CA1310743C - Dual port video memory system having a bit-serial address input port - Google Patents

Abstract

RCA 84,124 Abstract of the Disclosure A dual port digital memory system, which includes integral memory sequencing circuitry, is controlled by a bit-serial address and control signal. The address and control signal, which includes a data read address, a data write address and a control value, is serially loaded into a shift register. The address sequencing circuitry loads the read and write address values into integral read and write address registers and, based on the control value, initiates respective read and/or write operations.

Description

~3~7~
-1- RCA 8~,124 DUAL PORT VIDEO MEMORY SYSTEM HAVING
A BIT-SERIAL ADDRESS INPUT PORT

This invention relates to a digital memory system which has separate data input and data output ports and which has an additional serial input port for specifying the addresses used to access data and the type of access to be provided.
The design of a memory system used for field or frame based video signal processing in a consumer television receiver is subject to several constraints affecting both the performance and cost of the memory system. First, the memory system should accept and supply pixel samples at standard video rates. A commonly used video sampling rate in NTSC receivers, for example, is 14.32 M~z, or 4fsc, four times the frequency, fsc (3.58 MHz), of the color subcarrier signal component of the composite video signal. This sampling rate is preferred because it simplifies the processing of the color signal components of the composite video signal.- As a second --constraint, the memory system should include a sufficient number of storage elements to hold pixels representing one field of the video signal. This number of storage elements depends on the signal standard and on the sampling rate.
For example, in the NTSC system 238,875 pixel storage elements are needed to represent one field of a video signal that is sampled by a 14.32 MHz sampling clock signal. Third, the memory system should be able to be configured as delay element or as a random access memory.
When the memory system is configured as a delay element, the pixels provided at the output port are delayed by a ixed amount of time, for example, one field period, with respect to the samples applied to the input port. In this mode, the memory system may be part of a field progressive scan system, frame comb filter or temporal noise reduction system. When the memory system is configured as a random access memory, blocks of pixels may be written or read in an arbitrary seguence. This mode is useful for features -2- 1 31 ~ ~ ~ 3 RCA 84, 124 which produce an on-screen displa~ or for producing picture-within-a-picture (pix-in-pix) or magnified picture (zoom) displays. A fourth constraint which affects the design of a video memory system is its cos~. Desirably, the memory system should be inexpensive and should not need expensive peripheral circuitry to integrate it into the television receiver.
Previously, random access memory (RAM) integrated circuits developed for computers and other data processing applications have been adapted for video signal processing.
These integrated circuits are usually designed for the random access of individual multi-bit data words using a single data bus for both input and output. To use these memory circuits in a video memory system, it is necessary to use peripheral circuitry to control the sequence of addresses applied to the memories and to schedule the read and write requests -to avoid conflicting uses of the data bus.
One attempt to design a memory integrated circuit - 20 specifically--for-video signal processin~ applications i5 -- - -described in the paper entitled I'A l-M bit DRAM with 33-MHz Serial I/O Ports" by Ohta et al., IEEE Journal of Solid State Circuits, Vol. SC-21, No. 5, October 1986. The integrated circuit described in that paper has two four-bit serial ports, one each for data input and data output. In the memory circuit, four-bit data storage elements are arranged in blocks of eight. Four-bit nibbles representing portions of pixel values are applied to the input por-t at a 33 M~z rate. The nibbles are loaded into a serial-to-parallel input buffer which collects eight nibbles before writing them into a block of storage elements. Similarly, a block of eight nibbles is read from the memory into a parallel-to-serial buffer which pro-vides eight four-bit nibbles representing portions of pixel values in a serial stream at the 33 MHz rate. The read and write operations may be interleaved to achieve continuous data input and output at the 33 MHz rate. Since this memory system stores _3_ ~3 ~ ~ 3RCA 84,124 and provides data i~ four-bi-t nibbles, two of these memory systems are needed for processing eight-bit pixel values at the 33 MHz rate.
Although the memory system described in the above-referenced paper includes circuitry to control data bus contention resulting from coincident read and write requests, it does not significantly improve the memory address interface over that which is provided by random access memory circuits designed for data processing applications. Accordingly, the integrated circuit described in the above-referenced paper may need extensive peripheral circuitry to control the proper sequencing of address values applied to the memory circuitry. In addit.ion, because the address values are applied to the integrated circuit in parallel, it is contained in a 40-pin package. This is a relatively large pac~age which may be too costly to use in an inexpensive consumer television receiver.
The present invention is embodied in a digital data memory sys-Lem which includes a block~oriènt~ed data' ~``'~' ''''' storage array where each block has a uni~ue address value.
The memory system has a single address input terminal by which at least a read address value and system control values are applied to the memory system. The memory system also includes means for storing the applied address value.
Control circuitry, internal to the memory system is responsive to the control values for selectively addressing the stoxage array with the stored address value to provide the data held in the addressed block at an output port of the memory s~stem.
Brief Description of the Drawing FIGURE 1 is a block diagram of a memory system which includes an embodiment of the present invention.
FIGURES 2A and 2B show timing diagrams of several signals that are useful in explaining the operation of the memory system shown in FIGURE 1.

