View VP520S_8206178.PDF datasheet online --- IC-ON-LINE

Datasheet File OCR Text:

VP520S 1 VP520S pal/ntsc to cif/qcif converter advance information supersedes version in june 1995 digital video & dsp ic handbook, hb3923-2 ds3504 - 3.2 october 1996 features n lower power, pin compatible replacement for vp520 n converts ccir601 luminance and chrominance to cif or qcif resolution, and vice versa, using a 27mhz system clock. n luminance and chrominance channels have their own sets of horizontal and vertical filters with on chip line stores n each filter set may be configured to either decimate or interpolate. n ntsc line insertion or removal mode n produces / expects cif/qcif data in macroblock for- mat. n 120 pin qfp package fig 1 : simplified block diagram description the VP520S is designed to convert 16 bit multiplexed luminance and chrominance data between ccir601 and cif/ qcif resolutions. vertical and horizontal fir filters are pro- vided, with the vertical filters supported by on chip line stores. the coefficients used by the filters are user definable, and are down loaded from an independent host data bus. an internal address generator supports an external dram frame store, and also provides line to macroblock conversion. when producing cif or qcif video the horizontal filters precede the vertical filters, and are provided with between 8 and 16 taps. the vertical filters are provided with four cif line delays which allow a 5 tap filter to be implemented. when producing qcif the available ram is used to provide six line delays, which thus allows 7 tap filters to be used. when the device is producing ccir601 video, the incom- ing data must be in macroblock format, and the vertical filters precede the horizontal filters the inputs are firstly written to a external cif sized frame store, and are read out in line format. the VP520S will support two complete frame stores, and allows the cif/qcif data to be read out twice in order to produce two interlaced fields of video. the VP520S supports the conversion between cif/qcif and ntsc video. an extra line is produced for every five lines when producing cif data, and one line in six is removed when producing ntsc video. poly phase filters are used to provide the correct decimation and interpolation ratios. associated products n vp510 colour space converter n vp2611 h261 encoder n vp2615 h261 decoder n vp2612 video multiplexer n vp2614 video demultiplexer frame start / ready 16 bit frame store bus ram address generator supporting line to block conversion horizontal filters vertical filters four line delays filter block input/ output fifo muxing filter block filter block VP520S 8 bit luminance cref 8 bit macroblock bus coeff store host bus address control sync generator href vref mux 8 bit chrominance macroblock strobe req blks muxing
VP520S 2 video compress mode ( decimate ) this mode is used when ccir601 video is to be converted to cif or qcif spatial resolution prior to compression. incom- ing luminance and chrominance data does not need any prior buffering, but must meet the timing requirements given in figure 2. a bit in control register 1 allows the cb component to precede the cr component if necessary. this data is passed through vertical and horizontal decimating filters before it is stored in an external frame store. when a complete field has been decimated it is read out in macroblock format and transferred to the next system component. in this mode href, vref, and fref are normally inputs which are used to reference active video with respect to video synchronization pulses. the active going edges are used internally, and these must meet the set up time with respect to the system clock as given in figure 2. stable inputs are needed with no jitter due to asynchronous pixel clocks, but when this is not possible an external fifo can be used plus two extra signals as described later. the reference inputs need only stay active for one system clock period. note that the active going edges for href and vref can individually be defined to be high going or low going, through two bits in control register 0. also note that cref is always an input and is used as a qualifier for sclk. the actual edges of cref are not used. the internal sync generator can still be used in this mode, if there is a need to supply sync to the video source. the href and vref pins are then used to output hsync and vsync. composite sync is supplied on the csync pin. in addition the clmp pin provides a pulse [13 sclk's wide] which can be used to dc restore the black level in an a/ d converter. it is active high during the back porch. the horizontal blanking output (hblank) defines when the device expects the first pixel in a line to be supplied, and is derived from the user supplied href input. the delay between href and hblank is user definable in multiples of cref periods. if the defined value is zero then the href input must be horizontal blanking with the minimum set time speci- fied. the hblank output is then not defined. all data changes are referenced to the system clock. the edge actually used is indicated by the cref input signal, which has a period of double the clock period. the VP520S will strobe in data on the rising edge of the system clock which occurs whilst cref is high. pin description name type function y7:0 i/o luminance input or output bus c7:0 i/o chrominance input or output bus m7:0 i/o macroblock input or output bus d15:0 i/o 16 bit data bus for dram frame store a7:0 o multiplexed address bus to the dram a8 o most sig address bit or second cas ras o row strobe for the dram's cas o column strobe for the dram's r/ w o read/ write signal to the dram's href i/o horiz. reference in or horiz. sync out vref i/o vertical reference in or vertical sync out cref i/o cref in or cref out fref i/o field indicator in or out hblnk o horizontal blanking output csync o composite sync output in free run mode clmp o defines a black level clamping period for a/d converters vrst frame start identifier. if frst is low then a low going edge will reset the internal sync generator. frst field identifier reqyuv i request macroblocks from encoder mclk i/o macroblock i/o strobe fsig i/o frame start/ ready signal sclk i system clock. 27mhz in pal/ntsc systems hd7:0 i/o host data bus ha3:0 i host controller address bits rd i an active low host read strobe wr i an active low host write strobe cen i an active low enable for the strobes rst i power on reset tdi i jtag i/p data tdo o jtag o/p data tms i test mode select tck i jtag clock trst i jtag reset toe i when high all o/p's are high impedance note: "barred" active low signals do not appear with a bar in the main body of the text. fig 2 : luminance and chrominance inputs in the decimate mode. 10ns min 10ns min hblnk o/p 2ns min 10ns min href input programmed delay input clock first cb comp. 2ns min first cr comp. chrom input second i/p first i/p lum input 10ns min 2ns min cref input
