//Previos thread: 8364R7 Amiga Paula dissection.

This thread is about a transistor level dissection of the Rockwell R6545 CRTC (CRT controller),
brought to you by Frank Wolf and ttlworks.

In the beginning, there was the Motorola MC6845 CRTC,
and back then it was quite the thing for displaying text and graphics.
//Hitachi had the HD46505.
//During the cold war time, CM607 was a Bulgarian MC6845 clone.

The MC6845 even was used in early graphics cards for the IBM PC: MDA, HGC, CGA...
So as long as there are IBM compatible PCs which happen to have graphics cards
we can expect that there is something either in the hardware or at least in the software
which convincingly can emulate a MC6845, but I'm getting off topic.

The 6545 was intended to be a compatible drop in replacemend for the MC6845, featuring some additional functionality.
And that's where trouble starts: there were 6545 chips from different vendors, adding _different_ functionality.
André Fachat has a nice overview about the differences
and about the CRTC internals.

;---

From the list of features, to us the Rockwell R6545 was the most interesting 6545 variant,
because it offered things like interlaced mode and horizontal scrolling.

Metal traces in the R6545R1 chip looked like they were tacked together by an intern
on Monday morning before he had his first cup of coffee,
so the vectorisation of the silicon took Frank a bit longer than usual.

Note:
The circuitry inside the R6545 which is driven by CCLK uses dynamic latches,
so you better make sure that CCLK doesn't go slower than maybe 2kHz
(what should not happen in a typical application).
But this and the limited address range makes the R6545 a really bad choice
if you are in seek of a DMA chip for playing audio samples...

Note:
For consistence with Frank's notation, low_active signals are named foo#, not /foo.

Orientation for all the chip pictures: MA7 pad is North.

;---

Datasheets:
Bitsavers: Motorola MC6845 datasheet
Bitsavers: Motorola AN-0851 "MC6545 simplifies video display controllers", 1981
Bitsavers: Motorola AN-0834 "using the MC68000 and the MC6845 for a color graphics system", 1981

MOS 6545-1 CRTC controller datasheet, Nov. 1981
MOS 6545-1 CRTC controller datasheet, recreated

Rockwell R6545-1 CRTC datasheet, Dec. 1980

Bitsavers: Synertek 1981-1982 data catalog:
SY6545, page 3-139, PDF page 238
SY6545-1, page 3-155, PDF page 254

CRT controller handbook, Gerry Kane, 1980

Eagle 6.4 schematics for my schematic pictures in this thread,
just in case if somebody needs them.

Note: KiCad is supposed to be able to import these schematics,
unfortunately it doesn't seem to be possible to disable the layers 'name' and 'value' in KiCad schematics,
so making my schematics look nice and clean in KiCad will require some work, sorry.

A picture of the R6545R1 silicon, with the interesting areas marked according to the cheat sheet.



Just as a reference, another picture of the R6545R1 silicon without the markings.



Microscopic picture of the R6545R1 silicon:

R6545R1 cheatsheet:



Actually, the cheatsheet above was the second try.
The first try didn't work out well:

First:
sorry that we lack the capacity for interpreting the results of the dissection,
also we lack the capacity for writing the usual detailed text.

From our experience, best practice for dissecting a peripheral chip
is to do the pads (plus their drivers) first,
then to go for clock generation and reset
(because clock and reset signals will be used all over the chip),
and then to dissect the bus interface plus the address decoder.

Once you have the address decoder, you can tell which of the register Bits
inside the chip goes where.
And then to go for "the rest" between the register Bits and the pads.

;---

1abcdeh)

So now some pads:

1f) RA0
1g) RA4

These pads have a little bit more of logic attached to them:






;---

1i) MA multiplexer control

The edge detector for the light pen also had ended up in there,
we somehow had to cut that chip into pieces,
and it isn't always obvious which part does what...

2) CCLK clock generator

CCLK character timing clock is the time base for all of the counters.

We have the usual clock generator built around a RS flipflop.
RS flipflop basically is built from two NOR gates.

3) R/W#, PHI2, RS, CS#

Now for a closer look at the logic which is attached
to the R/W#, PHI2, RS, CS# pads:

4) read write logic

It is attached to 3), and it decides whether there is:
generic register read, generic register write,
status register read, address register write.

5a) address register
5b) address decoder

The address register basically is a 5 Bit half_static transparent latch,
attached to the North end of the address decoder.

The address decoder then selects a generic register to be read/written.

6) RST#, LPEN input

//The TEST pad is not available outside the chip.

RST# resets the counters in the chip.
Note, that RST# works a bit different from the reset input we had in the previous 65xx chips:
it is intended to synchronize the 6545 frame timing with line frequency,
or with "a master 6545" for implementing split screen and such.

LPEN is the light pen input.

Note:
LPEN input active blocks RES#.
RES# input active blocks light pen detection.

7) VSYNC output and logic

8) CRSR, DEN output

CRSR (cursor active) and DEN (display enable) outputs
can be delayed by one CCLK cycle,
and the implementation for the delay is quite minimalistic:

9) AP counter and logic

Unlike all of the other 6545 variants,
the Rockwell R6545 features an additional counter for CPU display memory access,
which basically can be incremented by register 31 reads/writes.

To me it looks like something like a kludge,
added to the 6545 circuitry by Rockwell,
note that additional clock generator.

//Since MA (memory address) already was taken,
//I had named that counter AP (address pointer).


;---





;---

10a) register 0 //horizontal total
10b) register 1 //horizontal displayed
10c) register 2 //horizontal SYNC position