-4~ ~3~7~ RCA 84,124 In the drawings, broad arrows represent busses conveying multiple-bit parallel digital signals and line arrows represent connections conveying single bit digital signals. Depending on the processing speed of the devices, 5 compensating delays may be required in certain of the signal paths. One skilled in the art of digital memory circuit design would know where such delays are needed in a particular system. The memory system described below accepts and provides four-bit nibbles at the NTSC sampled data video rate of 14.32 MHz. To simplify the explanation of this memory system, these four-bit nibbles are referred to as pixel values. A memory system which operates at 14.32 MHz and uses eight-bit pixel values may be built by configuring two of these memory systems in parallel.
FIGURE 1 is a block diagram of a video memory system which may be realized as a single 18-pin integrated circuit. The pin designations correspond respectively to signal names used in FIGURE 1 as shown in TABLE I.
TABLE I
`2G Pin/Signal Name Function DI0~DI3 4 data input signals WE write enable signal (active low) SAD serial address signal SAS address clock signal 25 TAS address transfer signal (active low) CLK system clock signal GW write clock gate signal (active low) GR read clock gate signal (active low) OE output enable signal (active low) 30 DOo~D3 4 data output signals The two remaining pins are for coupling the integrated circui-t to a source of opera-ting potential and a source of reference potential (e.g. ground).
The memory system shown in FIGURE 1 may include, for example, a sufficient number of bit storage elements (cells) in the-memory cell array 50 to hold-262,144 (2183 four-bit pixel values. This number of cells is sufficient ~5~ ~ 3 ~ 3 RCA 84,124 -to hold, for example, one field (238, 875 four-bi-t pixel values) of a sampled da-ta NTSC video signal having a sampling frequency of 14.32 MH~ (4fc~.
This memory system is able to accept and supply four-bit pixel values at standard video rates. The pixel values applied to the memory system may be written into storage cells addressed by a write address value while other pixel values held in s-torage cells addressed by an independent read address value are provided by the memory system. The write and read address values are incxemented as the pixel values are accep-ted and provided in order to maintain continuous input and outpu-t data streams. The address values may be changed and read or write operations may be enabled or inhibited by loading address and control values into the memory system via the serial address input terminal SAD. Another function of the control value loaded via the SAD terminal is to allow the memory read and write operations to be synchronized to the signal TAS.
A structural overview of the memory system is - 20 pLesented first, foll-owed by -a more detail-ed ~escription. - -The latter part of this detailed description includes an example of the internal sequencing of the memory illustrating its operational modes.
Four-bit pixel values, applied to the data input terminals DIo through DI3, are stored se~uentially in respective serial-to-parallel converters 38B through 38E.
The serial-to-parallel converters 38 collect, for example 32 pixel values during 32 successive periods of a write clock signal, WCK, and then transfer the collected pixel values as a single block of data into buffer latches 40B
through 40E respectively. This transfer occurs in a fraction of the next successive clock period. The pixel values are transferred from the latches 40 to the memory cell array 50 during several of the ensuing 32 clock periods while new pixel values are loaded into the serial-to-parallel converters 38. Similarly, a block of 32 pixel values is read from the memory cell array 50 over several periods of a read clock signal, RCK, and stored in -6- ~ 3 ~ ~ 7 ~ ~ RCA 84,124 output buffer latches 52A -through 52D. From the latches 52A through 52D, the block of pixel values is transferred to respective parallel-to-serial converters 54A through 54D
during a portion of one of the read clock cycles. The parallel-to-serial converters 54 then provide the individual four-bit pixel values, synchronous with the read clock signal, RCK, via the output terminals DOo through DO3 over the next 32 periods of the signal RCK.
The addresses applied to the cell array 50 and the control signals, which determine when the data are transferred from the serial-to-parallel converters 38 and latches 40 to the memory cell array 50, and from the cell array 50 to the latches 52 and parallel-to-serial converters 54, are generated by circuitry internal to the memory system shown in FIGURE 1. This circuitry, the memory sequencing circuitry 22, is controlled by externally provided address and control signal values as outlined below.
Read and write address values, each of which 20 specifies-a-particular ~lock of data ce-lls--in-the-m-emoLy~~~ --~-~ ~~
cell array 50, are applied to the SAD input terminal of the memory system as a bit-sequential signal. The signal applied to the SAD input terminal is separated into one read address and one write address and s-tored in the internal read and write address registers 24 and 26 respectively.
The memory system also includes an automatic refresh function which may, for example, restore all of the data in a predetermined portion of the memory cell array in one operation. The refresh address is internally generated and stored in an internal refresh address register 28.
The memory sequencing circuitry 22 may, for example, sequentially perform a read, a write and a refresh operation on the memor~ cell array 50 during one 32 clock cycle interval. These operations are timed by the sequencing.circui~y..22.to.pre.vent any conflicting.access.. .. .
to data in the memory cell array. Generally, during each of the read, write or refresh operations, the address value - _7_ ~ 3 ~ 3 RCA 84,124 used to access the memory is incremented by one in address incrementing circuitry 32 and s-tored into the read, write or refresh address register as appropriate.
Thus, the memory system shown in FIGURE 1 is capa~le of reading and writing pixel data using the same se~uence of successive block address values or using respectively di~ferent sequences of successive block address values. When the addresses used for reading and writing the pixel data are the same, the memory system may appear to be a one-field-period delay element suitable for use in field or frame based video signal processing circuitry. When different addresses are used for reading and writing, the memory system may appear as a delay element providing a time delay of less than one field period or as a "random access" memory where the pixel data are accessed in blocks of 32 four~bit pixel values.
The following is a more detailed description of the memory system shown in FIGURE 1. As set forth above, the read and write address values and the control value are 20 applied as a bit-serial signal to -~he serial address input ``
terminal S.~D. In the present embodiment of the invention, each of the address values are 13-bit values and the control value is a three-bit value. The bit-serial signal is loaded into the address input buffer lO in synchronism with a serial address clock signal -SAS. The address input buffer lO may, for example, be realized as a 29 bit serial input, parallel output shift-register. The read and write address values and the control value are applied to the serial input port and provided via parallel output ports of the buffer lO. In the bit-serial signal SAD, used in -this embodiment of the invention, the bits of the read and write address values are interleaved. The most significant bits of the two address values are the first two bits of the signal followed by the next most significant bits of the two address values and continuing on to the least significant bits. -The-three-bit control signal is appended to the interleaved addresses following the least '~` 8 ~ 3 ~ ~ 7 ~ 3 RCA 84,124 significant bits of the two address values. The signal TAS
is coupled to the read and write address latches 16 and 14 to control the transfer of the read and write address values from the address input buffer 10 to -the respective latches 16 and 14. Each of the latches 16 and 1~ may include, for example, 13 data-type flip-flops arranged as a parallel-in-parallel-out register. The address values provided by the buffer 10 are loaded into the latches 16 and 14 on the negative-going edge of the signal TAS.
The timing of an address load and transfer is shown in FIGURE 2. At time TAo, in this example, the memory system contains pixel data but has no outstanding read, write or refresh operations scheduled. The first bit of a new pair of address values ARo and AWo is applied to the address input terminal SAD at time TAo. The individual bits of the read address value, ARo, and write address value, AWo, and the three-bit control value are sequentially loaded into the address input buffer 10, one bit per period in the 29 periods of the signal SAS
~~ 20 ` followi'n'g ti'me TAo'." At time TA1, the address input buffer 10 contains a 13 bit read address value, a 13 bit write address value and a three-bit control value. On the negative going edge of the signal TAS, occurring at TA1, the read and write address values are transferred to the respective read and write address latches 16 and 14 and the three-bit control value is transferred to the memory sequencing circuitry 22.
The three control bits, designated RF, WF and MF, from the input buffer 10, and the signal TAS are applied to the memory sequencing circuitry 22 to control the operation of the memory system. The bits RF and WF control whether data are read from or written to the pixel storage elements addressed by the respective read and write address values in the respective latches 16 and 14. The bit MF determines whether these read and write operations are synchronized by circuitry internal to the memory system or by the externally supplied signal TAS. The function of the control signals RF, WF and MF are summari~ed in Table 2.

