Atari ST music vs Amiga music.

[oky2000]

I write Amiga music and sample size is only limited to chip ram (between 512kb and 2mb). I wish people would get their facts right. I can load a 512kb instrument even on my A1000.

 

CPU has nothing to do with sample playback, it can only be used to give virtual 8 channel sound etc..

 

Also 2 channel 14bit sound playback is possible. Like I said blank sheet of paper.

[atarian63]

...and we still like ST sound and we don't complain.

http://ym-digital.i-demo.pl/

Those are excellent especially the STe demo

[Lynxpro]

I'd have to agree with the consensus...and having owned both machines,

I'd say that the Amiga's sound was generally better.

 

Having said that though, I still like a lot of the ST's sounds. The

pipes from SOTB were mentioned. I've played the game on both, and

there's no doubt the Amiga version is better, but...I *like* the ST

versions music in the intro. Also, I love the pounding soundtrack to

StarDust on my Mega STe (of course that's an STe, but SD was brought

up). Several arcade games with the "chiptune" sound were fine by me,

I like the music to "Venus, the Flytrap", for example. Ultima 3's

dungeon music is unforgettable. I love the bard's music from the

Bards Tale, etc, etc,...

 

Keep in mind - a trained musician, I'm not, just "IMHO" and "YMMV".

 

 

 

It's too bad Atari Corp. couldn't get the AMY sound chip working and included in the ST. It [probably] would've cleaned the Amiga's clock in terms of musical abilities.

 

Although several years later, the Motorola DSP in the Falcon ruled. I read somewhere that it was "now" being used in the iPod Shuffle from a few years back...

[atarian63]

If strictly comparing music, I'd take the Amiga over the ST any day (as with the C-64 over the Atari 8-bits).

 

I'm sure the ST is capable of some decent music, but some games are just an abomination. The music in

is terrible compared to the Amiga version. Even if you compare it to the console ports, it's actually closer to the Master System version than it is to the Genesis/Mega Drive.

Amiga sound in games is better later on but early on it was much closer. as for A8 vs c64, I take the A8 any day next to the buzzy bee generator

[Lynxpro]

Have you ever listened to the Softsynth tunes on the A8? The waveforms were calculated via CPU and played writing to all 4 "DAC"s. due to the real 4 channels, no interferences happen there.

 

 

The source of such tune is slightly higher than "normal" POKEY tunes. But the waves are "free" . No need for Megabytes of DMA. Just the CPU could do faster there.

 

 

 

Wow. Now just for the sake of argument, what do you think that artist could do with the same tune using Dual Pokeys?

[kool kitty89]

I write Amiga music and sample size is only limited to chip ram (between 512kb and 2mb). I wish people would get their facts right. I can load a 512kb instrument even on my A1000.

 

CPU has nothing to do with sample playback, it can only be used to give virtual 8 channel sound etc..

 

Also 2 channel 14bit sound playback is possible. Like I said blank sheet of paper.

Umm, no, Paula has to loop the samples and set the samples for PAULA to read, it's not a hardware sample synth chip, it's a hardware DMA PCM decoder chip with variable sample rates. All it can do is stream and modify playback rate AFIK. The CPU has to set all the loops, etc.

 

Now, for normal, long (ie close to/over 1 second or a fairly large fraction of a second) samples, that CPU resource is negligible, but for the tiny, tiny samples mentioned above, we're talking thousands of loops per second. (ie what the PCE/TG-16 does in hardware, or full sample synth hardware -like Ricoh's 8 channel PCM synth chip or Sony's SPC700 unit in the SNES- though the latter can also use samples of any length rather than only 32 byte/word waveforms)

 

Longer samples use LESS CPU resource, but tiny samples (in the context of true chip synth emulation) will take significant CPU resource, ever closer to software PWM.

 

 

 

 

 

 

 

 

 

 

Paula is a different animal entirely really, and it would have been extremely limited in an older machine with limited RAM. (granted there's some pretty impressive 8k MOD out there ) That's one of the reasons you didn't see more samples on 8-bits, not just the CPU resource, but the already constrained memory. (and for demos in titles and such, CPU resource is a non-issue, hence why you saw MODs in a fair amount of ST demo screens but not MOD/sample stuff on 8-bits -other than the occasional low quality fixed pitch percussion/chime/etc sample -like in stormlord on the C64)

 

 

 

Have you ever listened to the Softsynth tunes on the A8? The waveforms were calculated via CPU and played writing to all 4 "DAC"s. due to the real 4 channels, no interferences happen there.

