Method and Apparatus for HSYNC Synchronization

1. Field of the Invention

The present invention relates to a method for achieving line-to-line synchronization. The invention allows for extremely accurate horizontal synchronizing signal (Hsync) Synchronization while not interfering with clock generation. The method overcomes line-to-line synchronization problems in a simplified design, while providing near-zero line-to-line placement error.

2. Description of the Prior Art

A critical parameter which affects the output quality of any raster scanning device is the accuracy of the line-to-line synchronization. Raster scanning devices such as laser printers or video displays typically supply a horizontal synchronization signal (HSYNC) as the laser, or illuminating beam, crosses the plane of a fixed detection means, usually located before, and near the beginning of each new scan. The HSYNC signal generated on each new scan serves as an indication that the beam is crossing a fixed horizontal reference point. Digital systems such as laser printers, must monitor the HSYNC signal and use it as a reference for the serialization of digital video data for the scan. Since digital systems are clocked at fixed intervals with no synchronous relationship to the physical generation of the HSYNC signal, line-to-line horizontal synchronization error can result.

For example, a digital system with a memory bitmap in which a single bit corresponds to a physical 1/600".times.1/600" area of the printed page might be clocked at a period equal to the time required for the laser to sweep 1/600" across the imaging surface. Therefore, the leftmost serialized bit of any scan is placed at a fixed number of clocks after the clock which first sampled an active HSYNC. Since the HSYNC signal can occur at any position between the clock edges, but the digital logic can produce new bits of video data coincident with the clock edges, the physical placement of the first pel, (a pel being one bit of a memory bitmap used to represent an image in a digital memory system,) may be in error by as much as 1/600". Any inaccuracies in line-to-line synchronization cause misalignment of pels which result in visible artifacts.

The bit is translated by the imaging device into a physical space corresponding to the image resolution, such as 1/600".times.1/600". The time required by the imaging device to sweep across that area is referred to as the pel time. PELCLK is a clock used to operate portions of a digital imaging system whose period is one pel time.

A slice is generally the number of discrete periods into which each pel is subdivided. By operating on a portion of a pel, various duty-cycles and print improvements may be obtained. The slice time corresponds to the pel time divided by the number of slices which make up the pel. SLICECLK is a clock used to operate portions of a digital imaging system whose period is one slice time.

There are three primary methods of reducing line-to-line placement error.

The first involves a high speed slice clock. A design can address line-to-line synchronization by running an internal clock, SLICECLK, at a higher frequency than the clock that is synchronous with the pel generation (PELCLK). The PELCLK signal is usually generated via divider logic which produces a new active edge of PELCLK every n counts of a counter. Using such an approach, it is possible to sample the HSYNC signal at a higher frequency than PELCLK. After sampling an HSYNC signal with the SLICECLK logic, the logic must restart PELCLK for each scan by resetting the divider logic which generates PELCLK. The foregoing method produces a first edge of PELCLK which has a maximum error equal to the period of SLICECLK. For example, if the SLICECLK runs at 4 times the frequency of PELCLK, the placement error can be as much as 1/4 pel, or 1/2400" in a 600 dpi (dot per inch) printer.

While the foregoing method provides more line-to-line synchronization accuracy than simply using PELCLK, PELCLK must either be stopped at the end of a scan and restarted on HSYNC, or it must contain some discontinuity at the point where the PELCLK divider is reset by HSYNC.

A second method utilizes high speed slice clock with PELCLK synchronization on either edge of SLICECLK. Previous Lexmark.TM. printer designs have employed a method using a high frequency clock that samples HSYNC on both rising and falling edges. When an HSYNC is sampled, control logic then generates a SLICECLK which is either the "true" or "inverted" version of the sampling clock dependent on which edge first sampled the active HSYNC signal. As a result, the first SLICECLK edge is within 1/2 of a clock period from the HSYNC signal.

