Differences in grayscale options for the SE/30

[ph4z]

Hi all, this is my discussion post to the TD forums (welcome here)

For my SE/30 quest I’m trying to understand the different SE/30 greyscale options.

Let me first set out what I know and then come to the question. Currently available options are (in no particular order)

1. 30Video GS - Internal grayscale for SE/30 by Zigzagjoe
2. SE/30 Internal Grayscale cards by Bolle
3. Micron Xceed

What I know: All use the same neckboard design (1&2 are clones of 3). [EDIT] But the cards of 1 & 2 are new designs and different from 3. 1 & 2 are modern reproductions available, 3 is hard to find.

What I don‘t know: How do the different “video”cards compare. In the FS listing Bolle mentions:

[2] doesn't run off the logicboard sync signals but instead generate them ourselves. This means there's no switching between onboard video and video generated by the card but instead both modes 1bit and 8bit are generated on the card. Not being dependent on the original sync signals lets us bump the resolution to 512x384 giving access to a wider range of applications that require that larger resolution.

But it is not entire clear what this means. From the text above it sounds like options 1 and 3 are limited in the internal resolution (to the stock 512x342?).

Can anyone help me explain how these cards compare in terms of image quality and performance for the internal display (not interested in dual monitor)?

Thanks!

 

[zigzagjoe]

Hi all, this is my discussion post to the TD forums (welcome here)

For my SE/30 quest I’m trying to understand the different SE/30 greyscale options.

Let me first set out what I know and then come to the question. Currently available options are (in no particular order)

1. 30Video GS - Internal grayscale for SE/30 by Zigzagjoe
2. SE/30 Internal Grayscale cards by Bolle
3. Micron Xceed

What I know: All use the same neckboard design (1&2 are clones of 3). 1 & 2 are modern reproductions available, 3 is hard to find.

What I don‘t know: How do the different “video”cards compare. In the FS listing Bolle mentions:



But it is not entire clear what this means. From the text above it sounds like options 1 and 3 are limited in the internal resolution (to the stock 512x312?).

Can anyone help me explain how these cards compare in terms of image quality and performance for the internal display (not interested in dual monitor)?

Thanks!

My card (30Video) and Xceed support optionally driving an external monitor while keeping the internal display active in black and white. Internal resolution remains 512x342. Bolle's card does only the internal display but at a higher vertical resolution of 512x384. That's the main difference between the cards, the rest is details.

To clarify though neither my card or Bolle's card are clones of the Xceed; that term is typically reserved for a reproduction of a vintage design. Bolle reverse engineered a vintage Formac video card and redesigned the timing, and my card is a new design with a newer video controller. Detail of the development of my card here , and the same for bolle's card here.

Image quality should be identical, I've never noticed a difference between the 30video and Xceed. I expect Bolle's would be the same. The Xceed card does not do gamma correction so it will have a slightly darker picture. I do not know if Bolle's card performs gamma correction or not.

Xceed is technically the fastest card, but that really is splitting hairs as these are all unaccelerated framebuffers so practical video performance actually depends more on your CPU accelerator rather than the video board.

My grayscale card is now the same card as used in my LCD kits, introducing 16 bit support, however as there is not a point in 16 bit grayscale (indistinguishable) it's only really relevant to external monitors. As it's the same card however there is an upgrade path to the LCD if you preferred.

 

[ph4z]

My card (30Video) and Xceed support optionally driving an external monitor while keeping the internal display active in black and white. Internal resolution remains 512x342. Bolle's card does only the internal display but at a higher vertical resolution of 512x384. That's the main difference between the cards, the rest is details.

To clarify though neither my card or Bolle's card are clones of the Xceed; that term is typically reserved for a reproduction of a vintage design. Bolle reverse engineered a vintage Formac video card and redesigned the timing, and my card is a new design with a newer video controller. Detail of the development of my card here , and the same for bolle's card here.

Image quality should be identical, I've never noticed a difference between the 30video and Xceed. I expect Bolle's would be the same. The Xceed card does not do gamma correction so it will have a slightly darker picture. I do not know if Bolle's card performs gamma correction or not.

Xceed is technically the fastest card, but that really is splitting hairs as these are all unaccelerated framebuffers so practical video performance actually depends more on your CPU accelerator rather than the video board.

My grayscale card is now the same card as used in my LCD kits, introducing 16 bit support, however as there is not a point in 16 bit grayscale (indistinguishable) it's only really relevant to external monitors. As it's the same card however there is an upgrade path to the LCD if you preferred.

Thanks ZZJ for the clear explanation. I appreciate the clarification that neither cards are clones of the Xceed, I was aware and was only talking about the neckboards (although I could be mistaken here), but the fact that the cards are new/redesigned was not not made explicit in the original post, I added a sentence for completeness and prevent confusion. This is really impressive work you have done, so it shouldn’t have been glossed over like that!