So what, that doesn't change anything I said before.

 

4-bit linear volume sounds a bit better than the logarithmic channels on the AY/YM chips, but both are limited by the low resolution and both take a ton of CPU resource.

And the higher the sample rate, the worse it sounds compared to 8-bit (or higher res) playback. If you compared 16 kHz 4 channel MOD on POKEY vs 16 kHz 4 channel software mixed 3-channel to 8-bit (or higher) look-up hacked MOD on the ST, the latter will sound significantly better. (as it would compared to AY hardware channels as well)

You get the added noise from 3 separate updates (intermediate values), but that's the same sacrifice you'd need to make for 6-bit PCM on POKEY. (except it would sound worse due to the lower res and 4 vs 2 or 3 updates on the AY -2 channel hacking is coarser than 3 but can still manage OK 8-bit and a good improvement over single hardware channels, at least at higher samples rates -at 4 kHz it probably isn't worth the effort or the added space for 8-bit vs 4-bit samples)

 

The source of such tune is slightly higher than "normal" POKEY tunes. But the waves are "free" . No need for Megabytes of DMA. Just the CPU could do faster there.

And no, those are very resource intensive, at least with the 68k. The 6502 is very good at doing fast/simple tasks (the PC Engine would be better at software MOD than the ST by a good margin because of that -DAC methods aside), and that's where it shines compared to the 68k or some others. (for tightly coded demos it would be more competitive though)

On top of that, you're talking demos, not easily implemented realtime in-game stuff: with the 6502, interrupts are a realistic option as 6502 interrupts are very light and also takes care of timing in hardware (for easy multitasking), but the 68000 has very heavy interrupt overhead by comparison. (hence why you don't often see software PWM stuff on the ST vs what you do with POKEY -at least in homebrew stuff)

Interrupt driven modulation on the A8 is very practical due to the CPU architecture, on the ST it eats tons of CPU time that's already pressed tight for graphics. (and low sample rate PCM is more practical in-game even, not MOD stuff, but fixed rate ~4 kHz stuff; you could do wave modulations at that same frequency -ie limited pitch- or a little higher with similar CPU respurce as PCM)

 

The use of the 6502 arch CPU on the Lynx also made it very useful for software modulation and PCM playback with reasonably little overhead.

 

 

Again, you CAN do the same thing with modulation on the ST's AY (albeit 3 channels), but it's intensive and no different than doing PWM on a beeper other than having hardware volume control and you have 3 hardware channels plus noise and intermittent use of modulation and hardware synth. http://battleofthebits.org/arena/Entry/down+and+dirty/3657/

 

There's nothing special about POKEY that makes it better for software modulation other than the built-in timers. (and only if the system doesn't have other interval timers to use -hell, in the A8, if you had at least 1 interval timer separate from POKEY you could use the GTIA channel as a 5tn hardware channel for PWM -but not other modulations like triangle or saw since there's no volume control, but pulse is the most significant in sound and simplest to do -and is in the ST example above and many SID tunes -SID does PWM/variable width pulse wave in hardware and POKEY sort of does but only for periodic noise)

It's also worse than Paula for software tiangle/saw due to the limitations of 4-bit output vs 8-bit. (the logarithmic volume of the AY would also be limiting in that respect) PWM (pulse/square wave) has no use for such resolution beyond normal volume control and will sound just as good on a 1-bit beeper toggle as a 16-bit DAC.

That's also why the Lynx is good for CPU modulation: a bunch of interval timers and 8-bit DAC mode (or at least 8-bit linear volume) on the 4 PSG channels.