As was the case in the previous method, the PELCLK is generated by logic which divides SLICECLK. The logic turns SLICECLK off at the end of a scan and restarts it when either edge of the high frequency sampling clock sees the HSYNC signal. Since the divider that generates PELCLK is also reset at the end of the scan, the first SLICECLK of a new scan restarts the divider, thus synchronizing PELCLK to within 1/2 of a SLICECLK. For example, if the SLICECLK runs at four times the frequency of PELCLK, the placement error can be as much as 1/8 pel or 1/4800" in a 600 dpi printer.

The foregoing method has the advantage of providing more line-to-line synchronization accuracy than simply using PELCLK or SLICECLK. The accuracy is potentially two times better than using a single edge of SLICECLK. However, the method has disadvantages. The SLICECLK must be stopped at the end of a scan and restarted on HSYNC, thereby producing a discontinuity in SLICECLK both at the end of the scan when attempting to stop the clock, and at the beginning of the scan when attempting to restart a new one. Such discontinuities can result in clock pulse widths that violate timing specifications for clocked components. Since the SLICECLK is high frequency, pausing and restarting the clock without generating glitches becomes more difficult.

Furthermore, as laser printer speeds continue to increase, so must the clock frequencies required to accurately sample HSYNC. At sixteen pages per minute print speed on the Optra.TM. family of printers, a clock frequency of 53 mHz is required to provide 1/4800" line-to-line synchronization accuracy. At 24 pages per minute, a frequency of approximately 85 mHz is required to provide the same accuracy. With today's available silicon geometries, it is impractical to clock an entire complex pel generation logic block at such frequencies.

A third method involves clock selection. U.S. Pat. Nos. 5,438,353, 5,109,283 and 5,122,883 describe similar methods of achieving line-to-line synchronization. These patents describe a system which produces the lower frequency PELCLK and achieves synchronization with the HSYNC signal by restarting the PELCLK with its first edge nearly coincident with HSYNC. The designs generate several clocks at the pel frequency. Each clock is successively offset in phase from its predecessor by a delay element. As the HSYNC signal becomes active, the logic examines all clocks and selects the clock which has the closest phase relationship to HSYNC. The selected clock then becomes PELCLK for the scan, and the process is repeated on successive scans. FIG. 1 illustrates such a method.

The clock selection method is advantageous by allowing near perfect accuracy in the phase relationship between HSYNC and the PELCLK. The phase difference between HSYNC and PELCLK can be as small as one delay element. Also, there is no need for high frequency logic which samples HSYNC.

The clock selection method has disadvantages. Most printer systems require the ability to subdivide a pel into slices to perform such actions as pel duty cycle modulation and pel edge shifting. This requires the use of a clock that is higher frequency than PELCLK. If each scan starts a new PELCLK with various phase relationships, it may be difficult or impossible to use higher frequency SLICECLK since it has no consistent phase relationship with PELCLK. In order to use this approach in a system which requires both a PELCLK and a SLICECLK, the design may need to select among phase-shifted SLICECLK's on each new scan and then generate PELCLK by dividing SLICECLK. With a clock selection method, it becomes more difficult to stop and restart SLICECLK due to its higher frequencies.

The foregoing and other deficiencies of the prior art are addressed by the present invention which is directed to a method of achieving line-to-line synchronization having a high degree of synchronization accuracy. The synchronization task is left to a specific high-speed block of the logic which signals the beginning of scans to the PELCLK-based logic at pel boundaries, and provides sub-pel shifting as it passes data between the pel generation logic and the laser control signal.

An object of the present invention is to make the synchronization error imperceptibly small. The invention may be used to control synchronization error to a level equal to the propagation delay through a single multiplexer element. Such delays can be controlled to values less than 1/2 ns in presently available silicon geometries.

Another object of the present invention is to eliminate the need to stop and restart the clocks.

Another object of the present invention is to clock all complex pel generation logic with the relatively low frequency PELCLK without regard to the synchronization task.

