US 3793717 A
In the manufacturing of a series of Cermet thick-film resistors deposited on insulating substrates wherein the wet-printed resistor composition units are dried and then passed through a furnace on a belt to fuse glassy components, drift in resistance away from the target resistance value of the fired units emerging from the furnace, which occurs due to a variety of causes, is compensated for by controlling and changing either or both peak temperature and belt speed in accordance with a measured relationship between resistivity and time-temperature in furnace.
Description (OCR text may contain errors)
United States Patent 191 Degenkolb et al.
[ METHOD OF CONTROLLING RESISTANCE VALUES OF THICK-FILM RESISTORS Inventors: Robert Stephen Degenkolb,
Indianapolis, Ind.; Trevor Richard Allington, Hempstead, England; Yin I-Iuai Wang; Martin Oakes, both of Indianapolis, Ind.
Assignee: RCA Corporation, New York, N.Y.
Filed: Apr. 1, 1971 Appl. No.: 130,316
US. Cl 29/610, 29/620, 148/129 Int. Cl H0lc 17/00 Field of Search..... 29/620, 621, 610, 593, 407;
References Cited UNITED STATES PATENTS 7/1966 Maissel et al..... 29/620 7/1969 Ireland et a1 29/620 Feb. 26, 1974 3,520,051 7/1970 Topfer et al. 29/620.X 3,607,386 9/1971 Galla et al. 29/620 X 3,665,599 5/1972 .I-lerczog 29/620 Primary Examiner- Charles W. Lanham 4ri q t EJQKK 'MklA-LQLRQEL. mwr Attorney, Agent, or F irm-W. S. Hill; G. H. Bruestle 57 ABSTRACT 7 Claims, 6 Drawing Figures ev l/Wyn? 14/420 tflA Vfiiii 1 METHOD OF CONTROLLING RESISTANCE VALUES OF THICK-FILM RESISTORS BACKGROUND OF INVENTION So-called hybrid integrated circuits of thick-film type normally include a pattern of conductors deposited on an insulating substrate and passive components such as resistors, capacitors and inductors also deposited as thick films on the substrate and connected to the conductors. Active components such as transistors and diodes and even more complex circuit portions on semiconductor chips, are usually mounted on terminals in the circuit.
A common type of resistor is that made of a viscous ink comprising particles of metals and/or metal oxides, glass frit, organic or inorganic binders and organic solvents. Oils are also sometimes included. This type of resistor is usually referred to as a Cermet resistor.
Cermet resistors are usually deposited by screenprinting a carefully controlled quantity of the resistor ink on a ceramic substrate. In making a production run of many thousands of circuits, a machine successively prints the resistor units on a succession of small ceramic plates over a period of time which may cover a production shift or several shifts.
After the resistor units are printed on the substrate, they are dried in air (usually at elevated temperature) to permit the solvent to evaporate, then subjected to a firing operation to fuse the glass frit and burn off the binder. Most Cermet resistor compositions contain particles of silver and palladium as the metallic portion. When the composition is fired, some of the palladium alloys with the silver and the excess palladium is oxidized to a state which depends upon time and temperature in the furnace. There are other commonly used resistor compositions such as those based on ruthenium The firing may be carried out by placing the ceramic substrates on a metal belt which moves at a slow, constant speed through a furnace which has a particular heating profile. When the glass frit fuses, it freezes the oxidation state of the palladium oxide, thus stabilizing one of the most important factors contributing to the resistance value of the fired resistor.
The engineer who designsthe circuit knows the resistance value he must have in a particular resistor. He also knows how much tolerance he can permit as percent deviation from the desired resistance value. Knowing also what area he has available to allot to the resistor, he selects a resistor ink that is designed to produce a resistor having the value of resistivity required to provide the desired resistance value when fired at a specified temperature. There are many other factors, however, which enter into a selection of a resistor ink for a particular circuit application. There are electrical factors to be considered such as wattage the resistor will have to carry, voltage gradient'across the resistor and noise factor. An ink must be selected that has been designed to carry the wattage required, without burnout or change, produce noise below a certain desired level and operate properly at the voltage gradient selected.