 

 

POKEY would, for all intents and purposes, be just as limited for hacking as the AY with some trade-offs. (single channel PCM would generally sound better on POKEY, but the difference for similarly optimized samples won't be extreme -and both will get noticeably staticy at higher sample rates -though have less low sample rate artifacts) For chip synth, the AY has higher frequency resolution (12 vs , hardware envelope (also hacked for low-res saw or triangle -usually saw, without CPU modulation), and stereo (though that's moot), but it lacks the fully general purpose 4 hardware channels (just 3 tone channels plus noise) and lacks the periodic pulse wave noise of POKEY (and TIA). Both are much better than the SN76489. (which itself is better than TIA for music, but generally worse for SFX )

 

 

 

There's almost no use of 3 channel sample playback systems on the ST using hardware channels, so it's really hard to compare, and the one openly available tracker is weaker than the ones for the Spectrum and CPC.

 

Compare the best Spectrum sample stuff (3 channel hardware via AY) to 4 channel POKEY MOD stuff at similar sample rates and you'll see only a moderate improvement over the AY. (it's also up to proper formatting of the samples and good preprocessing)

 

 

 

 

 

POKEY is also incapable of simulating 8-bit PCM (or higher) like the AY can, and that's important for higher sample rate stuff, you could do 6-bit max, and that's combing all 4 channels with sequential writes: thus you'd get the added noise the 3-channel AY hack also suffers from

Edited February 6, 2011 by kool kitty89

[emkay]

So what, that doesn't change anything I said before.

 

4-bit linear volume sounds a bit better than the logarithmic channels on the AY/YM chips, but both are limited by the low resolution and both take a ton of CPU resource.

And the higher the sample rate, the worse it sounds compared to 8-bit (or higher res) playback. If you compared 16 kHz 4 channel MOD on POKEY vs 16 kHz 4 channel software mixed 3-channel to 8-bit (or higher) look-up hacked MOD on the ST, the latter will sound significantly better. (as it would compared to AY hardware channels as well)

You get the added noise from 3 separate updates (intermediate values), but that's the same sacrifice you'd need to make for 6-bit PCM on POKEY. (except it would sound worse due to the lower res and 4 vs 2 or 3 updates on the AY -2 channel hacking is coarser than 3 but can still manage OK 8-bit and a good improvement over single hardware channels, at least at higher samples rates -at 4 kHz it probably isn't worth the effort or the added space for 8-bit vs 4-bit samples)

 

With POKEY there is not this or that. You can do combinations of all features... from time to time...

 

A standard setting for the ST (for example] is to have 2 16 Bit channels and two software Channels for Digitals...

 

Another feature is the 1.79MHz channel on the A8. The fast pulses of the filtered channel results in "new" waves at lower frequencies. With this features you can build waves as done with AdLib. The resulting waves get an analog style .

Actually, this is superior to PAULA but only in one channel.

Edited February 6, 2011 by emkay

[Rybags]

It's a no-contest really - the Amiga could easily enough do everything the YM could do... not that you necessarily want to emulate that chip anyway. About all that's missing is a psuedo noise generator, but when you can just play back the sound of a real bullet or explosion, who cares?

 

Small samples, I don't see the drama. Rather than bothering with thousands of tiny loops, all you need is to repeat the sample data to get the interrupts happening less often.

Doesn't it have a one-shot and repeat mode anyway?

 

Amy would have made a huge difference, but it just creates complex waveforms by joining together simple ones. Sure, it would have levelled the playing field a bit, but I doubt it would have won the day.

It wasn't able to play arbitrary sample data either, doing so would have involved massive CPU intervention.

[frank1974]

I write Amiga music and sample size is only limited to chip ram (between 512kb and 2mb). I wish people would get their facts right. I can load a 512kb instrument even on my A1000.

 

CPU has nothing to do with sample playback, it can only be used to give virtual 8 channel sound etc..

 

Also 2 channel 14bit sound playback is possible. Like I said blank sheet of paper.

 

Not in a single DMA driven operation. AUDxLEN is limited to 65535 words (128k max per sample) before you need to reload them.

[emkay]

I write Amiga music and sample size is only limited to chip ram (between 512kb and 2mb). I wish people would get their facts right. I can load a 512kb instrument even on my A1000.

 

CPU has nothing to do with sample playback, it can only be used to give virtual 8 channel sound etc..

 

Also 2 channel 14bit sound playback is possible. Like I said blank sheet of paper.

 

 

You can change the sample content for variations via CPU while PAULA is playing them ....

I'm quite sure, it was done in the past already with "Jam Cracker" the tunes run on OCS AMIGAs an sound extremely synthlike.