These and other attributes and objects of the present invention will be described with respect to the following drawings in which:

FIG. 1 is an example of a clock selection algorithm according to the prior art;

FIG. 2 is a waveform diagram representing the signals according to the method of the present invention;

FIG. 3 is a block diagram illustrating a delay chain for implementing the method of the present invention, according to the present invention;

FIGS. 4a-c are diagrams of elements of the delay chain of FIG. 3 shown in terms of digital logic elements, according to the present invention;

FIG. 5 is a block diagram showing the delay chain of FIG. 3 with cascaded clock-based synchronizer elements of FIG. 4, according to the present invention;

FIG. 6 is a waveform diagram of clock-based synchronization while looking for HSYNC, according to the present invention;

FIG. 7 is a waveform diagram of clock-based synchronization while printing, according to the present invention;

FIGS. 8a-c are diagrams of elements of a time-based delay chain in terms of digital logic elements, according to a second embodiment of the present invention;

FIG. 9 is a block diagram showing the time-based delay chain with cascaded clock-based synchronizer elements of FIG. 8, according to the second embodiment of the present invention; and

FIG. 10 is a waveform diagram of time-based synchronization while looking for HSYNC, according to the second embodiment of the present invention.

Referring to FIG. 2, according to the method of the present invention, as a new scan line begins, the HSYNC signal goes active. At the next rising edge of PELCLK, the active HSYNC is seen by the pel-based logic. The time between the rise of HSYNC and the next active edge of PELCLK is labeled t.sub.error. The parameter t.sub.error can be between 0 and the period of PELCLK, and is unknown to the pel-based logic.

After seeing an active HSYNC, the pel-based logic waits a specified number of PELCLKs to establish a left margin. In this example, the time elapsed for the left margin, t.sub.left margin, is the time required for 3 edges of PELCLK. After the expiration of the left margin, the pel-based logic then outputs the first pel of the scan on its VIDEO.sub.-- IN signal. If the VIDEO.sub.-- IN signal were actually used to drive the laser, the total time between HSYNC and the first VIDEO.sub.-- IN is:

The parameter t.sub.error actually represents the line-to-line synchronization error and varies between 0 and t.sub.PELCLK, (the time for one cycle of PELCLK). If the t.sub.error factor were eliminated, perfect line-to-line synchronization could be achieved. Referring to FIG. 2, if the VIDEO.sub.-- IN signal were sent through a chain of delay elements, whose output is VIDEO.sub.-- OUT. Now the total time between activation of HSYNC and VIDEO.sub.-- OUT would be represented as:

The parameter t.sub.full delay is the time required to propagate through the full chain of delay elements. Both t1 and t2 still include the t.sub.error factor which represents the line-to-line synchronization error. If it were possible to reduce the time required to propagate through the delay chain by t.sub.error, the t.sub.error could be eliminated, illustrated by the following equation:

From the foregoing equation, t3 achieves perfect line-to-line synchronization by elimination of the t.sub.error factor. The elimination of the t.sub.error factor is the crux of the method of the present invention and will be described with reference to FIG. 3.

The delay chain 100 depicted in FIG. 3 is the chain that contributes to the t.sub.full delay factor for t2 and t3. Since the laser is not imaging on a printable section of its scan when HSYNC is detected, VIDEO.sub.-- IN does not need to be passed through the chain while the logic is searching for HSYNC. Therefore, the design of FIG. 3 passes an HSYNC itself into the delay chain during that time. On the first PELCLK which detects a HSYNC, the depicted "snapshot logic" measures the distance HSYNC has shifted into the delay chain. Later, when serializing the scan, the "snapshot logic" uses only the remaining delay elements to provide the delay between VIDEO.sub.-- IN and VIDEO.sub.-- OUT.