Because there are also many variables in the resistor manufacturing process, the resistivity or resistance value that will be obtained with a given ink, given firing temperature and time, given film thickness, given screen mesh, given printing pressure, and given wire diameter, can be predicted only approximately. Some of these variables are: stretching of the screen, changes in screen wire tension, wearing of the screen, gradual thickening of the composition due to evaporation of the solvent, and variations in composition and purity of materials occurring from lot to lot and bottle to bottle. After samples are run and measured, additional adjustments must be made to bring the resistance of the initial part of a production run within the desired tolerance range.
With all normal care given to each step of the manufacturing process, two different effects always occur as a production run continues. One of these is the production of some resistors outside the tolerance range that must either be discarded or corrected in resistance value. Resistance values are commonly raised by removing a part of the resistance material. Values can be lowered by shorting out part of the current path, but this is not usually done in present production processes. The correcting process is tedious and expensive and it is of advantage to eliminate the necessity for it in as many circuits as possible.
Another effect that'occurs is a gradual drift of the mean or normal resistance value, of all units produced, away from the desired value. This drift occurs because of the production variables mentioned above.
It is often time consuming and expensive to bring the resistance value back to optimum by correcting the factor or factors responsible for the change. Wearing of the printing screen, for example, would entail frequent replacement with a new screen. Evaporation of solvent would require accurate measurement of solvent loss and frequent addition of more solvent which would then have to be thoroughly blended with the composition.
In order to lessen the production problems, a considerable effort was made previously to obtain inks which would undergo a minimum of final resistivity change with firing time and temperature. It was thought that this was the best way to minimize resistivity changes occurring in the final product.
It was found, however, that this was not a good solution to the problem. To obtain high yields of usuable circuits, designs of resistors were biased so that vmost of the resistor values would be lower than the desired value after processing. Removal of resistance material was necessary to raise most of the resistor values to the desired resistance. The many other variables in processing and materials, mentioned above, far outweighed the effects of variations in resistivity that occurred during normal firing operations due to slight changes in time andtempera'ture of firing.
It is highly desirable to have available some process variable which can be easily and quickly controlled to change the fired resistance bya known and accurate amount to compensate for resistance drift. It is an object of the present invention to provide such a control.
THE DRAWING FIG. 1 is a temperature profile from entrance to exit, of a laboratory size belt furnace used in making experimental runs to test the method of the present invention;
FIG. 2 is a graph of resistivity (p) for a particular fired Cermet resistor ink composition vs. peak firing temperature for a constant time in a furnace of the type having a profile like that of FIG. 1; l
FIG. 3 is a graph of normalized resistance (R vs. time in furnace at constant temperature profile for Cermet resistors of a particular composition, using a furnace having a profile similar to that of FIG. 1;
FIG. 4 is a graph of resistivity vs. belt speed for six different inks fired at the same temperature;
FIG. 5 is a schematic diagram of apparatus for computer control of belt speed in a thick-film resistor firing operation; and
FIG. 6 is a temperature profile of a production type furnace used in making production runs of hybrid circuits utilizing the methods of the present invention.
DESCRIPTION OF PREFERRED EMBODIMENTS In the method of the present invention, the peak firing temperature of the furnace and/or the time the resistor is in the furnace are controlled to produce resistors having a resistance value within a desired tolerance range. Changes in resistance value away from the desired range, due to the many processing variables referred to above, are compensated for by changing belt speed or peak firing temperature (or both) by a known amount.
In practicing the present method, it is first desirable to select a resistor ink with a time/temperature vs. resistivity curve with a slope (either plus or minus) which is nearly linear within the maturing (working) temperature range of the ink. The term resistivity, in this case, means sheet resistivity which is measured in ohms per square. The working temperature is that range of firing temperatures at which the glass frit fuses and the metal oxide content becomes established at a desired value. The maturing temperature may have a range of about 30 C.