[analmux]

[...]

POKEY also can't do the high-res DAC hack that the AY can via look-up tables: using all 4 channels at 5-bit resolution and optimized look-up, you can reasonably approximate 8-bit resolution output or a bit better than that even before the granularity gets too nasty. (ie to the point where mixing to 8-bit and dropping sample res is preferable to mixing to higher res -I think 10 bit might be reasonably possible)[...]

No problem for Pokey either.

 

Only using channel 1 & 3 (at 1.79Mhz) we can generate a sawtooth wave. Using the lowest volume (choosing 1 out of 0-15), varying pitch & managing the timing of the start of the waveform, and also choosing upwards or downwards sawtooth, we have a method for finetuning the 4/5 bit PCM. Then it's similar to an effective 8 or 9 bit sampling quality.

 

In this case, channel 3 can be used as Volume-only-mode channel, for coarse (4 bit) PCM / sampling. Channel 1 can be used to do finetuning / smoothening.

 

Then we still have channel 2 and 4 for doing other things, f.e. a standard synth voice, using simple 50Hz pokey-programming.

Edited February 6, 2011 by analmux

[kool kitty89]

With POKEY there is not this or that. You can do combinations of all features... from time to time...

 

A standard setting for the ST (for example] is to have 2 16 Bit channels and two software Channels for Digitals...

Digitals and software modulation is NOT going to happen in-game aside from very sparing and/or low sample rate stuff. (pulse wave basslines would work, or mid/low pitch notes, or low rate FX/drum/etc via interrupts)

Same for AY though, and you only need 1 channel for digital stuff if you're doing fixed pitch playback: interleave/multiplex mixing would have basically no overhead over using multiple hardware channels. (multiple channels is what you'd want for notes without scaling)

And can you really do volume modulation on 2 POKEY channels with 2 16-bit mode paired/slaved channels? (you couldn't use the timer interrupt method, but the ST has added interval timers to push that)

 

POKEY does have flexibility over the AY, but it's all trade-offs: tons of cases where one does better than the other. by default the AY has 3 12-bit square wave channels with the ability to apply a (limited) envelope along with noise generation (mixed into any of the 3 channels -a feature that's mainly useful if stereo is used). POKEY has 4 8-bit res square wave channels (aside from additional octave control) and noise generation (including periodic pulse) on any of those channels. (both have additional hacks via the envelope or such, and more via software stuff, but high frequency interrupts are not a realistic option on the ST as they are on the A8)

 

Another feature is the 1.79MHz channel on the A8. The fast pulses of the filtered channel results in "new" waves at lower frequencies. With this features you can build waves as done with AdLib. The resulting waves get an analog style .

Actually, this is superior to PAULA but only in one channel.

Not like Adlib, more like AMY or various additive synthesizers. Adlib uses FM OR additive synth: ie 2 waveforms added together (of a limited selection of waves derived from sine -most Yamaha synths used true sine only; the OPL2 and OPL3 are the exceptions), or with frequency modulation using 1 wave to modulate the other. (the latter being FAR more powerful and how a true square wave -or very close to it- could be generated from 2 sine oscillators where 4 would be needed via additive synth) Of course, FM is FAR more powerful with more operators and a variety of algorithms. (like the 4-op YM2151 of the arcade, Genesis's YM2612, or the 6-op DX-7 synthesizer -among many others)

Plus additional ADSR on top of that. (and additive synthesis of the resulting per-channel waveforms by pairing channels in software -often done on various PSGs as well: harmonizing/additive synthesis)

 

 

 

 

 

 

 

Only using channel 1 & 3 (at 1.79Mhz) we can generate a sawtooth wave. Using the lowest volume (choosing 1 out of 0-15), varying pitch & managing the timing of the start of the waveform, and also choosing upwards or downwards sawtooth, we have a method for finetuning the 4/5 bit PCM. Then it's similar to an effective 8 or 9 bit sampling quality.

And that's not using any interrupts, right? (vs software modulation -very technically possible on the ST, but impractical for most games)

Edited February 7, 2011 by kool kitty89

[emkay]

And can you really do volume modulation on 2 POKEY channels with 2 16-bit mode paired/slaved channels? (you couldn't use the timer interrupt method, but the ST has added interval timers to push that)

 

Why should this be a problem? When joining 16 bit, both channels play sounds depending on the other voices setting. Then you could set the wanted channel to volume only and use it for digitals.

