What is the autoguess feature? (sorry, I'm too lazy to dig through 13 pages of posts) Is it the way DOSBox decides whether a game should be passed through the composite emulation?

It would be really, really nice if we could at least override it with a hotkey or dosbox.conf setting. I've got a TV connected to the TV Out of my computer and would like to try testing some composite mode games on it using DOSBox (if this is even possible).

it's the behavior that makes kq1 display the correct colour at startup in composite mode.

it writes an unused bit. Which caused on some cga clone NewRisingSun was using when design the algorithm to produce a shift in the colours.

As the game did this write intentionally with a clone in mind. (so the behavior is "better" when you have clone). That's my primary reason to keep it in.

But the guess behavior is non standard and not present on the "official" cga card. That's why newrisingsun wants it out.

Autoguess tries to automatically guess which "hue" setting is correct by looking at unused register bits of port 0x3d9. It's not how the actual hardware works (hence 'unused'). Qbix wants to keep it because he likes how this incidentially makes KQ1 look like on MobyGames without any user adjustment. Of course you could also just set the default hue to -15 and keep the decoder itself intact, but why so simple if you can do it the complicated way? 😀

I personally don't care if it's incorrectly implemented on purpose in DosBox or not, as I'm not going to use DosBox anyway, but others just might care.

Autoguess tries to automatically guess which "hue" setting is correct by looking at unused register bits of port 0x3d9. It's not how the actual hardware works (hence 'unused'). Qbix wants to keep it because he likes how this incidentially makes KQ1 look like on MobyGames without any user adjustment. Of course you could also just set the default hue to -15 and keep the decoder itself intact, but why so simple if you can do it the complicated way?

I personally don't care if it's incorrectly implemented on purpose in DosBox or not, as I'm not going to use DosBox anyway, but others just might care.

Very interesting considering that MobyGames' KQ1 PC Booter screenshots were taken using the real CGA. Does anyone know which clone or clones this is supposed to work on? Not any of the important (PCjr., Tandy, Plantronics, 5140) clones capable of composite video, I'll wager. However, if it doesn't affect those games that assume a true IBM CGA is in the system, keep it.

What does (non-) interlacing have to do with this? CGA underscans, NES doesn't. Simple.
The slowness I'm talking about is not the slowness of a 16-bit/24-bit video mode, but the slowness of the gaussian filter used by the code.

I am correct in saying that the CGA does not doublescan for a composite color monitor or TV, it outputs 320x200 or 640x200, not 640x400. I would like to see Servo's capture card capture NES output.

How slow is the gaussian filter? Is it slow enough to bring a powerful system to its knees?

Ok, I dove into my closet of insanely numerous machines and can offer the following observations:

Yes, but I don't believe anyone has confirmed actually hearing it via Tandy DAC yet (although I could be mistaken)

I tried this out on a Tandy 1000 RL; the "Where am I?" speech and other sampled sounds are all there. Sounds pretty good, too, the same as a SoundBlaster.

For King's Quest PC Booter, I dug out a true IBM PC 5150. I have two CGA cards (shown below) I tried in this system, one apparently a PC Peacock by MA Systems (anyone heard of it?) and an IBM CGA card (perhaps someone can verify this is in fact an IBM CGA from the picture). Comparing the two cards, the clone looks identical to the MobyGames screenshot. The IBM card has very similar colors, but the image is a bit brighter; the hue might be slightly different, but didn't appear drastically so; the brighter image was the most noticable distinction. Looks like that hue setting does work with an IBM CGA ? I tried to take some pictures, but they didn't turn out terribly well (the colors are actually a lot closer to the Moby shot than they appear in that picture). Interestingly on my capture card the colors only look correct if I adjust the hue slightly; most games look correct with no hue adjustment.

I would like to see Servo's capture card capture NES output.

There's a few of those lying about...

The gold-brown PCB card is the real IBM CGA. I would expect the clone to offer additional functionality for a second board. The CGA outputs some very nice video quality, I would imagine. Also, if the picture you took is of the IBM CGA card, the berries in the picture look more red, whereas they look purple on your MobyGames screenshots. I would have assumed that red was the color Sierra was going for, but someone showed a magazine photo where the color was more purple. I think the brighter image, being more livelier and nicer to look at, should be the goal.

NewRisingSun wrote:

SirGraham wrote:

Incidentally, running the game without the switch will cause a crush sooner or later

I wasn't aware of such power.

