DeskJet printer chassis and mechanism design - technical

Larry A. Jackson

DeskJet Printer Chassis and Mechanism Design

THE CHASSIS OF THE DESKJET PRINTER is an injection molded plastic part that supports the mechanical and electrical systems (Fig. 1). Besides meeting its own objectives, the chassis design helps accomplish some of the overall objectives for the printer. Part count is minimized by the large amount of functionality built into this single part, and ease of assembly of the other parts to the chassis is a feature of the design. Both factors help reduce the cost of the printer.

Two important criteria for the chassis design were material selection and tooling. the material needs to be very good structurally, have good dimensional stability, and help dissipate electrostatic charge created by the paper motion. It also needs to be a good bearing material and a good snap material. Requirements for the tooling are that it be simple, fast, and durable (i.e., good for 200,000 parts).

The chassis is designed as one large part that takes the place of many parts and functions as the main structure for the printer mechanism. The chassis also integrates many of the functions of the printer.

The following is a list of the part attachment details of the chassis:

* Pressure plate spring locators

* Pressure plate location and bearing surface

* Adjustable wall and lever location and bearing surface

* Pinch roller location and bearing surface

* Drive roller location and bearing surface

* Transmission location and bearing surface

* Gear train location and bearing surface

* Paper motor screw holes

* Head driver cable and ferrite core location and holding snap for core

* Belt tensioner assembly location

* Head driver board location and retaining snap

* Prime pump bottom and location for the other parts of the pump

* Right wall retaining snaps

* Motor cable routing details

* Flex cable retaining detail

* Preloader assembly and location detail

* ESD clip and carriage rod retainer location detail

* Carriage rod location detail

* Carriage guide and paper guide location and screw hole

* Pump tube location.

The following are functional details of the chassis:

* Paper path guide surface

* Part of the input paper tray

* Envelope loading guide and surface

* Details for six grommets that mount the mechanism to the case parts

* Right corner separator used for picking a sheet of paper

* Details to reference the mechanism to the assembly pallet.

As the assembly of the product begins, a pin for the paper drive motor gear cluster is pressed in, pump tubing is put in place, and the head driver cable with its ferrite core is snapped into place. The chassis is placed on a pallet which locates on details within the chassis. The chassis then travels down the assembly line on a belt and all the other parts are assembled to it.

Material and Tooling

The material selected for the chassis is a polycarbonate with a 15% milled carbon fiber filler. This material is dimensionally stable and structurally strong enough to hold up well in the operating environment. The carbon filler helps conduct electrostatic charges from other parts and the chassis itself to ground. The material also works very well for the snap details that retain various parts and makes assembly easy. With the proper selection of mating part materials, the chassis material wears very well.

Even though the chassis is complex, the injection molding tool is fairly simple, containing only two slides and a few shutoffs. The tool is made from P20 steel, which is about as easy to machine as aluminum, but much more durable. Because the part is fairly big and complex we opted to build two gating systems into the tool, so it can be filled by a single sprue or four pin gates. As the initial parts were molded we used the single gate, but with experience, better parts were formed using the four-pin-gate system.

During the design and development of the chasis, four or five engineers were able to work on the part at the same time. There were at least two reasons for having more than one engineer involved. First, the engineers were working on parts that would eventually mate to different parts of the chassis, so it made sense for them also to develop the chassis details to fit their mating parts. Second, since the chassis is complex (fourteen E-size sheets are required to describe it), it was the pacing part in the schedule. To shorten the schedule, other engineers helped do some of the cross sections and dimensioning so the chassis could be tooled sooner. This multiple-engineer design was made possible by the HP ME Series 10 CAD system.

Paper Handling System

In the design of the DeskJet paper handling system, the primary goals were to pick a single sheet of paper from the input stack, present it to the print cartridge with the precision demanded by the 300-dot-per-inch resolution specification, and eject the sheet to an output bin. These goals were to be accomplished with a reliable, compact, and easy-to-use mechanism, and for minimum cost.