The present inventors have found that, in order that the slope of peak firing temperature vs. resistivity be effective as a processing control, the slope should be within the range of about 1-5 percent. That is, for every degree (C.) change in temperature there should be a change in the resistivity, in ohms per square, of about l-5 percent. If the slope is less than about 1 percent it is too shallow to be used as a control because it requires too great a temperature change to provide a given change in resistivity. If the slope is greater than about 5 percent, a change of 1 degree in temperature produces too great a change'in resistivity to bereadily used for control purposes. Preferably, the slope of the curve should be 22.5 percent.
FIG. 1 shows a typicaltemperature profile in a laboratory size belt type furnace, designed to properly fire a particular resistor ink. The furnace is about eight feet long. When the resistor ink is purchased from the manufacturer, a preferred firing profile is supplied for producing a desired resistivity. The preferred profile is different for each different ink composition. For the inks used as examples herein, nominal firing time suggested by the manufacturer is minutes from the furnace entrance to its exit.
The circuit designer must first decide which particular ink to use, from those available, to accomplish a particular result. He selects an ink that will permit the geometry of the resistor to fall within a certain range of geometries and must also take into account restrictions on electrical characteristics such as noise factor permissible and voltage gradients and wattage that the resistor must handle. I-Ie picks a resistivity in accordance with a geometry dictated by the other factors and which will therefore produce a resistance of desired circuit value in a resistor having certain given dimensions.
In addition, in accordance with the present invention, the circuit designer should also select a resistor ink having a peak firing temperature vs. resistivity curve slope as discussed above.
In printing the resistor on a substrate, other factors which must be considered include mesh of the printing screen, pressure of printing, and screen thickness. In the examples herein the dried film thickness (before firing) was 33-39 microns and the fired thickness was about 24-30 microns. Printing was done using a mesh screen.
When a run of circuits is being made, due to the many variables which have been discussed above, the resistance values of a printed and fired resistor may beginto drift away from the allowable range and finally reach a distribution outside tolerance. In accordance with the present invention, peak firing temperature can be adjusted to bring the resistances back to a desired value. This is done by reference to a curve such as shown in FIG. 2. This curve is a plot of the resistivity in ohms/square vs. peak firing temperature obtained by firing twenty-five samples of duPont resistor ink No. 8023 at peak temperatures of 725 C., 740 C. and 750 C. The curve was drawn through the means values. If, say, 740 C. is selected as the initial peak temperature at which the run is made, at this temperature and fifty minutes firing time, a mean value of 9,700 ohms/square should be obtained in the resistors produced. As the run progresses and the screen becomes worn and other changes occur, the mean resistivity may change and become, say, 11,000 ohms/square. To compensate for this change, the operator reads off the curve the temperature change needed to bring the resistivity back' down to 9,700 ohms/square and makes the adjustment. In this case the temperature change needed is a reduction of 10 C.
Although changing the peak firing temperature is one way to adjust for drifting resistivities when a long run of circuits is being produced, it has been found, in practice, that this factor is most convenient to use to make rather large changes in resistivity. For making smaller changes in resistivity it is much more convenient to change belt speed. The present inventors have found that fired resistivities of Cermet resistor inks vary regularly with time in the furnace and that, using this single control variable, they can compensate for nearly all of the resistance changes that normally occur due to the influences discussed above.
One example of how the resistance (in ohms) of resistors made from a particular resistor ink, varies with time in the furnace, is shown in FIG. 3. In this FIGURE, the time axis shows time decreasing from left to right.
The curve illustrates that, for one composition, resistance decreases at a changing rate as firing time decreases.