Well, it says so on spacequest.net and I also witnessed it myself - if I started the game without the -N switch, sooner or later it asked for a missing 992.V56 or 993.V56 file. It's surprising that you're not aware of this, since you know SQ4 so well.

NewRisingSun wrote:

SirGraham wrote:

So, I don't see why it shouldn't be included in the next release of DOSBox

It's not finished, it requires 16-bit or 24-bit color, and it's too slow; that's why.

Oh, ok. Do you have any plans of making it workable with DOSBox?

HunterZ wrote:

He's concerned that if AGI games detect the presence of Tandy sound hardware when DOSBox is not in Tandy mode, there will be no way to experience the IBM PC Speaker sound that those games can output on non-Tandy machines. I don't think anyone has addressed this question yet.

Yes, that's what I asked, and wd replied "It's off by default for machines!=tandy", and that's what I didn't understand. Will we still get PC speaker sounds out of games that auto-detect Tandy?

NewRisingSun wrote:

AGI simply detects a Tandy by looking for the string "Tandy" in the BIOS segment, and a PCjr by checking for byte FD at FFFFE.

So when machine is not set to Tandy, but Tandy speaker is enabled, the AGIs won't detect it?

NewRisingSun wrote:

The slowness I'm talking about is not the slowness of a 16-bit/24-bit video mode, but the slowness of the gaussian filter used by the code.

Games like Ulysses and Serenia don't even have animation, how slow can it be?

Qbix wrote:

Well The hue control resides nicely on my harddisk. It will make it into cvs builds eventually but I'm in no hurry.
(It works nicely btw)

You haven't succesfully talked me out the cheap autoguess feature.

It's nice to know that the hue control wasn't forgotten.
There's something I don't understand - how can DOSBox incorporate both the hue control and the auto-guess feature? Aren't they contradictive?

Servo wrote:

I tried this out on a Tandy 1000 RL; the "Where am I?" speech and other sampled sounds are all there. Sounds pretty good, too, the same as a SoundBlaster.

Ok, that settles it then. But how can DOSBox emulate the Tandy DAC without the ability to produce sampled sounds?

> Well, it's not really a shame, it's just that other than this small problem
> DOSBox provides an almost perfect Tandy emulation (at least for the
> games I'm playing).

Well then get it fixed and submit the patch. Just as easy as that.

> Will we still get PC speaker sounds out of games that auto-detect
> Tandy?

Why should they auto-detect a tandy if the machine is set to !=tandy??

> So when machine is not set to Tandy, but Tandy speaker is enabled, the
> AGIs won't detect it?

Unless you forced the tandy sound to be enabled in dosbox.conf you will
not get any tandy sound for machines!=tandy. If it is forced for all
machines, it depends on the game whether it provides support for
tandy sound on non-tandys.

Why is this a concern? Tandy Graphics = Tandy Sound. CGA or EGA Graphics = PC Speaker or Sound Card. I suppose the only time it would come into play is when the game supports Tandy Graphics but only PC Speaker sound. In that case, I would expect to hear PC Speaker sounds. If a game supports Tandy Graphics, it will gain nothing by going to EGA (except if the EGA's default resolution is 640x200/350.) and will lose Tandy Sound.

Yes, exactly.
The only problem with tandy3v sound (NOT the tandy DAC) is that
the 2nd dma controller uses the same port as the tandy. Thus
tandy sound is disabled by default for machines!=tandy, and when
tandy sound is forced-enabled for all machines the 2nd dma
controller is closed (effectively downgrading the sb16 to sbpro).

Great Hierophant wrote:

Very interesting considering that MobyGames' KQ1 PC Booter screenshots were taken using the real CGA.

Yes, but you don't know the hue setting on his TV card back then.

Great Hierophant wrote:

I am correct in saying that the CGA does not doublescan for a composite color monitor or TV, it outputs 320x200 or 640x200, not 640x400.

I didn't say "doublescan", I said "underscan", meaning leaving black borders to compensate for TVs that cut off the picture. Interlaced/noninterlaced is irrelevant to that discussion.
Servo has captured NES on MobyGames, look at Super Mario Bros. Most TV capture cards don't know progressive NTSC from interlaced NTSC anyway; they'll just capture it interlaced always.

