@Hypex

Quote:

Quote: Depends. Nowadays, one would just use the capabilities of graphic cards. Now days, yes, but I was thinking of back then.

Then, repeating in other words what I've already written, yes. It's graphics being in different memory areas and being combined only when the video signal is generated (i.e. without memory writes) that give free parallax.

Quote:

Anyway, the planar and chunky concepts are not mutually exclusive - again, think of byteplanes.Quote: It's only a boring matter of setting up the palette - all the rest just stays the same. Examples for a 4+4 dual playfield (for simplicity, let's ignore bank and LOCT switching): Okay so a bitplane colour trick it seems. But what about scrollng?

As already said, the rest just stays the same

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@Hypex

Not exactly. It worked in non HAM modes as well, the idea was just to "expand" the colour resolution from 4 bits per gun to 5 bits per gun. Using HAM was a pure experiment to see if one could emulate an RGB555 display with the trick. The work involved in eliminating the fringing effects was too much for it to be usable except for entirely still images.

The trick, in essence is best under stood by considering a single channel. So let's pick green. Your OCS/ECS palette only defined 4 bits for this. That gives you 16 levels, as you know.

Imagine you take your single pixel and split it into two halves (which is what the interlace is for). You can create shades of green between these 16 levels by setting the upper and lower halves to be one shade apart. So, on a scale of 0-15, if the upper pixel is 0 and the lower pixel is 1, you effectively have 0.5, which is basically like adding an extra bit of precision. It's error diffusion, really.

Doing this across all the intervals gives us 31, rather than 32 shades but it's close enough. The plan was to alternate the way the pixels were split so that the brighter "half" would follow a checkerboard pattern as this improved the quality somewhat.

The biggest problem was that the interlace alone would drive you insane without some sort of deinterlacer/Flickerfixer.

_________________
Doing stupid things for fun...

@Hyoex

Interestingly, I suppose the technique would have worked better if it were done using alternating fields at 50fps, it just never occurred to me to try it that way. That would be temporal rather than spatial averaging.

_________________
Doing stupid things for fun...

@saimo

Quote:

As already said, the rest just stays the same

Sure, but along with combination, two layers need scroll offsets as well. So would this be done usng the odd and even plane offsets are they are? Just taking dual mode out of the equation?

@Hypex

Quote:

Hypex wrote: @saimo Quote: As already said, the rest just stays the same Sure, but along with combination, two layers need scroll offsets as well. So would this be done usng the odd and even plane offsets are they are? Just taking dual mode out of the equation?

I'm not taking anything out of the equation: the BPLCON1 register (which contains the delays for both odd and even bitplanes) works the same way whether the Dual Playfield mode is active or not. Once again, the Dual Playfield only makes it easier to set up the palette (and, as a consequence, also makes some color registers free for use for sprites).

Last edited by saimo on 25-Jul-2018 at 05:31 PM.

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@Karlos

Ineresting. I think with the flicker, all you need is the right image, like a real image. Or one rendered so it works well as an interlaced image. Look at TV images, thet are all laced, but they manage it without flickerache in your eyes. Well CRT images I mean. Even when they had TV computer graphics, they still looked fine. Bit different to loading up an interlace mode on the Amiga and then, oh what the hell, what is this flick? We had MagicTV. But I think the Amigas problem with interlace was the OS didn't accommodate it very well. Or soften the edges. It needed anti-alasing. But that might have been too much work. Still, even a softner that halved the palette for one frame so the top edge wasn't in your face, wouldn't have gone astray.

Your idea reminds me of the audio modes I am thinking up. There is a virtual 9-bit mode combining two 8-bit samples, added together to form a range from -256 to 254 as it treats each strereo channel as one value. Symphonie does this.

I am thinking of an AM mode that uses volume along with sample amplitude to produce in betweens up to effective resolution of 14-bits. So if we multiply it up, volume by samples, 64x256=16384. Which gives us a step of 1/16384 or 0.0000610352. By comparison a 16-bit value of 65536 gives us steps of 0.0000152588. I thought 15-bit may be possible for some resson but the math doesn't go that far so don't know why.