In the example shown in FIG. 3, HSYNC has shifted three elements into the delay chain at the active edge of PELCLK. For this scan, these three elements of delay correspond to the t.sub.error described. The snapshot logic then eliminates the t.sub.error factor by inserting VIDEO.sub.-- IN into the fourth element. It is important to note that t.sub.full delay must always be greater than the period of PELCLK so there are always some remaining delay elements for insertion of VIDEO.sub.-- IN. In the method of the present invention, the delay chain can either be a high speed clocked shift register or simply a series of analog logic delays. In either case, the t.sub.error factor is always accurately eliminated, even if the speed of the delay elements vary.

The digital gate-level implementation of the present invention is comprised primarily of similar groups of logic functions, called synchronizer elements. The synchronizer elements may be cascaded in a serial manner to form a means for positioning the HSYNC signal accurately within a given sampling period (such as pel time). Each synchronizer element includes two main parts, a delay stage and a memory element. The purpose of the delay stage is to provide a known amount of delay from input signal to output signal. The function of the memory element is to remember, for an entire scan, if the HSYNC signal at the start of the scan passed completely through the delay stage prior to the beginning of the next sampling period.

FIGS. 4a-c illustrate the memory elements as digital logic symbols. FIG. 4a shows the synchronous memory element 110, which includes a simple multiplexer 112 and data flip-flop 114.

When the Sample control is set, Data In is passed through to the Memory Out on an active edge of Clock. If the Sample signal is not set, the previous Memory Out is retained.

FIG. 4b shows a clock-based delay stage 120, which uses a multiplexer 122 and a flip-flop 124 as well. The Source Select chooses whether the data comes from the previous delay stage or from the input video data. On an active edge of the clock, this data is passed through to the data out.

The memory element and the delay stage are combined, in the embodiment shown in FIG. 4c, to form a clock-based synchronizer element 130. The synchronizer element is generally operated in one of two modes, synchronization mode and print mode. When in synchronization mode, the delay stage Source Select is set to choose only "Data from Previous Delay Stage". The memory element then periodically samples to determine the state of the delay stage Data Out. At the end of synchronization mode, the state of the Memory Out output is then used during print mode to determine how the Source Select should be controlled for the next Synchronization Element.

The clock-based synchronizer elements, when cascaded, form a synchronization chain 140 as shown in FIG. 5. The incoming HSYNC signal is buffered by a flip-flop 142 prior to use in the synchronization chain 140. The buffered version, Buf.sub.-- Hsync, is then sent into the first synchronization element of stage 1 as shown, and it is also sent to control logic to signal that an HSYNC pulse has arrived. When the circuit is in synchronization mode, Look.sub.-- for.sub.-- Hsync is active. After the HSYNC pulse has been recognized by the control logic, Look.sub.-- for.sub.-- Hsync goes inactive, forcing the synchronization chain 140 into print mode. The Start.sub.-- of.sub.-- Pel signal goes high for the first SLICECLK period of each PELCLK. On the Start.sub.-- of.sub.-- Pel in synchronization mode, the chain of memory elements 110(denoted the "Hold Chain" 150) sample the state of the clock-based delay elements. While in synchronization mode, these samples will occur on a once per pel basis until the HSYNC pulse arrives.

Once HSYNC enters the ASIC, it will propagate through some number of synchronization elements 130 until a Start.sub.-- of.sub.-- Pel occurs. The number of synchronization elements 130 passed through prior to the Start.sub.-- of.sub.-- Pel depends on the temporal relationship between HSYNC and PELCLK. On the first Start.sub.-- of.sub.-- Pel after HSYNC, the Hold Chain 150 samples to determine how far down the Sync Chain 140 the HSYNC has propagated. At the same time, control logic, not shown, recognizes that an HSYNC has been received (by looking at Buf.sub.-- Hsync) and shuts off the Look.sub.-- for.sub.-- Hsync control forcing the circuit into print mode. Once Look.sub.-- for.sub.-- Hsync is removed, the Hold Chain 150 is prohibited from sampling for the remainder of the scan. Also with Look.sub.-- for.sub.-- Hsync removed, the Source Selects for each clock-based delay element 120 in the Sync Chain 140 are no longer forced to look at the output of the previous stage. Instead, the chosen data source for each delay element 120 is now determined by the state of the previous stage's Hold status. Thus, if a given stage saw HSYNC while in synchronization mode, then the next stage will choose the direct Video.sub.-- In during print mode. A stage which did not see HSYNC while in synchronization mode will force the following stage to select its data output, rather than incoming Video.sub.-- In. Therefore, the Video.sub.-- In will be selected as the data source for all synchronization elements 130 along the path that the HSYNC pulse had covered during synchronization mode. The last element whose source is Video.sub.-- In will propagate its output down through the remaining synchronization elements 130, effectively bypassing all prior synchronization elements 130. The bypassing of these stages corresponds to the elimination of the t.sub.error.