The worldwide market for this printer requires that the paper handling system work equally well with papers of different sizes (U.S.A. A-size and legal-size and metric A4), weights (U.S.A. basis weights of 16 to 24 pounds), compositions (photocopier through fine cotton bond), and textures (smooth through rough). In addition, the system must work equally well in a variety of environmental conditions, with temperatures ranging from 10[deg.]C thru 40[deg.]C and relative humidities ranging from 10% to 70%. it should be no surprise that the engineering properties of paper change significantly across those ranges. It was also a goal to allow the user to hand-feed envelopes to satisfy occasional needs.

A major constraint on the paper handling system comes from the use of inkjet technology. The ink is sprayed onto the paper wet, and requires a short drying time before it can be handled. As a result, the printed surface cannot be touched by either the printer mechanism or by another sheet of paper until the ink is dry. If it is touched too soon, the ink may smear or blot.

The paper handling system has three major functions: the picking of a single sheet of paper from the input stack, the movement of that sheet past the print cartridge, and the ejection of that sheet into an output bin. Fig. 2 shows the elements of the system.

Paper Pick

The goal that the printer be easy to use requires that paper be easy to load and that adjustments for the various sizes be minimal and simple. Paper is loaded into the front of the printer by selecting the appropriate width (typically once in the lifetime of the printer), sliding the backstop of the paper input tray out, inserting up to 12.7 mm (approximately 100 sheets) of paper into the tray, and sliding the backstop forward until it is against the back of the stack of paper. There is a small tab on the inside of the backstop, 12.7 mm off the floor of the paper input tray, which prevents the backstop from being fully seated if more than 12.7 mm of paper is inserted. This gives instant feedback to the user if there is too much paper. The adjustment for paper width is accomplished by moving a two-position (U.S.A./metric) front-panel lever to the appropriate position. The lever is attached to a sliding wall on the left side of the paper input slot. The sliding wall, a fixed wall on the right side, and a plate on the bottom form the paper input tray. The adjustment for paper length is accomplished by sliding the paper input tray backstop forward until it just touches the stack of paper in the input tray. to accommodate normal cutting tolerances ([plus-or-minus]0.7 mm) and environmentally induced tolerances ([plus-or-minus]2.0 mm), the left wall of the paper input tray is equipped with a spring-loaded guide, which takes up this tolerance and encourages the stack of paper to press against the fixed right wall. The right wall is used as an edge reference between the paper and the printer. The printer logic circuits expect the edge of the paper to be at that position and command the print cartridge to begin printing at a point just past it (allowing for an appropriate margin).

The plate that forms the bottom of the paper input tray is pivoted and spring-loaded such that, if unrestrained, the edge supporting the top of the paper (the edge inserted into the printer first) would pivot up. The plate is held down by a cam on the right side which rotates on demand from the transmission (described later). As the cam rotates, the plate rotates about its pivot and the top edge of the input stack of paper rises. The top sheet is simultaneously forced into the drive roller and into two corner separators, one at each top corner of the sheet. The drive roller is a set of three medium-soft rubber rollers (one each near the right and left edges of the paper and one centered) on a single shaft, which is rotating whenever a sheet of paper is being picked. When the top sheet of paper comes in contact with the rotating rubber drive roller, it is pulled forward and wraps around the drive roller. To ensure that only one sheet is picked, the corners of the sheet are forced to buckle over corner separators. This buckling force acts as a restraining force on the sheets of paper in the stack. Because the coefficient of friction between the drive roller and the top sheet is greater than that between the following sheets, and because the normal force is the same, the drive roller can impose enough force to overcome the buckling on the top sheet only. Because of the difficulty in modeling the engineering parameters of various papers in various environmental conditions, the geometry of these corner separators (essentially small triangles of plastic overlapping the corners of the sheet) was optimized by building a prototype printer with screwdriver-adjustable corner separators. This model was tested with the various papers and under the various environmental conditions until the best geometry was obtained. While the geometry selected is not radically different from the original design, this testing and adjusting improved the range of reliable picking considerably. Finally, after the sheet is picked, the cam continues to rotate and forces the plate, which is supporting the input stack of paper, down and away from the drive roller.

Envelopes are loaded manually by the user, bypassing the automatic pick system. The user simply presses the Load Envelope buttons on the keypad. This starts the drive roller rotating. The user then inserts an envelope into a convenient slot until it contacts the drive roller. The envelope follows the same path as an automatically picked sheet of paper.