VP520S 3 the first video line to be filtered and stored will be derived from the vertical reference input ( vref). the user can choose the number of transitions of the href input which must occur, after vref has gone active, before starting the filter opera- tion. data is then not written to the dram until after the pipeline delay through the filters. the VP520S only expects to use one field of ccir601 video, which can be selected by the fref input or internal logic. a bit in control register 1 ( internal / external field detect ) determines which option is to be used. an additional field select bit determines whether the field selected should correspond to fref being high or low. when the field select bit and the input are at the same logical level then that field is used. note that fref transitions must be coincident with active going vref transitions. internal logic is provided which determines the field ( field 1 ) in which vref goes active in less than half a line period after the href input last went active.the half line period is determined by vref going active between 1 and 432 cref qualified sclk edges after href went active (1-429 in ntsc mode). note that coincident vref and href edges will indicate this field on the first cref qualified sclk edge. this logic is used, rather than the fref input, when the internal / external field detect bit is low. field 1 is selected when the field select in control register 1 is low, and field 2 is used when the bit is high. in the split screen mode this logic is overridden, and both fields are actually used. external logic is assumed to switch between two sources of video, one for each field. the internal dram address generator is modified such that half area pictures from the centre of each source are actually stored as cif/qcif data. the first line used in each field will be 72 line delays in addition to the number which has been defined by the user. the split screen option is not supported in the qcif mode of operation, and a reset is needed after a mode change in cif. the VP520S will insert zero's into the line delays during vertical blanking. this ensures that all the filter accumulators are cleared and the edges of the picture are correctly proc- essed. the horizontal filters always give the required results since four decimated values are ignored at either side of the picture. incoming luminance data could have a black level of 16, which will be shifted if the filter coefficients are not chosen to exactly give a gain of unity. a control bit is thus provided, which when set causes 16 to be subtracted from incoming fig 3 : macroblock output timing luminance. a black level of zero will then stay as zero through- out the filter operation. at the output of the filters 16 is always added to the results, regardless of the state of the control bit. saturation logic ensures that these addition / subtraction operations do not produce negative results or values greater than 254. a control bit is also provided which selects between colour difference inputs and true cr cb chrominance values. cr cb values are 8 bit positive only numbers, with black levels of 128. these must be converted to two's complement signed num- bers by subtracting 128, thus giving a black level of zero through the filters. the outputs of the filters are always converted to positive only cr cb values by adding 128 to the results, regardless of the state of the control bit. coping with sync jitter when input syncs to the VP520S have jitter, due to the use of a composite video decoder which does not produce a line locked clock, it is necessary to use an external fifo line buffer. for this reason the VP520S supports a system in which external line buffer writes are controlled by the video source and line reads are controlled by the VP520S. the VP520S in the decode loop is assumed to be supplying sync to the VP520S in the encode loop, but the sync generator must be reset at the start of a frame to be in step with the video source. two pins have been supplied to support this situation, namely: vrst - pin 34, and frst - pin 36. the falling edge of vrst (frame start identifier) when frst (field identifier) is low identifies the start of the frame. these two inputs can typically be supplied by the brooktree bt812 composite video de- coder. note that host address 3 must be programmed with the value 02 hex to enable the reset operation. cif/qcif macroblock outputs when producing decimated cif/qcif data in macroblock format, the device raises a flag when a frame of data is ready for reading from the frame store ( fsig ). the fsig pin is automatically configured as an output in the decimate mode, but will only stay active (high) for the time given in figure 3. if a request macroblock response (reqyuv) is not obtained during this period, then fsig will be taken low and the frame of data presently available will be ignored. it will go high again when a new frame of data is available. fsig o/p reqyuv i/p 20ns max 10ns min first o/p valid o/p valid mclk o/p data o/p 2ns min 20ns max 20ns max 20ns max 60 sysclk max , 10sysclk min stays high for 11440 (ntsc) or 13284 (pal) sclks if reqyuv not received 20ns max sysclk 33ns min
VP520S 4 when it receives a reqyuv response from the next system component, it starts to output a macroblock by using an output strobe derived by dividing down the clock input. detailed timing is given in figure 3. this strobe only occurs when data is available at the output pins and at a rate of sysclk/4. the 'request macroblock' flag must go inactive and then active again before a further macroblock is made available. the frame ready flag is only available on the output pin if the frame enable bit is set in control register 1. through this control bit a host controller is able to determine whether a new frame is to be compressed and transmitted. in an alterna- tive arrangement the control bit can be permanently set, and the frame ready flag is then used as an interrupt to the host controller. it then generates a signal which is used as the frame ready signal for the next device. the following sections describe this interface as it applies to the vp2611 h261 video encoder. transfering macroblocks to the vp2611 when the VP520S has stored a complete field of deci- mated video in the dram, it raises a frame ready flag ( fsig ). if the bit in control register 1 does not inhibit the output, this flag becomes the frmin input on the vp2611. this responds to the frmin input by generating a request for macroblock data ( reqyuv ). the vp2611 must then receive a com- plete macroblock ( 384 bytes ) within 1870 cycles of the system clock. when the VP520S is producing decimated cif/ qcif data, writing line data to the dram has priority, and only four macroblock read operations are possible in every 32 clock cycles i.e. one read takes eight cycles. these, however, are 16 bit word operations and it thus requires 384 x 8/2 = 1536 cycles to output the