Sample waveforms for clock-based synchronization are shown in FIG. 6. The goal of the circuit is to determine how many slices occurred between the start of a pel and the occurrence of an HSYNC. For simplicity, a ratio of three slices to comprise every pel is shown in FIG. 6. Therefore, the Start.sub.-- of.sub.-- Pel signal is asserted once every three SLICECLKs. Following the Start.sub.-- of.sub.-- pel in period 1, Look.sub.-- for.sub.-- Hsync began. Note that an HSYNC was recognized by the circuit as an asynchronous input which occurred sometime during period 4, thus Buf.sub.-- Hsync went high on the start of period 5. Buf.sub.-- Hsync then begins to propagate through the Sync Chain 140 setting to a high logic state Sync1, Sync2, and Sync3 on successive clocks 6, 7, and 8, respectively. Since Buf.sub.-- Hsync goes high on the start of period 5, the next Start.sub.-- of.sub.-- Pel is not recognized by SLICECLK until period 8. On the rising edge of SLICECLK to start period 8, Start.sub.-- of.sub.-- pel and Look.sub.-- for.sub.-- Hsync are both high, allowing the Hold Chain 150 to sample the state of the Sync Chain 140. The Hold Chain 150 sampling of the Sync Chain at the beginning of period 8 causes the Hold Chain 150 to go to a binary state of b`1100000` if 8 synchronization elements are used. (As shown in FIG, 6, the Hold element of the final synchronization stage need not be included. Thus the positioning of Video.sub.-- In for an 8 element Sync Chain can be determined with a 7 element Hold Chain 150.)

Once the HSYNC signal has been recognized, control logic will force Look.sub.-- for.sub.-- Hsync to go low, putting the synchronizer circuit into print mode during period 8. Once in print mode, the Hold Chain 150 will "hold" its previous value b`1100000`, for the duration of the laser scan. As stated previously, in print mode, each element of the Sync Chain 140 chooses Video.sub.-- in if the previous Hold Chain element 140 was set, otherwise it chooses the output of the previous clock-based delay element 120. The first element in the chain always chooses Video.sub.-- In as it has no previous delay element from which to choose. In FIG. 6, a Hold Chain value of b`1100000` will cause two synchronization stages (in addition to the first stage) to select Video.sub.-- In as their data source. Since in FIG. 6, the first three synchronization elements choose Video.sub.-- In as their input source, the first two stages are effectively ignored while printing. All stages beyond the third stage will then choose the output of the previous stage as their data input. Since the first two stages are ignored while printing, the circuit has the effect of calculating t.sub.error =2*t.sub.SLICECLK, and removing this error from the resulting Video.sub.-- Out. The effect on the video output is illustrated in FIG. 7.

FIG. 7 shows that the first three elements in the Sync Chain 140 all choose Video.sub.-- In as their data source. All remaining elements in the Sync Chain 140 choose the output of the previous stage. Effectively, then the first two Sync elements Sync1 and Sync2, are ignored or "subtracted out" of the delay chain, producing the desired effect. From FIG. 7, each clock-based delay element 120 has the effect of delaying Video.sub.-- Out by one SLICECLK period. If 8 elements are used for an operating point in which 3 SLICECLKs occur per pel, then the final 5 stages of delay are added to each scan, regardless of the position of HSYNC. Such a delay adds a 5 SLICECLK absolute shift to the left margin which can be adjusted out by changing the pel-based left margin delay setting. In FIG. 7, Video.sub.-- Out is delayed from Video.sub.-- In by six SLICECLKs. Had the Hold Chain 150 been set to b`0000000`, then Video.sub.-- Out would have been delayed from Video.sub.-- In by eight SLICECLKs. The two SLICECLK periods removed correspond to t.sub.error.