A large number of experimental runs were made with a number of different duPont resistor inks to determine how changing belt speed affected the sheet resistivity of fired Cermet resistors. Results of one series of such runs is shown in FIG. 4. In this FIGURE, data are shown for six different inks having different resistivity ranges. All firing was done in the same furnace which was set to have a peak firing temperature of 730 C. Belt speeds of 3.5, 3.75, 4.0 and 4.25 inches per minute were used. The duPont inks used were numbers 7,822, 7,823, 8,022, 8,023, 8,024 and 8,025.
The graph shows that for all six inks, resistivities decrease with increasing belt. speed (decreasing time in furnace). Thus, for any one ink, to make a belt speed correction to compensate for relatively small variations in mean resistivity during a production run, one consults the original curve, reads off the speed correction necessary to effect a particular resistivity change and then either slows down or speeds up the belt drive motor accordingly. This is true, however, only when the initial sample run and the production run are close together in time and when the production run is short.
If the production run is long, the slope of the resistivity vs. belt speed curve changes considerably. In this case the original curve cannot be used to find the correction in belt speed required to bring about a particular change in mean resistivity. The new curve slope must be found.
In accordance with this invention it has been found that change in resistivity vs. belt speed follows a curve of the general form y (resistivity) a bx 0x where x is belt speed and a, b and c are constants which can be calculated after plotting the distribution of a number of samples for several different belt speeds. For small rate changes, this curve approaches a straight line and, for making rapid calculations it can be assumed to be a straight line. It may further be assumed that rates of change will be small if the process is well under control.
When the operator finds, during a production run, that mean resistivity, as calculated from a number or individual resistivities, has changed considerably, he uses this information to find one point on the new curve. He then can change ,belt speed by a known amount and make another sample run and calculate the new mean resistivity for the new belt speed. From these two points he can draw a straight line curve which has the new slope. Using this new curve he can find what belt speed is indicated for the means resistivity he actually wants.
In this method a computer is used to great advantage. The computer can continuously calculate mean resistivity of all resistors being produced, can, by having a voltage change converted to a binary number, sense a change in mean resistivity, and can calculate change in belt speed needed to bring mean resistivity back to its desired value. Belt speed can either be changed manually by an operator or automatically.
Peak temperature and time can both be used to control resistivities in making production runs of resistors in hybrid circuits. Temperatures can be used to make coarse adjustments. Together, these adjustments can be used to compensate for changes in ink from lot to lot or jar to jar and also changes in print thickness on a long time-constant basis. Thus, changes in print dimensions due to screen wear or other screen changes,
snap-off effect due to changes in screen wire tension and many other variables which occur on a long timeconstant basis, met with in the screening process, can be compensated for on a real time, in-process, control basis.
The present method does not completely eliminate the need for final correction of resistance values by abrasive or shorting out methods, but it greatly lessens this task.
It has been noted above that the present inventors have found that fired resistivities of the Cermet resistor inks of the palladium-palladium oxide-silver type vs. belt speed follows a curve of the type y a bx 0x This is a second order polynomial regression equation. If precise control of mean resistivity is desired, the predicted curve can be plotted for a particular ink and this curve, rather than a straight line approximation can be used to make corrections in belt speed. The numerical coefficients a, b and c are found by recording the resistivities of each unit of a rather large sample run (i.e., about units) for several different belt speeds at a particular temperature and feeding the information to a computer for solution. At each belt speed, the resistivities will vary over a considerable range.
In order to maintain practical production speeds without interruption, while testing and calculating is done, it is highly desirable to use automatic testing equipment and computer storage of measurement data, computer calculation of mean resistivity and computer calculation of the change in belt speed needed to keep drifting parameters within target range. An example of how this can be done is as follows.
The equipment '(FIG. 5) may comprise resistance measuring means 2 which includes sets of test probes (not shown) which can be lowered to contact pairs of metal pads connected to the ends of all resistors in the circuit. A scanner (not shown) is used to read values from each resistor in turn, after each circuit emerges from the furnace 4 on belt 6 and a digital resistance meter (not shown) (Vidar 521) is used to convert the values which are read, to ohms.
