Just out of curiosity, what are you comparing DOSBox's output to in order to determine its accuracy?

From here (which appear *very* close from what I remember on my old 'Turbo' XT...🤣):

http://www.mobygames.com/game/pc-booter/ms-pa … man/screenshots

The 'old' composite method used in DOSBox (Last official release), was much closer to colors than they are now (Latest CVS). It really looks horrible now.

-Trebor

How does it look like in the "latest CVS"? Screenshot please.

Also, keep in mind that a real NTSC monitor/TV has a "hue" control, which is adjustable in the original code as well. I understand it is not in DosBox's implementation (many things are missing from the original code). I suspect the color difference is because DosBox's "hard-coded" hue setting does not match the setting you had on your TV with the "Turbo-XT" back then.

I understand what you are saying. If the hue adjustment is the issue, then let me rephrase my original statement:
"Nice job with the update to the composite mode so Burgertime looks a lot better in Composite mode.

However, that same fixed made the *hues* in Atarisoft Ms. Pac-Man look horrible. Ms. Pac-Man produced much more accurate *hues* in Composite mode before the changes which improved Burger Time's composite *hues*". 😁

Ok, here's the screenshots. I'll do 2 post - since five are the max number of attachment. This first set is what the latest CVS "hues" are producing:

-Trebor

Here's the second set of three - Before changes were made to the "hues":

These hues look all right to me. I thought you had like a pink Pac-Man or something.

I understand what you are saying.

No, you don't understand. The algorithm produces "accurate" colors in that the resulting RGB colors are what you see on a NTSC-compliant TV with the "hue" control at "default" position. If your memory tells you something else, it's because you didn't have your TV in the "default" position. The MobyGames screenshot was not taken at the "default" position, the screenshot author adjusted the "hue" control on his TV card to what HE thought was correct-looking. The original algorithm featured a "hue" control as well so people who don't like the "default" position can adjust the hues, just as TV owners can adjust the output to what they "feel" is right (which is usually wrong, see below).

Or, to make it simple: DosBox is right, and your memory as well as the MobyGames screenshot are wrong. 😀
You obviously have never played the arcade version of the game. The hues as they are presented now may not be accurate with regards to what you remember, but they are certainly accurate enough with regards to what the programmer most likely intended the game to look like. The arcade version's level 1 maze is a desaturated orange with red borders, not a wine-red (or solid purple) with purple burders.

It wasn't a TV, it was a RGB/Composite monitor set to the default factory position and never adjusted for *any* game. 😁

Never played the arcade version? How old do you think I am? Ha! I am an arcade fanatic - Especially for the 80's titles - Spent way too much of my youth in Arcades...🤣. And if your memory of the Arcades is faded (my hasn't thankful - Arcades now are *nothing like they use to be) - There is MAME.

Furthermore, concerning your mentioning on how the programmers "intended" the maze color for the 1st board would look in the arcade. Start up Ms. Pac-Man in MAME and play with the Gamma settings. Higher gamma settings produce a washed-out pink color for the 1st maze (Similar to the original composite settings - The output is just more saturated). Addtionally, look at Ms. Pac-Man ported to any console - including Atari 2600, 5200, 7800, NES, SMS, Genesis, SNES... What color do you see for the 1st board? Is it a desaturated orange, or a pink color?

Anyhow, look at those hues again. You really think that the current hues are all correct? Looking at the 1st board, the maze is too orange. I am not sure where you see a purple for the 1st board. The only purple maze is for the 2nd board reflecting the new composite adjustment. The old composite settings has the 1st maze as a bright pink color. I agree though the pre-adjustments were perhaps too pink. But now it's way too orange!

Looking at the hue of the ghost monsters...You really think that's what the programmers intended? Sorry, the hue of the ghost monsters are way off now. They were most closer to their arcade counterparts prior to the adjustments.

I'm not saying to the old hues were completely accurate. However, the current set is even further removed from accuracy. If you believe the hues are at their correct default values - fine...They sure don't appear accurate 😒

-Trebor

RGB/Composite monitor

Well, which one? RGB or composite?

Ms-Pac Man writes the same value (0x0f) to the color select register 0x3d9 that Burger Time writes. It follows that Ms. Pac-Man's colors must be exactly the same as Burger Time's. If you wanted the maze to be pink, the beans in Burger Time would be pink as well. Beans are not pink. The bun would be a pinkish red. Burger buns are not pinkish red. It's as simple as that. Ok?
If you want a pink maze, the game would have to write 0x3f to the color select register 0x3d9. It doesn't do that, however, it writes 0x0f.

