Why did the Atari ST designers use such a CRAPPY sound chip?

[Foebane]

Originally Posted by Nishicorn

This is at least debatable.

What does 'better' mean? If you want four channels of sample-filled extravaganza, sure, Paula wins.

Yes, it does.

If you want quirky realtime 'personalized' synth composing and sound, Paula doesn't quite compete, as it has to be either stuck with the samples or the cold, manufactured synth sound that's not the same as a real synth chip, like YM2149-F.

I want samples from my sound chip, not synths.

Why is it so hard pill to swallow for many to appreciate and recognize this marvellous chip as a 'good sound chip for music and interesting sounds'? Anyone doubting YM2149-F's abilities have clearly not listened to the songs that are captured from the real Atari ST.

Samples will always be better than synths for sound accuracy, and I had enough of this synth stuff back on POKEY.

Don't be prejudiced and judge a good sound chip based on some streamlined emulation that cuts some corners and removes some details, subtleties and energy of the original.

1. Not everyone has access to (potentially expensive) original hardware.
2. No-one wants to listen to big audio clips that take up lots of disk space.
3. If the emulated sound accuracy isn't there, that's the programmers' fault.

Paula is also limited to being 8-bit (although I know it can reach at least 14-bit these days, not even sure how that's done) and 22kHz basically, whereas YM2149F, being a real synth, doesn't suffer from this kind of limitation. This means good-sounding high-pitched sounds, for example.

What do you mean, "these days"? Paula could do 14-bit from the start, it was well-documented, but the storage space wasn't there to utilise it at launch.

Samples are good for many things, but they're always a bit inflexible and static compared to a 'living synth sound' that can move in quirky ways as the song plays.

I want to hear computer music that sounds like REAL music, and Paula's samples do a great approximation of it in 8-bits.

I do love Paula, SID, OPL chips and YM2149F, so I don't really have hatred for any of these chips, I think they complement each other. Samples are good for drums and maybe 'adding channels' by using chords in one channel, and such, but try to give me a good, 'living leadsound' in a regular Amiga 500 Protracker song.. it either requires a long sample, or lots of tiny ones that you switch between and modify and it's still never gonna sound as organically and instantly good as, let's say a filtered PWM on a real SID chip. You have to really work hard for that.

And synths will NEVER EVER sound as good as real musical instruments. I just don't get this love for that sort of sound. My opinion, since it's my post.

You can't just sample some 'nice lead sound' sample and make a melody with that and expect it to be competitive. So samples are not 'better' for everything.

Samples may be flawed sometimes, but I prefer them to synth stuff. I noticed I only liked the Atari ST sound chip when I heard samples on them, in a MOD. Funny, that.

In any case, my point is, YM2149F deserves its place in computer music history, it can do amazing sounds that SID and Paula can't really compete with (well, Paula can of course sample that, but it's not quite the same, and it would degrade in quality anyway - playing high-pitched sounds in 8-bit 22kHz samples isn't going to sound as good as the original high-pitched sound on a real Atari ST), and I stand by that statement.

If you like that sort of sound, but as I said, I don't.

Of course SID can perform miracles that YM2149F can't reach or even dream of, but the same is true on the other direction as well. Only someone that hasn't heard the songs (and some others) that I recommended earlier, can claim otherwise. Don't just listen, but listen analytically, listen to the bass-sounds, the drum sounds, and for example, Timbral's quirky effects, where the pitch is moved while the volume is rapidly cut that sounds so cool. Easy to do on SID and OPL3 in theory, very hard to make sound as good and cool in practice.

Not my cup of tea.

My statements are not just statements, they are based on my experience and my painstaking experiments in trying to re-create this stuff. I must've created like 30 different 'Atari ST Bass'-instruments on my SID and more on OPL3, none of which sound even close to as cool, sharp and good as just the simple saw-like bass in Timbral's song, let alone Tao's masterpieces.

The only Tao I like is the one in Mysterious Cities of Gold, to be honest.