9 ~ 3~ ~ 7~ RCA 84,124 Control sit Value Effect RF 0 read operation enabled 1 read operation disabled WF 0 write operation enabled 1 write operation disabled MF 0 read and/or write synchronized to TAS
1 read and/or write operation internally synchronized In the example shown in FIGURE 2A, the three control bits, RF, WF and MF provided at time TA1 all have values of logic zero. Accordingly, data applied to the memory system via the input terminals DIo through DI3 after time TAl will be ~ritten into the pixel storage elements addressed by the ne~ly supplied wri-te address value, AWo; data provided via the output terminal DOo through DO3, after a suitable delay (e.g. 32 periods of the signal CLK3 will be data ~rom the pixel storage elements addressed by the newly supplied read address value, ARo, and the read and write operations using -~ 20 these address values~re~in1tiated by the nega~ e~oing edge of the signal TAS.
Referring to FIGURE 1, the read and write address values held by the latches 16 and 14, respectively, are applied to respective first input ports of the multiplexers 20 and 18. Second input ports of the multiplexers 20 and 18 are coupled to the output port of the address incrementing circuitry 32. The multiplexers 20 and 18 are controlled by the respective control signals SELR and SELW
provided by the memory sequencing circuitry 22. When new addxess values are available in latches 16 and 14, the memory sequencing circuitry 22 pulses the signals SELR and ~ELW, two and one-half master clock pulses after the negative golng edge of the signal q`AS to condition the respective multiplexers 20 and 18 to apply the address values ARo and AWo held by the respective latches 16 and 14 to the input ports of the read address register,24 and write address register 26, respectively. When new address values are not available in the latches 16 and 14 the -10~ ~ 3~7L~ RCA 84,124 signals SELR and SELW condition the multiplexers 20 and 18 to apply the address value provided by the address incrementing circuitry 32 to the input port of the respective read and write address registers 24 and 26. The 5 read address register 24 and write address register 26 are controlled by respective signals LDl and LD2 provided by the memory sequencing circui-try 22 to load new read and write address values, respectively, when the signals SELR
and SELW are pulsed or when respective incremented read and 10 write address values are available at the output port of the incrementing circuitry 32.
In addition to the 13~bit read and write address values held by the registers 24 and 26, a 10-bit refresh address value, held by a refresh address register 28, is 15 applied to the multiplexer 30. The input port of the refresh address register 28 is coupled to the output port of the incrementing circuitry 32~ Th~ ten least significant bits of the value provided by the incrementing circuit 32 are stored in the refresh address register 20 coincident with a pulse of the signal LD3 provided by thc memory sequencing circuitry 22.
The multiplexer 30 is controlled by a two-bit signal MC provided by the memory sequencing circuitry 22.
The signal MC conditions the multiplexer to apply either 25 the read address value from register 29~, the write address value from register 26 or the refresh address value from register 28 to the memory address register 31. The selected address value is loaded into the register 31 coincident with the positive-going edge of signal LD4 30 provided by the memory sequencing circuitry 22.
The 13-bit address value held by the memory address register 31 is applied to the address incrementing circuitry 32. The circuitry 32 may include, for example, a 13-bit binary adder (not shown) having the value held by 35 the register 31 applied to one input port and a value of one app]ied to the other input port. The addxess value available at the output port of the incrementing circuitr~
32 is applied to the refresh address register 28 and to the ~ 3 ~ RCA 84,124 second input ports of the respective multiplexers 18 and 20 as set forth above.
To understand how the indlvidual da-ta storage elements are addressed by the address value held in the regis-ter 31, it is convenient to regard the memory cell array as a matrix of blocks of pixel cells. The matrix has 21 ~1024) rows and 23 (8) columns. Each unique combination of a row and column address value corresponds to a respectively different block of 32 four-bit pixel storage cells. To access a particular block of cells, the ten least significant bits of the address value held by the memory address register 31 are applied to row address decoding circuitry 34 and the three most significant bits are applied to column address decoding circuitry 36. The lS row decoding circuitry 34, which may be of conventional design, decodes the ten least significant bits of the address value to select one of the 1024 rows, and the column decoding circuitry 36, which may be of conventional design, decodes the three most significant bits of the - 20 address value-to select-one of the-eight columns. -~he block of pixel storage cells selected by the row and column decoding circuitry 34 and 36 is gated onto a 32 x 4 (128) bit data bus, I/O, by conventional circuitry. The data bus I/0 is coupled to the data input gating circuits 42A
~5 through 42D and to the data output latches 52A through 52D.
Data are provided to the memory cell array 50 via the data input terminals DIo through DI3, serial-to-parallel converters 38B through 38E, data input latches 40~ through 40E and data input gating circuits 42A
through 42D.
Data applied to the write enable input terminal, ~E, and to the four data input terminals DIo through DI3 are stored in the respective serial-to-parallel converters 38A through 38E. A new bit of data is stored in each of the five serial-to-parallel converters coinciden-t with negative-going edge of each pulse of the clock signal WCK
provided by the memory sequencing circuitry 22. The signal WCK may be, for example, the logical AND of the system ~ 3 ~
-12- RCA 84,124 clock signal CLK and an inverted -~ersion of the write clock gate signal GW. When the signal GW is a logic one, the clock signal WCK is disabled and, so, data values applied to the serial-to-parallel converters 38 are not stored.
When the signal GW is a logic zero, however, the signal WCK
is enabled and data are stored in the converters 38 synchronous with the negative-going edges of the clock signal CLK.
In this embodiment, the memory sequencing circuitry 22 counts pulses of the signal WCK, as set forth below, to determine when the data held in the serial-to-parallel converters 38 is to be transferred to the five 32-bit latches 40A through 40E. The latches 40A
through 40E, each of which may include, for example, 32 data-type flip-flops configured as a parallel-in-parallel-out register, are controlled by the memory sequencing circuitry 22 to load the 32 four-bit pixel values and the corresponding 32 bits of the signal WE
coincident with the positive-going edge of a signal WLD.
The output values from the 32-bit latches 40B
through 40E are applied to respectively different ones of the data input gating circuits 42A through 42D. Each of these gating circuits includes, for example, 32 three-state gates (not shown). Each of the three-state gates in a gating circuit is connected to a respectively different output terminal of its associated data input latch. The output terminals of the three-state gates are coupled to respectively different bit lines of the data bus I/O. The 32 three state gates in each of the ga-ting circuits 42 are controlled by the signals generated by respectively different ones of the 32 OR gates 41. The OR gates 41 combine a signal ENW and a corresponding bit value of the 32-bit signal WE stored in the register 40A in a logical OR
function to generate 32 control signals applied -to the 32 three-state gates of each of the data input gating circuits 42.
The memory sequencing circuitry 22 provides a logic zero value for the signal ENW to -transfer, to the ~13- ~3~7~3 RCA 84,124 memory cell array 50, -the four-bit pixel values held in the latches 40B through 40E, as masked by the values of the signal ~ Held in latch 40A. The signal WE masks the pixel values as follows. When a bit of the signal WE is a logic zero, the corresponding three state gates in each of the gating circuits 42A through 42D are enabled to apply the corresponding four-bit pixel value to the data bus I/O of the memory cell array 50. When a bit of the signal WE is a logic one, however, the corresponding three-state gates in each of the input gating circuits 42A through 42D present a high impedance to the data bus I/O. Under these conditions, the value currently held in the corresponding pixel value storage cell of the addressed block in the memory cell array 50 is re-written into that storage cell.
By this operation, a logic one of the input terminal WE
effectively masks the corresponding pixel data value applied to the input terminal DIo through DI3 by preventing it from being written into the memory cell array 50.
A memory read operation proceeds as follows in this embodiment of the invention. Thirty-two four-bit pixel values are transferred from an addressed block of pixel value storage cells in the memory cell array 50 to the data output latches 52A through 52D when the signal ENR, supplied by the memory sequencing circuitry 22, has a value of logic-one. Each of the data output latches 52A
through 52D may include, for example, 32 data-type flip-flops configured as a parallel-in-parallel-out register. The 32 four-bit pixel values held in the latches 52A through 52D are transferred into four 32-bit parallel-to-serial converters 54A through 54D, respectively, coincident with the positive-going edge of a pulse of the signal RLD, supplied by the memory sequencing circuitry 22. Each of the parallel-to-serial converters 54A through 54D may include, for example, 32 data type flip-flops configured as a parallel-in-serial-out shift register.
Each of the parallel-to-serial converters 54 provides one-bit of the 32 four-bit pixel values ~3~7~3 -14- RCA 84 ,12D~