Paper Motion

The key component of the paper motion system is the driver roller. Paper wraps around this roller as soon as it is picked and remains in contact with it until it is ejected to the output bin. The diameter of the drive roller is ground to the precision required to maintain the 300-dot-per-inch print resolution, and it is important that the paper remain in intimate contact with it. Two sets of deformable pinch rollers, one located after the pick zone and one located near the print zone, and a set of leaf springs located immediately before the print zone, ensure this intimate contact. The goal of minimum cost motivated this single-roller design, but the conflicting requirements for paper pick and precise paper feed complicated the design process. Pick rollers are generally soft, which makes the diameter difficult to control, while feed rollers are generally hard and much easier to control. This dilemma was resolved by using a large-diameter roller (the absolute diametrical tolerance is relatively constant over a wide range of diameters, so the percentage tolerance is reduced as the diameter is increased) and by careful selection of the roller material and hardness.

As the paper feeds around the drive roller and approaches the print zone, it is pushed into a sheet-metal platen and forced to peel off the drive roller and through a slot formed by the platen and a parallel piece of sheet metal, the carriage guide. The bow caused by this change in direction stiffens the paper considerably and forces it to lie flat against the platen. The platen is spring-loaded against the carriage guide so that the printable surface of any thickness of paper will be touching the bottom surface of the carriage guide and will be parallel to the platen. This is very impoartant with inkjet printing technology, because the distance from the print cartridge to the paper (nominally 1.0 mm) must be carefully controlled over the entire print zone (8 inches wide by 1/6 inch deep). Using the carriage guide as a reference, the carriage is able to reference the print cartridge accurately to the paper. The printable surface of the paper is not touched by the printer after entering the print zone, and the quality of the document is protected while the ink dries.

While the printer logic circuits can assume that one edge of the sheet of paper is referenced to the right wall of the paper input tray, no similar assumption can be made about the top and bottom edges of the sheet. The out-of-paper switch is located so it can detect both the top and bottom of the sheet and signal the logic circuits. The switch is located between the pick zone and the print zone and consists of a lever which trips an optical switch when paper is present. Noting when paper is first detected, and when it is subsequently absent, and knowing the distance from the out-of-paper switch to the print zone, the logic circuits can calculate the location of both edges of the sheet. Furthermore, because paper is detected well before it enters the print zone, the unprintable region at the top of the sheet is limited only by minimal tolerances. The unprintable region at the bottom of the sheet is somewhat more limited by the distance between the leaf springs and the print zone because the paper cannot be driven precisely after it passes these springs, but this distance is small.

Paper Eject

After the sheet of paper is printed, the ink is wet for a short while and must dry undisturbed, or it will smear or blot. rather than wait for each sheet to dry before starting the next, the mechanism includes a one-sheet buffer which allows the first sheet to dry undisturbed while the next is printed. This is accomplished by sliding the sheet onto a set of output rails as it is being printed. When the sheet is complete, the transmission selects the eject cycle, which rotates the platen down, releasing the paper from between the carriage guide and the platen. As the platen reaches the bottom of its rotation, two tabs press against mating tabs on the output rails, pushing them back from under the sheet of paper. Suddnely unsupported, the paper drops into the output bin to dry. By the time the following sheet is printed and dropped, this first sheet will be dry. When the print job is complete, the output will be stacked in the output bin, conveniently facing the user.

Prototypes were built with fixed output rails that did not pull back from under the paper, and in most cases, the paper dropped into the output bin successfully as soon as it was no longer pinched between the carriage guide and the platen. With certain particular graphics patterns, and under certain environmental conditions, however, heavy bars of wet ink would swell the paper slightly, effectively forming stiffening ribs in the paper. This reinforced paper was stiff enough to support itself with the minimal support offered by the fixed output rails, and the paper would not drop. While this result was rare and not disastrous (the following sheet would knock the troublesome sheet either into the output bin or onto the floor), the moving rails eliminated the problem completely.