 

 

 

Another feature is the 1.79MHz channel on the A8. The fast pulses of the filtered channel results in "new" waves at lower frequencies. With this features you can build waves as done with AdLib. The resulting waves get an analog style .

Actually, this is superior to PAULA but only in one channel.

Not like Adlib, more like AMY or various additive synthesizers. Adlib uses FM OR additive synth: ie 2 waveforms added together (of a limited selection of waves derived from sine -most Yamaha synths used true sine only; the OPL2 and OPL3 are the exceptions), or with frequency modulation using 1 wave to modulate the other. (the latter being FAR more

 

 

Not sure if we talk about the same stuff. Amy is using "digital" waveforms and puts them together. Well, as we know, every wave can be build with sinewaves.

 

I'm talking about the generation of the wave.

AdLib produces high frequency pulses to get the base wave created.

In 1.79MHz mode and filter, it is very similar. The ultra-high waves build a low frequency "flanger" which results in a new Waveform. It's only depending on the programming how the shape of the wave has to "look" . The result is very "analog"

 

 

This is a screenshot from the emulation. The source is the melody of the pause mode in my game Admirandus.

 

Actually, the emulation has some problems with it. POKEY does it better.

 

[+rdemming]

A shame Atari didn't use the YM2203 though, that would have been pretty awesome, even on the STe along with the DMA sound. (definitely more of an advantage over the Amiga, especially in the right hands) Fully backwards compatible with the YM2149/AY and similar footprint too, so could have made sense to introduce later on even. (even if the price wasn't an issue, it was new in '85 and may not have been available -NEC probably had top priority for the PC8801 series)

 

IIRC, I read somewhere (old magazine) that Atari was planning to use a successor of YM2149 (seems to be the YM2203) but it didn't make it in the ST because the chip wasn't available at the ST launch date. Can someone confirm this?

 

Robert

[analmux]

[...]

Only using channel 1 & 3 (at 1.79Mhz) we can generate a sawtooth wave. Using the lowest volume (choosing 1 out of 0-15), varying pitch & managing the timing of the start of the waveform, and also choosing upwards or downwards sawtooth, we have a method for finetuning the 4/5 bit PCM. Then it's similar to an effective 8 or 9 bit sampling quality.

And that's not using any interrupts, right? (vs software modulation -very technically possible on the ST, but impractical for most games)

Of course it uses interrupts. The Pokey PCM / sampling features will not run by themselves, as Pokey doesn't have active DMA control. So, we already need an interrupt scheme, as in the classical case we also do 4-bit PCM sampling by using an interrupt (or fixed timing scheme). Doing some extra sawtooth modulation (also by interrupt/timing scheme) on top will help us getting a finer audio level control. Let's say, effectively 8 or 9 bit.

[kool kitty89]

A shame Atari didn't use the YM2203 though, that would have been pretty awesome, even on the STe along with the DMA sound. (definitely more of an advantage over the Amiga, especially in the right hands) Fully backwards compatible with the YM2149/AY and similar footprint too, so could have made sense to introduce later on even. (even if the price wasn't an issue, it was new in '85 and may not have been available -NEC probably had top priority for the PC8801 series)

 

IIRC, I read somewhere (old magazine) that Atari was planning to use a successor of YM2149 (seems to be the YM2203) but it didn't make it in the ST because the chip wasn't available at the ST launch date. Can someone confirm this?