data. in addition there is a maximum delay of 60 clock periods from receiving reqyuv to producing the first output strobe (mclk). this is still well within the time available. the four 16 bit words are stored in the VP520S and transmitted to the vp2611 as eight bytes using a strobe ( mclk ) derived from the system clock. this is only present when valid data is available, and it drives the pclk input on the vp2611. it takes the vp2611 almost exactly all the available time at 30 hz frame rates to process all the macroblocks. after a field time ( half an interlaced frame ) the VP520S will start to write new data to the dram, and data could be overwritten during the last macroblocks. since there is available space in the dram, a small address offset is used between video fields to avoid this problem. interpolate mode in this mode the VP520S expects to receive cif/qcif data in macroblock format, which it then writes to an external frame store. this is then read back in line format and passed through vertical and horizontal interpolating filters to produce two fields of ccir601 video. detailed input timing is given in figure 4. fsig automatically becomes an input which is used to identify the start of a frame and to reset the internal address counter. fsig must stay high until a complete cif/qcif frame has been received ( internal logic counts macroblocks ). if fsig goes low early then the complete frame will be ignored, and the previously received frame will continue to be dis- played. an input strobe, derived by dividing the system clock by four, must also be provided in order to input data. this must only be present when valid data is available on the input pins. incoming macroblocks are byte wide, and these are internally buffered to allow four 16 bit words to be written to the dram every 32 system clock cycles. this is equivalent to a byte input rate of sclk/4 which must not be exceeded. the cif frame store is double buffered such that a new frame can be received whilst the previous one is being displayed. in fact the use of 256k x 16 dram's gives sufficient capacity for more than three complete cif frames, and the internal address generator will simply roll around to make full use of the available space. once a complete cif/qcif frame has been received, it will normally be used to generate two interlaced pal or ntsc fields. these fields continue to be re-generated until a com- plete new cif frame has been received. the rate of receiving frames depends on the transmission bandwidth, but the maximum rate is 30 hz. the changeover to the newly received frame will occur when the VP520S has finished generating any one of the pair of interlaced fields for display, it does not first cr component valid first o/p valid hblank o/p input clock lum output cref o/p 20ns max second o/p valid chrom o/p first cb valid 20ns max 20ns max 20ns max 20ns max 20ns max 20ns max 33ns min mclk i/p fsig i/p second i/p first i/p 40ns min 40ns min 4sclk min 10ns min 2ns min data i/p fig 4 : macroblock input timing fig 5 : luminance and chrominance output timing table 1. external dram timing requirements t rac access time from ras - 105ns or under t cac access time from cas - 25ns or under t rp ras precharge time 50ns or under - t cp cas precharge time 12ns or under - t ras ras pulse width 80ns or under - t cas cas pulse width 50ns or under - t ref time between complete refreshes - 4 ms or over (8 ms with 256k x n) symbol parameter minimum maximum n.b. all times are quoted assuming 27mhz operation. for lower clock frequencies increase the above values proportionately.
VP520S 5 necessarily have to have generated two interlaced fields from the received frame. if the VP520S is receiving frames at the full cif 30 hz frame rate but only displaying pal frames at 25 hz, then periodically one of the pal frames ( comprising two interlaced fields at 50 hz ) will be generated from two received cif/qcif frames. an incoming cif/qcif frame will always be used since the interlaced field rate is always greater than 30 hz in either pal or ntsc. the data is read from the frame store such that interpo- lated data becomes available after programmed delays refer- enced to the vref and href signals. six bits are available to define the line delay, and ten are provided to define the delay from href in cref periods. the actual delays are greater than the programmed values because of the internal pipeline delays, which are also mode dependent. href and vref can either be user supplied inputs, or are generated internally from a pal/ntsc timing generator. a bit in control register 0 determines this option, and when the internal generator is specified the href pin becomes an output which supplies horizontal sync and the vref pin supplies vertical sync. a composite sync output is also pro- vided for system level use. in this mode the vref and href signals used internally are effectively vertical and horizontal sync, and the programmed delays should be chosen to reflect this condition. the signals provided from the internal timing generator allow the VP520S to drive the vp510 colour space converter and an rgb monitor. detailed output timing is given in figure 5. note that the chrominance order can be changed. alterna- tively they can be used to drive off the shelf composite video encoders. external chrominance data can have a zero colour differ- ence value of either 0 or 128. this is defined using the chrominance control bit. where 128 is the zero colour difference value, 128 will be subtracted from incoming chrominance data and 128 will be added to output chrominace data. output values will be limited to lie in the range 16 to 240. external luminace data can have a black luminace level of either 0 or 16. this is defined using the luminace control bit. where 16 is the black value, 16 will be subtracted from incoming luminace data and 16 will be added to output luminace data. output values will be limited to lie in the range 16 to 235. the data stored in the cif frame store will not contain the black levels normally present during horizontal and vertical flyback. this is inserted by the VP520S at the appropriate times in order to ensure that the correct filter operation occurs at the edges of the picture. in addition to these black levels during flyback, a bit in control register 1 allows all active video to be replaced by a fixed colour. this colour is user definable through yuv values in three registers. frame store interface all read and write operations to the external dram frame stores are based on the use of fast page mode with 13.5 mhz cas cycles. internally a 54 mhz clock is produced from the 27mhz system clock, and this determines the minimum time interval which can be used in the generation of pulses and defining precharge times. any dram used must meet the timing constraints given in table 1. reading and writing rates dictate the need for a 16 bit data interface, and line data is re-organized to