Transmission

The Deskjet printer is designed with low cost in mind. One of the ways of keeping costs down is to get the maximum use out of the motors. Three mechanical operations are required of the Deskjet mechanism in addition to positioning the paper and the printhead. First, the pressure plate must be raised and lowered to load a sheet of paper into the print zone. Second, the platen needs to be rotated down and the output rails opened to eject a sheet into the output tray. Finally, the peristaltic pump is operated to prime the pen. A low-cost multiplexing device to supply power for these three independent operations from one of the motors was a design goal of the Deskjet mechanism.

There are several constraints placed on the design of this multiplexing transmission. Minimal cost is a primary constraint. Cost must be below the cost of adding additional motors, clutches, or solenoids and be sufficiently lower in cost to warrant the added development costs. Laser-quality print requires that loads placed on the carriage servo be minimal, so the paper drive motor is the motor of choice for supplying power for the three operations mentioned. The carriage motor can be used, but only as an actuation device with light loads. This constraint takes some of the burden off the development of the carriage servo system.

Another constraint on the design of the multiplexing transmission is that, because the paper motor must accurately position the paper while paper is in the print zone, no additional loads are allowed on the paper motor while printing. A final constraint is to implement the design without the use of additional electronic components such as sensors, switches, or solenoids. The constraint is intended to keep costs down and can be removed if that appears to be the lowest-cost alternative.

The Deskjet transmission provides the desired functionality within the constraints listed. Power is delivered to each of the three sy stems through three gear trains. The transmission takes the power from a gear driven by the paper motor and transfers it to one of the three gear trains. The selection of a particular gear train is done by an actuator on the printhead carriage. The eject operation occurs after printing is complete. The priming sequence is used before printing starts, and the paper load operation is timed to be complete before the paper is positioned for printing.

Transmission Design

The transmission consists of five parts plus the carriage actuator (Fig. 3). The five parts are a segmented pinion gear, a clutch gear, a follower, a trigger, and a detent spring. The carriage actuator consists of a spring with an effector attached to its end.

The segmented pinion gear consists of three gear segments placed next to one another, one for each gear train. One of these gear segments meshes with a gear driven by the paper drive motor. The gear segments are spaced apart with axial hubs. Two of the hubs are offset from the gear centerline to act as cams. The follower rides on these two cam hubs. The carriage actuator slides along the upper surface of the follower. The follower has three arms, one for each gear train. Each arm has two ledges, an upper ledge and a lower ledge. These ledges are used to position the triggers at the limits of their travel when not actuating a gear train.

There is one clutch gear for each gear train (three total). The clutch gear teeth have a face width that is twice as large as the gear segments on the pinion gear. These two gears are positioned so the pinion gear segment uses only one side of the clutch gear. The clutch gear has several teeth cut away so the pinion gear cannot drive the clutch gear. For the pinion gear to drive the clutch gear, the clutch gear must be engaged by an external driving element, which is supplied by the trigger (discussed further below). Once the clutch gear has engaged the pinion gear, the clutch gear is driven for one revolution. At that time, the missing teeth prevent the clutch gear from continuing. The clutch gear is detented into position by the spring.

The final part of the transmission is the trigger. There is one trigger for each gear train (three total). There are three details on the trigger. At the top is the ledge, which the carriage actuator uses to lift the trigger and engage the clutch gear. At the bottom is the hook, which mates to a detail on the clutch gear. The third detail is a ledge in the middle of the trigger. This ledge is used by the follower. When the follower is at the upper limit of its travel, the trigger is lifted to pull the clutch gear into its detented position. When the follower is at the lower limit of its travel, the trigger is pulled down to ensure that it will hook the clutch gear as it completes its cycle.

The transmission activites a specific gear train when actuated by the carriage. The gear train remains in motion for one revolution of the clutch gear. The system is self-initializing and requires no additional electronic input beyond the two motors. The components described met cost requirements. An additional feature of the design is that additional gear trains can be driven by adding one more clutch gear and trigger per gear train.

Paper Drive Motor

The objective for the DeskJet printer's paper drive motor and gear train was to attain laser-printer quality at 300 dots per inch with a quiet, low-cost, high-torque, easy-to-assemble drive system.

