PC still playing Amiga catchup

[amigaksi]

The 1 second periodicity was only an example. You could trigger it every 10 seconds. What matters is whether the game responds to the 1ms wide pulse whenever it is issued. If it doesn't then the game simply isn't sampling at a rate that guarantees a sample is read within a 1ms window - ie it isn't sampling at 1kHz. By gradually increasing your pulse width to the point where you get a consistent response (eg 9 in 10 pulses trigger the action) then you know you are in the right approximate resolution of the application's sample rate.
...

I do have some source code for games that just poll the joystick without waiting for VBI.  In the case of River-raid, it does not let you fire again until the missile hits its target or is off the screen.

>As for the rest, they could but that still does not demonstrate that any particular game title samples this quickly. You need to actually test with real applications that were written, not software you write purely to demonstrate that it is possible.

They are both useful.  See if they exist in existing applications and to show that it's more accurate to sample at higher rates.

>PS, see my other crazy idea to allow 1kHz sampling an analogue joystick. In case you can't find it, a quick recap:

>Use the X/Y potentiometers to adjust the left/right volume of a simple hardware oscillator running at several kHz that is sampled by the soundcard at 44.1 kHz. Calculate the RMS power for frames of 441 samples (one frame for left and one frame for right) and look up the RMS value in a suitable calibration table to determine X and Y position.

Computing an on-going RMS may slow things down though.  How about, just looking for a max/min value every millisecond or so and using up 4 channels to get a digital joystick out of it.

[amigaksi]

You can claim it all you like, but that doesn't make it true. It's approximately 558ns. An accurate figure, determined by a high precision scope might record 555, 559. I totally agree with you that it doesn't matter in system since the circuitry will all be running at whatever speed the clock gives, but the claim it is accurate to 558ns is not true at all. You have to quote a margin of error for any real measurement.

...

There's some ppm rating on the crystals but in the general case it's more accurate than even the rating.  I know the NTSC/PAL crystals are pretty solid when it comes to timing since they have an exact formula to work with: 227.5*262.5*60/1.001 = NTSC color burst signal (3579545.4545Hz) and timing based on that should give 558.73ns for Copper.  Are you claiming that TV's are skipping frames once in a while?

>For example, some years ago, Redrumloa was experimenting with underclocking the main oscillator on the motherboard as he observied it was giving a speed up as based on common amiga benchmarking programs where the native chipset wasn't being used for video etc.

There are cases where there's manufacturer defect.  They do use the T=1/f even on PCs when they use the timer to update system time every 55ms or so.

[Karlos]

>PS, see my other crazy idea to allow 1kHz sampling an analogue joystick. In case you can't find it, a quick recap:

>Use the X/Y potentiometers to adjust the left/right volume of a simple hardware oscillator running at several kHz that is sampled by the soundcard at 44.1 kHz. Calculate the RMS power for frames of 441 samples (one frame for left and one frame for right) and look up the RMS value in a suitable calibration table to determine X and Y position.

Computing an on-going RMS may slow things down though.  How about, just looking for a max/min value every millisecond or so and using up 4 channels to get a digital joystick out of it.

For a single pot value, doing 441 multiplications, 440 additions (most of which could be optimised using multiply add style instructions) and a square root might take at most a few hundred cycles. You can even get rid of the square root all together and just have the values in your calibration table squared and look up the nearest match. However, even with the square root, on a GHz class machine that's peanuts, you are looking at microseconds. Given that searching through the 441 samples will require as many memory accesses, which are by far the slowest operation in the task for a modern CPU it probably wouldn't be faster.

[bloodline]

For a single pot value, doing 441 multiplications, 440 additions (most of which could be optimised using multiply add style instructions) and a square root might take at most a few hundred cycles. You can even get rid of the square root all together and just have the values in your calibration table squared and look up the nearest match. However, even with the square root, on a GHz class machine that's peanuts, you are looking at microseconds. Given that searching through the 441 samples will require as many memory accesses, which are by far the slowest operation in the task for a modern CPU it probably wouldn't be faster.