allow a 16 bit word to consist of either two luminance values or two chrominance values. this gives compatibility with the macroblock require- ments since a sub block is either all chrominance or all luminance data. reading or writing macroblock data requires jumps between pages, but four words can always be read or written using fast page mode. read and write operations must be timeshared to meet the requirements of the system. this time-sharing is based on the use of 16 cycles of the 13.5 mhz clock. when reading or writing line data to the store, 10 cycles are used for eight words, and six cycles are left free for four exchanges with the encoder or decoder. the additional cycles are needed when trs data valid tri tac tas tah thz tlz tsh address chip select read strobe data out tws data valid twi tds tas tah tdh tsh address chip select write strobe data in twa write cycle read cycle characteristic addresss set up time address hold time cip select set up time chip select hold time strobe inactive time data access time delay to o/p's low z delay to o/p's high z characteristic addresss set up time address hold time chip select set up time chip select hold time strobe inactive time strobe active time data set up time data hold time note is the period of the input clock symbol tas tah trs tsh tri tac tlz thz min 10ns 10ns 10ns 2ns ns 2ns max 20 +3ns 25ns symbol tas tah tws tsh twi twa tds tdh min 10ns 10ns 10ns 2ns 1ns 3ns 10ns 10ns max fig 6 : host interface timing advanced data
VP520S 6 using fast page mode in order to guarantee ras precharge times and ras to cas delays. the above time partitioning gives a line rate of 6.75 mhz, which meets real time cif requirements. the exchange rate with the encoder or decoder is only half of this, but is adequate for cif data at 30 hz frame rates. in the decode mode the VP520S produces two fields at 60 hz rates from every 30 hz received frame, thus writing need only be half the rate of reading. in the decimate mode the VP520S produces a cif frame using line rates which could have supported two 60hz fields, but only one is used. thus reading rates need only be half writing rates since the spare field time is available. in the interpolate mode two complete cif frame stores are required, which dictates the use of 256k word dram's. the a8 pin then provides the ninth address bit needed for such devices. in the decimate mode only one cif frame store is required, and a control register bit allows the user to select either 256k word dram's, or 64k x 16 devices. in the latter case two such devices are needed, and the a8 pin now supplies a second cas strobe to enable the second device. refresh cycles generate cas before ras sequences. host interface the VP520S employs a conventional memory mapped host interface using a data bus and an address bus. to minimize on pin count the VP520S only uses four address lines, and all internal ram is addressed through counters. all data is validated with a read or write strobe, and an active low enabling signal. these strobes can be asynchronous to the 27 mhz clock, but the latter must be present to move the data through several pipeline delays. strobes must thus be valid for several clock periods. timing is shown in figure 5. in the worst case mode ( qcif to ntsc video ), the device must store 40 horizontal coefficients and 210 vertical coeffi- cients. internal storage must thus be provided for a total of 250 eight bit coefficients, and this is split into four blocks. these consist of storage for 24 horizontal luminance coefficients; storage for 16 horizontal chrominance coefficients; storage for 70 vertical luminance coefficients; and finally 140 vertical chrominance coefficients. each block of ram has its own internal address counter, and all counters are simultaneously reset with a write to address f hex. each ram area has an associated address as listed below, and a read or write using that address will increment the relevant counter. attempts to use more addresses than are applicable to a particular area will cause undefined behaviour. address allocations are given below; addr function 0 reserved 1 r/w horizontal luminance coefficients. max 24 2 r/w horizontal chrominance coefficients. max 16. 3 normally 00 hex. when 02 hex the sync generator can be reset with the frst and vrst pins. 4 reserved for internal use 5 r/w vertical luminance coefficients. max 70. 6 r/w vertical chrominance coefficients. max 140. 7 set to the normal operating value of 01 hex by reset. when loaded with 21 hex an encoding plus a decoding VP520S can be connected 'back to back' for test purposes or coefficient investigations. no other values must be used. 8 control register 0. see below. 9 control register 1. see below a line delay from vref to first active line. 6msbs only a/b pixel delay from href to first active pixel 2 bits from a plus 8 from b to give a 10 bit value. bit a1 is the msb c blanked screen y value d blanked screen u value e blanked screen v value f clear all address counters the bits in control registers 0 and 1 are used individually, and are defined below. where necessary the action caused when changing a control bit is delayed until the start of a new field. register 0 (address 8) bit function 0 interpolate if high, decimate if low 1 pal if low, ntsc if high 2 qcif if high, cif if low 3 if low subtract 16 from y, add 16 back after filtering 4 if low subtract 128 from chrominance i/ps, add 128 to o/ps 5 if low generate sync, if high lock to href and vref 6 if low then active edge of vref is low going. 7 if low then active edge of href is low going. register 1 (address 9) bit function 0 if low then u inputs precede v inputs and outputs 1 if low use the internal field detect logic 2 field select. see text. 3 if low use 64kx16 dram ( encoder only ) 4 when high specifies split screen mode (encoder only) 5 when low the frame ready flag is enabled 6 when high the screen is blanked (colour defined in addresses c, d, e) 7 when high dram writes are disabled use of address 7 by loading hex 21 into host address 7 it is possible to connect the encoding and decoding filters into a back to back configuration. this is useful for test purposes or for evaluating the filter coefficient values, and it avoids the need for a 'frame start' signal into the filter in the decode path. in normal operation address 7 should contain 01 which is the default after a reset operation. loading coefficients the following tables show the coefficient storage locations for different modes. the filter sections below describe the use of coefficient sets. within a set, coefficients are stored in ascending order, ie. c0, c1, c2 etc. note that some locations are shown as not used. however, since each store is loaded sequentially, the data stream used to load the coefficient stores must contain padding values corresponding to the unused addresses. note also that only the address range shown in the tables have to be loaded with data.