synchronous with a read clock signal RCK, provided by the memory sequencing circuitry 22. The signal RCK may be, for e~ample, the logical AND of the master clock signal CLK and an inverted version of the read clock gate signal GR.
In order to maintain a con-tinuous stream of four-bit pixel values from the parallel-to-serial converters 54, the memory sequencing circuitry 22 may, for example, include a counter ~not shown) which counts the pulses of the signal RCK. This counter is reset when new pi~el values are loaded into the parallel-to-serial converters 54. When the counter has counted 32 pulses of the signal RCK (indicating that the 32 pixel values have been shifted out of the converter 54), the memory sequencing circuitry 22 pulses the signal RLD to load 32 new pixel values into the converters 54 and resets the RCK
counter (not shown).
The output terminals of the four parallel-to-serial converters 54A through 54D are coupled to respective three-state gate circuits 56A through 56D.
The three-state gate circuits 56 are controlled by an externally supplied output enable signal OE. When the signal OE has a value of logic one, the three state gate circuits 56 present a high impedance at the output terminals DOo through DO3. When the signal OE has a logic zero value, however, the circuits 56 present the logic values, applied to their input terminals, at the output terminals DOo through DO3.
In addition to controlling the reading of data from, and writing of data into the memory cell array 50, t~le memory se~uencing circuitry 22 also controls the periodic refreshing of data values stored in the memory cell array 50. To understand how the refresh operation occurs, it is helpful to first understand why the data values are refreshed. In the embodiment of the invention shown in FIGURE 1, the memory cell array 50 is composed of 22 (1,048,576) dynamic bit-storage celJ.s`arranged as 213 (8,192) blocks of 32 four-bit pixel storage cells. A datum is stored in one of these dynamic bit-storage cells as a ~L31~7f~3 -15- RCA 84,12 charge or absence of a charge on a capacitor (not shown).
This charge is no-t stable and tends -to dissipate over time.
Consequently, the memory system includes an automatic refresh operation which periodically restores the charge on all of the capacitors to maintain the data in the storage cells. To understand how the refresh operation is performed, it is helpful to know the general structure of the memory cell array. The memory cell array 50 used in the present embodiment of the invention may be considered to be organized as a matrix having 1024 rows, 256 columns and four bit planes. The four bit planes correspond, respectively, to the four bits of pixel values. The 256 columns on each bit plane are organized in eight groups where each group includes 32 columns, all 32 columns in a given group have the same column address. In this memory organization, each of the 256 columns in each of the four bit planes has a separate sense amplifier (not shown).
When the sense amplifier is gated onto the bus I/O, the charge on the capacitor of the addressed bit cell may be read onto the bus or written from the bus (i.e-. changed to equal the new bit value applied via the bus I/O). In a refresh operation, the bit values provided by the sense amplifiers are not gated onto the data bus, consequently, all eight column select lines from the column decode circuitry 36 may be activated for each 10-bit row address provided to the row address decoder 34. Thus, each refresh operation restores 256 pixel values (1024 bits of data).
All of the data in the memory cell array 50 may be refreshed in 1024 refresh operations (i.e. a refresh operation for each row in the matrix).
The automatic refresh system used in the memory system shown in FIGURE 1 refreshes all of the bit-storage cells in the memory cell array in less than 5 milliseconds.
To do this, one row of the memory cell array is refreshed every 64 cycles of the clock signal CLK. A typical refresh operation proceeds às follows. Once a refresh address has been established, the memory sequencing circuitry 22 pulses a signal REF applied to the column decoding circuitry 36 ~3~7~
-16- RCA 84,124 and to the memory cell array 50 to activate the sense amplifiers in all 256 columns on all four-bit planes of the memory cell array 50 and to inhibit the gating of the sense amplifiers to the data bus. The sense amplifiers read and restore ~he data held in each of the dynamic bit storage cells in the addressed row.
The structure of the individual memory bit storage cells and of the sense amplifiers depends on the technology (e.g. NMOS or CMOS) used to implement the memory system and may also depend on factors such as the desired data access time and the power dissipation for the memory system. Exemplary circuitry for implementing the memory storage cells, sense amplifiers and address decoders is shown in an article by K. Kimura et al. entitled "Power Reduction Techniques in Megabit DRAMIs", IEEE Journal of Solid-State Circuits, Vol. SC-21, No. 3, June 1986, pp.