The equipment may also include two computers such as Digitalv Equipment Corporations PDP8, one of which (not shown) stores data on magnetic tape. The other computer 8 receives data measurements from the resistance meter during production and issues commands to reject or accept circuits and to vary the speed of furnace belt 6, if necessary.
' Prior to an actual production run of a particular circuit, certain values are recorded and punched on a paper tape. For each resistor in the circuit there is a target mean resistance value that the operator will try to attain. A low and a high action limit is then set on both sides of the mean. For a run to be acceptable, the mean should be kept within these limits.
Knowing the tolerance that will be permitted by circuit demands for a particular resistor, the reject high and the reject low limits can be set. If any single resistor 'is outside its allowed range, that entire circuit must be rejected.
From experience it is known how much a resistor can be increased in value by adjusting after firing to bring it within tolerance range. This adjust value must be recorded.
The computer must also be informed of the number of resistors in the circuit and the number of the control resistor which is to be used for recommending belt speed changes. The control resistor selected should be one with the closest tolerances.
Since a typical circuit may have many resistors of widely different resistance values, more than one ink composition will usually be involved. Preferably, all inks used in the same circuit should have similar characteristics. That is, resistors made from all of them should change in resistivity in the same direction and to about the same extent with changes in peak firing temperature and change in time-in-fumace. This is preferable so that measurements taken on one of them can be applied to control all of them.
The constants required for the belt speed for the control resistor, should also be recorded on the punched paper tape as should also the maximum number of outof-tolerance circuits acceptable for the sample.
A range constant is also entered for each resistor on the circuit identifying the range to which the digital measuring equipment is to be set.
The PDP8/l computers used in hybrid circuit runs had two core banks, Bank and Bank 1. Each bank of cores was divided into 40 octal pages of 177 octal words. Since this was a 12-bit word machine, indirect addressing was required for crossing page boundaries and for linking the banks.
Bank 0 the system bank, was built of:
a. Interrupt service routines b. Fortran called arithmetic and 1/0 routines c. Applications routines permitting the Fortran to address peripheral equipment via subroutines. Bank 1 the user bank contained:
a. Initialization routines b. Work waiting list c. Teletype service routine d. Arithmetic and calculation programs for measuring, sampling and predicting changes in production e. Data storage programs for historic analysis.
To start the measurement system, a command is typed to begin and to indicate that a new batch of circuits, possibly a different circuit configuration, is about to be measured. This is followed by typing the production circuit number, batch number, and then feeding in the above mentioned punched paper tape of circuit constants.
Meanwhile the circuits are going through the furnace on a belt and, as they emerge from the furnace, probes are lowered to contact the pads connected to each resistor on each circuit in succession. The scanner reads a voltage value across each resistor in turn and the digital resistance meter converts the reading to phms.
The computer then prints out the measured value of each resistor. If all the resistors in the circuit are within tolerance range, the circuit is recorded as good and is directed to a particular bin holding all the good circuits. If any single resistor is below the reject low limit or above the reject high limit, the entire circuit is directed to a reject bin. If any single resistor is below the adjust value but above the reject low limit, that circuit is directed to the adjust bin. The readings of resistance for each resistor are accumulated to calculate the mean value of each one for some selected number of consecutive circuits.
If, when the control resistor is measured, the mean resistance value of the lot moves outside the tolerance range, the operator is notified by a typed statement of belt speed change recommended. To do this the computer must be told the number of circuits to be added together for sample purposes and the frequency of sampling.
Additional data items that must be fed into the computer to effect a change in belt speed are:
1. Measurement of standard components periodically to check on the accuracy of the measuring bridge. 2. Initial belt speed which is an estimate based on samples run previous to the production run.
When the required number of readings have been accumulated for the sample, the mean resistance value of each resistor is calculated and sample totals are typed out.