The foregoing synchronizing method determines the position of HSYNC with respect to a pel-based PELCLK in terms of slices, or full periods of the sub-pel SLICECLK. A second method for determining the position of HSYNC, according to the present invention, is to measure the position of HSYNC in units of analog delay elements.

In the preferred embodiment of the invention, the clock-based chain described above is used to determine the position of HSYNC relative to PELCLK in units of slices. In addition, a more finely resolved method of synchronization is used to determine the position of HSYNC relative to SLICECLK in units of delay elements. These two methods work in parallel to determine the position of HSYNC relative to a pel and relative to a slice in synchronization mode. The resulting delay chains are then connected serially during print mode to provide a calculated amount of delay in units of slices as previously described, plus a calculated amount of delay units of delay stages as described below. The effect of combining both "layers" of synchronization is to provide synchronization accuracy resolved to a single delay stage time while saving the significant number of logic gates that would be required if the time-based chain were used alone.

In the second embodiment, the basic building blocks of a time-based delay chain are very similar to those of the clock-based delay chain discussed previously. The time-based delay chain uses an identical synchronous memory element 110 to store the position of HSYNC relative to the clock. The time-based delay stage is similar to the clock-based delay stage except that the method of determining the delay differs. In the clock-based delay stage, the total delay incurred is equal to one period of the SLICECLK. In the time-based delay stage 160, the total delay incurred is equal to the total silicon speed delay through the multiplexer and the analog delay line. Thus, the total delay through a delay element is a variable number which changes depending upon actual circuit layout, supply voltage, ambient temperature of the device, and variations in delay are effectively calibrated-out by the operation of the circuit. The two foregoing building blocks are combined in FIG. 8 to form a Time-Based Synchronizer element 170.

In a manner similar to the clock-based synchronizer elements 130, the time-based synchronizer elements 170 are cascaded as shown in FIG. 9. The time-based synchronization chain 180 also has a synchronization mode and a print mode as determined by the state of the Look.sub.-- for.sub.-- Hsync signal. When Look.sub.-- for.sub.-- Hsync is active, the Sync Chain 180 is sampled by the Hold Chain 150 on every SLICECLK active edge. Thus when an HSYNC enters the ASIC, it propagates some number of delay elements. On the next SLICECLK, the state of the Sync Chain 180 is sampled by the Hold Chain 150 and the sample control for the Hold Chain 150 is shut off by Buf.sub.-- Hsync.sub.-- N, an inverted version of the Buf.sub.-- Hsync signal discussed previously. Once the Hold Chain 150 has stored the HSYNC propagation information, the control circuitry takes away the Look.sub.-- for.sub.-- Hsync control signal, causing the synchronization chain to flip into print mode. Once in print mode, the Hold Chain status is used to determine how many elements to subtract off of the total delay.

Sample waveforms for time-based synchronization are shown in FIG. 10. The circuit is initially in synchronization mode, as shown by a high level on Look.sub.-- for .sub.-- Hsync. Sometime during the first SLICECLK period, HSYNC is received. HSYNC propagates through Sync1 and Sync2 outputs prior to the next rising edge of SLICECLK. Upon the active edge of SLICECLK, HSYNC is recognized by the control circuitry (via Buf.sub.-- Hsync.sub.-- N) and causes Look.sub.-- for.sub.-- Hsync to shut off on the following SLICECLK. The low state of Buf.sub.-- Hsync.sub.-- N prohibits an additional sample of the Hold Chain 150 from taking place on the third SLICECLK shown. Therefore, the Hold Chain 150 retains the value it stored after the second SLICECLK or b`110`. This value indicates that the HSYNC signal propagated completely through the first two time-based delay elements 170 before the next SLICECLK rising edge. Thus, as with the clock-based delay elements, the first two time-based synchronization stages will be effectively ignored, subtracting off t.sub.error =2*t.sub.d, where t.sub.d =t.sub.mux +t.sub.delay element.