3~1-389.
FIGURES 2A and 2B are signal timing diagrams which are useful for illustrating the operation of the memory system under the control of the memory sequencing circuitry 22. The timing values shown in these figures are specific to this embodiment of the invention. They are provided to illus-trate one possible implementation of the invention.
The master clock signal CLK used in this embodiment of the invention has a frequency of 14.32 MHz, 4fc in the NTSC system. The other clock signals, SAS, WCK
and RCK are all derived from this master clock signal.
The timing diagrams in FIGURES 2A and 2B show a series of memory functions which illustrate the operation of the memory sequencing circuitry 22. As an aid in explaining the timing diagrams time values are marked across the top of the FIGURES 2A and 2B. The time values TAo, TAl, TA2 and TA3 designate t,imes at which memory address events occur, either the start or finish of an - address load operation.- The time values TC0 through,TC13 mark the boundaries of internal memory cycles, each memory cycle spans eight periods of the master clock signal CLK.

-17- ~L 31 ~3 7 l~ 3 RCA ~34 ,124 Memory operations which affect the memory cell array 50 (i.e. memory read, write and refresh) occur during respectively different memory cycles and are initiated on the memory cycle boundaries The time values TBo, TBl and TB2 mark times at which data are transferred from the serial-to-parallel converters 38 to the data input latches 40 and/or times at which data are transferred from the data output latches 52 to the parallel-to-serial converter 54.
Briefly, between times TAo and TA1, a read address value ARo and a write address AWo are loaded into the memory system. Read operations using ARo and ARo+l are initiated at times TCo and TC5, respectively. A write operation using memo~y address AWo is initiated at time TC4. Between times TA2 and time TA3, a read address value ~Rl and write address value AWl are loaded into the memory system. A memory write operation using address AW1 is initiated at time TC8 and a read operation using addr~ss ARl is initiated at time TCg. Memory refresh operations are initiated at times TCl and TC1o. These timing diayrams and the explanation set forth below define the function of the memory se~uencing circuitry 22 sufficiently to allow one skilled in the art of digital logic design to assemble circuitry from conventional components which may be used as the memory sequencing circuitry 22.
To understand the sequencing of the memory system, it is convenient to separate the circuitry which processes the address and control signal values from the circuitry which performs the read, write and refresh operations. As set forth above, at time TAl, the read and write address values ARo and AWo are stored in the respective read and write address latches 16 and 14. One master clock period after the signal TAS has been pulsed, the memory sequencing circuitry 22 pulses the signals SELR
and SELW to condition the respective multiplexers 20 and 1~
to apply these address values to the read and write address registers 24 and 26, respectively. Shortly-after-the positive going edges of the signals SELR and SELW, the circuitry 22 pulses the signals LDl and LD2 to load the 7 ~ ~
-18- RCA 84,124 respective address values ARo and ~W0 into the respective read and write address registers 24 and 26.
As set for-th above, the MF bit of the control value applied to the memory sequencing circuitry 22 at time TAl is a logic zero. Conse~uently, the read and write operations which uses the address values supplied at are synchronized to the signal TAS. In the present embodiment of the invention this is accomplished by resetting the counters (not shown) in the memory sequencing circuitry 22 which count pulses of the respective signals WCK and RCK, and by scheduling a memory read operation. To indicate that a read write or refresh operation is scheduled, the memory sequencing circuitry 22 may include, for example, three set-reset latches (not shown). The first, second and third latches are set by the circuitry 22 to schedule a read, a write and a refresh operation, respectively and are reset bv the circuitry 22 when the corresponding operation is initiated. To prevent any conflict when more than one operation is scheduled, a priority is established such that a write operation has the highest priority followed by a read operation and then a refresh operation. This priority assignment is somewhat arbitrary although it is generally desirable for refresh operations to have the lowest priority.
Since the counter (not shown), which tallies pulses of the signal WCK is reset by the signal TAS, the 32 four-bit pixel values, applied to the data input terminals DIo through DI3 over the 32 periods of the signal WCK
following the negative going edge of the signal TAS are sequentially stored in the serial-to-parallel converters 38B through 38E. To simplify the explanation of the memory sequencing circuitry 22, it is assumed that the input terminal WE is held at a logic zero value so that none of the four-bit pixel values are masked from being written into the memory cell array 50 as set forth above. The en-try of pixel data into the serial=to-parallel converters 38 is started at time TA1 when the value of the externally 19- 13~7~ RCA ~4,124 supplied write clock gate signal ~W is changed to a logic zero.
The time TCo is at the start of the first memory cycle shown in FIGURE 2A. At -time TCo, a read operation and a refresh operation are scheduled. According to the priority scheme set forth above, the read operation is initiated first.
The first step in the read operation is to change the signal MC to condition the multiplexer 30 to apply the read address value ARo from the read address register 24 to the memory address register 31. One clock period after time TCo, the signal LD4 is pulsed to load the read address value into the memory address register 31. The address value held by the register 31 is applied to the row decoding circuitry 34 and column decoding circuitry 36 to select a block of pixel storage cells from the memory cell array 50. Two and one-half periods of the signal CLK after time TCo, the signal ENR is pulsed for five periods of the master clock signal CLK. While the signal ENR has a logic one value, the data output latches 52 are gated onto the bus I/O. Coincident with the negative going edge of the signal ENR, the 32 four-bit pixel values held in the addressed block of pixel storage cells are transferred to the output latches 52. Seven periods of the signal CLK
~5 after time TCo, the signal LDl is pulsed to load the incremented read address value from the incrementing circuitry 32 into the read address register 24.
The data held in the latches 52 at time TCl is not transferred to the parallel to serial converters until a time 32 periods of the signal CLK after time TAl. This time is shortly after time TBo in FIGURE 2A. The delay of 32 master clock periods between the completion of the read operation and the availability of the pixel data at the output terminals DOo through D03 allows the memory sequencing circuitry 22 to initiate the read operation, which was initiated in the present example at time TCor either at time TCo, TCl or TC2. This flexibility ensures that any potential conflict between read, write and refresh ~31~3 RCA 84,124 operations can be resolved before the data read from the memory cell array 50 is expected a-t the output port.
At time TCl, the portion of the read operation which accesses the memory cell array 50 is complete. At this time, the refresh operation, which has been scheduled since time TCo, is initiated by the memory sequencing circuitry 22. The circuitry 22 first changes the state of the signal MC to condition the multiplexer 30 to apply the refresh address value from the refresh address register 28 to the memory address register 31. The circuitry 22 then pulses the signal LD4, one master clock period after time TCl, to load the refresh address value into the register 31. On the third clock period after time TC1, a pulse having a width of three clock periods is applied, as the lS signal REF, to the column decoding circuitry 36 and to the memory cell array 50, to refresh the addressed row of bit-storage cells in each of the four bit planes of the memory cell array 50. Seven master clock periods after the time TCl, the signal LD3 is pulsed to load the ten least significant bits-of-the incremented refresh address value, as provided by the incrementing circuitry 32, into the refresh address register 28. At time TC2, the refresh operation is complete.
Starting at time TA2, a new address and control value is applied to the serial address input terminal SAD
and loaded into the input buffer 10 synchronous with the clock signal SAS. This address load operation occurs in the same sequence of steps as the one set forth above and is not described in detail. At time TA3, a new read address value, ARl, and a new write address value, ~Wl, are loaded into the read and write address latches 16 and write address latches 14, respectively, and new control values RF, WF and MF, having respective values of zero, zero and one, are loaded into the memory sequencing circuitry 22.
The effect of these new address and control values on the operation of the memory `system is described below in relation to the description of the memory system in the time following time TC5.