 

Robert

I believe the YM2203 was launched in 1985 close to the time the YM2151 did, though preproduction samples probably would have been available earlier (for various potential buyers to test), however the initial quantities may have been limited and Yamaha had several other buyers: there were several arcade companies using it (including Sega and Capcom), but more so, NEC likely would have had priority for the new models of the PC-8801 adding the YM2203. (possibly the new PC-9801 as well)

 

So that would have left it as a later option, but Atari didn't push for that either it seems. (making it standard on the 1040 and MEGA might have made sense, or retroactive across the board for all ST models -again, better if expansion provisions were made and there were several other features -like scrolling and DMA sound- that also could/should have been addressed fairly early on to promote support as standard features -simple scrolling would have been faster/simpler/cheaper to add than a full blitter and could have been embedded into the SHIFTER while a simple DMA sound channel also should have been relatively simple to add much sooner than the STe -especially a simple fixed/limited frequency mono channel, though pushed to 48-50 kHz like the STe or Tandy-1000 would have been nice, especially for multiplex/interleave mixing)

 

So a scrolless shifter and plain YM2149 could have been stop-gaps for early models. (a bare 8-bit DAC port would also be a nice stop gap for software samples sound prior to DMA -mainly for demos and deciated speech/music programs though, or low sample rate in-game stuff)

 

There was also the AY-3-8930 with hardware variable duty cycle pulse wave and 3 envelopes (so closer to the SID, but without filtering or ring modulation -and having to use the envelope to hack low res saw or triangle waves). The YM2203 also didn't support those features, so it wouldn't have been a clean upgrade as it could be for the YM2149 or plan AY-3-8910/12/13.

 

 

 

 

 

And can you really do volume modulation on 2 POKEY channels with 2 16-bit mode paired/slaved channels? (you couldn't use the timer interrupt method, but the ST has added interval timers to push that)

 

Why should this be a problem? When joining 16 bit, both channels play sounds depending on the other voices setting. Then you could set the wanted channel to volume only and use it for digitals.

OK, but you wouldn't be using POKEY interrupts to drive those PCM channels. (you could use external interrupts or cycle timed loops though -the latter being difficult to implement in-game though)

And, again, that really doesn't change things a whole lot from what the AY already offers in the ST, at least for the PCM side of things. (and anything in-game would be very limited as it is now -especially anything interrupt driven)

 

Yes, POKEY is more flexible in some areas, but they both have trade-offs.

 

 

In any case, it's a bit moot since Atari should have aimed at at least having rudimentary DMA sound like the MAC but perhaps at a higher sample rate (with or without an additional PSG), and that would have likely been cheaper in the long run too, if they removed the PSG at least -they could cut the YM and one of the ACIAs in favor of a VIA or CIA -and have a few "bonus" I/O lines from those chips. (it could have meant a delay in the release of the ST though, same with adding scrolling registers to the SHIFTER -both things should have been relatively simple and inexpensive to add, but there probably would have been a trade-off in development time and release date . . . albeit they could have included better provisions for expansion from the start and added those simple -but very important- features by '86 if not late '85 with relatively easy drop-in replacements for early models -maybe have a direct write DAC port in the early models and an added DMA chip added when it was complete)

An FM synth chip would have been nice, but they'd just become available in '85 and would be in limited supply as well as less favorable price wise. (if they did keep the YM2149, the 2203 would have been a very nice replacement later on)

 

 

 

 

Not sure if we talk about the same stuff. Amy is using "digital" waveforms and puts them together. Well, as we know, every wave can be build with sinewaves.

The Amy comment was on the issue of adding the POKEY channels together, more of a rough analogy though. (just additive synthesis)

 

I'm talking about the generation of the wave.

AdLib produces high frequency pulses to get the base wave created.

In 1.79MHz mode and filter, it is very similar. The ultra-high waves build a low frequency "flanger" which results in a new Waveform. It's only depending on the programming how the shape of the wave has to "look" . The result is very "analog"

How is that similar to the YM3812/OPL2? Do you mean in the way the additional (non sine) base/single operator waveforms are created? (aside from 2-op additive or FM synth)

 

That's something unique to the OPL2 and OPL3 (and OPLL) with all other Yamaha FM synths using pure sine waves. (including the original YM3526 OPL) I don't think it was very often used for Adlib stuff, or OPL3 stuff for that matter. (at least I've gotten that impression)

 

 

Do you mean analog as in analog synthesis sounding? (digital synth like Yamaha FM can sometimes have that sound a bit too -especially if trying to emulate analog synth sounds, but I've come to associate the more distinct nature of digital FM separately -the general crisp sound from FM synth -aside from implementations that low pass filter the crap out of it- is one of the distinctive differences -albeit that's more tied to analog synth often using heavy filtering, and without that it would be more like simpler chips sound -pure square/triangle/saw/pulse with or without ADSR)

 

 

 

It's interesting that most POKEY chiptunes seem to avoid using those tricks. I think I've heard a few using sawtooth stuff, but mainly square/noise/periodic pulse/noise and software pulse modulation. (the AY's saw hack seems to get used much more often -though PWM is rather rare by comparison)

I'm not actually sure this is using saw, but it sounds like it might:

http://battleofthebits.org/arena/Entry/pokey+mann/3639/ (pulse waves of certain duty cycles can sound a bit like saw -especially at low frequencies)

 

 

 

 

 

 

[...]