I think encoding the x/y signals in a single channel, while efficient and the "right way" to do it, is overly complex for this application... just have two square wave oscillators, and alter the amplitude depending upon position... the Audio interface I referred to has 10 channels, but even the simplest audio input has two channels... x on left, y on right.. simples

[Karlos]

I wasn't encoding them into a single channel, I was suggesting the attenuation of the output of one mono oscillator separately into left and right (the volume levels for left and right representing representing X and Y respectively) and then analysing the RMS values of each 441 sample frame for left and right independently to infer the X/Y pot values.

The reason for RMS is to calculate a representative value since you'll never get an exact number of oscillator cycles (regardless of waveform shape) in a single frame.

If you are going to go to the trouble of 1kHz sampling of your analogue joystick, there's no point in doing it half arsed, eh?

[bloodline]

I wasn't encoding them into a single channel, I was suggesting the attenuation of the output of one mono oscillator separately into left and right (the volume levels for left and right representing representing X and Y respectively) and then analysing the RMS values of each 441 sample frame for left and right independently to infer the X/Y pot values.

The reason for RMS is to calculate a representative value since you'll never get an exact number of oscillator cycles (regardless of waveform shape) in a single frame.

If you are going to go to the trouble of 1kHz sampling of your analogue joystick, there's no point in doing it half arsed, eh?

Well... I can't really see the point of a joystick sampling at 1Khz... so it wouldn't make much difference, in reality...

[Karlos]

Well... I can't really see the point of a joystick sampling at 1Khz... so it wouldn't make much difference, in reality...

What? Are you insane? :p

[bloodline]

What? Are you insane? :p

:lol:

[cecilia]

[Karlos]

That was a fine looking machine, to be fair

Pity Commodore never thought to market the Amiga comparatively. Idiots.

[koaftder]

Amiga; "cats up"

[smerf]

Hi,

@Karlos & Amigaski,

I am probably an idiot, but everything I found on joystick polling on the Amiga says that it operates at 1mhz. I looked at schematics, at building boards, etc. etc.etc. even checked my CD32 Schematics, I see 1mhz. Please correct me if I am wrong, also while looking at the schematics I see a diode and cap in circuit with joystick plug, could this possibly be for joystick contact bounce. I could be entirely wrong, but just wanted to see what you would say.

Sitting on a joystick so much fun, just move the sprite by wiggleing your bun

smerf

[persia]

The world is moving towards more distance between the programs and hardware.  You have several cores each running at 3 GHz, a bus running in the GHz range, USB 2 running in the 100 MHz range, the hardware has evolved to a point it responds faster than a human can react with it.

The speed of the equipment has also eliminated the need for ports that do only one thing, nowadays your USB port can do everything, so you can plug anything from a monitor or printer to a joystick or mouse into them, no more worrying about not having the right port or the right number of ports.  I would never dream of going back to the old days.

Living in 2009 and loving it!

[DamageX]

Because PCs were originally business machines

Originally, they weren't. The first IBM PC shipped with 16KB RAM, a cassette tape interface, and BASIC in ROM. They accidentally failed in the home computer market, and succeeded in the business microcomputer market. They then tried again for the home market, with the PC jr. but failed. The PC jr. didn't really sell until they introduced a proper keyboard and an upgrade module that enabled it to be used for the same stuff that PCs were being used for.

[amigaksi]

Originally, they weren't. The first IBM PC shipped with 16KB RAM, a cassette tape interface, and BASIC in ROM. They accidentally failed in the home computer market, and succeeded in the business microcomputer market. They then tried again for the home market, with the PC jr. but failed. The PC jr. didn't really sell until they introduced a proper keyboard and an upgrade module that enabled it to be used for the same stuff that PCs were being used for.

Yeah, PC jr failed but IBM PC didn't have the colors, sprites, blitter, sound DACs, etc. and it's price was way too expensive for gaming or for home use.

[warpdesign]

Yeah, PC jr failed but IBM PC didn't have the colors, sprites, blitter, sound DACs, etc. and it's price was way too expensive for gaming or for home use.

Then they tried again with the PC/AT and blew everything away. Welcome to 2009