Learn some humility, learn to listen, receive, experiment for yourself, use authentic sound, and start to appreciate instead of just simply bashing based on your faulty and incomplete understanding of what the Atari ST's sound chip really is.

Authentic sound like samples? I don't care about the Atari ST sound chip, and neither did the designers, since they favoured MIDI over it and used it as almost an afterthought.

Remember, this is all my opinion, since it's my post. Oh, and my thread, too.

[Nishicorn]

Originally Posted by jizmo

The same thing kinda happened to me with Adlib as well: I found the soundscape tinny and plastic at the time, but these days I really enjoy a song that does not shy away from the characteristic Adlib sound, but rather embraces it to the maximum!

Well said.

Yes, observations change, and the initial impression might actually be based on faulty information, limitations that have nothing to do with the chip or the chip's power or capabilities themselves, and so on.

AdLib's sound had two-fold effect in me initially; first I wasn't too impressed with it, as I had heard the Amiga's glorious sound and SID's wonderful basslines and melodies already. 'Lemmings' on AdLib just sounded like a very poor man's version of anything. Ecch.

Second, a bit later, when game developers were more familiar with the chip and started creating inspirational and even orchestral musics with it, I started fantasizing about what it would be like to compose music using that splendidly MULTI-CHANNEL chip. Sure, it doesn't seem that multi-channel if you consider Amiga's one channel can have a multi-channel chord in it going on, and still have three channels free for other stuff..

.. but still, 9 channels of pure, live, synthesized sound instead of the lifeless, static samples that always sound different on different pitches due to the length changing (I think there were a couple of trackers and such that tried to remove this effect, like TFMX and some Protracker version maybe, but this is basically how pitches were creted with samples - making the sample play slower or faster, which created annoying problems that synth sound simply never had) was a dreamy think to think about.

I started to really like Adlib's sound during early 1990s, when the games started to mature their soundworld, and playing some exciting games on my friend's PC and completing Sierra adventures together, and such things kind of installed an admiration and love for this sound into me, that I always missed, when that era was over. It was a unique sound, a personality I never heard on the Amiga. Lucasfilm/LucasArts Games games and Wing Commander just sounded so much better on the OPL2.

The thing is, little did I know how much the OPL2 (let alone OPL3) was capable of! Almost no games ever really explored OPL3's capabilities although almost all Sound Blasters had that chip for years, because OPL2 would be more compatible, and in any case, they never 'banged the hardware' but instead just used a 'generic system' where they could just compose the song once, and then the different drivers would play it for different soundcards.

This leads to Rob Hubbard's songs sounding very non-specific and generic - if you listen to 'Budoukan' (I know it's spelled Budokan in the west, but that's just wrong), and pay attention to the instruments, you can realize how generic and simple they are - nothing like the quirky stuff Mr. Maestro did on the SID back in the day. He had become complacent and lazy, and started composing 'one song' that was then just 'converted' to different drivers, so individual instruments were never even touched by Hubbard's hand on the OPL2, let alone an OPL3 instrument being created..

So as much as we heard back in the day, we NEVER heard the full capabilities of OPL2, and almost any capabilities of OPL3. There were a few games, but I never bumped into those back in the day, and often they were made too late anyway, when 'real soundcards' had started to conquer the markets. Audiences don't ever demand 'live synth', they just demand 'what's the newest and coolest' without understanding the details.

So OPL3 was never utilized almost at all, and whatever was utilized was very generic. Almost no one these days knows fully what that chip is (or would be) capable of (in hands of geniuses), not even me, and I own two of them.

It's sad we never got 'Hubbard-effect on OPL3', where a genius like that would've taken OPL3 and composed something as quirky and wonderous as Hubbard composed on the SID (among some other composers as well).

Think if Budoukan's (I refuse to spell it 'Budokan') instruments were as painstakingly crafted to sound as authentic, as japanese, as good, as quirky and interesting, as personalized, as Hubbard's SID instruments were... they would sound so different.