If the mean of the control resistor is outside the action limit, the required belt speed change is calculated as follows:
RT= Resistance target value x Mean resistance value of sample dR/ds= Coefficient of change in resistance with belt speed Sn Present belt speed.
The change in belt speed can be made automatically. The information output of the computer 8 is fed to a digital-to-analog converter 10 which changes the voltage across the motor of belt drive means 12 to either speed it up or slow it down.
After the change in belt speed has been made, an interval is allowed for the circuit to begin to arrive at the test point, which have been fired under the changed conditions.
A small part of an actual production run is described below as a working example. The circuit had eight resistors of widely differing values. Resistors numbers 1 to 6 were made of the same resistor ink which was made up of about a 50-50 blend of Du Pont No. 8279 giving a nominal value of sheet resistivity of about 015 KQ/square when fired at a peak temperature of 740 C and corrected to a thickness of 1 mil, and No. 8281 giving a nominal value of about l.8 K-Q/Square under the same conditions. Resistors Nos. 7 and 8 were made of an ink giving a resistivity of about KQ/square.
Table I shows design value and design tolerance for each resistor.
TABLE I Resistor No. Design Value Design Tolerance R 39000 :20% R 10 K!) 15% R; 100 0 +30% -40% R 180 0 :IO% R I00 0 +20% R 220 9 (Ratio of RJR, .022i8%) R, 2.2 M!) :20% 220 K!) r2096 Most of these resistors were actually designed as pairs of resistors of equal value and, by either being connected in series or parallel, designed to give the values listed above. However, resistors Nos. 1 and 6 were simple rectangles with the following dimensions:
No. l X0.09 inch X 0.335 inch No. 6 0.083 inch X 0.38 inch The resistors were printed on the substrate to a thickness such that the final thickness, after firing, was 0.9:.1 mil. The printing head used was a Presco unit adapted to a Jade Co. mechanical transport. The dried thickness, before firing, was about 30 percent greater than this value. All resistors made of the same ink were printed in a single pass. Inthis example, resistors 7 and 8 were printed first and dried by being run through an infra-red drier for about 5 minutes. Resistors l to 6 were then printed and the drying operation was repeated. The drying operation drives off substantially all solvents from the ink compositions. The time and temperature should be long enough to make certain that drying is complete.
9 The units were then fired in an E. I. Hayes Co. F-4 production type furnace. The furnace had a length of 35 ft. and had a temperature profile as shown in FIG. 6. No temperature figures are given for the final 8 ft. of the furnace since the temperature is decreasing and immaterial to the resistance values of the fired units.
Table II gives a summary'of values for a 100 unit sample run through the furnace.
TABLE II Num LowAdj Good Hi Vhi Mean 1 1 9s 0 1 .3702 E 01 2 0 65 34 1 0 .9258 E 01 3 0 0 100 0 0 .9555 E -01 4 0 19 so 0 1 .1716 E 00 5 1 87 0 2 .9114E-0l 6 0 0 100 0 0 .2110 E 00 7 0 11 86 3 0 .2002 E 04 s 0 3 92 5 0 .2129 E 03 BS 12304 New TS 12228 CKTG=28 A=6l K=1 R=l0 tance of the control resistor closer to the design value Table II, above, is a copy of a printout from the computer. The first column gives the resistor number. The second column gives the number of resistors that were two low to be adjusted to fall within tolerance limits. It will be noted that only one circuit had one resistor (No.
5) in this category. The circuit that had this resistor had outside the tolerance limits on the high side. Since there was no convenient way to reduce the resistance values of these resistors, the circuits containing them were rejects.
Column 6 lists the number of each resistor that had very much too high values. These were also rejects.
Column 7 gives the statistical mean resistance value in K!) for all 100 resistors measured. After the E is the number of places the decimal point must be moved to get the true value in K9. in the case of resistors No. 3 and 5, a minus sign precedes the number after the E. This means that the decimal point must be moved to the left. In the other cases it is moved to the right.
