With the introduction of Core 2.00.26 today, a new ZX Spectrum Next graphics mode is being unleashed to the wild thanks to the hard work of Allen Albright (Alcoholics Anonymous)
Below, please find the initial documentation, so you can start developing for it!
The tilemap is a hardware character oriented display that comes in two resolutions: 40×32 (320×256 pixels) and 80×32 (640×256 pixels).
The display area on screen is the same as the sprite layer, meaning it overlaps the standard 256×192 area by 32 pixels on all sides. Vertically this is larger than the physical HDMI display which will cut off the top and bottom character rows (making the visible area 40×30 or 80×30) but the full area is visible on VGA.
The obvious usage of the tilemap is for super fast, clearly readable and wide multicoloured character display; The non-obvious usage is to make fast and wide full colour backgrounds that combine with your Layer 2 graphics and sprites to fill the void between ULA and LoRes, expanding it for resolution.
The tilemap is defined by two data structures.
The first data structure is the tilemap itself which indicates what characters occupy each cell on screen. Each tilemap entry is two bytes so for 40×32 resolution, a full size tilemap will occupy 2560 bytes, and for 80×32 resolution the space taken is twice that at 5120 bytes. The tilemap entries are stored in X-major order and each two-byte tilemap entry is stored little endian:
bits 15-12 : palette offset bit 11 : x mirror bit 10 : y mirror bit 9 : rotate bit 8 : ULA over tilemap (if the ula is disabled, bit 8 of tile number) bits 7-0 : tile number
The character displayed is indicated by the “tile number” which can be thought of as an ASCII code. The tile number is normally eight bits allowing up to 256 unique tiles to be defined but this can be extended to nine bits for 512 unique tiles if the ULA display is disabled.
The other bits are tile attributes that modify how the tile image is drawn. Their function is the same as the equivalent sprite attributes for sprites.
Bits will apply rotation and mirroring and colour can be shifted with a palette offset. If the ULA display is enabled, bit 8 will determine if the tile is above or below the ULA display.
B. Tile Definitions
The second data structure is the tile definitions themselves.
Each tile, identified by tile number, is 8×8 pixels in size with each pixel four bits to select one of 16 colours. A tile definition occupies 32 bytes and is defined in X major order with packing in the X direction in the same way that 4-bit sprites are defined. The 4-bit colour of each pixel is augmented by the 4-bit palette offset from the tilemap in the most significant bits to form an 8-bit colour index that is looked up in the tilemap palette to determine the final 9-bit colour sent to the display.
Tiles are therefore defined using 16 colours with the tilemap palette offset able to act as index into the tilemap palette to vary the display colour.
Memory Organization & Display Layer
The tilemap is a logical extension of the ULA and its data structures are contained in the ULA’s 16k bank 5. If both the ULA and tilemap are enabled, this means that the tilemap’s map and tile definitions should be arranged within the 16k to avoid overlap with the display ram used by the ULA.
The tilemap exists on the same display layer as the ULA. The graphics generated by the ULA and tilemap are combined before being forwarded to the SLU layer system as layer U.
Combining ULA & Tilemap
The combination of the ULA and tilemap output is done in one of two modes:
the standard mode and the stencil mode.
The standard mode uses bit 8 of a tile’s tilemap entry to determine if a tile is above or below the ULA. The source of the final pixel generated is then the topmost non-transparent pixel. If the ULA or tilemap is disabled then they are treated as transparent.
The stencil mode will only be applied if both the ULA and tilemap are enabled. In the stencil mode, the final pixel will be transparent if either the ULA or tilemap are transparent. Otherwise the final pixel is a logical AND of the corresponding colour bits. The stencil mode allows one layer to act as a cut-out stencil for the other.
Programming Tilemap mode
(R/W) 0x6B (107) => Tilemap Control bit 7 = 1 to enable the tilemap bit 6 = 0 for 40x32, 1 for 80x32 bit 5 = 1 to eliminate the attribute entry in the tilemap bit 4 = palette select bits 3-0 = Reserved set to 0
Bits 7 & 6 enable the tilemap and select resolution. Bit 4 selects one of two tilemap palettes used for final colour lookup. Bit 5 changes the structure of the tilemap so that it contains only 8-bit tilemap entries instead of 16-bit tilemap entries. If 8-bit, the tilemap only contains tile numbers and the attributes are instead taken from nextreg 0x6C.
(R/W) 0x6C (108) => Default Tilemap Attribute bits 7-4 = Palette Offset bit 3 = X mirror bit 2 = Y mirror bit 1 = Rotate bit 0 = ULA over tilemap (bit 8 of tile id if the ULA is disabled)
If bit 5 of nextreg 0x6B is set, the tilemap’s structure is modified to contain only 8-bit tile numbers instead of the usual 16-bit tilemap entries. In this case, the tile attributes used are taken from this register instead.