VP520S 7 a: horizontal luminance store this is a 24 byte ram and coefficients will be stored as follows. the full sequence is obtained by writing to address 1, twenty four times and supplying the required data. mode addresses coefficient set ccir -> cif 0-7 1 ccir -> qcif 0-15 1 cif -> ccir 0-5 1 6-11 2 qcif -> ccir 0-5 1 6-11 2 12-17 3 18-23 4 b: horizontal chrominance store this is a 16 byte ram and coefficients will be stored as follows, by writing to address 2 the required number of times. mode addresses coefficient set ccir -> cif 0-7 1 ccir -> qcif 0-15 1 cif -> ccir 0-3 1 4-7 2 qcif -> ccir 0-3 1 4-7 2 8-11 3 12-15 4 c: vertical luminance store this is a 70 byte ram and coefficients will be stored by writing to address 5 the required number of times. mode addresses coefficient set 625 line -> cif 0-4 1 525 line -> cif 0-4 1 5-9 2 10-14 3 15-19 4 20-24 5 25-29 6 625 -> qcif 0-6 1 525 -> qcif 0-6 1 7-13 2 14-20 3 21-27 4 28-34 5 35-41 6 cif -> 625 line 0-4 1, even field 5-9 1, odd field cif -> 525 line 0-4 1, even field 5-9 2, even field 10-14 3, even field 15-19 4, even field 20-24 5, even field 25-29 not used 30-34 1, odd field 35-39 not used 40-44 2, odd field 45-49 3, odd field 50-54 4, odd field 55-59 5, odd field qcif -> 625 line 0-6 1, even field 7-13 2, even field 14-20 1, odd field 21-27 2, odd field ocif -> 525 line 0-6 1, even field 7-13 2, even field 14-20 3, even field 21-27 4, even field 28-34 5, even field 35-41 1, odd field 42-48 2, odd field 49-55 3, odd field 56-62 4, odd field 63-69 5, odd field d: vertical chrominance store this is a 140 byte ram and coefficients will be stored by writing to address 6 the required number of times. mode addresses coefficient set 625 line -> cif 0-4 1 525 line -> cif 0-4 1 5-9 2 10-14 3 625 line -> qcif 0-6 1 525 line -> qcif 0-6 1 7-13 2 14-20 3 cif -> 625 line 0-4 1, even field 5-9 2, even field 10-14 1, odd field 15-19 2, odd field cif -> 525 line 0-4 1, even field 5-9 2, even field 10-14 3, even field 15-19 4, even field 20-24 5, even field 25-29 not used 30-34 1, odd used 35-39 not used 40-44 2, odd field 45-49 3, odd field 50-54 4, odd field 55-59 5, odd field qcif -> 525 line 0-6 1, even field 7-13 2, even field 14-20 3, even field 21-27 4, even field 28-34 5, even field 35-41 6, even field 42-48 7, even field 49-55 8, even field 56-62 9, even field 63-69 10, even field 70-76 1, odd field 77-83 2, odd field 84-90 3, odd field 91-97 4, odd field 98-104 5, odd field 105-111 6, odd field
VP520S 8 112-118 7, odd field 119-125 8, odd field 126-132 9, odd field 133-139 10, odd field qcif -> 625 line 0-6 1, even field 7-13 2, even field 14-20 3, even field 21-27 4, even field 28-34 1, odd field 35-41 2, odd field 42-48 3, odd field 49-55 4, odd field horizontal filters chrominance data is assumed to have already been decimated down to half the horizontal sampling rate of the luminance data, before it is applied to the VP520S. when producing cif data both luminance and chrominance are then both decimated by two, when producing qcif data they are both decimated by four. simulations with actual video have shown that 8 tap cif filters and 16 tap qcif filters give more than adequate performance in the decimation mode. in the interpolation mode these same simulations have shown the need for longer filters in the luminance channel. the hardware thus supports a 12 tap filter when interpolating luminance from cif inputs, but only 8 taps are provided for each chrominance channel. even longer filters are needed when qcif data must be interpolated, and the luminance channel is provided with 24 taps, and each chrominance channel with 16 taps. note that when interpolating by two the output rate is double the input rate, but every other input will be conceptually zero. similarly when interpolating by four there are three zero's between every data point, even though the output rate is four times the input rate. thus during any clock period only one half or one quarter of the coefficients are actually in use, and the computational burden is no greater than when doing the equivalent decimation. since all the coefficients are not in use during any clock cycle, it is convenient to refer to two smaller sets of coeffi- cients. thus the 12 tap cif luminance filter, for example, can be considered to have two sets of 6 coefficients, and the 24 tap qcif luminance filter to have four sets of 6 coefficients. the addressing in the coefficient rams uses this concept of sets. vertical filters the vertical filters are designed to produce cif with the spatial relationship shown in figure 7, and qcif with the spatial relationship shown in figure 8. original pal or ntsc video contains lines of coincident luminance and chromi- nance, but the cif specification requires that the decimated chrominance information is shifted such that it lies mid way between two luminance lines. this is achieved by choosing the centre outputs from the filter which best fit the require- ments. the filter outputs actually used by the device are shown by the arrows in figures 7 and 8, and are optimal when the even field provides the original video. it is assumed that one of the interlaced fields has been discarded prior to the VP520S, and thus no further decimation occurs when producing cif luminance from pal ( ntsc in 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 10 11 0 2 4 6 8 10 cif lines lum chrom pal lines even odd ntsc lines even odd repeat repeat repeat fig 7 : cif spatial relationships
