TMS9918A / TMS9929A / F18A Aspect Ratios

[briantw]

Regarding a recent discussion on the various aspect ratios available to the stock console (or emulations thereof), there seems to be some clarification needed on what the actual aspect ratios are of the original TI output (both 60Hz and 50Hz). I say "needed" because, firstly, this info is very hard to find, and, secondly, there is a belief among some that the TMS9918, at least, has a 4:3 aspect ratio addressable pixel area with square pixels. This is not true, and, consulting the datasheet to solidify the numbers in my own mind, it is clear that neither the TMS9918A nor the TMS9929A employ use of square pixels. Most of the emulators I've seen use square pixels. Hopefully these numbers will be of use to someone else at some point.

 

On page 5 of a document entitled TMS 9918A 9928A 9929A VDP Preliminary Specification 1981, there's this section:

 

 

The hasty handwritten figures aren't mine; that's how I found this document on the web, and boy, it is not an easy document to find. Here it is for posterity, and, from this datasheet, plus the F18A documentation here, I've put together a side-by-side comparison of the different aspect ratios offered by each VDP.

 

Remember that all of these are designed to send their output to a 4:3 screen. With the TI's addressable pixel area being 256x192, or also 4:3, you may think that this is a great match: 4:3 addressable area to a 4:3 screen = 4:3 aspect ratio. But no, the timings of the various analogue video systems have to be respected, so extra pixel rows and columns are required to border the addressable area. The 60 Hz VDP outputs an area of 284x243, although only 256x192 pixels are addressable, while the rest of the 'pixels' are set to a solid background colour. The 50 Hz VDP outputs an area of 284x294, although, again, only 256x192 pixels are addressable. Of course, there are no actual 'pixels' in the border area, but the pixel clock is running, nonetheless, and a pixel width is used as a reference in the above datasheet for the amount of vertical or horizontal space occupied by this area.

 

The F18A outputs square pixels, as asserted by the chip's designer. However, square pixels (1.333:1) are a slightly different 'squeeze' to what you'd get from an NTSC console (1.52:1), and very different to what you'd get from a 50 Hz console (1.82:1).

 

Given the numbers provided by the creators of all three products, the side-by-side comparison of aspect ratios looks like the following diagram:

 

 

Personally, I most prefer the NTSC shape. It's a nice fit for the available screen area. The 50 Hz (PAL) that I have access to I.R.L. is too squashed in the vertical direction (or "short"), while the F18A is too "tall" for my liking.

 

That being said, my previous sentence is merely my opinion, and not a slight on anyone nor any reason to take offense. I'm just stating a preference, and y'all are welcome to your own!

 

Also, please note that results may vary, depending on the knobs and settings on your old CRT. On some of them, you can adjust H-WIDTH, and even V-HEIGHT is often a possibility. Both of those will alter the aspect ratio by squeezing or stretching the image. LCDs do a better job of remaining faithful to the input signal and not chopping anything off, but, if you fiddle with the menus, you can usually override that as well, depending on the monitor.

 

If I've made any errors, please let me know!

 

 

 

 

[Willsy]

Can't add any technical information to your very informative post, but I can tell you that on PAL systems using the RF modulator (I don't know about NTSC system) *at least* the first three characters all the way down the left hand side of the screen are not visible. This seems to be a problem caused by the modulator as using a monitor and composite/analog (I.e not RF) does not cause this issue. That's why TI BASIC uses a 28 column screen. Not sure if the above info is of value or not. For what it's worth...!

[TheMole]

For what it's worth, I never cared about the aspect ratio of the pixels, I only cared about maximizing my screen real estate. I used a philips monitor that, as you point out, had h-width and v-height controls and always calibrated the picture to take up as much screen real estate as possible without losing rows/columns to overscan. I guess that means I've always had square pixels (4:3 ratio resolution filling up a 4:3 ratio screen) on my PAL console, and I assume that at least those of us who were using crt monitors back in the day must've done something similar. Heck, I can remember doing this for all my computers over the years, from old 8-bitters to more 'recent' PC's with VGA monitors.

[+OLD CS1]

Can't add any technical information to your very informative post, but I can tell you that on PAL systems using the RF modulator (I don't know about NTSC system) *at least* the first three characters all the way down the left hand side of the screen are not visible. This seems to be a problem caused by the modulator as using a monitor and composite/analog (I.e not RF) does not cause this issue. That's why TI BASIC uses a 28 column screen. Not sure if the above info is of value or not. For what it's worth...!