I made my own experiments with creating OPL3 instruments, and it CAN be made sound much closer to japanese instruments (even by me, let alone someone like Hubbard) than you can hear in that pathetic Budoukan song of Hubbard.

It's sad that there were so many different soundcards and drivers, that no musician could ever compose for every chip separately and manipulate each instrument to a perfection - they were on deadline, they were lazy, it would've been unfairly too much work anyway, and generic midi-style composing was the perfect solution, especially if the paycheck is more interesting than instrumental/musical/sound expression you -could- tweak out of the OPL2 and OPL3 - but won't.

Well, I might be the only individual in the Universe to want and try to rectify this situation, I intend to explore OPL3 to its fullest and do as much and as good-sounding stuff with it as I possibly can, and experiment in all ways I can.

And even STILL I couldn't even come near to being able to re-create the sharp, cool bass-sounds you hear in those Atari ST songs...

[no9]

Originally Posted by AmigaHope

Paula is also limited to being 8-bit (although I know it can reach at least 14-bit these days, not even sure how that's done) and 22kHz basically, whereas YM2149F, being a real synth, doesn't suffer from this kind of limitation. This means good-sounding high-pitched sounds, for example.

Samples are good for many things, but they're always a bit inflexible and static compared to a 'living synth sound' that can move in quirky ways as the song plays.

I do love Paula, SID, OPL chips and YM2149F, so I don't really have hatred for any of these chips, I think they complement each other. Samples are good for drums and maybe 'adding channels' by using chords in one channel, and such, but try to give me a good, 'living leadsound' in a regular Amiga 500 Protracker song.. it either requires a long sample, or lots of tiny ones that you switch between and modify and it's still never gonna sound as organically and instantly good as, let's say a filtered PWM on a real SID chip. You have to really work hard for that.

Yes, and Paula chip with all it's limitations opens wastly wider range of sound timbres than any narrowly specialized synthesis based sound chip. I actually don't like to reduce Amiga sound just to Paula, because Paula is a chip which resides in certain hardware environment and thus enables it to produce wide pallete of sounds. So it is more the sound of Amiga, not just Paula. But I also often use just 'Paula' term for the sake of simplicity. We all here are pretty aware how it works.

So while Amiga/Paula is sample based it is not limited to samples if those short samples are treated as sound generators, oscillators. AHX by Abyss proves that by reserving reasonably small system resouces you can have pretty nice modulations with those.
[ Show youtube player ]

There are more synthesis based trackers on Amiga https://github.com/bryc/code/wiki/Amiga-synth-trackers

Many of them are not to much composers friendly, but it shows that not only Protracker defines the sound of Amiga/Paula. As you mention, Protracker was also capable to modulate sound with a little bit of effort, but what doesn't? To program interesting sounds with synth-chips you also need to put some work.

Btw. Paula doesn't relies on fixed frequency. It even doesn't play exact 22,5kHz. Also 8-bit samples is not what you get from Paula because there is a sample volume register which let's you playback quiet sounds without introducing a noise. And I'm not even talking about all this 14-bit stuff.

So if one is looking for widest possible palette of sounds the answer is clear. This doesn't prevent anyone from enjoying any synth chip or even prefer it over Amiga/Paula. Like for example... one can prefer listening to a solo guitar over the whole orchestra.

[Bruce Abbott]

Originally Posted by Nishicorn

This is at least debatable.

Not really. The YM22149 generates square waves with up to 32 levels. Paula can easily reproduce that with enough fidelity that you would not be able to tell that it wasn't a YM, and play several channels of PCM sound at the same time. So it is objectively better than the YM, but uses a lot a lot more resources to do it.

Why is it so hard pill to swallow for many to appreciate and recognize this marvellous chip as a 'good sound chip for music and interesting sounds'? Anyone doubting YM2149-F's abilities have clearly not listened to the songs that are captured from the real Atari ST.

