Compact Mac Video Adapter

[retr01]

Hello!

I acquired a SE/30 in excellent condition. I have had a strong computer background, especially Apple since I was a kid. I am happy to be back into tinkering.

I saw on YouTube that Adrian Black showed how to do a video out from a Macintosh Classic from its analog board through RGB2HDMI. I realized that in the Mac SE and SE/30, the analog board (AB) was designed differently than in the Mac Classic and Classic II. Adrian's video is based on Classic’s AB. The SE and SE/30, as well as older compact Macs, do not have that AB design.

Thank you, @Stephen, for doing this! You did splendidly, and I appreciate that. I want to have the video out from the SE/30 like it was done back in the day to TTL output while I am waiting for @Zane Kaminski and the folks at Garrett's Workshop to come up with a new PDS video card for the SE/30 with a second PDS connector for a network card that I have. I would like to have this output set up.

So, I purchased the Power R Model 2703 Video Adapter for Apple Macintosh SE or SE/30 off eBay. @Stephen, according to Power R's defunct website via the Wayback Machine (scroll down to the bottom of the page), this model has the exact specifications as 2702, only different connectors. Connecting the male end to the SE or SE/30 logic board (LB) and the cable from the female end from LB to the AB is straightforward.

​

All I have to do is plug-in male to male dupont jumper cables in the black TTL header and solder one end to the RGB female connector as shown in the schematic created by @Stephen. Is that correct? Then, I would get the RGB2HDMI adapter with Pi Zero.

@Stephen can you update your GitHub page to include a clone version for the SE and SE/30 since it needs those types of connectors, as shown in the photo herein? I would be happy to let you know the result after doing this and if the results are favorable?

By the way, thank you, @JDW, so much for your beautiful videos on YouTube! I appreciate your clear explanations and showing the process. Those helped me a LOT to tear down my SE/30.

 

[Stephen]

@Stephen can you update your GitHub page to include a clone version for the SE and SE/30 since it needs those types of connectors, as shown in the photo herein? I would be happy to let you know the result after doing this and if the results are favorable?

I'm not opposed — but one of the advantages of open source is that other people can spin off their own version! If no one wants to take a stab at this then I will make an updated version just for you .

 

[retr01]

I'm not opposed — but one of the advantages of open source is that other people can spin off their own version! If no one wants to take a stab at this then I will make an updated version just for you .

@Stephen, I am willing to try my best for the spin-off. Thank you.

 

[-SE40-]

Just wondering, will this conversion / adaptation also work similar for a classic II board?

 

[retr01]

Morning, @-SE40-!

Yes and no. It depends on various compact Macs' analog boards (AB) and logic boards (LB). Based on the differences, adapting to produce a 1-bit black and white (B/W) video output via (TTL) is done differently. One way is to use a vintage or modern adapter. Another way is to wire the connection based on the required matching pinouts. First, the differences between the ABs will be shown. Second, the location where the adapters are plugged differ. Third, the alternative wiring connection differs as well.

1. Here are the general differences between the ABs of the compact Macs. The first two ABs do not have taps to produce the B/W video output as it is done on the LB. However, the third AB has video output taps.

​	​	​
128k, 512k, Plus​	SE, SE/30​	Classic, Classic II​

2. Vintage closed-source and modern open-source adapters send the 1-bit B/W video output via TTL to the DE-15 (DB-15) or DE-9 (DB-9) RGB connector. From there, a vintage RGB monitor can be connected, or a modern monitor can be connected through modern adaption such as Extron or RGB2HDMI. Pictures, links, and explanations are currently in this thread.

3. Alternative wiring connections can be made after identifying the correct video pinouts between the connections. For example, Adrian has shown an excellent method for the Classic and Classic II on a YouTube video:

As for COLOR video output, the 128k, 512k, Plus, SE, Classic, and Classic II generally require a SCSI external video adapter, as an example explained in this message. The SE/30 is the only 030 compact Mac with a 030 PDS slot capable of utilizing color video cards. As pointed out in another message earlier, the SE/30 can display 256 grays on the internal CRT monitor with an adapter.

