables themicroproce~sor to readily identify and expunge rom the
table sequences and subtot~ls which have no further util-
ity in a giYen search operation. ~he instruc~ion 2~2 also
causes the previous subtotal to be stored in the ~earch
memory 36 and indexes the pointer 136 downwardly to the
next scale in the weight 14.
The ~earch continues by adding t:he wei~ht of the
next scale as indicated at instruction 168 and repe~ting
subroutines P0, Pl or P2 as de~cribed above.
After the lowest priority scale has been tested
in a combination, the next movement of the pointer 136
causes the test 170 to divert the program to subroutine Pl
shown in Fig. lOB. Under these conditions, the search is
either finished because all combinations have been tried
or the search has reached an $ntermediate staga in the
sequence where a subcombination has been exhausted~ Sub-
routine Pl cauRes the pointer 156 to inde~ upwardly
through the sequence memory table 20 to either the initi~l
S0 when the search is finished or ts some intermediate
address in the table where the sequence entry identifies a
new unexhausted ~ubcombination.
Accordingly~ having determined that the lowest
priority scale has been examined, the first instruction
210 of the subroutine Pl clear~ the sequence entxy at the
current address of the sequence table and index~s the
pointer 156 upwardly to the preceding addres~ as indicated
at instruction 212. The sequence entry is te~ted accord-
iog to the instruction 214, and the test is carried out at
~- -2~;
21~ to determine if ~he seque~nce entry is either a posi-
tive ~equence numbe~, a negative punctuation mark or zero.
The negative number indicate~ that the combination associ-
ated with that previous entry was exhau~ted, and in that
event, th~ en~ry is ~leared and the pointer 1~6 is ~tepped
further upward through the sequence table by again follow-
ing ~he instructions 210-21~. If the sequence entry i~
positive at test 216, then the ~ubcombination is not ex-
hausted and the program proceed~ ~o instruction 218.
Also, if the entry is zero, indicating that the address S0
has been reached in th~ ~able, ~he same instruction 218 is
pursued. Instruction 218 moves the pointer 156 one step
further upward in the table 20 to reach the unexhausted
combination i~ a search i~ not over. To determine whether
the search i~ over or not, the result~ from the test 216
are re-examined at the test 220. ~s~uming that the search
is not over, the sequence entry will be positive and the
search is thus ~ontinued to examine other combination~,
includinq the unexhausted combination through subroutine
Rl a~ indicated at 198.
Eventu~lly, all of the ~bcombinations are ex-
hausted and the results of the test 220 ;ndicate that the
polnter 156 has reached address S0~ The ~ero entry in the
sequencP number location is identified and in that event,
instruction 222 cau~es the largest addend to be entered in
the ~LAST ADDEND" location of the search memoxy 36, ~nd
the ~ubroutine is exited at 163. The microproces~or util-
izes the last addend figure in decodin~ the best sum com-
bination from th~ ~equenc~ number. The large~t addend is
indicative of bow ~any ~cal~ were utilized in the search
. .~
-25~
~nd is a key element in decoding the sequence number.
In summary~ a combination weighing system has
been described in which a novel scale calibration sy~tem
i~ used to establish calibrated measurements from the
scal~ between weighing operations. The signal produced by
the scale is used in a combination weiyhing system having
a plurality of similar scales~ and in order to improve the
reliability and accuracy of the measurements, the weight
signals are averaged after multiple samples are taken. To
determine the best combination of the scales approaching a
target weight, combinations of ~he weights from the scales
are examined on the basis of a predetermined se2rsh se-
quence, a~d certain steps of the sequence are omitted to
shorten the search operation.
While th~ present invention has been described in
a preferred embodiment, it should be understood that nume-
rous modifications and ~ubstitutions can be made without
departing from the ~pirit of the invention. For example,
it is not essential to utilize the precise scale str~cture
shown in Figs. 2-5 in the combination system shown in Fig.
1. Other scale~ providing welght ~ignals may also be
used~ The averaging technique improve~ reliability of the
weight signals, but it not es~ent~al to the described
searching operation~ The searching process may be carried
out with the same number of ~cales in each cycle of ope~-
a'cîon, or the micropros:essor may omit from successive
cycles tbose ~cales that have not been reloaded.. Accord-
ingly, the present invention has been described in a pre-
ferred embodiment by way oP illustratiorl s:ather than
limitation.
--2~-