Only using channel 1 & 3 (at 1.79Mhz) we can generate a sawtooth wave. Using the lowest volume (choosing 1 out of 0-15), varying pitch & managing the timing of the start of the waveform, and also choosing upwards or downwards sawtooth, we have a method for finetuning the 4/5 bit PCM. Then it's similar to an effective 8 or 9 bit sampling quality.

And that's not using any interrupts, right? (vs software modulation -very technically possible on the ST, but impractical for most games)

Of course it uses interrupts. The Pokey PCM / sampling features will not run by themselves, as Pokey doesn't have active DMA control. So, we already need an interrupt scheme, as in the classical case we also do 4-bit PCM sampling by using an interrupt (or fixed timing scheme). Doing some extra sawtooth modulation (also by interrupt/timing scheme) on top will help us getting a finer audio level control. Let's say, effectively 8 or 9 bit.

And high speed interrupts will kill the ST, that's why you see very limited in-game PCM and little to no software modulation.

 

You don't need an interrupt scheme either, it could be done all with cycle timed code, but that's tougher to manage outside of demos (the Apple II, CoCo, and Speccy had to do it though -and the Master System, among others). The Z80 in the Genesis has to do it that way too, but that doesn't have to heavily multitask either (only simple sound register updates, if that), and you wouldn't want to use interrupts anyway as they eat up CPU time like crazy compared to good cycle counted code. (interrupts would be OK for lower rate stuff, but not anything advanced like >16 kHz playback along with sound management, let alone scaling notes for multi-channels stuff or software decompression)

The 3 channel digital tracker on the Spectrum 48k should be all software timed code, and I'd expect the better software MOD players on the ST to be the same. (or A8 for that matter, though the gain isn't nearly as dramatic as other platforms due to the fast 650x interrupts -plus you'd only get 3 channels via interrupts anyway)

 

 

This topic has already come up a few times, and also popped up over at Sega-16.

http://www.sega-16.com/forum/showthread.php?p=276570#post276570

http://www.sega-16.com/forum/showthread.php?p=276847#post276847

http://www.sega-16.com/forum/showthread.php?p=277067#post277067

http://www.sega-16.com/forum/showthread.php?p=277537#post277537

Edited February 8, 2011 by kool kitty89

[kool kitty89]

Hmm, another thing is that any interrupt driven effects would be impossible on the 7800. (no IRQ on the cart slot and not even RIOT interrupts -which would also be useless for any effects needing interrupts synced to POKEY's timers but useful for other things -including TIA hacks) Plus, the 7800's DMA set-up could create situations where you wouldn't have reliable interrupts. (ie a lot of missed/delayed responses to IRQ -or you could make sure MARIA left enough DMA time for the CPU in active display)

I don't think that's ever an issue on the A8/5200 since ANTIC/GTIA can't saturate the bus like that.

[VaderXS]

The fact that the father of the 64 moved to Atari tells the whole story. The Yamaha was a supern chip on the Atari ST, even used by Madonna. BTW, there's no bad chips, only bad coders. ST all the way!

[jefffulton]

To answer your question @ATARI7800fan. The Sampled music at the beginning of Turrican II is in the mono ST version opf the stereo Amiga tune -  and sounds great. The developers did a fantastic job porting Turrican and Turrican II to the ST, fitting in everything they could from the Amiga version within the 512K memory barrier for ST releases (self imposed by most software houses to sell to the greatest number of Atari owners).   They created a multiplexed YM sound driver and ported the Amiga music using the YM's chip tune capabilities. If you see a video that has sampled stereo music during ST gameplay, that one is a fake, unless someone has ported the game to the STE and made adjustments to the sound.