 

Like that for a lot of NTSC TVs, too. I remember on my little 10" B&W I could see everything, but on the main 21" color TV I could not see the outer two columns (1 and 32,) and on a big color I picked up at a garage sale I could see everything. Strange things.

[Stuart]

Can't add any technical information to your very informative post, but I can tell you that on PAL systems using the RF modulator (I don't know about NTSC system) *at least* the first three characters all the way down the left hand side of the screen are not visible. This seems to be a problem caused by the modulator as using a monitor and composite/analog (I.e not RF) does not cause this issue. That's why TI BASIC uses a 28 column screen. Not sure if the above info is of value or not. For what it's worth...!

That's caused by a timing bug in the VDP chip. It gets a mention in Karl Guttag's stuff on spatula-city.org.

[+OLD CS1]

That's caused by a timing bug in the VDP chip. It gets a mention in Karl Guttag's stuff on spatula-city.org.

 

Definitely good reads here. In particular I like VDP Debacle Memo and how two years later there are still problems. Some of the things he complains about at TI are eerily similar to complaints of engineers at Commodore, especially where the Amiga was concerned.

 

In another document, message on modification to 9918 from 1982, there is a mention of a 64 byte color table in graphics 2 mode. This is bit-map mode, correct? I have never read about using a 64 byte color table in bit-map, so I assume the "logic error" was never corrected; and what would be the implications of a 64 byte color table, anyway?

 

EDIT: And bit-mapped ("super pattern") mode only required an additional 50 transistors?!

[Asmusr]

In another document, message on modification to 9918 from 1982, there is a mention of a 64 byte color table in graphics 2 mode. This is bit-map mode, correct? I have never read about using a 64 byte color table in bit-map, so I assume the "logic error" was never corrected; and what would be the implications of a 64 byte color table, anyway?

 

The 64 byte color tables, I guess, is what you get when you set VR3 to 0 in bitmap mode (see http://www.unige.ch/medecine/nouspikel/ti99/tms9918a.htm under Hybrid bitmap modes). Unfortunately reducing the color table to 64 bytes also reduces the number of unique patterns to 8 because the color mask also affects the pattern mask (this sounds like the bug mentioned in the note). Otherwise it could have been useful for creating color reduced bitmap graphics that could almost have been double buffered in VDP RAM. I say "almost" because the first 8 patterns would be occupied by the color table of the other buffer and vice verse. As it is you would have to use the undocumented and slightly awkward bitmap text mode if you wanted to double buffer a full bitmap screen.

[matthew180]

Standard CRTs (not talking about widescreen formats) are physically a 4:3 (1.333) ratio, which covers TVs for both NTSC and PAL. The active display area of the 9918A, 9928, and 9929 are all 256x192, which is 4:3 (256 / 192 = 1.333). If you use the hsize and vsize controls on an analog TV and adjust the display (assuming the TV will actually cooperate) such that the active area fills the physical visible area, the pixels will maintain the 4:3 ratio regardless of the frame format (NTSC or PAL).

 

TI knew that TVs have a wide range of capabilities so they added the borders to try to keep the active display visible on the physical surface of the CRT.

 

For NTSC and PAL there are a set number of vertical lines and a refresh rate. NTSC has 262.5 lines at 60Hz and PAL has 313 lines at 50Hz. That means each horizontal line is:

 

NTSC: 63.49us per horizontal scan line (1/60 = 16.666ms per frame / 262.5 lines = 63.49us)

PAL: 63.89us per horizontal scan line (1/50 = 20.000ms per frame / 313 lines = 63.89us)

 

A CRT has no horizontal resolution, thus no aspect ratio. It is not until the horizontal scan line is arbitrarily sliced up that a pixel ratio can be determined. The NTSC specification designated a 4:3 aspect ratio for broadcast TV (I don't know about PAL). The max horizontal "resolution" is a limitation of electronic and physical characteristic of the tube itself and the quality of the circuits.

 

Note that the horizontal and vertical times do not speed up or slow down at any time during a frame, they are constant. However, their physical size (amplitude and offset of the voltage oscillator driving the coils) can be adjusted.