The function provided by this circuit allows for extremely fine resolution on HSYNC synchronization. The size of the delay element can be chosen to provide the desired accuracy in synchronization. The finest accuracy would be provided if the delay element were left out entirely, making t.sub.d =t.sub.mux +0. In a 0.5 uM technology such as the Oki.TM. Semiconductor MSM98R family, HSYNC resolution of about 500 picoseconds (or half a nanosecond) would be achieved. However, note that more delay stages are required if the delay provided by each stage is less. The minimum number of delay stages required is given by the following equation:

In this equation the minimum time delay per stage is the minimum propagation delay possible through a delay stage under best case environmental and process conditions. If fewer stages are used, the time-based synchronization chain will not be long enough to accurately measure all HSYNC pulses.

Pulse width modification is a primary concern when utilizing a long combined delay path to create accurate output pulses. With a long combined path, the additive effects of the differences between the low-to-high and high-to-low propagation delays of the components can cause a perfect input pulse to be greatly stretched or shrunk by the final output stage. Some small amount of pulse width modification is unavoidable, but the effects can be minimized. To minimize the possibility of pulse width modification, components were chosen so that differences in propagation delays would cancel each other out. Multiple tap points along the delay chain are provided, so that while running faster SLICECLKs, an earlier position along the delay chain can be chosen so as to skip the latter portion of the chain, thereby avoiding any pulse width modifications incurred during that portion of the chain.

As mentioned previously, the total delay through a delay element is a variable number which changes depending upon the actual circuit layout, supply voltage, ambient temperature of the device, and variations in silicon fabrication processing which effect speed. Variations in environmental and process conditions will be automatically compensated out by the time-based synchronization circuit. For example, if the silicon is 20% fast, then if 5 stages would normally comprise t.sub.error, then the circuit will compensate during the synchronization phase to select 6 stages for t.sub.error.

Since the same logic gates are used both as a means for measuring HSYNC and as the method for providing the necessary delay to video data, the effect of variations in circuit layout are minimized. For example, assuming that the third delay stage is 10% longer than the other delay stages, since the same logic gates are used for the third delay stage for both synchronization and printing, then the resulting accuracy of the circuit varies only by 10%. If, however, separate logic chains were used for synchronization and printing, the errors might be additive so that a 10% increase in the third stage of the synchronization coupled with a 10% reduction in the third stage of print delay would produce an overall (110%/90%)-100%=22% variation in synchronization accuracy, worst case.

For the circuit to operate successfully, a couple of points must be addressed. Once the HSYNC is detected, the higher-level pel-based logic must be informed by the synchronizer. For reliable operation, the signaling to the pel-based logic must always occur on the same active edge of the pel-based clock. If variation exists in signaling the pel-based logic, then the synchronization function will vary by an entire pel time.

Tap points may be provided along the clock-based chain as the selected VIDEO.sub.-- OUT whenever the number of slices per pel is less than the maximum number possible. For instance, if the current operating point has three slices per pel, the output of the third synchronization element may be chosen as VIDEO.sub.-- OUT rather than the nth element as shown in the figures. That minimizes the absolute delay applied to all data in three slices per pel mode.

Some methods of tuning the video pulse width allow the pulse width to be tuned to a finer resolution than the slice rate. If these methods required use of synchronous logic, then these functions would need to be inserted between the clock-based synchronization layer and the time-based synchronization layer. For example, video pulse widths may be tuned in a special mode based on either edge of SLICECLK. In another mode, video pulse widths can be tuned using delay lines to achieve even finer video pulse width control. Both methods for tuning video pulse widths are performed logically between the clock-based and time-based synchronization chains.