I'm aware of the existing 14-bit routines but I don't see how setting volume to one on one channel helps. I suppose I'd have to look at the math in the routines involved. But I like the idea of using AM. Just seems pure with 8-bit samples multiplied by 6-bit volume for 14-bit precision.

@saimo

Forgot about this.

Quote:

Footnote: since the day, when I was a teenager, it almost blacked me out, I quite simply wish that interlacing was never invented (I had been pixelling a logo on an interlaced 640x512 screen for quite some time and then, all of a sudden, I almost fell off the chair and lost consciuosness). The only thing I would ever use it for is a smooth sub-pixel vertical scroller (where everything scrolls, otherwise the flickering would be visible).

This sounds like that condition they warn against with computer games and fast motion. Like epilepsy. In the recent past I was on my AmigaOne XE doing my usual nightly routine. Now by this stage I had a "normal" LCD and a good non-interlace screen mode. I was typing away when suddenly I felt this kind of light headness that happens when you are about pass out. I got up, fetched some water, then sat back down and settled myself down until it passed. Blood red can trigger it but don't recall seeing something like that kind of red on screen.

@Hypex

Quote:

I'm aware of the existing 14-bit routines but I don't see how setting volume to one on one channel helps. I suppose I'd have to look at the math in the routines involved.

It's very simple. In a 16 bit sample, the upper byte "weighs" 256 times more than the lower byte. In theory, one could play the upper bytes of a waveform on one channel at full volume and the lower bytes on another channel at 1/256th of the volume to obtain the same effect as playing the 16 bit samples directly (provided that the channels are properly mixed, which is something that Paula seems to do acceptably). Unfortunately, Paula only allows 1/64th of the full volume, hence the low byte "weight" has to be divided by 4, which is what causes the loss of the 2 least significant bits and gives a resolution of 14 bits.

(Fun coincidence: besides doing parallax scrolling without Dual Playfield, SkillGrid - the game I'm working on - also uses 14 bit playback for music.)

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@Hypex

Quote:

Quote: Footnote: since the day, when I was a teenager, it almost blacked me out, I quite simply wish that interlacing was never invented (I had been pixelling a logo on an interlaced 640x512 screen for quite some time and then, all of a sudden, I almost fell off the chair and lost consciuosness). The only thing I would ever use it for is a smooth sub-pixel vertical scroller (where everything scrolls, otherwise the flickering would be visible). This sounds like that condition they warn against with computer games and fast motion. Like epilepsy.

Exactly. And I do add that warning to my games (several of which are quite hectic and flashy). I don't suffer from epilepsy or, if I do, it's a very shallow problem. In fact, the mentioned episode is the only one I've ever had, but flashing patterns and lights definitely are not pleasant to me.

Quote:

In the recent past I was on my AmigaOne XE doing my usual nightly routine. Now by this stage I had a "normal" LCD and a good non-interlace screen mode. I was typing away when suddenly I felt this kind of light headness that happens when you are about pass out. I got up, fetched some water, then sat back down and settled myself down until it passed. Blood red can trigger it but don't recall seeing something like that kind of red on screen.

I'm no expert at all, so I might be saying nonsense here, but I wonder if it was just the refresh - we don't notice it, but it's there. However, if the screen was static and the only changing parts were the characters you were typing, I guess it must have been something else. Maybe the monitor wasn't exactly stable, i.e. subject to minimal flickering / brightness variations?

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@Hypex

The 14-bit replay is pretty well defined these days. The point is that Paula has a 6 bit volume per channel that is independent of the actual DAC resolution. I am not entirely sure if this 6-bit volume is applied after a strictly 8-bit DAC (i.e. a digital control over analogue volume) or if it's a fixed precision multiplier into a higher resolution DAC, but the effect is that at a volume of 1, you basically get a full, or nearly full 8-bit resolution. I believe that the 14 bit routines send the most significant 8 bits of a 16-bit sample to a channel at full volume. The least significant 8 bits are then right shifted 2 bits* and sent to the other channel at minimum volume. By simple summation the output manages to achieve 14 bit.

* it's probably a bit more involved as the sign of the overall 16-bit sample probably matters and assuming the summation is basically in the analogue domain, the non-linearity of the 8-bit DAC also is a factor. I distinctly remember going through the cybersound calibration tool, eliminating all the excess tone from every band with OCD like tenacity. Ironic since it's that very same bass-heavy, nonlinear, aliased sound that I ended up using my machines for when making music.