 

In the case of the 9918A, TI decided to divide the horizontal time into 342 slices. This means that the amount of time spent drawing a pixel is fixed to 185.649ns per pixel (for NTSC). As soon as they did that, they established a horizontal pixel width compared to the number of lines (pixel height). They also chose to ignore NTSC's half a scan line, knowing TVs would not have a problem keeping sync (although some modern TVs do have a problem with this). Thus you have:

 

NTSC: 342 horizontal pixels / 262 scan lines = 1.305 which ~= 1.333 or 4:3 ratio

PAL: 342 horizontal pixels / 313 scan lines = 1.092 (not sure the ratio)

 

On an analog TV you can adjust the hsize and vsize separately which changes the physical distance the electron beam travels, so you can affect the appearance of the pixels to some degree. However, those changes do not affect the timing of the horizontal or vertical periods, nor the physical 4:3 ratio of the glass CRT. Adjusting the 256x192 active display (which has nothing to do with the arbitrary borders) to the physical area of a CRT will give the same pixel ratio of 4:3 on an NTSC and PAL system.

 

I suspect in reality you cannot adjust the vertical borders off the screen on a PAL TV (you might be able to do it on an NTSC TV though), so the pixels will have a different physical look. Back in the TV days, everyone would perceive a different ratio depending on how their TV was adjusted.

 

Modern monitors give us the benefit of perfectly filling the physical flat display, so our physical pixel ratio is always fixed to that of the monitor.

 

640x480@60Hz "VGA" has 525 lines and a horizontal scan line period of 31.469us (basically twice NTSC), however the spec divides the horizontal scan line into 800 divisions, or roughly 40ns per pixel. That means:

 

VGA 640x480: 800 horizontal pixels / 525 scan lines = 1.523 ratio.

 

It appears the F18A would have a slight ratio problem. However, when used with a 4:3 monitor (not wide-screen) the 640x480 active display is stretched to fill the physical monitor screen, just like you would with the hsize and vsize of and old CRT TV. So the pixel area the F18A is working in is a forced 4:3 aspect ratio. From that the F18A uses fat pixels (2x the VGA pixels) and margins to center the active 256x192 display of the original VDP. VGA 640x480@60Hz is also considered a 4:3 aspect according to Google.

[briantw]

Hey Matt

 

I've never been able to adjust horizontal or vertical size on old CRT televisions (not without opening them up or sticking a screwdriver through the back of the TV console), nor would I have been advised to when I was sharing the family TV with everyone else who would like to watch their sitcoms in the correct aspect ratio. So, the PAL I was subjected to as a kid was always very squashed.

 

It is possible, to a degree, with CRT monitors, but because the squeeze was so extreme in 50Hz, it's not always possible to get to square pixel territory. I managed it with a Philips CM8833-II, but not with my Sony KX-14 CP1. The latter monitor, which I still use, has no H-size option, and the V-size allows for slight variation.

 

 

VGA 640x480: 800 horizontal pixels / 525 scan lines = 1.523 ratio.

That's very close to the 1.52 rounded off in the figure above!

 

I think I may be mistaken in my recollection of the V9t9 emulator (not discussed above). I believe it was running at 320x200, not 320x240, so the image was even taller than the F18A, i.e., on the opposite extreme of the 50Hz chip. I need to fire V9t9 up again after almost 20 years, but I don't have the hardware to support it. However here's a video I shot in the late 90's. You can hear the difference between the emulator at 60Hz (left) and the TI at 50Hz (right). You can also tell the difference in corrected aspect ratio (monitor on the right vs. big TV that has no adjustment possibilities externally). I'll post the link below as I'm going to have to download the vid from FB and up to YouTube.

 

V9t9 does look a little extreme in the vid.

[briantw]

Here's the link: http://youtu.be/-M50p14KFx4

 

My bad memory - the Philips monitor isn't running an original TI in PAL; it's also a PC running an emulator. The only real TI there is the one on the TV. Aspect ratio difference still visible though.

Edited January 30, 2015 by briantw

[briantw]

640x480@60Hz "VGA" has 525 lines and a horizontal scan line period of 31.469us (basically twice NTSC),

 

Agreed, but only 480 are actually active and addressable. In order to avoid confusion, I must reiterate that, in my diagrams above, I am ONLY discussing the lines and pixels in the signal that are actually active and visible. Front porch, back porch, blanking and sync are not taken into consideration; only the active image area.

 

It's pretty clear from the datasheet, but, for NTSC this means 243 active image lines on a screen that shows the entire field, and 294 in PAL. The 262.5 and 312.5 lines for NTSC and PAL, respectively include inactive lines that are not shown on the screen in anyway, are not part of the display, and have thus been removed from consideration.