~21- ~ 3~ 3 RCA 84,124 At time TC2, a read operation and a refresh operation have been performed on the memory cell array 50 and no write operation is scheduled since only 19 four-bit pixel values have been loaded into the serial-to-parallel converters 3~. Since no memory operations are scheduled at time TC2 or at time TC3, none are performed. At time TBo, the 32nd pixel value is loaded into the serial-to-parallel converters 38. Coincident with the positive-going edge of the next successive pulse of the signal WCK, the memory se~uencing circuitry 22 pulses the signal WLD to transfer the 32 four-bit pixel values from the converters 38 to the input latches 40. At the same time, the circuitry 22 resets its internal counter (not shown~ which tallies the pixels applied to the serial-to-parallel converters 38 and schedules a memory write operation.
In the example shown in FIGURE 2A, the time at which the contents of the serial to-parallel converters 38 are to be transferred to the latches 40 is 32 periods of the master clock signal CLK after time TA1. This is also the time at which the pixel data values read from the the block memory cell array 50 having the address ARo are to be supplied via the output terminals DOo through DO3.
Accordingly, at the same time that the circuitry 22 pulses the signal WLD, it also pulses the signal RLD to transfer the contents of the output buffer latches 52 to the parallel-to-serial converters 54. At this time, the circuitry 22 also schedules the next read operation.
Simultaneously, the externally supplied read clock gate signal GR and output enable signal OE are changed to have logic zero values, enabling the signal RCK and the three-state gates 56. Synchronous with the signal RCK, the 32 four-bit pixel values held in the parallel to serial converters 54 are provided via the output terminals DOo through DO3.
At time TC4 a read operation and a write - - -- - operation are scheduled.- Since- the write operatioIl has a higher priority it is ini-tiated by the memory sequencing circuitry 22. At time TC4 the signal MC is changed by the -22- ~ 3~07~ RCA 84,124 circuitry 22 to condition the multiplexer 30 to apply the write address value AWo, held in the wri-te address register 26, to the memory address register 31. One master clock period after time TC4, the circuitry 22 pulses the signal LD~ to load the address value AWo into -the register 31.
Two master clock periods after time TC4, the memory sequencing circuitry 22 changes the signal ENW to a logic zero value, for five periods of the master clock signal, to gate the 32 four-bit pixel values held by the latches 40 lQ onto the data bus I/O. During the time interval that the signal ENW has a logic zero value, these 32 pixel values are written into the addressed block of memory storage cells. Seven master clock cycles after time TC4, the signal CK2 is pulsed to load the incremented write address value from the incrementing circuitry 32 into the write address register 26.
Referring to FIGURE 2B, at time TC5, the memory write operation is complete and a memory read operation is scheduled. The read and write address values, AR1 and AW1, supplied at time TA3, one-half of one master clock cycle before time TC5, have not yet been propagated to the read and write address registers 24 and 26, respectively.
Consequently, the read operation initiated at time TC5 uses the incremented read address value (i.e. ARo+l) from the read address register 24 rather than the newly provided read address value AR1.
The address value ARo+1 is transferred from the read address register 24 to the memory address register 31 by the signal LD4 emitted by the memory sequencing circuitry 22 one master clock period after time TC5. Two and one-half master clock periods after time TC5 (or three clock periods after time TA3) the newly provided read and write address values are loaded into the respective read and write address registers 24 and 26 by the memory sequencing circuitry 22 by pulsing the signals SELR, SELW, LD1 and,LD2 as set forth above. - - --The read operation initiated at time TC5 proceeds in the same manner as the read operation which was -23- ~3~74:3 RCA 84,124 initiated at time TCo except -that the memory sequencing circuitry 22 does no-t pulse the signal LD1 seven master clock cycles after TC5 to load the incremented address value into the register 24. The memory sequencing circuitry disables the signal LD1 and LD~ to prevent the newly loaded address values AR1 and AW1 from being overwritten by the incremen-ted address value from a read or write operation which may be in pxogress as the new address values are supplied. The signals LDl and LD2 are only disabled when the control value MF of the newly loaded address and control signal is a logic one. In this instance, the read and write operations using the newly supplied address values occur only after any currently scheduled read and write operations. When the control value MF is a logic zero, the sequencing circui-try 22 schedules a new read operation which overrides any currently scheduled read operation. In addition, it resets the internal counter which tallies the pulses of the signal WCK, postponing the next memory write operation.
Time TBl is 32 mas-ter clock periods after time TBo. In the time interval between the times TBo and TBl, 32 four-bit pixel values have been stored in the serial-to-parallel converters 38 and 32 four-bit pixel values have been provided by the parallel-to-serial converters 54. Consequently, coincident with the positive going edge of the first pulse of the signal CLK following time TB1, the memory sequencing circuitry 22 pulses the signals RLD and WLD to transfer, respectively, the 32 four-bit pixel values stored in the output la-tches 52 into the parallel-to-serial converters 54 and the 32 four-bit pixel values held in the serial-to-parallel converters 3~
into the input latches 40. The memory sequencing circuitry 22 also schedules a read operation and a write operation, one master clock period after time TB1.
The four-bit pixel values applied to -the data input terminals DIo through DI3 between time TBl and TB2 are written into the block of pixel value storage cells which is addressed by the write address value AW1. In this ~3ln7~3 -24- RCA 84,124 example, these pixels are to be subsampled in a one-to-two ratio. This subsampling may be performed, for example, to compress the lines of pixel data stored in the memory cell array 50 by one-half to produce a compressed image for a pix-in-pix feature. The subsampling is achieved by applying a bilevel signal, having a frequency that is half the fre~uency of the master clock signal, as the write clock gate signal GW. As set forth above, the signal WCK
is the logical AND of the master clock signal CLK and an inverted version of -the signal GW. The signal WCK
determines which pixel values are loaded into the serial-to-parallel converters 38. It is also used by the memory se~uencing circuitry 22 to determine when to trans~er the collected pixel values to the input latches ~. The signal WCK after time TBl is a clock signal having a 25~ duty cycle and a fre~uency that is one-half of the ~e~lency of the signal CLK. Responsive to this signal, ~e serial-to-parallel converters store only every other one of the four-bit pixel values applied to the input terminals DIo through DI3. The pixel values that are stored are those occurring coincident with the negative-going edges of the clock signal WCK.
At time TC8, a write request and a read re~uest are pending. According to the priority scheme set forth ~5 above, the write request is initiated at time TC8. The 32 ~ r-bit pixel values applied to the memory system in the in~erval between T~o and TB1 are written into the block of pi~el storage cells that are addressed by the write address value AWl loaded at time TA3. This write operation is ~a pe~ormed in the same manner as the one that was initia-ted ~t time TC4. Accordingly, it is not described in detail.
Time TC8 is 64 master clock cycles after time T~o ~ the time that the previous refresh opera-tion was s~eduled. Since the memory system is designed to refresh ~5 one ro~ of bit storage cells every 64 master clock cycles, a refresh operation is scheduled at time TC8. -- -~
At time TCg, the memory sequencing circuitry 22initiates a read operation to transfer the 32 four-bit 13~L~7~3 -25- RCA 84,124 pixel values, addressed by the read address value, AR1, from the m~mory cell array 50 into the output latches 52.
This read operation is performed in the same manner as the one that was initiated at time TCo and, so, is not desc~-ibed in detail.
The refresh operation, which was scheduled at time TC8, is performed at time TC1o. This refresh operation proceeds in the same manner as the one initiated at time TC1, consequently, it is not described in detail.
l~ The time TB2 is 32 periods of the clock signal RCK after time TB1. At time TB2, all the 32 four-bit pixel values which had been transferred to the parallel-to-serial converters 54 shortly after time TB1 have been provided by the converters 54 via the output terminals DOo through DO3.
Coincident with the positive-going edge of the first pulse o ~le master clock signal after time TBl, the memory se~llencing circuitry 22 pulses the signal RLD to transfer t~e 3~ four-bit pixel values held in the latches 52 into e parallel-to-serial converters 54. In this example, ~0 these values are to be provided at the ouiput terminals DOo through DO3 at one-half the rate at which -the prior pixel values were supplied. This may be desirable, for example, to expand the lines of video samples provided by the memory system to produce a magnified image. The expansion of the ~5 lines of samples is accomplished by changing the read clock ~ate signal, GR, to be a bilevel signal having a frequency that is one-half the frequency of the signal CLK. This si~nal, G~, changes the signal RCK to have a 25% duty cycle and a frequenc~ one-half of the frequency of the signal 3~ CLK. Responsive to this signal, RCK, the parallel-to-s~rial converters 54 provide the 32 four-bit pixel values ~-t ~l~ desired rate.
Since the memory seguencing circuitry 22 schedules read and write operations after counting 32 ~ulses of the respective signals RCK and WCK, no new write ~eration will be scheduled until a time (not shown) 6~ --master clock cycles after TBl and no new read operation ~ 3~7~3 RCA 84 124 will be scheduled until a time ~not shown) 6~ mas-ter clock cycles after TB2.
As set forth above, the read clock gate signal GR
and write clock gate signal GW can be used to control when, and at what rate, data are respectively read from and written into the memory system shown in FIGURE 1. The clock signals which load -the values into the serial-to-parallel converters 38 and provide values from the parallel-to-serial conv~rters 54 are synchronized to the master ciock signal because transfers occux only on the edges of master clock pulses. However, the actual data input and output may be otherwise asynchronous. Data may be applied to or provided by the memory system at any fre~uency that is a rational fraction of the master clock frequency.
While this memory system is described in the context of video signal processing, it is contemplated that it may also be used in computer data processing applications. For example, the ability to simultaneously 2n and semi-synchronously write and read continuous data streams at respectively different data rates may make this memory system desirable as a data buffer between a microprocessor and a peripheral device, such as a printer, or as an emulator for a mechanical mass storage device, e.g. as a `'RAM DIS~" device.