I am not sure if there are LC PDS slot color video cards for the Color Classic and Color Classic II to drive external color video.

 

[-SE40-]

Good morning to you too….although its evening here

Thank you very much for your extended explanation.
This is much appreciated!
I think others also will find this info very usefull too

Somehow I did come to option 3.
I saw the vid of ADB, and got me thinking…..
Would it be possible to combine the CL II board with the RGB to HDMI board and a nice LCD and make Hartmuts laptop what never was.
No analog board …. is that possible?
Adding a Lipo pack should last a while on mobile use.
This board is recapped and working fine, so that is a good start.
I could then recreate Hartmuts “laptop” enclosure, and make it fully functional.

Would that not be supercool?

 

[retr01]

Yeah! Super cool!

@-SE40-is that the Classic II logic board? I can see the memory SIMM fashion and the 030 CPU. What are the wires for?

As for one of Hartmut's portable Mac designs, I think that is possible with the right logic board and video TTL output. Fabricate the case, keys, LCD housing, hinges, etc., via 3D printing.

Might the keyboard be one of the most expensive things, though? I have read that custom keyboards are insanely costly, not because of the keys but the switches?

Is that a project you want to do as a new thread? Do we need a new category for "the Vaporized Macs" here on Tinker Different?

 

[-SE40-]

Hi, retro01!

indeedClassic II board.
This is the board of my first Mac.
I like it for its compactness, and so usable for this purpose.
At least if I can get it to work with a flat panel display without too many extra parts.

its not a project yet, until its electronics side is made /proven to work.

its actually a similar sort of challenge as with
Erics Color Classic.
But with another gen board.

For the keyboard I have Conversion of a original keyboard with custom made keys in mind.
The enclosure I would love to create a mould for from high quality 3D printed shells….but thats still a long way to go.
A man can dream…..

The wires are temporarely for a 2032 battery soldered on BTW.

 

[retr01]

Hi, retro01!

Hello, @-SE40-. That is retr01 (it's a zero and one, no letter O).

Its not a project yet, until its electronics side is made /proven to work.

Ah, cool!

 

[-SE40-]

My bad…retr01

As I am no electronics wizzard, any help with this upcomming conversion is welcome…

Thanks in advance!

 

[retr01]

Hey @Stephen,

Can you show us a picture of the modern clone of the Power R 2702 that you assembled based on your GitHub repo, please?

 

[alxlab]

​

Here's a conversation with Zane about using a Pico Pi to convert the video signal from a compact Macintosh (128K, 512K, SE, SE/30, Classic, Classic II) to something more usable by modern screens. Maybe this will give some inspiration to someone.

alxlab-zone66x — 02/07/2022​

Was contemplating the use of a Pi Pico for outputting the video of a compact b/w Mac

1. [12:52 PM]
   the ADC on the pico should be good enough to read the h amd v sync
   
   
2. [12:52 PM]
   the video signal is just 1-bit
   
   
   

3. Zane Kaminski — 02/07/2022​

   Well for the Mac it’s all digital so no need for an ADC
   
   
4. [12:52 PM]
   Yeah
   
   
5. [12:53 PM]
   But there remains the problem of when to sample the pixel
   
   
6. [12:53 PM]
   How do you get your 15.6672 MHz clock exactly lined up with the video output so that you sample in the middle of the pixel
   
   
   
   

7. alxlab-zone66x — 02/07/2022​

   Yeah that's where my knowledge of how this stuff is handled fails me
   
   
   

8. Zane Kaminski — 02/07/2022​

   You can tap the board or build a PDS card but that’s not as good

9. alxlab-zone66x — 02/07/2022​

   there must be some hacky cheap way to handle the straight output

10. Zane Kaminski — 02/07/2022​

    My best idea for the Mac is to sample at 50 MHz and your program knows that there are 50/15.6672 sampled pixels per actual pixel and you go by that factor, adjusting one pixel forward or back according to when your software sees a transition
    