VP520S 9 fact needs some interpolation - see the relevant section ). chrominance, however, is decimated by two. when produc- ing qcif data the luminance channel is decimated by two, and the chrominance by four. when the VP520S is used to derive interlaced ccir601 video, the internal address generator will read the cif/qcif frame store twice in order to produce the two fields. each field has its own set of coefficients. internal ram is provided which will support four cif line delays for both chrominance and luminance. five tap filters are thus possible for cif conversions. with a qcif system the internal ram could theoretically be used to provide eight qcif line delays. in practice, however, little benefit is obtained by using vertical filters with more than seven taps, and thus only six line delays are used. polyphase filters are used to support the spatial conver- sions. pal conversion is relatively simple and only requires a set of coefficients for each mode. ntsc conversion requires several sets of coefficients since the 240 lines in a field must be converted to 288 lines of cif. one line is repeated in every five to produce six lines which are then filtered with their own coefficients. the generation of interpolated outputs requires cif / qcif data to be repeatedly read from the frame store at various line intervals. this is all handled by the internal address generator, and is transparent to the user. the device then produces coincident luminance and chrominance data which has been interpolated from data in the frame store. the first line will be produced to match the delay from the vref input which has been pre-defined. this delay must be greater than the internal pipeline delay, which itself is mode dependent ( delay yet to be determined ). the device introduces black lines at the top and bottom of the fields. thus the first and last lines in the interpolated field will be filtered with varying amounts of black information. pal vertical filtering when producing cif data the five tap filters provide outputs for every line at the 6.75 mhz decimated line rate. every filtered luminance line is used but every other filtered chrominance line will be discarded. filter outputs correspond- ing to odd numbered pal chrominance lines in any field being at the centre are used to provide the cif chrominance lines. this is shown by the arrows in figure 7. when decimating down to qcif seven tap filters are used, which provide outputs for every line at the 3.375 mhz line rate. only every other filtered luminance line, and every fourth chrominance line are actually stored in the frame store. different pal lines are used to produce the offset luminance and chrominance lines as indicated by the arrows in figure 7. when interpolating from cif the luminance channel con- ceptually uses a 10 tap filter, with every other input line containing only zero's. thus only five coefficients are actually used when producing interpolated lines for the even field, and five different coefficients are used when producing the odd field. the device thus stores two sets of five coefficients; one set for each field produced by reading the cif frame store twice. the chrominance filter is conceptually a 20 tap filter with three lines of zero's for every actual input. thus each chromi- nance channel needs four sets of five coefficients; two sets are needed to produce one field, and two sets are needed for the other field. the same chrominance data is read twice for a given pair of luminance lines, in order to provide inputs for the filter. thus the internal line delays contain the same set of chrominance data on two consecutive lines supplying data to the filters. when interpolating from qcif, seven coefficients can be 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 0 2 4 qcif lines lum chrom pal lines even odd ntsc lines even odd fig 8 : qcif spatial relationships
VP520S 10 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 0 2 4 qcif lines lum chrom ntsc lines even odd fig 9 : interpolating from qcif to ntsc used in each set since six line delays are provided. the luminance filter conceptually contains 28 taps ( four sets of seven coefficients with two sets used to produce each field ). similarly the chrominance filter consists of 56 taps arranged as eight sets of seven coefficients with four sets needed for each field. in order to provide data for the filters each lumi- nance line is read twice, and each chrominance line is read four times to produce each field. ntsc vertical filtering one field of ntsc video consists of 240 chrominance and luminance lines, which must be converted to 288 lines of cif luminance and 144 lines of cif chrominance. the luminance increase is mechanized by repeating the first line in every five to produce six lines, which are then applied to the vertical filters. a different set of coefficients is used for each line, requiring a total of 30 to be stored within the device. the line repeat causes one set of line data to be used twice, but each time different coefficients are used by the filter. this technique is equivalent to interpolating the data by six, and then decimat- ing by five. the required coefficients for each of the six sets can be derived by conceptually using this approach. the line repeat requires an additional fifo line delay before the four delays used by the filters. by reducing the horizontal blanking time it is possible to read six lines ( one is repeated ) from the fifo in the time taken to acquire five lines of video with blanking. chrominance data also passes through the input fifo and one line in every five is repeated. this is done in order to avoid differential delays with the luminance data. three chromi- nance lines are only needed, however, for every five original lines. they are produced by using three sets of five coeffi- cients and discarding two filtered lines in every five. the three selected filter outputs are chosen such that the centre line of the filter is closest to the cif line number needed. the centre lines which are actually used are shown in figure 8, and result in a sequence of two chosen outputs then a gap followed by one output then a gap. simply using every other output would not give the best fit. a simplified approach is used when decimating down to qcif resolution, and the input fifo is not used. six luminance lines are derived from ten ntsc lines by choosing the six outputs produced when the centre line in the filter is closest to the qcif line that is needed. overall this results in a