Great Hierophant wrote:

How slow is the gaussian filter? Is it slow enough to bring a powerful system to its knees?

Rendering one picture takes half a second on a low-end (PII) system. It'll take less on a high-end system of course, but still too long to get 60/60 fps.

Servo wrote:

For King's Quest PC Booter, I dug out a true IBM PC 5150

Unless you calibrated both your TV set and capture card with something like video essentials, your TV, capturing device and digital camera are likely to change the colors more than another CGA card can. You wrote that in real life it looks more like MobyGames than on that digital camera picture; well that ought to tell you something about the color fidelity of your camera. Now if you imagine that your TV probably is just as unfaithful about reproducing the proper colors, maybe even more so because of NTSC...

SirGraham wrote:

Well, it says so on spacequest.net and I also witnessed it myself - if I started the game without the -N switch, sooner or later it asked for a missing 992.V56 or 993.V56 file. It's surprising that you're not aware of this, since you know SQ4 so well.

Forget it; it went right over your head. 🙄

SirGraham wrote:

Games like Ulysses and Serenia don't even have animation, how slow can it be?

I'm assuming that DosBox redraws the entire screen every frame, not just the parts that change.

SirGraham wrote:

But how can DOSBox emulate the Tandy DAC without the ability to produce sampled sounds?

Huh?

wd wrote:

(effectively downgrading the sb16 to sbpro).

Isn't port 0xc0 just needed for DMA channel 4? The default 16-bit DMA is 5, and the SB16 can also set to use the 8-bit DMA channel for 16-bit I/O.

> Isn't port 0xc0 just needed for DMA channel 4? The default 16-bit DMA is
> 5, and the SB16 can also set to use the 8-bit DMA channel for 16-bit I/O.

Tandy uses the range from 0xc0-0xc7 (the last four attributed to the
tandy DAC). Not sure how it was when one of the special tandy
sound cards for a non-tandy computer (or was it a regular card?)
was put into a normal PC, they have additional ports at 0x1e0-...
and maybe "overruled" ports 0xc0,0xc1 and 0xc4-0xc7 (some games
don't use the ports at 0x1e0 on non-tandy PCs).

Unless you calibrated both your TV set and capture card with something like video essentials, your TV, capturing device and digital camera are likely to change the colors more than another CGA card can. You wrote that in real life it looks more like MobyGames than on that digital camera picture; well that ought to tell you something about the color fidelity of your camera. Now if you imagine that your TV probably is just as unfaithful about reproducing the proper colors, maybe even more so because of NTSC...

Yeah, I think the camera shots probably turned out pretty useless in the end. The camera auto adjusts depending on conditions, and the room lighting and whatnot also throw it off not to mention add glare to the screen, etc...

Mentioned earlier was the use of unused bits to alter the hue on some CGA cards; I tried to see if my two cards exhibited any noticable differences. Not terribly conclusive as there are many other CGA's out there, but it was late and I thought could be interesting to try out.

Great Hierophant wrote:
Very interesting considering that MobyGames' KQ1 PC Booter screenshots were taken using the real CGA.
Yes, but you don't know the hue setting on his TV card back then.

right, I couldn't even tell you that. Remember my original Ms. Pac-Man shots were off quite a bit (also capture on real CGA), though the KQ1 shots look to be pretty close: almost exactly right for the CGA clone, and a bit too dark with very minor hue difference for the IBM CGA. This is comparing to a set that has been calibrated with video essentials. It's really just coincidence this happened and is a result of whatever settings my capture card had at the time (keep in mind I could capture of the real CGA but have the results look like pretty much anything if the hue/brightness/contrast aren't set correctly on my capture card).

I've been using two tv's for comparisons in this operation; one a DLP set that has been calibrated with video essentials and should be reasonably faithful, the other a 13" Best Buy special which has the default settings it shipped with and is noticably off a bit but at least not wildly so. When comparing two cards on the same tv it shouldn't matter too much; both cards will be uniformly off so differences should still appear (if any exist).

Games like Ulysses and Serenia don't even have animation, how slow can it be?

Other games do have animation though...