Not so hard for me to swallow at all. In fact while writing that post I was also listening to AY music on my Aquarius, using the expansion I made for it which currently has a YM2149 in it.

IMHO the AY songs in my collection are better than any that have been mentioned here so far. Part of that is no doubt my tastes in music, but another factor is that the ST only had mono output. Also many 'chip' tunes do not take advantage of the low frequencies it can produce, perhaps because typical amplifiers and speakers of the day could not reproduce them well. You have to listen to the YM2149 on a good stereo system to really appreciate it.

Paula is also limited to being 8-bit (although I know it can reach at least 14-bit these days, not even sure how that's done) and 22kHz basically, whereas YM2149F, being a real synth, doesn't suffer from this kind of limitation. This means good-sounding high-pitched sounds, for example.

This is more a matter of external filtering than the chips themselves. In my AY expansion I ignored the recommendation for a low pass filter because I didn't want to compromise the high frequencies. But my ears are 30dB down at 3kHz, so I need that just to hear the higher frequencies at all.

Most Amigas have a low pass filter that starts to have effect above ~7kHz, except for the A1200 which is nearly flat to 20kHz. When playing 'chip' music aliasing isn't an issue like it is with sampled sounds, so the filter just has to deal with the volume control PWM frequency (which is well above the audio band). What this means in practice is that Paula itself is quite capable of producing just as 'sharp' sounds as the YM when playing 'chip' music, though the A1200 might be the only Amiga that can do it (depending on which AY/YM system you are comparing it to).

Samples are good for many things, but they're always a bit inflexible and static compared to a 'living synth sound' that can move in quirky ways as the song plays.

Yes, though the 'staticness' can be gotten around by having more samples or generating synth sounds on the fly.

I do love Paula, SID, OPL chips and YM2149F, so I don't really have hatred for any of these chips, I think they complement each other.

I agree. The unique characteristics of each chip encourage production of unique music for each. So even though Amiga songs could sound the same as AY songs, they generally don't.

Of course SID can perform miracles that YM2149F can't reach or even dream of, but the same is true on the other direction as well.

That's what they tell me, but the SID music I have heard so far all seems to be variations on a theme. In comparison the AY/YM seems to be more flexible - or perhaps it is just that SID musicians like producing a certain type of sound.

[meynaf]

Originally Posted by Bruce Abbott

Not really. The YM22149 generates square waves with up to 32 levels. Paula can easily reproduce that with enough fidelity that you would not be able to tell that it wasn't a YM, and play several channels of PCM sound at the same time. So it is objectively better than the YM, but uses a lot a lot more resources to do it.

Errrh, no. You cannot "easily" reproduce YM2149 sound on Paula. Not accurately in any case.

First, the volume levels of Paula are (more or less) linear, whereas YM has logarithmic scale. This means not enough volume levels on Paula to impersonate as an YM, at least with small values. But this is probably your smallest problem.
Second, the YM has 3 outputs that get mixed on the ST in a strange way. They kinda fight against each other in the electronics outside the chip, leading to nonlinear mixing. This means, even though you can (at least in theory) output single channel with great accurary, output of 2 or more is NOT the simple addition of them. The output isn't an exact square wave either.
Third, hell will happen if the music on the ST uses envelope with small period, and you can expect many Hippel musics to do just that.

At the end, you can play many musics with simple emulation, but it will not sound like real ST and it's easy to tell the difference. And some musics will fail miserably.
More complex emulation, like what's done in emulators require output table of 4096 entries sampled from real ST, and takes half the power of 68030/50.
I know this, because i wrote both.

[Foebane]

In the end, it's down to personal preferences as to what constitutes a "good" sound chip.

In my case, it's the following:

Paula - Pure awesomeness
SID - OK
POKEY - A bit tone deaf, but great bass notes
ST - Just, ugh

[drHirudo]

Originally Posted by meynaf

More complex emulation, like what's done in emulators require output table of 4096 entries sampled from real ST, and takes half the power of 68030/50.
I know this, because i wrote both.