At the bottom of the Table, totals are given. Note that the total number of good circuits was 28. Sixty-one circuits had resistors that were out of tolerance, within adjustable limits. One circuit (the K category) was a reject because the ratio of resistors 2 and 6 was out of tolerance. There were 10 circuits that had to be rejected because one or more resistors were either too low or too high.
Resistor No. 2 was taken as the control resistor because it had the closest tolerance limits (Table I) of :Spercent. Note that, in this run, the mean value of the control resistor was outside its tolerance limits.
Below Table II the value BS 12,304 is a measure of the actual belt speed. The number represents the number of gear wheel teeth passing a transducer in a given time, in this case ten seconds. The actual belt speed was about 26 inches per minute. Since the control resistor was out of tolerance, the computer recommended that the belt speed be changed to a new target speed, TS of 12,228 in order to raise the mean resis- In this Example the belt speed adjustment was made manually by an operator turning an adjustment knob. The operator did not adjust the belt speed to the exact figure recommended by the computer because, at this stage of production, there were additional factors to be taken into consideration that were not programmed into the computer. v
It will be noted from Table II that the total number of rejects R at the bottom of the Table is not the sum of the figures in the Low, HI and VHI columns. This is because of such factors as that some of the circuits containing one resistor in the HI category may also have one or more other resistors in the same or other reject categories. A circuit that has more than one resistor in a reject category is only counted once in the total number of circuits rejected.
Table III, below, is a copy of the computer printout for the next 100 circuits run after the belt speed had been changed by the operator after the run of Table II.
TABLE III Num LowAdj Good Hi Vhi Mean 1 0 0 97 0 3 .3990 E 01 2 0 4 O l .9878 E 01 3 O O O 0 .1062 E 00 4 0 l 93 5 l .l873 E 00 5 0 0 99 0 l .1002 E 00 6 O O 99 O l .2l38 E 00 7 l l 77 ll l0 .2289 E 04 8 O 78 12 9 .2325 E 03 BS=12255 NewTS=l2ll7 CKTG=56 A=5 K=0 R=39 It may be noted, first, that the new actual belt speed BS was somewhat higher than that recommended by the computer. However, the decrease in belt speed (increase in time-in-fumace) that was used, resulted in a rise in mean resistance values of all resistors. The mean resistance value of the control resistor (Number 2) was brought up closer to the target value of K0 and it was now within tolerance limits.
An important gain achieved was the increasein the number of GOOD circuits from 28 to 56. This is a desirable change since the GOOD circuits do not have to go through the trimming step. This is more than just a decrease in production cost. Circuits that are not adjusted by shading away part of the resistor ink are less noisy, more stable, and hold up better on life tests than circuits which have been adjusted.
An undesirable change that resulted from the above described speed change was the raising of the number of reject circuits from 10 to 39. This was mostly caused by resistors 4, 7 and 8 having been raised above the target value of mean resistance. Because of the danger of this occurring, the operator did not decrease the belt speed as much as the computer recommended.
Because of factors such as the difficulties encountered in probing and making proper contact to the resistors to test them, some of the circuits indicated in the printed summary as R (rejects) were not actually rejects. When these circuits were re-tested, some turned out to be acceptable.
1. Ina method of manufacturing thick-film Cermet 5 resistors having a certain desired mean resistivity within a certain percent tolerance range, said resistors being of the type made by depositing a continuing series of resistor ink units on insulating substrates, permitting said units to dry by evaporation of solvent, and passing the dried units at a known rate of speed through a belt furnace having a particular temperature profile and peak temperature, to mature the composition, the steps of:
printing said resistors using the same resistor ink having a peak firing temperature vs. resistivity curve slope of about 1-5 percent, plotting a curve of various peak firing temperatures vs resistivity for fired resistors made from said ink, measuring the resistivities of at least a statistical sampling of said resistors as they emerge from the furnace, automatically comparing the mean resistivity of said fired resistors with the desired mean resistivity, and
if the mean resistivity of said sampling has changed from its initial value, changing said peak temperature by an amount read from said curve, to bring it back to the desired value.