Servo wrote:

Mentioned earlier was the use of unused bits to alter the hue on some CGA cards

I don't know how many "some" is, it just was that way on my cheap no-name far east CGA clone (I don't think a high-quality branded product would have given up on me just because I wrote a wrong value into a CRTC register...). Earlier in the thread I wrote that this was maybe intentional to make games look better, now I think it's just an artifact of that cheap product, because setting unused bits in the mode control register (0x3d8) messes up the color burst completely, and that's certainly not intentional.
It also was prone to produce slight image distortions whenever the hard drive (which is on an ISA controller card in that particular system) was accessed. I guess it's a problem of insufficient shielding or something. Oh well, now it's gone. I can't buy an original IBM CGA because I don't have an 8-bit slot, only a 16-bit ISA one, and of course you can't use that because of the IBM card's shape. The clone fit nicely into a 16-bit slot.

If you want to seriously check whether unused bits of port 3d9 have any effect... do you still have that program I sent you to do those color bar shots that you made for me? You wrote that several shots looked absolutely the same both on the IBM CGA and the Tandy. Well, they looked different on the clone, so this is what you want to look at.

I do still have the test program; I'll check that out. I'm now away for the Thanksgiving holiday for the next few days, though I do have internet access I don't have my machines handy so I'll confirm this when I get back...

Jfi: Tandy DAC seems to be working in the SQ3 intro.

Servo wrote:

I'll check that out

Thanks. Could you also make two screenshots of Microsoft Decathlon? One from the calibration screen, one from the title screen, preferably using the IBM 5150 rather than the Tandy, because the font really gets important here. I'm asking because I'm not sure how much luminance the CGA-generated colors should have; there are reasons to believe they should vary from color 1 thru 6, but the CGA patent (from 1984, should be expired by now) suggests they should be constant. This has a direct effect on how much distortion there is (see pictures). Since the RGB-to-Composite circuit has been patented by IBM, I could imagine that this too is an area where clones, attempting to avoid using patented designs, differ from the original card.

wd wrote:

> Well, it's not really a shame, it's just that other than this small problem > DOSBox provides an almost perfect Tandy emulatio […]
Show full quote

> Well, it's not really a shame, it's just that other than this small problem
> DOSBox provides an almost perfect Tandy emulation (at least for the
> games I'm playing).

Well then get it fixed and submit the patch. Just as easy as that.

If I were a programmer, I'd be happy to write patches for DOSBox. But since I'm not, all I can do is some beta-testing.

wd wrote:

Why should they auto-detect a tandy if the machine is set to !=tandy??

They shouldn't, and this is indeed the situation with 0.63, but in a CVS update from November 10 (by c2woody), it says: "src/: dosbox.cpp, hardware/tandy_sound.cpp, ints/bios.cpp: add support for tandy sound when machine is not set to tandy". According to this update, Tandy sound will work when machine!=tandy. That's why I asked about it in the first place.

wd wrote:

Unless you forced the tandy sound to be enabled in dosbox.conf you will not get any tandy sound for machines!=tandy. If it is forced for all machines, it depends on the game whether it provides support for tandy sound on non-tandys.

So there'll be a new option in DOSBOX.CONF to force Tandy? Cause I don't remember such an option in 0.63.

NewRisingSun wrote:

Forget it; it went right over your head. 🙄

I understand now that you were cynical, but cynicism is sometimes hard to detect on forum posts, especially when it's morning and you're tired and you're reading the post while alt-tabbing between it and work stuff. Should have thought about it though, knowing your experience with SQ4.

wd wrote:

Jfi: Tandy DAC seems to be working in the SQ3 intro.

Does that mean you heard Roger say "where am I"?

And on a different subject, here's a quote from Moby Games trivia for KQ3:
"If you had CGA and the wizard used a powerful spell, the entire screen shook. (This was an intentional special effect.) This was accomplished by tweaking the CGA registers to scroll the screen left and right rapidly."
Is it correct? I don't think this feature is emulated in DOSBox or MESS (though I didn't check thoroughly yet).

> If I were a programmer, I'd be happy to write patches for DOSBox. But
> since I'm not, all I can do is some beta-testing.

Then it's surely not a shame that nobody did yet find a solution for
some obscure aspect ratio issue, is it?

> Does that mean you heard Roger say "where am I"?

Try it?

For the Tandy Sound: there's a new option, it was already
present in earlier dosbox versions, but was removed in favour
of sb16 for vga machines (always enabled for tandy machines).
The default (tandy=auto) enables tandy sound (3v+DAC) only
when machine==tandy. The forced enable (tandy=on) will have
tandy sound for all machines, but only a few games really use it.