The delays between gates vary depending upon a number of factors including actual component placement and wiring within the die. The variability in placement and wiring can be reduced significantly if the subject logic is restricted to a small portion of the die. The present implementation includes floor planning of the silicon so that the SLICECLK based logic would be restricted to a small portion of the die, thereby minimizing variability due to placement and wiring. The circuit can thus achieve the highest obtainable silicon performance using semi-custom logic design methods.

The present invention, unlike conventional systems, performs synchronization on the data path as opposed to the clock path. The measurement of error between an synchronous event such as HSYNC and a system clock is achieved by allowing HSYNC to propagate into a chain of delay elements. Synchronization error is eliminated by delaying the video output by the remaining delay elements of the chain.

The synchronization method of the present invention allows the clock which controls the video conditioning logic to run continuously, thereby providing a clean, simple clock waveform crucial to modern high-level synchronous logic design methodologies.

The method of the present invention provides the capability for layering the accuracy of synchronization based upon the available clocks within the ASIC. For example, the specific implementation, set forth previously, provided the primary layer of synchronization relative to a pel boundary, and a secondary layer of synchronization relative to a slice boundary. The layering method allows highly accurate synchronization for an order of magnitude fewer logic gates than other approaches.

The accuracy of the synchronization provided is not limited by the speed at which clock based logic will run. Through the use of internal delay line chains, the synchronization concept can be expanded to virtually any accuracy needed, limited only by the silicon speed of the integrated circuit.

By providing selectable output tap point for choosing the source of the synchronized video, the minimal number of delay elements can be chosen based upon the frequency of the slice based clock. By allowing for a selectable tap point, the total variability introduced by the delay line chain length can be minimized.

The method of the present invention of synchronization further utilizes the same logic gates to serve as both the calibration detector while sampling the HSYNC signal and as the video stream delay while printing. By using the same logic gates, the effect of inaccuracies caused by the physical placement of the logic gates within the device and process variations across the die are minimized.

It is contemplated as part of the present invention that an asynchronous signal will be fed into a chain of clocked or analog elements, and the number of elements traversed into the chain at a next clock edge will be measured. Alternatively, measurement can be performed using a high speed digital counter/timer to measure time between an synchronous event and a discrete clock edge. In another embodiment an analog voltage ramp is initiated at an asynchronous event and the voltage at the next clock edge is then examined and remembered.

The delay can be adjusted by inserting data into the chain of clocked or analog delay elements, and eliminating error by bypassing a number of the delay elements which correspond to the error. Alternatively, the data can be inserted into a beginning of the chain of clocked or analog delay elements, and the error eliminated by bypassing a number of the delay elements at an end of said chain. The number of bypassed delay elements is derived from said error.

In another embodiment it is contemplated that data will be inserted into the beginning of multiple chains of clocked or analog delay elements, and each of the chains will have a different length. The error is eliminated by selecting an output of one of the chains depending upon the measured error.

The data may be inserted into a first-in-first-out (FIFO) structure, and extraction begun from the FIFO structure at a delay in time derived from the measured error. Yet another alternative inserts data into a chain of storage elements whose high-speed clock rate has been adjusted by an amount derived from the measured error.

The foregoing invention is based on the measurement of error as the relationship between the synchronous event and the next clock edge. With such a definition of error, the adjustment is performed by subtracting delay from the final output. The method can also be performed by reversing the definition, and measuring error as the relationship between a clock edge and the synchronous event. In such a case, the error could be eliminated by adding an appropriate delay to a final output.

Having described several embodiments of HSYNC synchronization in accordance with the present invention, it is believed that other modifications, variations and changes will be suggested to those skilled in the art in view of the description set forth above. It is therefor to be understood that all such variations, modifications and changes are believed to fall within the scope of the invention as defined in the appended claims.