In the investigation phase, many drive systems were assembled and evaluated. We first attempted to use a small 1.8-degree hybrid step motor commonly found in disc drives. The advantage of this drive was the high torque and high precision inherent in the hybrid step motor. Unfortunately, this system did not meet our needs because of the large amount of mechanical vibration the motor produced when overdriven. The motor had to be overdriven with a high voltage drive to deliver the large amount of torque necessary to drive the printer. For reasons of manufacturability, the printer has many loose-fitting snap-together parts. These loose parts amplified the motor's vibration so much that it sounded like a fire alarm. Since the DeskJet printer is supposed to be silent, we decided to switch to a permanent-magnet, 7.5-degree "tin-can" step motor. The advantages of the tin-can motor are low cost and no vibration, but with the sacrifice of resolution and accuracy.

Since the resolution of low-cost tin-can motors is limited to 7.5 degrees per step, we were forced to use a high gear reduction to attain the 300-dots-per-inch accuracy requirement. With the printer's large 2.04-inch-diameter drive roller, the gear reduction that yields 1/300 inch per step is 40:1. At 40:1, the step motor would have to rotate at 600 full steps per second to feed paper at the minimum desired form-feed rate. With tin-can motors, the available torque drops off rapidly with increasing speed, preventing the use of the 40:1 gear reduction. The compromise was to use a 20:1 gear reduction that yields 1/150 inch stopping resolution, and to use the firmware to shift the dots in the printhead to achieve the 1/300-inch drop placement resolution.

The final gear layout is illustrated in Fig. 4. One advantage of this drive system is the large diameter of the drive roller feeding the paper. The linefeed error because of runout of the drive roller and drive gear is relatively small because the error is inversely proportional to the drive roller diameter. This allows us to use inexpensive methods for manufacturing the roller. The biggest disadvantage to this system is its high susceptibility to runout in the tin-can step motor shaft. The error caused by shaft runout is: Error = A [integral] (FIM/2)sin([theta])d[theta] where A is a constant taking into account the gear reduction and drive roller diameter, and FIM is the full indicator runout of the shaft and pinion. This function is maximum at [theta] = [pi] radians. Unfortunately, the trade-off between speed and resolution forced us to design a gear train that rotates the motor shaft [pi] radians for a standard 1/6-inch linefeed. Therefore, we had to work with the motor manufacturer to minimize the large amount of runout inherent in tin-can step motors.

Problems and Solutions

Using one motor to pick paper, prime the pen, eject paper, and accurately position the paper required some design trade-offs. The motor has to supply a high torque at slew speed to drive all these functions. Life testing with the first molded parts revealed that the torque required to operate the paper-pick cam increased rapidly with time. The motor would stall after only 100 pages of print. The increase in torque was caused by wear between the plastic chassis posts and the nine plastic gears in the paper-pick gear system. After testing many different material combinations, we finally were able to come up with a combination of carbon, glass, Teflon, and silicon fillers for polycarbonate, polypropylene oxide, and acetal plastics that would not cause an increase in torque.

To measure and evaluate the linefeed errors, a vision system was developed. This allowed us to measure dot placement accuracy down to [plus-or-minus]0.0002 inch, which is necessary when attempting to evaluate a system with linefeed errors below 0.002 inch. The vision system allowed us to determine the periodic nature of the errors and pinpoint which gear was causing a problem. For example, a large error would occur every time the follower in the transmission reversed direction. The fluctuation in the torque load caused the drive roller shaft to lift up in its sloppy bushing, thereby preventing the drive roller from rotating the proper amount. The solution was to design in a preloader spring that keeps the shaft preloaded in one position within the bushing.

Quality Assurance

Since at least ten different tolerances can affect the linefeed accuracy, a method of monitoring the linefeed had to be implemented in production. The error is measured twice daily and plotted on control charts to ensure that the DeskJet printer maintains its laser-quality print.

Carriage Motion Control

The decisions made in selecting and designing the motion control systems for the DeskJet printer reflect the overall goals for the printer. The primary goal, excellent print quality with 300-dot-per-inch resolution, requires precise knowledge of the carriage position and the ability to position the paper precisely. Cost goals required that parts costs be kept to a minimum, and the need for silent operation dictated the use of quiet motors.