Also, you won't get anywhere with your "phenotypical" approach --- the algorithm is based on the inner workings of a NTSC color decoder, not on some deliberative adjustments as to what "looks right". If every game --- King's Quest, Burger Time, Bruce Lee, BC's Quest for Tires, Oil's Well, Winnie the Pooh, Mickey's Space Adventure ---, except that single one looks right, do you think it's the algorithm, or the game?

Doesn't the booter version of this game support EGA? I have the original disk on 5.25" floppy but no longer have a 5.25" drive so I can't test it. Maybe my recollection wrong and I was only ever able to play it in 4-color RGB CGA because I didn't get into PCs until EGA was standard.

However, if it does support EGA then it might provide a useful comparison.

booter dunno. Tried it only with machine is cga. Offers option for RGB and composite.
the game is quite old so I'm not sure if the original had ega.

HunterZ wrote:

Doesn't the booter version of this game support EGA?

No. Unless there was a later version.

Sorry, you're right. Looking at RGB mode screenshots has brought back memories of what I remember. Interesting that so many CGA-only games had the ability to look better than what my EGA hardware could display in CGA modes at the time. I guess the hardware companies figured that most software that used composite CGA would be updated to use 16 color EGA instead.

NewRisingSun:
The Monitor was both RGB and composite - It has seperate inputs for the RGB cable and for composite input. It used a toggle button between using the two signals.

My "phenotypical" approach was based on response by your claim to what the arcade version looked like and the programmers "intentions". Sorry, if your claim previously made little sense when countered.

Again, I am not stating the entire new hues are incorrect. However, it appears some are completely off. I'm not stating the maze *has* to be pink. I even stated in my previous post, that the pink did seem a little too brite. But the orange is the extreme on the other side of the coin.

And thank you for bring up Bruce Lee because it just supports my point (And I didn't try it out until you mentioned it). Those hues are off in areas too. The same goes for King's Quest.

Look, I'm not trying to argue here. To be honest, I'm sick of this going back and forth. I'm not trying to win any battles. Just look at the hues. I'm not saying they are *all* incorrect - but *some* of values are clearly wrong. Whatever the reasoning you want to use - I'm not denying - The proof (Or the hues) speak for themselves.

Well...I didn't get it at first. I did not understand why you where so passionate about going against what I was stating, and so strongly for what's currently implemented. Then I look over the "40 Column Text Mode Issues", the new composite hues are a result of an implementation of "NewRisingSun's algorithm". Look, I'm not insulting your work and I'm not trying to make you look bad - okay? You clearly have put a lot of time and effort into this.

You want to be declared the winner of this "discussion"? Fine, it's yours. However, as far as stating your algorithm is the say all be all of composite output? Sorry, no dice. Something is incorrect. But hey, I'm just an end users who has a strong passion and great memory for games. I'm dropping this thread as arguing is futile. I simply stated what was an honest observation and not a personal attack on someone or their work. Have a nice day.

-Trebor

I'm not sure if DOSBox CVS is using NewRisingSun's most recent algorithim.

However, I will say that NewRisingSun's are calculated according to RGB conversions of the colors as produced on the NTSC "color wheel." NTSC TVs have a tint control to correct errors in the NTSC transmission signal. NewRisingSun has also adjusted for the "regular" differences between the official NTSC standard and TV manufacturer's tweaks to it. I believe this is sometimes called the "red shift" in the US and the "yellow shift" in Japan to make the flesh tones of the majority of the inhabitants of those countries to look more natural.

Does the algorithim show the proper effect that the higher bandwidth luninence signal has on the lower bandwidth colorburst signal, (i.e.
are the resulting phase shifts correct?)

Do the phase shifts accurately select the hues of an average NTSC TV or composite monitor?

You want to be declared the winner of this "discussion"? Fine, it's yours.

What the hell are you talking about? I just explained to you about port register writes, how the same register write must result in the same colors produced (Burgertime colors = Ms. PacMan colors). Do you understand that? Please tell me if you do.
And please limit your comments to the technical discussion and cut the projectional psychobabble --- this is not Oprah.

And thank you for bring up Bruce Lee because it just supports my point (And I didn't try it out until you mentioned it). Those hues are off in areas too. The same goes for King's Quest.[...] Whatever the reasoning you want to use - I'm not denying - The proof (Or the hues) speak for themselves.