(R/W) 0x6E (110) => Tilemap Base Address bits 7-6 = Read back as zero, write values ignored bits 5-0 = MSB of address of the tilemap in Bank 5
This register determines the tilemap’s base address in bank 5. The base address is the MSB of an offset into the 16k bank, allowing the tilemap to begin at any multiple of 256 bytes in the bank. Writing a physical MSB address in 0x40-0x7f or 0xc0-0xff, corresponding to traditional ULA physical addresses, is permitted.
The value read back should be treated as a fully significant 8-bit value.
The tilemap will be 40×32 or 80×32 in size depending on the resolution selected in nextreg 0x6B. Each entry in the tilemap is normally two bytes but can be one byte if attributes are eliminated by setting bit 5 of nextreg 0x6B.
(R/W) 0x6F (111) => Tile Definitions Base Address bits 7-6 = Read back as zero, write values ignored bits 5-0 = MSB of address of tile definitions in Bank 5
This register determines the base address of tile definitions in bank 5. As with nextreg 0x6E, the base address is the MSB of the an offset into the 16k bank, allowing tile definitions to begin at any multiple of 256 bytes in the bank. Writing a physical MSB address in 0x40-0x7f or 0xc0-0xff, corresponding to traditional ULA physical addresses, is permitted. The value read back should be treated as a fully significant 8-bit value.
Each tile definition is 32 bytes in size and is located at address:
Tile Def Base Addr + 32 * (Tile Number) (R/W) 0x4C (76) => Transparency index for the tilemap bits 7-4 = Reserved, must be 0 bits 3-0 = Set the index value (0xF after reset)
Defines the transparent colour index for tiles. The 4-bit pixels of a tile definition are compared to this value to determine if they are transparent.
(R/W) 0x43 (67) => Palette Control bit 7 = '1' to disable palette write auto-increment. bits 6-4 = Select palette for reading or writing: 000 = ULA first palette 100 = ULA second palette 001 = Layer 2 first palette 101 = Layer 2 second palette 010 = Sprites first palette 110 = Sprites second palette 011 = Tilemap first palette 111 = Tilemap second palette bit 3 = Select Sprites palette (0 = first palette, 1 = second palette) bit 2 = Select Layer 2 palette (0 = first palette, 1 = second palette) bit 1 = Select ULA palette (0 = first palette, 1 = second palette) bit 0 = Enabe ULANext mode if 1. (0 after a reset)
The tilemap has its own pair of palettes for looking up 9-bit colours. Each tile definition pixel is 4-bits which is combined with the 4-bit palette offset from the tilemap entry in the most significant 8-bits. This 8-bit value is passed through the tilemap palette to generate the final 9-bit colour for the pixel.
(R/W) 0x1B (27) => Clip Window Tilemap bits 7-0 = Coord. of the clip window 1st write = X1 position 2nd write = X2 position 3rd write = Y1 position 4rd write = Y2 position The values are 0,159,0,255 after a Reset Reads do not advance the clip position
The tilemap display surface extends 32 pixels around the central 256×192 display.
The origin of the clip window is the top left corner of this area 32 pixels to the left and 32 pixels above the central 256×192 display. The X coordinates are internally doubled to cover the full 320 pixel width of the surface. The clip window indicates the portion of the tilemap display that is non-transparent and its indicated extent is inclusive; it will extend from X1*2 to X2*2+1 horizontally and from Y1 to Y2 vertically.
(R/W) 0x2F (47) => Tilemap Offset X MSB bits 7-2 = Reserved, must be 0 bits 1-0 = MSB X Offset Meaningful Range is 0-319 in 40 char mode, 0-639 in 80 char mode (R/W) 0x30 (48) => Tilemap Offset X LSB bits 7-0 = LSB X Offset Meaningful range is 0-319 in 40 char mode, 0-639 in 80 char mode (R/W) 0x31 (49) => Tilemap Offset Y bits 7-0 = Y Offset (0-191)
These are scroll registers for scrolling the tilemap area. As with other layers, the scroll region wraps.
(R/W) 0x68 (104) => ULA Control bit 7 = 1 to disable ULA output bit 6 = 0 to select the ULA colour for blending in SLU modes 6 & 7 = 1 to select the ULA/tilemap mix for blending in SLU modes 6 & 7 bits 5-1 = Reserved must be 0 bit 0 = 1 to enable stencil mode when both the ULA and tilemap are enabled (if either are transparent the result is transparent otherwise the result is a logical AND of both colours)
Bit 0 can be set to choose stencil mode for the combined output of the ULA and tilemap. Bit 6 determines what colour is used in SLU modes 6 & 7 where the ULA is combined with Layer 2 to generate highlighting effects.
Below you can find two illustrations of Tilemap mode memory usage without and with the ULA disabled kindly provided by Kev Brady.