Carriage Motor

A brush-type dc servo motor fit our requirements best and was chosen to drive the DeskJet printhead carriage. A hybrid step motor was considered, but a servo motor was chosen instead, for two reasons. The first reason is the relatively silent operation of the servo motor, which is important for desktop operation. The second reason is that the encoder needed to control the servo motor can also be used to provide position information for firing the printhead, ensuring accurate dot positioning over varying operating conditions. A brush-type servo motor was chosen over a brushless motor for its lower cost.

Carriage Mechanical Hardware

The DeskJet printhead rests in a carriage that slides on a stainless steel rod. Paper is fed along the bottom of a sheet-metal guide and the carriage slides along the top of the guide, so the spacing between the printhead and the paper is controlled. The carriage is held against the guide by gravity. The carriage motor is mounted on the right side of the chassis and drives the carriage via a toothed belt and pulley. The pulley has 21 teeth and a tooth pitch of 0.08 inch/tooth, giving it a circumference of 1.68 inches. Mounted under the motor is an encoder with 504 slots which, over 1.68 inches, give the encoder an effective resolution of exactly 300 dots per inch. This allows the encoder output to control the firing of the printhead.

Carriage Servo Electronics

Most of the electronics required to control the carriage servo are contained within the custom CMOS IC described on page 77. The principal components of the servo are a timer that sets the servo sampling rate, a quadrature decoder and 16-bit up/down counter that convert the two outputs of the encoder into a position that can be read by the microprocessor, and a pulse width modulator that converts an eight-bit output from the microprocessor into a pulse train with a duty cycle proportional to the processor output (Fig. 5).

The printer's Z80 microprocessor is interrupted by the timer, causing the microprocessor to read the carriage position out of the up/down counter. The microprocessor then applies a control algorithm to the position and computes an output, which goes to the pulse width modulator. The output of the pulse width modulator drives a high-power monolithic H-bridge driver, which drives the carriage motor. Pulse width modulation is used to drive the motor for its high efficiency. This keeps the power dissipation in the drivers low, eliminating the need for heat sinks or fans. The pulse width modulator operates at 19.2 kHz to prevent it from generating audible noise in the motor.

Servo Performance Requirements

The main goals for the carriage motor control system are accurate velocity and positioning control. It is important to have good velocity control while printing to maintain the print speed while also maintaining good print quality. If the carriage speed drops while printing, the throughput of the printer will decrease and printing will take longer. If the carriage speed increases while printing, the Maximum fire rate of the printhead can be exceeded and print quality will suffer. Good position control is necessary to move the carriage to the correct position before starting to print and to stop the carriage in the correct position.

A velocity control servo would have provided good velocity control while printing, but it would not have provided the positioning capability we desired. A position control servo gives us the ability to position the carriage accurately, and also gives us more accurate control of the carriage velocity while printing. A velocity control servo will have a small steady-state velocity error, but a position servo has zero steady-state velocity error.

Use of a position control servo does introduce problems that would not occur with a velocity servo, however. The first problem occurs when moving at a constant velocity. In a velocity control servo, the reference is a constant velocity. When using a position control servo, the reference is a constantly changing position, and the change in the position reference divided by the time between servo samples is equal to the desired velocity. The other problem created by a position control loop is the introduction of a dc pole, which makes compensation of the servo more difficult.

Physical Plant Model

The motor/carriage system is modeled as a second-order system, with perfect coupling between the motor shaft and the carriage. The transfer function for the motor/carriage system is: M.sub.1.(s) = W(s) / V(s) = K.sub.t / (sL + R)(sJ + D) + K.sub.e.K.sub.t where the voltage from the pulse width modulator amplifier, V(s), is the input and the shaft velocity, W(s), is the output. K.sub.t is the torque constant and K.sub.e is the voltage constant of the motor (these variables are equal when MKS units are used), L is the motor terminal inductance, R is the motor terminal resistance, J is the system inertia, and D is the system damping. However, this is a position servo loop, so M.sub.2.(s) = O(s) / V(s) = K.sub.t / s((sL + R)(sJ + D) + K.sub.e.K.sub.t.) where the position, O(s), is the integral of the velocity.