_________________
Doing stupid things for fun...

Read this if interested...

https://groups.google.com/d/msg/comp.sys.amiga.programmer/8lDYatiVBbA/P1UT25_1C7AJ

And what I understood from some other discussions Mimetics from 1986 was first one utilizing 14-bit audio... but not much information about it.

http://chiclassiccomp.org/docs/content/computing/Mimetics/Mimentics_SoundscapeProMIDIStudio_photocopy.pdf

_________________
Amiga Developer
Amiga 500, Efika, Mac Mini and PowerBook

@Karlos

Quote:

The least significant 8 bits are then right shifted 2 bits* * it's probably a bit more involved as the sign of the overall 16-bit sample probably matters

The low 6 bits values always have to be treated as positive. The correct 14 bit value is given by the signed 8 bits value multiplied by 64 plus the absolute value of the low 6 bits.
A few examples...

positive high 8 bits / positive low 6 bits:
00000001 000001 = 65 = 1*64 + 1

positive high 8 bits / negative low 6 bits:
00000001 100001 = 97 = 1*64 + 33

negative high 8 bits / positive low 6 bits:
11111111 000001 = -63 = -1*64 + 1

negative high 8 bits / negative low 6 bits
11111111 111111 = -1 = -1*64 + 63

If I remember correctly, the first time I implemented 14 bit playback, I had first assumed that I'd need asr to shift the low bits, but that resulted in a huge noise. But even if I don't remember correctly (it was 22+ years ago, after all) last year, when I returned to 14 bit playback, I made the same mistake and got the same result

Last edited by saimo on 26-Jul-2018 at 12:11 PM.

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@saimo

Quote:

It's very simple. In a 16 bit sample, the upper byte "weighs" 256 times more than the lower byte. In theory, one could play the upper bytes of a waveform on one channel at full volume and the lower bytes on another channel at 1/256th of the volume to obtain the same effect as playing the 16 bit samples directly (provided that the channels are properly mixed, which is something that Paula seems to do acceptably).

Okay I get what you are saying regarding the weight. I wonder if anyone tried the theory you are proposing. With the full 8-bits.

Quote:

Unfortunately, Paula only allows 1/64th of the full volume, hence the low byte "weight" has to be divided by 4, which is what causes the loss of the 2 least significant bits and gives a resolution of 14 bits.

I wonder if this is divided before or if the lower 8-bits are simply plonked into place with the volume at one? An issue I see is one of sign. As the MSB and LSB would usually be different they likely would be a different sign. So I too wonder if that would be corrected with the LSB scaled to 6-bits with the MSB sign added.

Quote:

(Fun coincidence: besides doing parallax scrolling without Dual Playfield, SkillGrid - the game I'm working on - also uses 14 bit playback for music.)

Yes I saw that mentioned. Do you use AHI or custom routines?

Quote:

I'm no expert at all, so I might be saying nonsense here, but I wonder if it was just the refresh - we don't notice it, but it's there. However, if the screen was static and the only changing parts were the characters you were typing, I guess it must have been something else. Maybe the monitor wasn't exactly stable, i.e. subject to minimal flickering / brightness variations?

From what I recall there wasn't much going on screen but I don't recall seeing flickering of any sort. Even subtle. I could have just gone weird. I tend to have a sensitive stomach. Though that is usually met with cramps. I don't know. I have these wierd problems. I have got some vertigo in the past as well. I ended up in the hospital one time after passing out when I woke up with a load of fliud and busting to get to the loo. Unfortunately I was gone before I hit the ground but at least I missed the bowl! They did basic tests but it was inconclusive.

@Karlos

Quote:

The 14-bit replay is pretty well defined these days. The point is that Paula has a 6 bit volume per channel that is independent of the actual DAC resolution.

Yes I understood that much, the rest just looked a bit strange. Or at least the way they went about it. Which wouldn't be too much work.

Quote:

I distinctly remember going through the cybersound calibration tool, eliminating all the excess tone from every band with OCD like tenacity.

I went through it as well, but I wondered why it just didn't do a standard 14-bit conversion. That is convert it in a generic sense. By standard we didn't calibrate normal Paula output.