 

I did learn one thing from this discussion, however: Apparently VGA 640x480 includes a border of 8 pixels at the top and 8 pixels at the bottom, as well as 8 pixels on the left and right, making the full displayable area 656x496. I did not know that. It's great when I can learn something from a discussion.

 

http://www.epanorama.net/documents/pc/vga_timing.html

 

 

 

2 lines front porch 2 lines vertical sync 25 lines back porch 8 lines top border 480 lines video 8 lines bottom border

 

Read more at: http://www.epanorama.net/documents/pc/vga_timing.html

[Davvel]

Can someone kindly send me the PAL circuit diagram for the TI 99/4A as I am having problems debugging an issue regarding the TMS9929 output (Black Screen long tone) which I have been working on for long now. I replaced the 10.7mhz crystal and got a fully tested and working tms9929 chip but to no avail.

 

I noticed that the circuit diagram I have shows that pin 40 is hooked to a 12pf ceramic capacitor which in turn is grounded to earth. In my circuit board I think it is wired differently (The big brown ceramic capacitor next to the crystal).

 

I also have an extra resistor (56 ohm) going to earth connected to Pin 39 which I could not find in the circuit diagram.

 

I have 2 identical TI 99 4A's which I can compare to. I removed the video chip from both the good and the bad computers and started checking some things on both.

 

Pin 39 & 40 on Good had 5v and logic probe made a high pitch sound.

Pin 39 & 40 had very low voltage (1v - 2v) and logic probe emits no sound at all.

 

All the other voltages in both computers (VCC,VDD,VBB,VSS) are identical.

 

I do not have oscilloscope yet.

 

Do you have any hints.

 

Thanks and best regards - Wishing all happy holidays.

 

David

 

David

[Willsy]

If you're getting a black screen and tones then I would say that the 9900 CPU is not initialising. The first areas I would look at would be the 256 byte RAM chip and the ROMs. If either of those has failed, then nothing will work. I'd check the RAM chip first. If those check out okay, check the 9900 cpu. I've never known a CPU to blow though.

Edited December 15, 2015 by Willsy

[Stuart]

Circuit diagram - page 3 of [ftp://ftp.whtech.com/datasheets%20and%20manuals/Hardware/TI%20Circuit%20Diagrams%20and%20Schematics.pdf] (but it sounds like you might have this already?). There are a number of links which are in place or removed depending on the type of VDP for PAL or NTSC output.

 

As Mark said above, if the console is giving a long continuous tone when you switch on, the processor is not resetting and running correctly. Other culprits might by the 3 GROMs (try reseating them), the 9901 and the 9904 clock chip.

 

The 56 Ohm resistor going to earth from pin 39 - sure it's not a capacitor? Can look very similar ...

[sometimes99er]

Nice thread. Thanks.

 

 

I've been running TI-99/4A emulation for more than 10 years.

 

Mostly I preferred a pixel-perfect 2x2 window.

 

Occasionally, while testing or playing, I went full screen. This make the pixels somewhat fuzzy, but overall a very good feel. Much more like, I'm sitting in front of a TI. - However the visuals are a bit bombastic. - And since having widescreen format, I was a little, what and okay.

 

Since reading this thread, with estimated aspect ratios of 1.5 for NTSC, 1.8 for PAL and 1.3 for F18A, I have had the occasional pixel-perfect 3x2 window in emulation. Expecting an aspect ratio of 1.5 (3/2). And becoming more and more happy about it.

 

Only now I calculated the aspect ratio of my widescreen monitor. Perfect 1:1. So the 3x2 windows are then actually perceived as an aspect ratio of 1.5. Good.

 

So I've been developing (having a bit of fun) with 1:1 for years. Now I will definitely be using 3:2 more.

 

[Omega-TI]

As long a circle on the screen comes out as a circle on the printer I'm happy.

[Paradroid90]

Hi Everyone, I apologise if this has been answered before. I am quite new to the TI994A and looking for some advice.

 

It seems that getting an image out of compontent out of my PAL UK unit is proving a real challenge.

 

I have 3 TV's which will accept Compontent. My Samsung Multisync confirms that it cannot display the signal, a small LCD confirms the same.

 

My main TV Technika displays the TI signal but the colours are off. On Parsec for example, the black background comes up blue.

 

Three quick questions if I may.