It seems like you consider yourself an authority on which hues are "correct", and I'm not sure why. King's Quest's hues match those on the MobyGames page exactly, if that's what you're going for. Which hues do you think are wrong in King's Quest? 🙄
If you look at the "40 column mode issues" thread, you'll find that the person who did the MobyGames shots now says DosBox' colors are more accurate as well.

My "phenotypical" approach was based on response by your claim to what the arcade version looked like and the programmers "intentions". Sorry, if your claim previously made little sense when countered.

You said the arcade version has a washed-out pink. That's not "pink", that's "peach", which is a desaturated orange, the blue component is not nearly the same as the blue component (which it would if it was really "pink"). If you lower the gamma in MAME, the maze becomes actually more orangy.
You also told me to check out console versions. It's true that the Genesis port is pink, but Namco's own NES port has a green maze, for example, so I'm not sure you can go for console ports to get an accurate opinion.

"red shift" in the US and the "yellow shift"

"boost", not "shift". Yes, it's for fleshtone color improvement, but I don't think even that could turn orange into pink. As I wrote previously, those MobyGames screenshots look as if color register 0x3d9 contained the value 0x3f. The game, at least the version I have, writes value 0x0f, which gives you the "orange" hues as required for Burger Time, not the "pinkish" hues. Either there is another version of the game out there, that writes 3F instad of 0F to the register, or the register changes contents magically by some other means --- maybe some BIOS call that accesses register 0x3D9 which is not emulated.

Does the algorithim show the proper effect that the higher andwidth luninence signal has on the lower bandwidth colorburst signal, (i.e. are the resulting phase shifts correct?)

Since, as you have pointed out, the horizontal frequency is in sync with the color burst reference carrier, the line is divided into 160 four-column groups. Column one when lit gives you +I, two +Q, three -I, four -Q. Apply a low-pass filter, and you get the appropriate hues. If all four columns are lit, you get +I + -I and +Q + -Q, thus, no hue. The algorithm, at least my version, correctly emulates that, and QBix' 256-color mode adaptation is good enough an estimate, it just has a little too much color bleeding.
he only thing my algorithm does not accurately reflect are the LUMA artifacts --- dot crawl and the like. But that's hardly the problem this discussion is about.

What the hell are you talking about? I just explained to you about port register writes, how the same register write must result in the same colors produced (Burgertime colors = Ms. PacMan colors). Do you understand that? Please tell me if you do.

I do understand what you are saying. Again, I'm not denying the technical specifics behind your claim. I never had or played Burgertime in composite mode on the original equipment/display. So, I have no way of making a distinction with that game and the accuracy of the colors/hues. I have played Atarisoft's Ms. Pac-Man in composite mode on an actual IBM XT compatible clone

And please limit your comments to the technical discussion and cut the projectional psychobabble --- this is not Oprah.

Sorry, Dr Phil. But you started that crap with the "'phenotypical' approach" analysis. 😁

It seems like you consider yourself an authority on which hues are "correct", and I'm not sure why. King's Quest's hues match those on the MobyGames page exactly, if that's what you're going for

I don't consider myself an authority on hues. Again, I am giving my observation from experience. Earlier, you stated the Ms. Pac-Man composite screenshots from Moby were not a reliable reference. However, the King's Quest screenshots from Moby are a reliable reference?

You said the arcade version has a washed-out pink. That's not "pink", that's "peach", which is a desaturated orange, the blue component is not nearly the same as the blue component (which it would if it was really "pink"). If you lower the gamma in MAME, the maze becomes actually more orangy.

Peach? That color doesn't look "peach" to me. I'm a little lost with your point of 'the blue component is not nearly the same as the blue componenet' statement. However, I do agree with you on the lower gamma settings point. If you turn the gamma down real low it does take on a slight orange look to it.

You also told me to check out console versions. It's true that the Genesis port is pink, but Namco's own NES port has a green maze, for example, so I'm not sure you can go for console ports to get an accurate opinion.

Come on. The Namco port(s)? Please. Namco butchered both the Pac-Man and Ms. Pac-Man ports to the NES. The *real* Ms. Pac-Man port - which is much more closer and strived for arcade accuracy is the Tengen one on the NES. Guess what color the first board is on the Tengen port of Ms. Pac-Man on the NES? Here's a hint - It is not "desaturated orange".