Claims (8)

1. A digital memory system realized as a single integrated circuit comprising:
a block oriented data storage array wherein each block has a unique address value;
an address input terminal for applying a bit-serial signal including at least a bit-serial read address value and a bit-serial system control value;
address register means, coupled to said address input terminal, for storing said bit-serial read address value;
means, coupled to said address register means for incrementing said stored read address value;
a memory output port; and control means, responsive to said control value, for selectively addressing the data storage array in succession with the stored read address value and the incremented stored read address value to condition said data storage array to provide the data held in the block having said stored read address value and the data held in the block having said incremented stored read address value sequentially at said memory output port.

-28- RCA 84,14

2. In a memory system including an array of data storage elements arranged as a plurality of separately addressable blocks of data storage elements where each block includes a sufficient number of data storage elements to hold B separate data values and where B is a positive integer, circuitry for retrieving data from said array of data storage elements, comprising:
address register means for holding an N-bit address value for addressing one of said separately addressable blocks of data storage elements;
a control signal input terminal for applying a bit-serial control signal which includes an N-bit read address value and an M-bit control value, where M and N are positive integers;
means, coupled to said control signal input terminal for storing the N-bit read address value of said control signal in said address register means;
memory reading means, coupled to said address register means and to said array of data storage elements and responsive to said M-bit control value for selectively conditioning said array of data storage elements to provide B data values held in a block of data storage elements which is addressed by the N-bit read address value stored in said address register means;
a source of data output clock signal; and means, coupled to said memory reading means and to said array of data storage elements, and including shift register means having a parallel input port and a serial output port, for storing, in parallel, the B data values provided by said array of data storage elements and for sequentially providing said B data values via said serial output port synchronous with said data output block signal.