    
11. [12:56 PM]
    So like an 01010101 pattern on the Mac will be stretched out into: 000111000111000111
    
    
12. [12:56 PM]
    But eventually there will be 4 in a row not 3
    
    

13. alxlab-zone66x — 02/07/2022​

    oh yeah the scaling is no prop
    
    
14. [12:56 PM]
    I would use 1024 x 768
    
    

15. Zane Kaminski — 02/07/2022​

    You sorta watch for that to keep your thingy in sync
16. [12:56 PM]
    No no not on the screen in the end
17. [12:56 PM]
    Like coming into your gizmo

18. alxlab-zone66x — 02/07/2022​

    ah ok for detection of the actual pixels you mean
    
    

19. Zane Kaminski — 02/07/2022​

    Since the pixels change at 15.6672 MHz but you sample at approximately 3x that. So your software sees 3-4 pixels for every one the Mac puts out
    
    
20. [12:57 PM]
    Yeah

21. Androda — 02/07/2022​

    Pi pico's PIO units ought to be able to sample the digital signal
    
    

22. Zane Kaminski — 02/07/2022​

    And you sorta watch and keep it synced
    
    

23. Androda — 02/07/2022​

    Those things are nuts, and sample every clock cycle
    
    
    
    
    

24. Zane Kaminski — 02/07/2022​

    Well if we can sample at 3x the pixel speed we build something like a UART to capture the pixels
    
    
25. [1:00 PM]
    In software though
    
    
26. [1:00 PM]
    Yknow how a UART doesn’t have the bit clock, you have to oversample and the UART figures out which bit is which and there’s the whole concept of tolerable baud rate error
    
    
27. [1:00 PM]
    That’s my “only an RPi” strategy for this
    
    
    
    

28. alxlab-zone66x — 02/07/2022​

    actually my knowledge on any of these protocols is like 0 but what your saying makes sense
29. [1:02 PM]
    what if we add an external clock that gives us the 15.66672?
    
    
30. [1:02 PM]
    would that make our life easier?
    
    

31. Zane Kaminski — 02/07/2022​

    It’ll wander
32. [1:03 PM]
    So basically your software aims to watch for hsync to indicate the start of the line, take in 1634 pixels (512 * 50/15.6672) and then sorta process that bitstream into 512 actual pixels
    
    
33. [1:04 PM]
    I guess it’s possible, just sorta keep track of your fractional pixel position. Shouldn’t need to adjust forward or back during a line
    
    

34. alxlab-zone66x — 02/07/2022​

    actually if we sample at 3x the pixel clock then could auto correct like you said
    
    
35. [1:04 PM]
    a pixel would have 3 of the same value
    
    

36. Zane Kaminski — 02/07/2022​

    Yes but you can never depend on it being exact
    
    
37. [1:04 PM]
    So it will be 2.9something or 3.0something
    
    
38. [1:05 PM]
    So since that’s weird already I figured exactly 3x doesn’t have much of a point and we can advance by 3.19 or whatever
    
    
39. [1:06 PM]
    and just round to the nearest pixel, plus detect if there is a change too close to the current sample position and skip one forward or backwards
    
    

40. alxlab-zone66x — 02/07/2022​

    yeah add a delay or something
    
    
41. [1:06 PM]
    it would be really interesting if it worked half decently
    
    

42. Zane Kaminski — 02/07/2022​

    I can’t figure out how to get the bits into the pi though
    
    
43. [1:08 PM]
    Like we want it to just be dumping what it sees on the wire straight into memory in the background of other processing but idk if it has a peripheral that can do it
    
    
44. [1:08 PM]
    If not then we have to involve a little FPGA to buffer it in a FIFO so the CPU can get a bunch of bits out all at once
    
    

45. alxlab-zone66x — 02/07/2022​

    well there would be some processing of the signal
    
    

46. Zane Kaminski — 02/07/2022​

    I just mean to read the oversampled 50 MHz pixels
    
    
47. [1:09 PM]
    Then we format those right and actually put em in video ram

48. alxlab-zone66x — 02/07/2022​

    can just use the standard io libs?
    