luminance sequence consisting of two outputs then a gap, followed by one output then a gap and is shown in figure 8. three chrominance lines are derived from the same inputs by using three sets of seven coefficients. the chromi- nance sequence is also shown in figure 7, and consists of an output then three gaps, followed by an output and two gaps. when interpolating from cif up to ntsc resolutions, it is necessary to read lines of data from the cif frame store with reduced blanking periods. the timing is calculated such that six lines are read in the time that five lines would have been read if they had the correct blanking period. these fast lines are continuously filtered using all the available information, and the results are written to an output fifo. this fifo is then read with the correct blanking period inserted in order to provide ntsc data at the output pins. thus five lines are read out in the time taken to load six lines ( one of which need not actually be written since it is never used )
VP520S 11 five sets of coefficients are used to produce the five lines which are actually stored, but the coefficients are different for the even and odd field generation. thus a total of ten sets of five coefficients are internally stored. in effect we have inter- polated by five and then decimated by three in order to produce the complete ntsc frame. each cif chrominance line is used to produce two filtered ntsc chrominance lines, and one filtered line in every six is then ignored. this is mechanized by reading each cif chromi- nance line twice for every pair of luminance lines. the same filtering and discard technique as used in the luminance channel is then applied, using five sets of coefficients for each field. ten sets are thus needed to produce two ntsc fields. we have effectively interpolated by ten and then decimated by three to produce 480 chrominance lines for the complete frame. when interpolating from qcif to ntsc the additional output buffering is not used. instead a sequence is used which will generate 10 ntsc lines in any field from six qcif luminance lines and three chrominance lines. figure 9 illus- trates how the first and fourth lines are used once and the second, third, fifth, and sixth used twice to produce qcif luminance. since this 1 - 2 - 2 sequence is used twice in every ten lines, only five rather than ten sets of coefficients are actually needed for each field ( ten sets in total ). the first and third chrominance lines are used three times, and the second line is used four times. thus ten sets of coefficients are needed for each field ( twenty sets in total ). each luminance and chrominance set consists of seven coefficients, since six line delays are provided for the filters. jtag test interface the VP520S includes a test interface consisting of a boundary scan loop of test registers placed between the pads and the core of the chip. the control of this loop is fully jtag/ ieee 1149-1 1990 compatible. please refer to this document for a full description of the standard. the interface has five dedicated pins: tms, tdi, tdo, tck and trst. the trst pin is an independent reset for the interface controller and should be pulsed low, soon after power up; if the jtag interface is not to be used it can be tied low permanently. the tdi pin is the input for shifting in serial instruction and test data; tdo the output for test data. the tck pin is the independent clock for the test interface and registers, and tms the mode select signal. tdi and tms are clocked in on the rising edge of tck, and all output transitions on tdo happen on its falling edge. instructions are clocked into the 3 bit instruction register (no parity bit) and the following instructions are available. instruction register name ( msb first ) 111 bypass 000 extest (inversion except for vref, href, csync and clmp) 010 sample/preload the tap controller used in this device does not support a separate intest instruction but allows extest to drive the internals of the device as well as to drive the output pins. output enables are thus present in the chain which are not connected to pins but which allow extest to be used to control the impedance of all the outputs. the toe pin, which can separately be used to control the impedance of all the outputs, can be monitored as an input through the scan chain but cannot be used to control the outputs through the tap controller. the signals controlled by the various enables are listed below: pad name signals controlled dram_oeb a8:0, ras, cas, r/w refs_oeb fref, vref, href csync_oeb clmp, csync, hblnk cgtout_oeb test function only d_oeb d15:0 m_oeb m7:0, mclk, fsig c_dec_b cref yuv_oeb y7:0, c7:0 cdata_oeb hd7:0
VP520S 12 advance data advance data characteristic output high voltage output low voltage input high voltage input low voltage input leakage current input capacitance output leakage current output s/c current notes on maximum ratings 1. exceeding these ratings may cause permanent damage. functional operation under these conditions is not implied. 2. maximum dissipation or 1 second should not be exceeded, only one output to be tested at any one time. 3. exposure to absolute maximum ratings for extended periods may affect device reliablity. 4. current is defined as negative into the device. absolute maximum ratings [see notes] supply voltage v dd -0.5v to 7.0v input voltage v in -0.5v to v dd + 0.5v output voltage v out -0.5v to v dd + 0.5v clamp diode current per pin i k (see note 2) 18ma static discharge voltage (hmb) 500v storage temperature t s -65 c to 150 c ambient temperature with power applied t amb 0 c to 70 c junction temperature 150 c package power dissipation 5000mw delay from output high to output high impedance test waveform - measurement level delay from output low to output high impedance h v 0.5v v 0.5v l 1.5v 0.5v 1.5v 0.5v delay from output high impedance to output low delay from output high impedance to output high v - voltage reached wh en output driven hi g v - voltage reached wh en output driven lo w h l units v v v v m a pf m a ma symbol v oh v ol v ih v il i in c in i oz i sc min. 2.4 - 2.0 - -10 -50 10 max. - 0.4 - 0.8 +10 +50 300 value typ. 10 conditions i oh = 4ma i ol = -4ma v dd - 1v for sysclk and mclk gnd < v in < v dd gnd < v out < v dd static electrical characteristics operating conditions (unless otherwise stated) tamb = 0 c to +70 c v dd = 5.0v 5% ordering information VP520S/cg/gh1r (commercial - plastic qfp package)