What if we look from the other side?

What CPU power will Atari ST need to be able to play simple .mod tracks, or Turrican I and II music with the use of it's YM chip? Will it be able at all?

[Thomas Richter]

Originally Posted by meynaf

Errrh, no. You cannot "easily" reproduce YM2149 sound on Paula. Not accurately in any case.

Thus, if I understand correctly, the problem is really at the analog end, and not really on the chip side.

[Thomas Richter]

Originally Posted by Foebane

In the end, it's down to personal preferences as to what constitutes a "good" sound chip.

In my case, it's the following:

Paula - Pure awesomeness
SID - OK
POKEY - A bit tone deaf, but great bass notes
ST - Just, ugh

Well... the chips were designed with different target applications in mind. SID is a musical instrument requiring minimal additional hardware resources. POKEY is a sound effects generator which can also play a bit of music, Paula is requiring more resources to generate sample-generated sound (which later became the norm anyhow), but it's coming from another chip generation anyhow.



The YM - well, I do not really know. It is a chip from the POKEY/SID generation requiring minimal resources and using a very simplistic design, but there is really nothing specific to the thing. No effects, no waveform synthesis, no sample DMA. It is just a very "generic" programmable square-wave generator without any specific extras that would make it stand out. Probably "the cheapest thing Atari could get in order to deliver the system in time".

[meynaf]

Originally Posted by drHirudo

What if we look from the other side?

What CPU power will Atari ST need to be able to play simple .mod tracks, or Turrican I and II music with the use of it's YM chip? Will it be able at all?

Yes it is able. It has been done for mods, both Turrican musics have been done, and even simple 68000 can do this. Don't expect very high quality, though -- and exact cpu usage will depend on the compromises you make here.

Originally Posted by Thomas Richter

Thus, if I understand correctly, the problem is really at the analog end, and not really on the chip side.

Yes. Same chip in different hardware may sound differently.
Still, the chip side can do things at very high frequencies and this is equally hard to emulate.

[jizmo]

Originally Posted by Nishicorn

Well said.

Yes, observations change, and the initial impression might actually be based on faulty information, limitations that have nothing to do with the chip or the chip's power or capabilities themselves, and so on.

After reading this I took some time to reflect on why these observations have changed and really had to take a trip down to the memory lane to the late 80s to get to the bottom of this.

My take is that it was due to the whole industry at the time still striving to reach the compact disc soundtrack quality that was going to be the silver bullet. Even if poor in today's standards, Amiga's punchy drums, and sampled instruments were certainly closer to that than what the FM based systems had to offer at the time, with the soundscape you'd never expect to hear on a commercial CD.

As the CD quality became commonplace and even banal in the early 90s, it was then easier to once again for us to appreciate the characteristic and recognisable soundscape of various sound chips of the past.

[chip]

For me computer music DOESN'T must reflect real music

I listen to it especially because it is DIFFERENT from the real music

I like it because it's synthetic and without words actually

And i absolutely don't care if the various sound chips are able or not to replicate CD quality ........ who cares, really ?

[jizmo]

Originally Posted by chip

For me computer music DOESN'T must reflect real music

I think you missed the point. I never said that computer music needs to reflect real music. What I meant is that Hi-Fi, CD quality sampled sounds and the soundscape is what people were really seeking after in the late 80s, so it was valued above anything home computers with FM-based synthesis had.

Take Jarre or Kraftwerk, or Tangerine Dreams for example. Both hobby and professional composers were trying to imitate the same richness and effects on the home computers of the era and it was the sampling approach that Paula brought that took the biggest step towards that direction and this is why Amiga's sound resonated so much better at the time for most people. Amiga could quite faithfully reproduce any sound fed to it from a compact disc or a high end Korg synthesizer.

I for one remember the very moment when I knew I just had to get an Amiga of my own.