    

49. Zane Kaminski — 02/07/2022​

    Not sure
50. [1:12 PM]
    Can we make it like do a 1634 bit spi transaction and dump it in memory for processing?
    
    
51. [1:12 PM]
    That’s my best guess
    
    

52. alxlab-zone66x — 02/07/2022​

    yeah guess it would be a matter of trying out stuff at this point
    
    
53. [1:15 PM]
    Gives me an idea that I'm not totally dreaming in color
    
    

54. Zane Kaminski — 02/07/2022​

    Yeah here we go
55. [1:15 PM]
    https://github.com/rsta2/circle/blob/master/sample/23-spisimple/kernel.cpp
    GitHub
    circle/kernel.cpp at master · rsta2/circle
    A C++ bare metal environment for Raspberry Pi with USB (32 and 64 bit) - circle/kernel.cpp at master · rsta2/circle
    
    
    
    
    
    
56. [1:16 PM]
    So at every line we basically start a really long SPI transaction and collect the results on the next line
    
    
57. [1:16 PM]
    Should work!

58. alxlab-zone66x — 02/07/2022​

    interesting!
    
    
59. [1:17 PM]
    so basically the hsynch would trigger that
    
    

60. Zane Kaminski — 02/07/2022​

    Yeah but there’s a gotcha
    
    
61. [1:17 PM]
    I think we have to start it on the previous line’s end
    
    
62. [1:18 PM]
    And get a bunch of garbage in the beginning
    
    
63. [1:18 PM]
    Otherwise we’ll miss the first few pixels
    
    
64. [1:18 PM]
    Because interrupt latency, etc.
    
    
65. [1:19 PM]
    But then there’s a new problem... we need to capture hsync in time with the SPI bitstream so we can know when the active video region starts
    
    
66. [1:23 PM]
    Well the rpi SPI can do 125 MHz so we can like double up samples with some external logic
    
    
67. [1:23 PM]
    Even bits are hsync, odd bits are pixel color
    
    

68. alxlab-zone66x — 02/07/2022​

    ah 2x spi interfaces on the pico
    
    

69. Zane Kaminski — 02/07/2022​

    Ooh yes maybe
    
    

70. alxlab-zone66x — 02/07/2022​

    but we have vsync as well
    
    

71. alxlab-zone66x — 02/07/2022​

    I need to verse myself more on how the different protocols work
    
    
72. [1:46 PM]
    right of course the hdmi out will take up some io also
    
    
73. [1:46 PM]
    https://github.com/Wren6991/Pico-DVI-Sock
    
    GitHub
    GitHub - Wren6991/Pico-DVI-Sock: A DVI Sock board for Pico
    A DVI Sock board for Pico. Contribute to Wren6991/Pico-DVI-Sock development by creating an account on GitHub.
    
    
74. [1:46 PM]
    Like the silkscreen "NOT HDMI just sparkling DVI"
    
    
    
    

75. Zane Kaminski — 02/07/2022​

    Vsync is easier since it's slow
    
    
76. [1:48 PM]
    2 chips
    
    
77. [1:48 PM]
    Here we go
    
    
    
    
    
    
78. [1:48 PM]
    This gizmo will mux the hsync and intensity on to the same wire
    
    
79. [1:49 PM]
    The catch is you don't know which will come first, the hsync or the video hahahahah
    
    
80. [1:49 PM]
    I can add some more to fix that
    
    
81. [1:50 PM]
    Yeah that's easy to fix
    
    
82. [1:50 PM]
    So basically you get the RPi to do a 125 MHz SPI transaction (fast but it can do it), send SCK in there, and then it alternates VID and HSYNC on MISO which goes back to the RPi
    