VP520S 13 pin 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 func gnd a8 a7 a6 a5 vdd gnd a4 a3 a2 a1 a0 vdd gnd rw vdd gnd ras vdd gnd m7 m6 m5 m4 pin 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 func m3 m2 m1 m0 mclk vdd gnd reqyuv gnd vrst fsig frst vdd rst tck tms trst tdi tdo toe vdd vref fref href pin 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 func cref gnd csync y0 y1 y2 y3 y4 y5 y6 y7 vdd gnd hblnk c0 c1 c2 c3 c4 c5 c6 c7 n/c gnd pin 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 func vdd sclk gnd vdd ha0 ha1 ha2 ha3 wr rd cen hd0 hd1 hd2 hd3 hd4 hd5 vdd gnd hd6 hd7 clmp vdd d15 pin 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 func d14 d13 d12 gnd vdd d11 d10 d9 d8 gnd vdd d0 d1 d2 d3 gnd vdd d4 d5 d6 d7 cas gnd vdd table 2: 120 pin qfp pin assignment
VP520S 14 jtag bit number 93 92 91 90 89 88 87 86 85 84 83 82 81 80 79 78 77 76 75 74 73 72 71 70 69 68 67 66 65 64 63 62 61 60 59 58 57 56 55 54 53 52 51 50 49 48 47 46 45 44 jtag bit number 145 144 143 142 141 140 139 138 137 136 135 134 133 132 131 130 129 128 127 126 125 124 123 122 121 120 119 118 117 116 115 114 113 112 111 110 109 108 107 106 105 104 103 102 101 100 99 98 97 96 95 94 direction out out out out out out out out out out out out out in out in out in out in out in out in out in out in out in out in out in in in out in out in out in out in out out out out in out in signal a8 a7 a6 a5 a4 a3 a2 a1 a0 rw ras dram_oeb m7 m7 m6 m6 m5 m5 m4 m4 m3 m3 m2 m2 m1 m1 m0 m0 mclk mclk m_oeb reqyuv fsig fsig rst toe vref vref fref fref href href cref cref refs_oeb c_dec_b csync csync_oeb y0 y0 y1 y1 signal y2 y2 y3 y3 y4 y4 y5 y5 y6 y6 y7 y7 hblnk c0 c0 c1 c1 c2 c2 c3 c3 c4 c4 c5 c5 c6 c6 c7 c7 yuv_oeb cgtout (n/c) cgtout_oeb sclk ha0 ha1 ha2 ha3 wr rd cen hd0 hd0 hd1 hd1 hd2 hd2 hd3 hd3 hd4 hd4 direction out in out in out in out in out in out in out out in out in out in out in out in out in out in out in out out out in in in in in in in in out in out in out in out in out in direction out in out in out in out out out in out in out in out in out in out in out in out in out in out in out in out in out in out in out in out in out out in in jtag bit number 43 42 41 40 39 38 37 36 35 34 33 32 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 signal hd5 hd5 hd6 hd6 hd7 hd7 cdata_oeb* clmp d15 d15 d14 d14 d13 d13 d12 d12 d11 d11 d10 d10 d9 d9 d8 d8 d0 d0 d1 d1 d2 d2 d3 d3 d4 d4 d5 d5 d6 d6 d7 d7 d-oeb* cas vrst frst cgtout (n/c) this pin is only used for gps test purposes and should not be used for system purposes. table 3: jtag register allocation

m mitel (design) and st-bus are registered trademarks of mitel corporation mitel semiconductor is an iso 9001 registered company copyright 1999 mitel corporation all rights reserved printed in canada technical documen t a tion - n o t for resale world headquarters - canada tel: +1 (613) 592 2122 fax: +1 (613) 592 6909 north america asia/paci?c europe, middle east, tel: +1 (770) 486 0194 tel: +65 333 6193 and africa (emea) fax: +1 (770) 631 8213 fax: +65 333 6192 tel: +44 (0) 1793 518528 fax: +44 (0) 1793 518581 http://www.mitelsemi.com information relating to products and services furnished herein by mitel corporation or its subsidiaries (collectively mitel) is believed to be reliable. however, mitel assumes no liability for errors that may appear in this publication, or for liability otherwise arising from the application or use of any such information, product or service or for any infringement of patents or other intellectual property rights owned by third parties which may result from such application or use. neither the supply of such information or purchase of product or service conveys any license, either express or implied, under patents or other intellectual property rights owned by mitel or licensed from third parties by mitel, whatsoever. purchasers of products are also hereby noti?ed that the use of product in certain ways or in combination with mitel, or non-mitel furnished goods or services may infringe patents or other intellectual property rights owned by mitel. this publication is issued to provide information only and (unless agreed by mitel in writing) may not be used, applied or reproduced for any purpose nor form part of any order or contract nor to be regarded as a representation relating to the products or services concerned. the products, their speci?cations, services and other information appearing in this publication are subject to change by mitel without notice. no warranty or guarantee express or implied is made regarding the capability, performance or suitability of any product or service. information concerning possible methods of use is provided as a guide only and does not constitute any guarantee that such methods of use will be satisfactory in a speci?c piece of equipment. it is the users responsibility to fully determine the performance and suitability of any equipment using such information and to ensure that any publication or data used is up to date and has not been superseded. manufacturing does not necessarily include testing of all functions or parameters. these products are not suitable for use in any medical products whose failure to perform may result in signi?cant injury or death to the user. all products and materials are sold and services provided subject to mitels conditions of sale which are available on request.

▲Up To Search▲

Price & Availability of VP520S

	To Download VP520S Datasheet File
If you can't view the Datasheet, Please click here to try to view without PDF Reader .