I don't remember anyone rating Atari's sound over the Amiga's in the late 80s; it was simply not even a competition at the time. But as CD sound became an everyday thing and quality and the polyphony surpassed the 8-bit fidelity Amiga had brought to the table a few years earlier, people learned to appreciate the characteristic and very recognisable lo-fi YM and SID sounds again.

[Bruce Abbott]

Originally Posted by meynaf

Errrh, no. You cannot "easily" reproduce YM2149 sound on Paula. Not accurately in any case.

Define 'accurately'.

First, the volume levels of Paula are (more or less) linear, whereas YM has logarithmic scale. This means not enough volume levels on Paula to impersonate as an YM, at least with small values. But this is probably your smallest problem.

Paula has 14 bits per channel, which should be plenty enough. But as you say, this is the smallest problem.

Second, the YM has 3 outputs that get mixed on the ST in a strange way...

Not that strange, it just sums the 3 DAC current outputs in a single 1k resistor. This causes compression at higher levels, with the level of each channel affecting the others. On my AY/YM board each output is loaded individually with 470 Ohms, which virtually eliminates compression and interaction.

The output isn't an exact square wave either.

On the ST this is true, since low pass filtering is applied with a cutoff frequency of ~2kHz.

At the end, you can play many musics with simple emulation, but it will not sound like real ST and it's easy to tell the difference. And some musics will fail miserably.

And those same musics could well 'fail miserably' on my own actual YM2149 (assuming I could even get a player for them).

More complex emulation, like what's done in emulators require output table of 4096 entries sampled from real ST, and takes half the power of 68030/50. I know this, because i wrote both.

This is where you went wrong. I never said anything about emulating an ST, but simply reproducing the 'sound' of a YM2149 - the same way we can reproduce the 'sound' of a piano or a guitar. Yes, there will be differences between that sound and various YM2149 implementations (which themselves have their own slightly different sounds) but unless you define 'better' to mean 'closer to the exact sound of an ST', the Amiga is objectively better because it can play AY/YM style music and more.

Making the ST's sound the 'gold standard' (that makes other platforms no better because they can't easily emulate it exactly) is silly because in the ST the YM2149 is crippled by its mixer circuit. It's not even as good as a YM could be, let alone as good as an Amiga.

[meynaf]

Originally Posted by Bruce Abbott

Define 'accurately'.

Something like "you would not be able to tell that it wasn't a YM".

Originally Posted by Bruce Abbott

Paula has 14 bits per channel, which should be plenty enough. But as you say, this is the smallest problem.

But then you have to change both the sample and the volume, which makes matters a little more complex than just altering volumes. Especially when fast volume changes need to be applied.

Originally Posted by Bruce Abbott

Not that strange, it just sums the 3 DAC current outputs in a single 1k resistor. This causes compression at higher levels, with the level of each channel affecting the others. On my AY/YM board each output is loaded individually with 470 Ohms, which virtually eliminates compression and interaction.

Strange or not, the end result is that you can not simply compute the output level for any given configuration. Or i want to see the formula !

Originally Posted by Bruce Abbott

On the ST this is true, since low pass filtering is applied with a cutoff frequency of ~2kHz.

Oh, and i forgot this one as well. But the output from the mixer, before filtering, is already not perfect square waves mixed together - that was the point.

Originally Posted by Bruce Abbott

And those same musics could well 'fail miserably' on my own actual YM2149 (assuming I could even get a player for them).

If it's in another machine type, that may indeed be true. I was very deceived when i attempted to get my old oric tunes to the ST. Same kind of chip (AY-3-8912), just 1Mhz diff so all periods should be multiplied by two -- but an end result not sounding at all like it should (note frequencies ok, but nearly all volumes gone wrong). Not that the mixing was linear there either...

Originally Posted by Bruce Abbott

This is where you went wrong. I never said anything about emulating an ST, but simply reproducing the 'sound' of a YM2149 - the same way we can reproduce the 'sound' of a piano or a guitar. Yes, there will be differences between that sound and various YM2149 implementations (which themselves have their own slightly different sounds) but unless you define 'better' to mean 'closer to the exact sound of an ST', the Amiga is objectively better because it can play AY/YM style music and more.