    
    

83. alxlab-zone66x — 02/07/2022​

    lol this is already more advance than my current level of electronic design
    
    
    

84. Zane Kaminski — 02/07/2022​

    the 74LVC1G57 is basically a mux
    
    
    

85. alxlab-zone66x — 02/07/2022​

    what the idea between using the same SPI vs two separate I/O? (edited)
    
    
    

86. Zane Kaminski — 02/07/2022​

    Well then you know for sure they are exactly lined up in time
    
    
87. [1:53 PM]
    Who knows how long the function to start an SPI operation takes, probably more than 1 Mac pixel time
    
    
    

88. alxlab-zone66x — 02/07/2022​

    the hsync front porch and back porch is supposed to be pretty long so maybe it's not a problem
    
    
89. [1:54 PM]
    http://www.waveguide.se/?article=compact-mac-video-adapter
    
    www.waveguide.se
    Compact Mac video adapter « waveguide.se
    [img=mac-se-video-bmva.jpg] For many years I have had an old Macintosh SE logic board that was scrapped out from an old Apple repair shop. Someone had broken the SIMM sockets at some point by snapping the retaining clips at the ends. A common problem u...
    
    
90. [1:54 PM]
    he mentions "The horizontal signal is also unusually long and extends well into the visible area of the video lines."
    
    
    

91. Zane Kaminski — 02/07/2022​

    Interesting!!
    
    
92. [1:55 PM]
    Well at least it's constant
    
    
    

93. alxlab-zone66x — 02/07/2022​

    on a normal vga signal you got lots of extra not drawn pixels at the begining and end of a line
    
    
    

94. Zane Kaminski — 02/07/2022​

    So the time-correlation of the hsync to the video signal is still important
    
    
95. [1:56 PM]
    We just have to find the hsync transition and then go backwards some fixed amount instead of forward
    
    
96. [1:57 PM]
    All this depends on starting at the end of the previous line of course
    
    
    

97. alxlab-zone66x — 02/07/2022​

    as long as the hsync isn't so far off as to go in between lines it should be ok
    
    
    

98. Zane Kaminski — 02/07/2022​

    Yeah shouldn't be
    
    
99. [1:59 PM]
    Ooh and the guy who wrote that article has a picture of the artifact we must avoid (edited)
    
    
100. [1:59 PM]
     His monitor is sorta skipping pixels wrong
     
     
101. [1:59 PM]
     
     
     
     
     
     
     

102. alxlab-zone66x — 02/07/2022​

     yeah that's kinda cool his monitor actually output something
     
     
     
     

103. Stephen — 02/07/2022​

     this solution is very comparable to the power-r adapter. Right?
     
     
     
     

104. Zane Kaminski — 02/07/2022​

     Yeah
     
     
     

105. alxlab-zone66x — 02/07/2022​

     mine just says hsync out of range
     
     
     

106. Zane Kaminski — 02/07/2022​

     Should just be a passthrough
     
     
     

107. elemenoh — 02/07/2022​

     @Zane Kaminski would your widget have digital or analog video out?
     
     
     

108. alxlab-zone66x — 02/07/2022​

     at first I was thinking VGA out but if you dealing with all digital why bother going analog?
     
     
109. [2:02 PM]
     seems simpler to just do dvi with a pi pico anyhow
     
     
     

110. Zane Kaminski — 02/07/2022​

     So umm, we start the SPI thing at the end of the line, wait until the next line, find the hsync transition, go backwards some fixed amount to find the first pixel, then sample each pixel by going forward 62.5/15.6672 (approx 3.989) in the buffer and copy all 512 of that line into the back buffer in preparation to frame-skip copy into the actual video buffer to avoid tearing
     
     
     
111. [2:03 PM]
     ^ that's the algorithm
     

 

[retr01]

That is an interesting conversation, @alxlab.