The encoder is modeled as a simple gain, K.sub.p = 4K.sub.enc./(2[pi]) where K.sub.p is the encoder gain in counts/radian, and K.sub.enc is the encoder line count in lines per revolution. The encoder outputs go to the position counter, which keeps track of position and is read by the microprocessor when it is interrupted by the timer.

The pulse width modulator amplifier is also modeled as a simple gain, K.sub.a = V.sub.s./K.sub.pwm., where V.sub.s is the motor supply voltage and K.sub.pwm is the pulse width modulator count corresponding to 100% duty cycle (full output).

The loop gain without the controller is the product of the three models, M(s) = (2K.sub.t.K.sub.enc.V.sub.s.)/([pi]K.sub.pwm.) / s((sL + R)(sJ + D) + K.sub.e.K.sub.t.)

This system has three poles. The first is at zero and is caused by the integration of velocity to position. The other two poles are located at about 4 Hz (the "mechanical" pole) and 225 Hz (the "electrical" pole).

Stability and Sampling

The design goal for stability of the servo was 30 [deg.] of phase margin. The process of sampling adds an additional phase shift to the loop. That phase shift is modeled as [theta] = e.sup.sT/2., where T is the sampling period. The phase shift at any given frequency increases as T increases, so it is desirable to keep the time between samples as short as possible to minimize the phase shift caused by sampling. The system microprocessor has many other tasks to perform while printing, so the amount of time that can be dedicated to servicing the servo is very limited. This requires that we limit the closed-loop bandwidth to obtain the phase margin that we require, and also limits the dc gain of the servo. Fortunately, when printing we want good velocity control, but precise positioning accuracy is not as important (we still know the position accurately, however). At those times when we do want to be able to position the carriage more accurately (e.g., when positioning the interposer arm in the transmission) the printer is not printing, so we can use some of the processor bandwidth that would ordinarily be used for processing dot data for servo control instead. This led us to implement a two-servo system: one for printing and the other for positioning.

Servo Design Methodology

Lead-lag compensation is used for both servos. A pole and a zero are placed at equal ratios above and below the desired crossover frequency to maximize the phase margin. The compensation was designed using classical control theory in the continuous domain and then converted to the discrete domain.

The Z80 microprocessor does not have much arithmetic capability, so an iterative design process was used to keep the compensation algorithm as simple as possible. The compensator transfer function was converted to a discrete control algorithm using a bilinear transform with prewarping at the crossover frequency, and then the coefficients were truncated to make the arithmetic easy. The new algorithms were then converted back to the frequency domain and compared to the original goals. This process resulted in simple algorithms that meet our performance requirements.

Low-Gain Algorithm for Printing

The low-gain servo operates at a sampling frequency of 300 Hz to minimize the demand on processor bandwidth when printing, while maintaining good velocity and position control. The design goals were for a crossover frequency of about 24 Hz (required for the low sampling rate) with at least 30 degrees of phase margin. The actual crossover frequency is about 20 Hz and the phase margin is about 54 degrees (Fig. 6).

High-Gain Algorithm for Positioning

A higher sampling rate (about 1200 Hz) is used when positioning the carriage in the transmission or in the service station. The higher sampling rate allows a wider servo bandwidth and higher gain, which in turn allows more accurate positioning. The initial design goal for the high-gain servo was a crossover frequency of about 72 Hz and 30 degrees of phase margin. The compensated loop gain for the algorithm actually implemented has zero dB gain at about 76 Hz, and has about 33 degrees of phase margin (Fig. 7).

Acknowledgments

Paul Harmon, Bill Huseby, Kevin Moon, and John Rhodes assisted with the design of the mechanism. Randy Krauter, Vance Stephens, and Bob Stavig provided manufacturing engineering support. Gene Frederick was responsible for plastic part tooling. Jeff Ward was responsible for sheet metal tooling, carriage motor procurement, and various design concepts. Jim Burruss at HP's Vancouver Division and Steve Witte and Mark Majette at HP's San Diego Division provided valuable advice and assistance in the development of the DeskJet servo.

COPYRIGHT 1988 Hewlett Packard Company
COPYRIGHT 2004 Gale Group