Making the ST's sound the 'gold standard' (that makes other platforms no better because they can't easily emulate it exactly) is silly because in the ST the YM2149 is crippled by its mixer circuit. It's not even as good as a YM could be, let alone as good as an Amiga.

Well then, we weren't talking about the same thing.
I haven't said the YM is in any way better than an Amiga, so far not. It has its distinctive sound, some like it, some don't, no more no less.
Nevertheless, emulating it is my problem, and if possible with small enough cpu footprint. Hopefully currently all my ST game ports are satisfied with simple emulation...

[Cyprian]

Originally Posted by meynaf

Strange or not, the end result is that you can not simply compute the output level for any given configuration. Or i want to see the formula !

the output voltage level was measured at the beginning of '90


Regarding formula, it is available, just check Hatari or Steem SSE source code.

[meynaf]

Originally Posted by Cyprian

the output voltage level was measured at the beginning of '90

I know. I have the 4096-entries table.

Originally Posted by Cyprian

Regarding formula, it is available, just check Hatari or Steem SSE source code.

They use the output level table. At least Hatari has the option to use "math model" instead, but the table is what gives the results closest to real ST.

[rabidgerry]

Originally Posted by manossg

https://youtu.be/u5VrqlLItco?t=21

Sounds amazing to me! I love the scratchy crazy drums on the left and those tractor beam/phone ring noises.

[manossg]

Originally Posted by rabidgerry

Sounds amazing to me! I love the scratchy crazy drums on the left and those tractor beam/phone ring noises.

Listening to it again now, what a thumping tune!

[ImmortalA1000]

The possibilities were limitless with Portia for 1985, the problem was lazy idiots who sampled naff waveforms from horrible FM synths to make totally jarring game soundtracks that have bugger all to do with the atmosphere of a game like Risky Woods. When used in the right hands like Jeroen Tel in Agony (the frequency range on my Spectrum Analyser is quite high on the piano sample) or David Whittaker's Shadow of the Beast 1 tunes (512k game!) the Amiga has awesome sound. Few people did tunes that good. C64 games had much more talent in the sound department, and the two best Amiga games tunes are from SID composers. I suppose Ghouls n Ghosts will be mentioned but that game is PD quality at best so who is going to buy a £25-30 just for the music?

End of the day Atari had to hit a price point, they had to pay other companies to do work that in the past MOS Technology did for Jack for basic salaried wage and Atari had to make a profit. 512k of RAM was very expensive in early 1985 too, sharp drop in price by 1987 when A500 turned up (it accounts for more than 50% of the price drop). It's the second best computer after the Amiga 1000 in 1985.

The Amstrad's 1 channel to the left, 1 to left and right, 1 channel to the right is an idiotic idea and to be quite honest 99.9999999% of people buying an Amstrad CPC had no choice but to listen to it via a crap 2cm tiny little speaker even if they used it on a TV via the modulator (if you could find one).
At least the ST sound comes out of a decent speaker if you have a decent TV (we had a TV with very decent stereo speakers in 1982 so no reason to hear the AY/YM sounds in 2cm speaker quality lol)

It is what it is, the 520STM + external disk drive bundle in Spring 1985 cost less than a Commodore 128 + 1571 disk drive by a significant margin.

I just kept my 1983 C64 and bought some games for my C64 and some for my 520STM in 1986/87 until I found a used Amiga 1000 for half the price of an Amiga 500

I never had any issues with getting audio out from the monitor port on my 520STM, didn't sound any more hissy than doing the same via the monitor port on my C64. Would never buy an amplifier to listen to ST music. Maybe the later 520STFMs for £299 RRP had duff sound, can't say.

I think the ST soundchip is there for all sorts of stuff (disk controller?) so it just fit their budget restricted design. It's not as good as the sound on a $4000 512k Mac of 1985 sure but then the Mac has shit graphics so there's that too.