@Hypex

Quote:

Quote: Unfortunately, Paula only allows 1/64th of the full volume, hence the low byte "weight" has to be divided by 4, which is what causes the loss of the 2 least significant bits and gives a resolution of 14 bits. I wonder if this is divided before or if the lower 8-bits are simply plonked into place with the volume at one?

Nope, one has to feed Paula with 6 bits data (that's the whole point).

Quote:

An issue I see is one of sign. As the MSB and LSB would usually be different they likely would be a different sign. So I too wonder if that would be corrected with the LSB scaled to 6-bits with the MSB sign added.

Luckily sign is not an issue - see my previous post.

Quote:

Quote: (Fun coincidence: besides doing parallax scrolling without Dual Playfield, SkillGrid - the game I'm working on - also uses 14 bit playback for music.) Yes I saw that mentioned. Do you use AHI or custom routines?

The game totally disables the OS and owns the hardware. It's all custom assembly code.

Quote:

From what I recall there wasn't much going on screen but I don't recall seeing flickering of any sort. Even subtle. I could have just gone weird. I tend to have a sensitive stomach. Though that is usually met with cramps. I don't know. I have these wierd problems. I have got some vertigo in the past as well. I ended up in the hospital one time after passing out when I woke up with a load of fliud and busting to get to the loo. Unfortunately I was gone before I hit the ground but at least I missed the bowl! They did basic tests but it was inconclusive.

Well, who knows, maybe - just maybe - some imperceptible screen-related factor coupled with some physical condition/trigger of yours. Or maybe not... impossible to say.

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@itix

Quote:

Read this if interested...

That was a big one.

I see this was described. So what's interesting here is that my same idea is proposed. To use AM mode and modulate 8-bits of sample data by 6-bits of volume. It just didn't point out what is isolated as the MSB and LSB when doing it as AM. I also see that my AM idea is the second obvious thing listed. However, in my mind, it is the first obvious thing to do.

Quote:

And what I understood from some other discussions Mimetics from 1986 was first one utilizing 14-bit audio... but not much information about it.

No just a smidge on page 157. Suggesting by the context that a 14-bit sample would have the 8-bit MSB modulated the the 6-bit LSB.

I wonder with all the methods if any one measured output? By any means. Sampled it into a 16-bit sampler or measured the levels with some sound oscilloscope. Using common sine, saw, tri and square waveforms. Then comparing with 16-bit variants from CD possibly. I've made this go way off topic.

@saimo

Quote:

saimo wrote: @Karlos Quote: The least significant 8 bits are then right shifted 2 bits* * it's probably a bit more involved as the sign of the overall 16-bit sample probably matters The low 6 bits values always have to be treated as positive. The correct 14 bit value is given by the signed 8 bits value multiplied by 64 plus the absolute value of the low 6 bits. A few examples... positive high 8 bits / positive low 6 bits: 00000001 000001 = 65 = 1*64 + 1 positive high 8 bits / negative low 6 bits: 00000001 100001 = 97 = 1*64 + 33 negative high 8 bits / positive low 6 bits: 11111111 000001 = -63 = -1*64 + 1 negative high 8 bits / negative low 6 bits 11111111 111111 = -1 = -1*64 + 63 If I remember correctly, the first time I implemented 14 bit playback, I had first assumed that I'd need asr to shift the low bits, but that resulted in a huge noise. But even if I don't remember correctly (it was 22+ years ago, after all) last year, when I returned to 14 bit playback, I made the same mistake and got the same result

Makes sense. I've always assumed the reconstruction of the 14-bit output happens as voltage summing the analogue domain after each component goes via independent 8-bit DAC outputs which is why calibration is necessary to get the best sound. This may not be correct, perhaps you know?

Assuming it is that, or something pretty similar, I've often wondered if it would be better to use a companded format in which a 16-bit sample is tramsformed into a format that allows one Paula channel to be used as a volume modulation for another. I am not sure how much latency there is in such an approach, although even if volume changes take tens of times longer than instantaneous sample value changes, one could still use the technique as a form of dynamic gain control.

This seems to be more or less the same thing Hypex is suggesting.

Last edited by Karlos on 26-Jul-2018 at 08:26 PM.