 

1 - Is there anything I can do to make the component cable to display on the other screens

2 - Can you install a NTSC chip in a pal machine to use composite

3 - Should I just bite the bullet and save for a F18?

 

Thanks in advance.

[blackbox]

Impressed that your Technika can give you something- does it by any chance have a colour difference (YUV) input somewhere?

 

The PAL TI99/4a does not have an RGB output which your other monitors are looking for.

 

You could possibly connect your PAL TI99/4a to give you a stable monochrome image,available between pins 6 and 2 (earth and luminance).

 

Getting good colour is either very expensive (there are a handful of monitors that have an analog colour difference input) - or not easy- find an analog tv set that can tune to Channel 33 and use the TI modulator. A charity shop or "antique" shop may help.

 

In theory the colour difference output was higher quality than an RGB output could offer.

 

The console output on the 6 pin connector is sound, B-Y, R-Y, Luminance, +12v and centre earth. The console output is happier feeding a high impedance - 600 ohms- and going as low as 75 ohms could damage the console.

 

The Einstein computer used a similar output and if you can find one an old Tatung Einstein monitor will probably work.

 

The F18 is a replacement video chip with more useable output for modern monitors. A handful of odd TI99/4a programs will not function with it due to a dropped mode.

 

regards

 

 

Hi Everyone, I apologise if this has been answered before. I am quite new to the TI994A and looking for some advice.

 

It seems that getting an image out of compontent out of my PAL UK unit is proving a real challenge.

 

I have 3 TV's which will accept Compontent. My Samsung Multisync confirms that it cannot display the signal, a small LCD confirms the same.

 

My main TV Technika displays the TI signal but the colours are off. On Parsec for example, the black background comes up blue.

 

Three quick questions if I may.

 

1 - Is there anything I can do to make the component cable to display on the other screens

2 - Can you install a NTSC chip in a pal machine to use composite

3 - Should I just bite the bullet and save for a F18?

 

Thanks in advance.

 

[Paradroid90]

Impressed that your Technika can give you something- does it by any chance have a colour difference (YUV) input somewhere?

 

The PAL TI99/4a does not have an RGB output which your other monitors are looking for.

 

You could possibly connect your PAL TI99/4a to give you a stable monochrome image,available between pins 6 and 2 (earth and luminance).

 

Getting good colour is either very expensive (there are a handful of monitors that have an analog colour difference input) - or not easy- find an analog tv set that can tune to Channel 33 and use the TI modulator. A charity shop or "antique" shop may help.

 

In theory the colour difference output was higher quality than an RGB output could offer.

 

The console output on the 6 pin connector is sound, B-Y, R-Y, Luminance, +12v and centre earth. The console output is happier feeding a high impedance - 600 ohms- and going as low as 75 ohms could damage the console.

 

The Einstein computer used a similar output and if you can find one an old Tatung Einstein monitor will probably work.

 

The F18 is a replacement video chip with more useable output for modern monitors. A handful of odd TI99/4a programs will not function with it due to a dropped mode.

 

regards

 

 

 

Thanks for the feedback, Reference the Technika I do not think it has, but as the unit is the front room, it is not very practical. Would a cheap option be to by an NTSC TI99 and run on composite on my Samsung Multi Sync? Also can you just swap the graphics chip to convert a Pal to an NTSC system.

 

Thanks

[Asmusr]

You cannot just switch the chip with an 9918A. You can switch it with an F18A but they are currently out of stock.

 

I had some success with a Sony Bravia TV:

http://atariage.com/forums/topic/172937-getting-s-video-from-the-ti-console/?do=findComment&comment=2759716

 

But the most practical solution, if you don't want to use a modulator, is probably to get an NTSC console. It works with the European power supply.

Edited September 23, 2018 by Asmusr

[Paradroid90]

You cannot just switch the chip with an 9918A. You can switch it with an F18A but they are currently out of stock.

 

I had some success with a Sony Bravia TV:

http://atariage.com/forums/topic/172937-getting-s-video-from-the-ti-console/?do=findComment&comment=2759716

 

But the most practical solution, if you don't want to use a modulator, is probably to get an NTSC console. It works with the European power supply.

Thanks for the feedback - Could not see that one for sale, and from what I could see the screen is 40 inch, bit too big for my office, but did see a Sony KDL-20S3000 20 in. LCD TV local to me on E Bay, with Component and S Video, so could be a good Retro Monitor what ever happens. Wish me Luck :-)