SuperMac Spectrum 8 Series II - No that's not a serial port

[David Cook]

In an earlier post, I expressed curiosity about an unpopulated connector on the original SuperMac Spectrum 8 video card. Someone mentioned that the Series II version of the card has a second connector that is populated. On eBay, I found two of the Series II cards, although they are often not labeled by that name by sellers.

Here's the card:



It does indeed have two populated connectors on the slot edge. One connector is the standard DB-15 Macintosh video out (they corrected the position so it fits in compact Macintosh II series slots). The other port sure looks like a Macintosh 8-pin mini DIN serial port.



If it is a Macintosh-compatible serial port, the pinout should look like this:



A quick check reveals that pin 4 on the SuperMac port is NOT ground. Here are the pins on the PCB which I referenced when tracing them back to components:

1. U21 pin 2 1Q output flip-flop
2. Ground->100 ohm->port output->330 ohm->U57 pin 4 q1 output serial in parallel-out shift register. Register input is U56 pin 19 output of identity comparator. Voltage divider clamps the voltage to 1.16V.
3. Unused
4. Ground->330 ohm (orange orange brown)->port output->220 ohm (red red brown)->+5V. So, steady 3V.
5. H2 on SM34061 video chip HSYNC
6. H1 on SM34061 video chip VSYNC
7. Unused
8. Ground

Ground, hsync, and vsync, start to make this look like some sort of video out. Not sure about the purpose of the digital outputs of pins 1 and 2. All I know is, there are no inputs and this is not a serial port.

BEADS?

There are some glass beads at the base of all of the radial tantalum capacitors on one of the boards. Not sure why. Anyone know?



BAD BOARD

Unfortunately, one of the boards has partially corrupt video regardless of the resolution. There is no corrosion or obvious damage. The ROM is valid. Notice that many of the pixels are correct. For example, the menu bar at the top and the icons on the right. I feel like this should be telling me something obvious.



RESOLUTIONS and ROM

The Series II is very similar in capabilities to the original board. However, both oscillators show up correctly on the SuperVideo control panel and they've added a 800x600 mode.



One board's ROM is labeled 1002941-1001A and internally contains v1.0d14. The other board's ROM is labeled 1002941-1001C v2.0 and internally contains v1.0d15. Odd that they skipped 1002941-1001B and that the v2.0 labelled ROM has v1.0d15 internally.

These ROMs are 16KB each (double the earlier board's 8KB) and are not inverted.

ROMs attached.

- David

 

[jmacz]

Any more obvious pattern in the bad video when you are in black and white?

 

[David Cook]

Any more obvious pattern in the bad video when you are in black and white?

No, not really. Here's what I've tried so far:

1. Set up two monitors (left: internal IIci, right bad board) 640x480 B&W. Opened up SimpleText and dragged a mainly white screen across the width. Took a screenshot in the Mac. It also sees a scrambled display. So, this is a memory/interface problem, not an output problem.



2. Swapped the TMS (U27) and SuperMac (U54) chips from the good board to the bad board. Good board still works, bad board still doesn't. So, not those chips.

3. Flir images show something potentially interesting. It looks like the memory chips are not being exercised on the bad board.
(Left: good, Right:bad)



Alternatively, the memory chips on the bad board might just run cooler, as they are from a different manufacturer.

- David

 

[olePigeon]

Those beads are weird. A while back I was recapping a VCR, and all the through-hole caps had little fabric covers on the legs. I don't know what their original purpose was, but they wicked up the leaking electrolytic goo and kept it off the motherboard. So that was an unintended bonus. Unfortunately the surface mount ones leaked everywhere anyway and ruined it.

 

[jmacz]

The symptoms remind me of the S24/V I never fixed and also on the memory/interface side. Curious what you find.

It’s cool you got to test the socketed chips. Mine are all soldered in and with so many pins, I didn’t want to risk taking a working chip off another card (yet) to rule chips out.

 

[David Cook]

The symptoms remind me of the S24/V I never fixed and also on the memory/interface side. Curious what you find.

I wasn't actually going to try to repair the bad board. But, I followed your S24/V repair saga religiously. So, now I feel obligated to make an attempt.

My SuperMac board features MT42C4064 dual-ported VRAM chips. Each chip only has 4 bits of output. So, SuperMac connected the same column input (CAS) into horizontal pairs to make it 8 bits (one byte). SuperMac connected the same row input (RAS) across the board to make a 32-bit output. All of the chips share the same eight address pins.



How this works is:
1. Set the address pins (all chips will see this) for the desired row address inside of the VRAM chip.
2. Trigger the RAS to lock in and select that address for only 8 chips (a specific row). The other 16 chips (other two rows) can now be considered 'deselected'. They don't get to play along because their row wasn't selected.
3. Now change the address pins for the desired column address inside of the VRAM chip.
4. Trigger the CAS of either one column for an 8 bit output or all 4 columns for a 32-bit output. Although there are 6 chips per column, remember only one of their rows was selected, so only 2 chips (4 + 4 = 8 bits) in that column get to play along.

On the SuperMac card, there are four 8-bit chips (SN74LS245 octal bus transceivers -- labeled 'column data' above) to hold the input or output data. 4x8=32 bits. So, that supports being able to operate on an entire row at a time.

There are only two address chips, supporting 4 bits of addressing each (labeled "Addr hi" and "Addr lo" above). This combined 8-bit address goes to all VRAM chips at the same time. Only the RAS/CAS selected chips care.

Since something is wrong with memory i/o on this board, I hooked up the oscilloscope to see memory addressing in action. I only have four channels on my scope, so I chose to view RAS (row), CAS (column), and two of the address lines (A7 and A0). Below we see that A7 is low and A0 is high when the row is selected (latches on low). Then A0 is made low before selecting the column (latches low). The signals stay stable for long-enough periods before the row and column are deselected. Beautiful.



Now let's look at the bad board. Purely through random chance of what address was being read, A0 is the first to go low and then A7, but otherwise the activity is the same. But, there is a huge problem: A0 doesn't really go low. It is sort of floating in 'no man's land' of voltages. That looks like between 1V and 2V.



The VRAM datasheet says a low voltage must be between -1.0V and 0.8V.


It turns out that addresses A4-A7 are all good, but A0-A3 don't go low enough. And what do you know? The good ones go to one chip and the bad ones go to a different chip. It seems this board has a dying 74F257. Depending on the address (the number of bits it needs to pull low), it can produce 'okay' to 'really bad' outputs. The VRAM is sometimes understanding the correct address and sometimes looking in the wrong place to read/write data.

I've got a replacement 74F257 on the way. We shall see if this fixes the board.

- David

 

[jmacz]

This is cool. I used a digital logic analyzer when I was debugging my S24/V which masks this as I'm seeing highs/lows only. I'll need to revisit my card and take a look at the raw voltages on my oscilloscope.