Moreover, I’m really glad to hear performance wise the cards are all close, that takes out one variable. That is performance among the cards though, but what is there a performance difference between stock monochrome and a greyscale conversion? Sorry if this is an ignorant question…

Very nice that you were able to unify the card for both the colour LCD and greyscale, I saw it in the posts. This leaves more options open!

I wonder how the extra vertical resolution affects the picture on the CRT, I can imagine it changes the aspect ratio or leaves black bars, it is 12% extra after all. I’ll need to read/watch reviews to understand this better and read up on the links you provided.

Edit: Okay I think I get it now. Both resolutions 512x342 and 512x384 (used in the LC) are used for 4:3 aspect ratio displays, the pixels then have different ratios. (With 384 actually resulting in 1:1 pixels).

 

[zigzagjoe]

Thanks ZZJ for the clear explanation. I appreciate the clarification that neither cards are clones of the Xceed, I was aware and was only talking about the neckboards (although I could be mistaken here), but the fact that the cards are new/redesigned was not not made explicit in the original post, I added a sentence for completeness and prevent confusion. This is really impressive work you have done, so it shouldn’t have been glossed over like that!

Moreover, I’m really glad to hear performance wise the cards are all close, that takes out one variable. That is performance among the cards though, but what is there a performance difference between stock monochrome and a greyscale conversion? Sorry if this is an ignorant question…

Very nice that you were able to unify the card for both the colour LCD and greyscale, I saw it in the posts. This leaves more options open!

I wonder how the extra vertical resolution affects the picture on the CRT, I can imagine it changes the aspect ratio or leaves black bars, it is 12% extra after all. I’ll need to read/watch reviews to understand this better and read up on the links you provided.

Haha, sorry, I might be a little touchy on the clones thing Yes, the neckboard I use is a derivative of Bolle's to be specific. Same basic circuit but bits moved around but functionally identical.

Yeah, there are some minor performance differences but the majority of what Mac OS is doing is changing single bytes or bits within a byte (depth < 8bit) so it spends more time in the QuickDraw algorithms rather than pushing pixels, outside of specific cases like blits. I think I posted some system info benchmarks on 68kmla somewhere... At the same bit depth (ie. B&W) any video card will be faster as the builtin video isn't particularly performant with only an 8 bit data bus. When using grayscale@8bpp it ends up roughly as a wash compared to the stock. Blits/Screen copies will be slower (simply due to more data to transfer) but normal drawing of lines and the like will be similar or faster. Still mostly one byte transfers but no need to mask off unused bits.

Aspect ratio isn't too bad from what I've seen with Bolle's. It's moving from a 3:2 aspect ratio to 4:3,. With my card and the xceed, if your screen is already well adjusted it shouldn't require anything but the cutoff adjustment done, but Bolle's needs the height adjusted a little IIRC. You'd want to take particular care to keep the pixels square while adjusting the CRT after installing.

For reference, here's the installation instructions I provide: https://github.com/ZigZagJoe/30Video-Support/blob/main/Installation Instructions HC-GS.pdf

 

[JDW]

The Xceed card does not do gamma correction so it will have a slightly darker picture.

That's the one thing which always irks me about my Xceed cards. (I have the MacroColor 30HR & Color 30HR.) When I compare with stock video, stock video is brighter, without being blown out or overly bright (where the black surround starts to brighten). So if the 30video can yield 256 grays at the same brightness level as stock video, that's saying a lot!

 

[zigzagjoe]

That's the one thing which always irks me about my Xceed cards. (I have the MacroColor 30HR & Color 30HR.) When I compare with stock video, stock video is brighter, without being blown out or overly bright (where the black surround starts to brighten). So if the 30video can yield 256 grays at the same brightness level as stock video, that's saying a lot!

I don't know that it's any brighter than stock video as far as pure B&W tones. I did notice that with the Xceed also. I believe it's a property of the video amp section of the neckboard. I would say on average it's probably still slightly dimmer than the stock video. I also noticed that the cutoff on the analog board "settles" a little as the machine warms up, so if the surround is slightly visible at power on, it fades out afterwards. This is why my instructions say to do brightness adjustments after the machine has run for a few minutes.

Gamma correction is for correcting the brightness of midtones rather than lows or highs. Practically, grays become somewhat brighter so it feels like the entire picture is brighter.

Why: CRT has a nonlinear response curve. Excerpt from DC&D.

I did some cursory testing with a brightness sensor and baremetal code with and without gamma correction. This setup is non-calibrated and not the most accurate, but you can see how the response is more linear with the correction applied.