I agree this is comparable to the Power R series of adapters, including 2702 and 2703 that @Stephen and myself have.

So, is it possible that simply using the Pi Pico as a video signal converter to become an alternative to RGB2HDMI with Pi Zero? My only concern is, can enough data be placed with on the Pi Pico to have full featured OSD menus show up on a modern display, many adjustments, varying video signals, etc? The Extron, RGB2HDMI, and other RGB to whatever video signal converters have OSD displayed on modern screens.

On the other hand, this is an excellent start to have a signal conversion that just works, requiring less components that are readily available, and cheaper without needing adjustment by the end user.

 

[alxlab]

That is an interesting conversation, @alxlab.

So, is it possible that simply using the Pi Pico as a video signal converter to become an alternative to RGB2HDMI with Pi Zero? My only concern is, can enough data be placed with the Pi Pico to have full featured OSD menus show up on a modern display, many adjustments, varying video signals, etc? The Extron, RGB2HDMI, and other RGB to whatever video signal converters have OSD displayed on modern screens.

Sorry I wasn't clear enough. This discussion was to make the cheapest purpose built converter to just output video from a compact Macintosh which is 512 x 384 and 1bit.

Using the analogy of slicing bread, sure you can use an expensive Swiss Army Swiss Champ XXL (RGB2HDMI) with a bunch of extra tools you won't need or you can just use a dollar store knife (PI Pico or whatever low cost microcontroller will do the job) for a fraction of the cost.

The RGB2HDMI has the harder job of trying to handle anything thrown at it. For the compact Macs we're dealing with a know quantity so hopefully we could try to make a less complex, plug and play and cheaper device.

 

[retr01]

Thank you, @alxlab, @Stephen, and @Zane Kaminski! You guys are awesome!

Is there a GitHub repo with a tentative schematic of this Pico Pi approach?

 

[alxlab]

Don't think anyone has attempted anything yet with the pico pi. There are at least a couple sdks for doing video out on the pico:

http://www.breatharian.eu/hw/picovga/index_en.html
https://picockpit.com/raspberry-pi/raspberry-pi-pico-video-output/

Ideally we would want 1024 x 768 output so we could just double the resolution of 512 x 384 so it will scale nicely with no weirdness.

 

[retr01]

Cool. Yes, I agree with the ideal resolution of 1024 x 768 output by upscaling 2 times of the native 512 x 384.

 

[-SE40-]

Pico VGA……wow!

This would be great as the regular pi’s seem hard to get.
Also the smaller size is a big plus for Hartmut’s laptop.
So any pico pi will do?
Sounds almost to good to be true

Ive found this list for the classic I and II 14 pin connector…
Would I be making connection directly for pin 4-5-6 that should do the job?

I sure have some reading todo

 

[retr01]

Ive found this list for the classic I and II 14 pin connector…
Would I be making connection directly for pin 4-5-6 that should do the job?

So, I have a question @-SE40-. Where on the logic board is this connector? Is it the same one that connects to the harness from logic board to the analog board?

 

[Zane Kaminski]

Thank you, @alxlab, @Stephen, and @Zane Kaminski! You guys are awesome!

Is there a GitHub repo with a tentative schematic of this Pico Pi approach?

I would say it's not a schematic-heavy kind of project. Regarding the algorithm I described:

"Start the SPI [transfer] at the end of the line, wait until the next line, find the hsync transition, go backwards some fixed amount to find the first pixel, then sample each pixel by going forward 62.5/15.6672 (approx 3.989) in the buffer and copy all 512 of that line into the back buffer in preparation to frame-skip copy into the actual video buffer to avoid tearing."

The first thing would be to create some kind of simulation to try out the method before actually implementing the hardware. We would want to test the algorithm against many combinations of small frequency and phase differences between the Mac and Pi clocks to make sure it recovers the pixels crisply in all cases.