Look, NewRisingSun, I have no hard feelings toward you, or any vendetta to being down what has been implemented. Looking at the "40 column Text" post has a new user "Servo" who just signed up after discovering these post and posted information which backups up what you are stating. If I'm wrong, I admit it - and I'm sorry. However, I am almost 100% certain I am not. I also did a new install/creation of Starflight and Wasteland for Composite mode graphics and again, some of the hues/colors/whatever do not look right there either. And no, it was not the EGA version of Starflight that I played. I had the original release version which came on two 5 1/4" 360Kb low density floppies. Your graphic option were CGA, Monochrome(Hercules), and Composite.

Maybe it was the graphics card in my system - I do not know, and sadly no longer have the system. Not that you may be interested, but it was an IBM XT clone made by Laser. The graphics card was Monochrome, 4 color CGA, and Composite. I had no hard drive and no 3 1/2" floppy. I had duel 5 1/4" 360Kb low density drives. It ran in the "normal" 4.77Mhz or "turbo" 10Mhz pressing a toggle button in front of the CPU case. My monitor was made by Magnovox, and as I said could toggle between CGA mode and composite mode by pressing a button on the control panel of the display.

The reason I am so "confident" of the colors is because the composite supported games seemed so few and far between - At least from what I owned and knew. Anytime I acquired a composite supported game, I was thrilled and those games left a burning image in my mind. Especially, with the alternative of only 4 color CGA support for games that did not support composite mode. Ms. Pac-Man, Bruce Lee, King's Quest, Starflight and Wasteland were some of favorites. Although admittedly I like Bruce Lee on the C64 better (I know heresay)...🤣.

In any event, new presective: Thank you NewRisingSun for your hard work on the Composite Output. IMHO, the hues do not look 100% correct. However, I'll blame it on my graphics card that I was using way back then (20+ years ago). Unless someone else backs what I am saying, I guess nothing else will be or can be done about it.

I believe there is some sort of "Font" issue concerning "Standard Fonts" or/and some odd or different fonts some graphics card may or did use for a particular graphic setting(s).[/?] Perhaps some graphic cards had the hues 'factory set' differently then others. Maybe some monitors had their hue setting 'factory set' differently then others. Just like there were different fonts implemented by different cards, perhaps the same is true concerning colors/hues.

Unfortunate and unique to me at the time was my setup. My friends and family either had Tandy 1000's, C64, Apple IIe or some other sort of machine. So my experience is unique to my machine I guess. Unless of course anyone has or knows of anyone who owned an Laser Turbo XT...🤣.

-Trebor

But you started that crap

You asked how old I think you are. Based on that comment, 10.

with the "'phenotypical' approach" analysis.

"phenotypical" means you look at the observable results (hence "pheno"), while I look at what's going on inside the color decoder, the behavior of which I am emulating with the algorithm.

Let's sum this up for you:
There are two ways how NTSC colors could be wrong: the algorithm, or the user-adjustable settings. You don't like the results with the default settings, based on your memories. Other people tell us that the default settings are correct. I tell you that by adjusting the user-adjustable hue setting, you can get your "pinkish" look.
Now what does that tell us? Should I change the algorithm, or should you adjust the user-adjustable settings?

And that concludes this discussion. 😀

You asked how old I think you are. Based on that comment, 10.

You really have a unique style with people, don't you? I never attacked or insulted you personally, but you clearly remain on the offensive. I just finished thanking you for your hard work, and even apologized for the misunderstanding of my observations and your analytical approach to the situation.

"phenotypical" means you look at the observable results (hence "pheno"), while I look at what's going on inside the color decode […]
Show full quote

"phenotypical" means you look at the observable results (hence "pheno"), while I look at what's going on inside the color decoder, the behavior of which I am emulating with the algorithm.

Let's sum this up for you:
There are two ways how NTSC colors could be wrong: the algorithm, or the user-adjustable settings. You don't like the results with the default settings, based on your memories. Other people tell us that the default settings are correct. I tell you that by adjusting the user-adjustable hue setting, you can get your "pinkish" look.
Now what does that tell us? Should I change the algorithm, or should you adjust the user-adjustable settings?

Well, thank you Mr. Webster for defining what I already know 'phenotypical' to mean.

And thank you, for reiterating my conclusion(s) in my last post. Essentially, as I stated, my monitor or/and graphic card probably had different default hue settings then what is currently implemented.

Have a nice day.

-Trebor

I am starting to think about closing this topic...

Mirek