 

[KennyPowers]

Haha, sorry, I might be a little touchy on the clones thing Yes, the neckboard I use is a derivative of Bolle's to be specific. Same basic circuit but bits moved around but functionally identical.

Can confirm that they're functionally identical as I happen to be using ZZG's neckboard with Bolle's video card.

 

[ph4z]

I am really impressed with the amount and quality of the response . I am learning a lot and am glad to have joined the forum. Thanks!

 

[sunvalleylaw]

This has been very helpful. Thank you for the question and post, and thank you @zigzagjoe for your answers.

 

[Paralel]

My card (30Video) and Xceed support optionally driving an external monitor while keeping the internal display active in black and white. Internal resolution remains 512x342. Bolle's card does only the internal display but at a higher vertical resolution of 512x384. That's the main difference between the cards, the rest is details.

To clarify though neither my card or Bolle's card are clones of the Xceed; that term is typically reserved for a reproduction of a vintage design. Bolle reverse engineered a vintage Formac video card and redesigned the timing, and my card is a new design with a newer video controller. Detail of the development of my card here , and the same for bolle's card here.

Image quality should be identical, I've never noticed a difference between the 30video and Xceed. I expect Bolle's would be the same. The Xceed card does not do gamma correction so it will have a slightly darker picture. I do not know if Bolle's card performs gamma correction or not.

Xceed is technically the fastest card, but that really is splitting hairs as these are all unaccelerated framebuffers so practical video performance actually depends more on your CPU accelerator rather than the video board.

My grayscale card is now the same card as used in my LCD kits, introducing 16 bit support, however as there is not a point in 16 bit grayscale (indistinguishable) it's only really relevant to external monitors. As it's the same card however there is an upgrade path to the LCD if you preferred.

Can it actually drive the internal CRT so that it actually has 65536 shades of gray on screen or is it limited by the neckboard to only 256? If so, that is something I would find very interesting. I've never seen 16-bit grayscale outside of professional radiography monitors. I've always wondered if a consumer monitor could be driven in a similar fashion if provided with the right hardware.

If it can't actually drive the CRT to 65536 shades of gray as a result of the neckboard, could said neckboard be upgraded to support 16-bit?

I apologize if you have addressed this elsewhere, or if it seems like a stupid question, but my knowledge in this area of hardware is very limited.

 

[zigzagjoe]

Can it actually drive the internal CRT so that it actually has 65536 shades of gray on screen or is it limited by the neckboard to only 256? If so, that is something I would find very interesting. I've never seen 16-bit grayscale outside of professional radiography monitors. I've always wondered if a consumer monitor could be driven in a similar fashion if provided with the right hardware.

If it can't actually drive the CRT to 65536 shades of gray as a result of the neckboard, could said neckboard be upgraded to support 16-bit?

I apologize if you have addressed this elsewhere, or if it seems like a stupid question, but my knowledge in this area of hardware is very limited.

The answer is no, in a complicated fashion.

The neckboard is fully analog and I can't particularly speak to the design of it. For the sake of this let's assume it and the CRT are ideal devices and will output a distinct shade given a particular voltage. From the video card side of things, "Thousands" video mode is a 15-bit color mode as far as Macintosh is concerned. Luminance mapping is done in hardware so given an image with sufficient bit depth under ideal conditions, 32768 shades would be possible. Practically this won't be achieved due to inconsistencies in the RAMDACs and accuracy limitations in the luminance mapping circuit. It's hard to say what the actual distinct number of shades are. Certainly far more than are of any use, even 256 grays is beyond practical limitations of the human eye.

In particular though 16 bit grays on my hardware will be worse than 8 bit mode as gamma correction is no longer possible at that depth; the actual "usable" bit depth would be further reduced due to the inherent non-linear response of the CRT.

 

[sunvalleylaw]

The answer is no, in a complicated fashion.

The neckboard is fully analog and I can't particularly speak to the design of it. For the sake of this let's assume it and the CRT are ideal devices and will output a distinct shade given a particular voltage. From the video card side of things, "Thousands" video mode is a 15-bit color mode as far as Macintosh is concerned. Luminance mapping is done in hardware so given an image with sufficient bit depth under ideal conditions, 32768 shades would be possible. Practically this won't be achieved due to inconsistencies in the RAMDACs and accuracy limitations in the luminance mapping circuit. It's hard to say what the actual distinct number of shades are. Certainly far more than are of any use, even 256 grays is beyond practical limitations of the human eye.

In particular though 16 bit grays on my hardware will be worse than 8 bit mode as gamma correction is no longer possible at that depth; the actual "usable" bit depth would be further reduced due to the inherent non-linear response of the CRT.

Well, when and if you do any more of them, I will be it like a ski bum on a powder day!