_________________
Doing stupid things for fun...

@saimo

Quote:

saimo wrote: @cdimauro Quote: and yes: plane 4 (fifth) was always set (it was the mask). Ah, right. This just made me realized I wrote some nonsense in my previous post (at some point I totally ignored the fact that one needs to add 16 to 32 to obtain 48 ). Fixed. BTW, did you use the blits on plane 4 for Blitter-based collision detection? Given the plane-by-plane blits, that came for free. Edit: on secound thought, that would halve the colors available to backgrounds, so I guess that the answer is no.

Indeed.

It was possible to use it for catching the collisions between the two characters, but it was anyway too much expensive due to the too high amount of "tiles" which composed the bodies.
Quote:

Quote: However the characters had the possibility to change some colors from some basic palettes, in order to allow 2 players to use the character, but with some different colors to distinguish them. So, you cannot directly blit the BOBs as they are. Hmmm... Do you perhaps mean that one of the plane 3 was used as an offset to address two separate palette ranges (16-23 plus 48-55 / 24-31 plus 56-63)?

I don't remember now if it was plane 3, or some combinations of plane 3 and 4. Anyway, this part was handled by the main coder (I was the additional coder. I've contributed to Fightin' Spirit graphic routines only because I had some ideas when I was working at USA Racing, where I was the main code. The game was never released).
Quote:

Quote: Unfortunately sometimes the frame rate dropped to 17fps when there was too much graphic to be drawn. Ah, I haven't seen it much in action (I've mostly seen screenshots), so I had not noticed that. Yeah, with so many planes, big bobs, and animations, that's understandable.

Exactly. However it didn't happened often to slow down to 17FPS, fortunately.

@Karlos

Quote:

Makes sense. I've always assumed the reconstruction of the 14-bit output happens as voltage summing the analogue domain after each component goes via independent 8-bit DAC outputs which is why calibration is necessary to get the best sound. This may not be correct, perhaps you know?

Nope, I don't know, but I've also assumed the same. On the other hand, the result is so good (at least, it is said to be so, but I don't know if anyone ever made measurements), that it seems more likely that Paula mixes everything digitally and then sends the higher precision results through the DACs (but maybe it is already known that the DACs are 8 bit, so there goes my theory... I'm pretty ignorant about the subject).
Maybe there's something about this in the article indicated by itix (as soon as I saw the wall of text, I closed the page ).
To be honest, I never gave it much thought: once I read that 14 bit output was possible on the Amiga, so I went home, sat at the table, and came up with my way of doing it (and then I always assumed that it was exactly the same way everybody was doing it, not seeing other possibilities) - it worked, and that was enough for me.

Quote:

Assuming it is that, or something pretty similar, I've often wondered if it would be better to use a companded format in which a 16-bit sample is tramsformed into a format that allows one Paula channel to be used as a volume modulation for another. I am not sure how much latency there is in such an approach, although even if volume changes take tens of times longer than instantaneous sample value changes, one could still use the technique as a form of dynamic gain control.

I can't say, I've never played with modulation. But I've been wondering if another possibility briefly mentioned in the Amiga Hardware Reference Manual has been fully explored: setting the period to a value below 123 so that Paula just repeats the value in AUDxDAT continuously, and the data in RAM is used for modulation... who knows, maybe there's some quirk that can make this useful.

(A dream anyway would be delaying the output of a channel by half period, so that it is possible to play on a pair of channels 8 bit samples at about 58 kHz without doublescan.)

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@saimo

Quote:

Nope, one has to feed Paula with 6 bits data (that's the whole point).

That's a shame since there is 8-bits of space. So it creates noise instead?

Quote:

Luckily sign is not an issue - see my previous post.

Okay it gets chopped out of the equation.

Quote:

The game totally disables the OS and owns the hardware. It's all custom assembly code.

No RTG support then?

Quote:

Well, who knows, maybe - just maybe - some imperceptible screen-related factor coupled with some physical condition/trigger of yours. Or maybe not... impossible to say.

It is hard to say. Come to think of it I recall it being matched to a ringing in my ears. I started to refer to this as a ringing attack. I still gets the ringing but just the noise. I've also has tinnitus the last few years so likely related.