2. In a method of manufacturing thick-film Cermet resistors having a certain desired mean resistivity within a certain percent tolerance range, said resistors being of the type made by depositing a continuing series of resistor ink units, made from the same resistor ink, on insulating substrates, permitting said units to dry by evaporation of solvent, and firing the dired units in a furnace having a particular peak temperature, to mature the composition and obtain resistors having a desired mean resistivity, the steps of:
firing a series of production resistor units in said furnace, measuring the resistivities of at least a statistical sampling of said production resistors after they emerge from the furnace and determining the production run mean resistivity thereof,
measuring the amount of drift, if any, of said measured production run mean resistivity away from the desired resistivity,
determining the relationship. between fired resistivity and peak firing temperature and time in said furnace for said sampling of resistors,
comparing the mean resistivity of said fired resistors with the desired mean resistivity thereof, and adjusting the belt speed of said furnace and the peak firing temperature in accordance with said relationship an amount such as to bring the mean resistivity of subsequently fired units back to the desired value.
3. A method according to claim 2 in which only said time in furnace is varied to bring mean resistivity back to a desired value.
4. A method according to claim 2 in which the peak firing temperature is about 720 C. and time in furnace is about fifty minutes.
5. in a method of manufacturing thick-film Cermet resistors of the type made by depositing a continuing series of resistor ink units'having substantially the same geometry, and made from the same ink, on insulating substrates, permitting the units to dry by evaporation of solvent, and passing the dried units through a belt furnace having a particular temperature profile and peak temperature, to mature the composition and obtain resistors having a desired mean resistivity, the steps of:
measuring the mean resistivity of fired resistors emerging from said furnace at a known belt speed,
measuring how much said mean resistivity differs from said desired mean resistivity,
determining the relationship between fired resistivity and belt speed for said resistors, and
setting a new belt speed in accordance with said relationship and resistivity difference which will produce resistors having said desired mean resistivity.
6. In a method of manufacturing thick-film Cermet resistors having a certain desired mean resistance within a certain percent tolerance range, said resistors being of the type made by depositing a continuing series of resistor ink units on insulating substrates, permitting said units to dry by evaporation of solvent, and passing the dried units at a known rate of speed through a belt furnace having a particulartemperature profile and peak temperature, to mature the ink composition, the steps of:
printing a series of said resistors having the same geometry and using the same ink and firing them in said furnace,
determining the relationship between fired resistance and time-in-furnace for said resistors,
measuring the resistances of at least a statistical sampling of said resistors as they emerge from the furnace,
automatically comparing the mean resistance of said fired resistors with the desired mean resistance, and
adjusting the belt speed of the furnace in accordance with said relationship an amount such as to bring the mean resistance of subsequently fired units back to the desired mean resistance value.
7. In a method of manufacturing thick-film Cermet resistors having a certain desired mean resistance within a certain percent tolerance range, said resistors being of the type made by depositing a continuing series of resistor ink units on insulating substrates, permitting said units to dry by evaporation of solvent, and firing the dried units in a furnace having a particular peak temperature, to mature the composition and obtain resistors having a desired mean resistivity, the steps of:
firing a series of test run resistor units having the same geometry and using the same ink in said furnace at different belt speeds and from the results,
determining the relationship between resistance and belt speed for said unit,
firing a series of production resistor units in said furnace,
measuring the resistances of at least a statistical sampling of said production resistors after they emerge from the furnace and determining the production run mean resistance thereof,
measuring the amount of drift, if any, of said measured production run means resistance away from the desired resistance, and
in accordance with the previously determined relationship between resistance and belt speed, adjusting the time in furnace to bring the mean resistance of subsequently fired units back to the desired mean resistance value.