-29- RCA 84,124

3. The circuitry set forth in Claim 2 wherein said means for storing said N-bit address value in said address register means includes:
a source of address load clock signal;
shift register means, having a serial input terminal coupled to said control signal input terminal and a parallel output port coupled to said address register means, and being responsive to said address-load clock signal for storing said bit-serial control signal; and control means, coupled to said shift register means and to said address register means, for transferring said N-bit address value from said shift register means to said address register means.

4. The circuitry set forth in Claim 2 wherein said memory reading means is responsive to one of the M bits of said M-bit control value to the exclusion of the other M-l bits of the control value for selectively reading the B data values held in the block of data storage elements which corresponds to the N-bit address value stored in said address register means.

-30- RCA 84,124 Canada

5. In a memory system including an array of data storage elements, arranged as a plurality of separately addressable blocks of data storage elements where each block includes a sufficient number of data storage elements to hold B separate data values and where B is a positive integer, circuitry for storing data into and retrieving data from said array of data storage elements comprising:
write address register means for holding an N-bit address value corresponding to one of said blocks of data storage elements;
read address register means for holding an N-bit address value corresponding to one of said blocks of data storage elements;
a control signal input terminal for applying a bit-serial control signal which includes an N-bit write address value, an N-bit read address value and an M-bit control value, where M is a positive integer;
means, coupled to said control signal input terminal for storing said N-bit write address value in said write address-register means and said N-bit read address value in said read address register means;
a source of data input clock signal;
input buffer means including shift register means having a serial input port and a parallel output port for storing B successive data values synchronous with B
successive pulses of said data input clock signal and for providing said B data values in parallel at said parallel output port;
memory sequencing means, coupled to said read and write address register means and to said array of data storage elements and responsive to said M-bit control value being one of a set of first predetermined values, for conditioning said array of data storage elements to store the B data values provided by said input buffer means into a block of data storage elements which corresponds to the address value held in said write address register means, and responsive to said M-bit control value being one of a set of second predetermined values for conditioning said -31- RCA 84,124 Canada Claim 5 continued array of data storage elements to provide B data values held in a block of data storage elements which corresponds to the address value stored in said read address register means;
a source of data output clock signal; and output buffer means, coupled to said memory sequencing means and to said array of data storage elements, and including output shift register means having a parallel input port and a serial output port for storing, in parallel, the B data values provided by said array of data storage elements and for providing said B stored data values via said serial output port synchronous with said data output clock signal.

6. The circuitry set forth in claim 5 wherein said means for storing said N-bit read and write address values in said respective read and write address register means includes:
a source of address-load clock signal;
shift register means, having a serial input terminal, coupled to said control signal input terminal, and first and second parallel output ports, coupled, respectively, to said write and read address register means, and being responsive to said address-load clock signal for storing said bit-serial control signal; and control means, coupled to said shift register means and to said read and write address register means, for conditioning said respective write and read address register means to store the N-bit write and read address values, provided at the respective first and second parallel output ports of said shift register means.

-32- RCA 84,124 Canada

7. The circuitry set forth in claim 5 wherein said memory sequencing means is responsive to a first one of the M bits of said M-bit control value for conditioning said array of data storage elements to store the B data values provided by said input buffer means and responsive to a second one of the M bits of said M-bit control value for conditioning said array of data storage elements to provide the B data values held in the block of data storage elements corresponding to said read address value.

8. A video memory system comprising:
pixel value storage means including an array of pixel value storage elements, arranged as a plurality of separately addressable blocks of pixel value storage elements, where each of said blocks includes a sufficient number of pixel value storage elements to hold B pixel values, where B is a positive integer;
a pixel value input port for applying pixel values to the memory system;
a source of pixel value input clock signal;
input buffer means having a serial input port coupled to said pixel value input port and a parallel output port coupled to said pixel value storage means for storing B successive pixel values applied to said pixel value input port in synchronism with said pixel value input clock signal and for providing said B pixel values to said pixel value storage means via said parallel output port;
a pixel value output port for providing pixel values retrieved from said video memory system;
a source of pixel value output clock signal;
output buffer means having a parallel input port coupled to said pixel value storage means and a serial output port coupled to said pixel value output port and responsive to an output data transfer signal for storing B
pixel values provided by said pixel value storage means and for supplying the pixel values provided by said pixel value storage means in a predetermined sequence at said pixel -33 - RCA 84,124 Claim 8 Continued value output port in synchronism with said pixel value output clock signal;
memory control means for controlling the transfer of pixel values from said input buffer means to said pixel value storage means and from said pixel value storage means to said output buffer means comprising:
input address register means for holding an N-bit address value for addressing one of the separately addressable blocks of pixel value storage elements of said pixel value storage means;
output address register means for holding an N-bit address value for addressing one of the separately addressable blocks of pixel value storage elements of said pixel value storage means;
a control signal input terminal for applying a bit-serial control signal which includes an N-bit input address value, an N-bit output address value and an M-bit control value, where N and M are positive integers;
a source of address-load clock signal;
shift register means, having a serial input terminal, coupled to said control signal input terminal, and first and second parallel output ports coupled to said input and output address register means, respectively, and being responsive to said address-load clock signal for storing said bit-serial control signal;
address control means, coupled to said shift register means and to said input and output address register means for conditioning said respective input and output address register means to store the respective N-bit input and N-bit output address values provided by said shift register means; and -34- RCA 84,124 memory sequencing means, coupled to said input and output address register means, to said input and output buffer means and to said array of pixel storage locations, and responsive to the M-bit control value in said shift register means having one of a first set of predetermined values, for transferring the pixel values available at the output port of said input buffer means to a block of pixel value storage elements of said pixel value storage means addressed by the address value held by said input address register, and responsive to the M-bit control value in said shift register means having one of a second set of predetermined values, for transferring pixel values, in a block of pixel value storage elements of said pixel value storage means addressed by the address value held by said output address register, to said output buffer means.