Jump to content
Sign in to follow this  
kool kitty89

Why didn't/couldn't Atari Corp use the AMY in the ST?

Recommended Posts

I've seen a lot of mixed info on AMY from some claims that Atari Corp simply "couldn't get it to work" but that seems to be outdated information and there's apparently more direct responses from the people involved that point to the chip being completed to manufacturing quality and reaching protype/demonstration stage (possibly preproduction) in the XEM. Given what's mentioned here:

http://www.atarimax.com/jindroush.atari.org/achamy.html

http://en.wikipedia.org/wiki/Atari_AMY

 

The XEM was dropped of course, with the A8 line declining, but why not repurpose it for the ST? I seem to recall that AMY was initially considered for the ST, but it was realized that it couldn't be done in time for the ST's planned release and got moved onto the XEM in the short term, but why not push it back to the ST after that?

Were the problems interfacing it directly with the 68k or was it just too expensive? (the original layout intended a dedicated 8-bit MCU/CPU to drive it, so that would be a bit excessive, and the external DAC would add to board space and cost -especially if full 16-bit was used, but couldn't they have made another revision after AMY-B with an internal DAC, even if they dropped to an 8-bit output resolution to save cost? -in the latter case that should also have dramatically cut pin count and package size to 24 or 28 pins -24 if the reserved pins weren't used for anything)

Share this post


Link to post
Share on other sites

I reckon they just couldn't get it working properly.

 

What docs I've seen of it seem to refer to interfacing with Intel processors, so maybe they used IBM machines in early testing.

 

Supposedly the engineers with the know-how got axed before they realised what they had on their hands.

 

As for the XEM - tragic that it never came out, but who knows - the CES demo unit may well have been a kludge using a hidden PC doing the grunt work. Wouldn't be the first time that sort of thing happened.

 

I don't know if it even did sample playback as we know it. From what I gathered (looking some time back), you programmed all of the individual step characteristics for the waveform you wanted and the chip pieced them together and played them.

Edited by Rybags

Share this post


Link to post
Share on other sites

From what i heard, atari (tramiel) sold or licensed out the design to a third party co. to get a working ST version, which apparently they succeeded in doing so, but like anything concerning deals with tramiel things didn't go smoothly the company who were doing the ST version threatened to go ahead an launch the ST version and tramiel/atari hit them with a lawsuit/legal action (and alas we ended up with no ST version, but doc's/schematics or even proto's must be out there...somewhere)

Edited by carmel_andrews

Share this post


Link to post
Share on other sites

I reckon they just couldn't get it working properly.

 

Supposedly the engineers with the know-how got axed before they realised what they had on their hands.

I think that's old info (it's on Atarimuseum, but a lot of stuff hasn't been updated there for quite a while), and there's directly conflicting information that states the engineers continued to work on the project as needed.

 

http://www.atarimax.com/jindroush.atari.org/achamy.html

--------------

Group leader was Gary Sikorski. Main architect was Scott Foster, helped by Steve Saunders. Sam Nicolino was hardware designer and translated the achitecture into an MSI simulation and then a chip. John Palevich did some of the software, Tom Zimmerman did even more. Some original software still exists on some demostration and development disks that came with one 65XEM prototype. The disks are available at John Hardie's site.

 

With the firing of the Advanced Technologies Group at Atari, the new engineering staff now under the leadership of the Tramiels attempted to wedge this chip into an Atari 8bit, hence the Atari 65XEM. Shown at the 1985 CES, the 65XEM was touted as having remarkable capabilities for sound/voice synthesis. However without the original designers, getting the chip to function properly became an expensive burden which the Tramiels could not afford. This information contradicts the one from Sam Nicolino (hw designer) who said: 'I worked with them as needed.', while finishing the chip design. Sam left Atari after he gave Amy-B step design to Leonard (I assume Tramiel, ed.).

---------------

 

Now those 2 things wouldn't necessarily be mutually exclusive is the issues weren't with getting a functional chip, but interfacing it in a satisfactory manner with what they had planned.

 

Also:

From the data sheet, my guess as to why Atari couldn't make it work has got to be either lack of manpower or just plain old ineptitude. As far as I can see, it's a pretty simple device.

 

That was Atari Corp. that couldn't get it to work right in what they wanted, not Atari Inc. Atari Inc. had already demonstrated it publicly, and had it planned for use in several different projects. Don't forget, Alan Kay's team had already moved on, as had much of the other advanced engineering groups. Jack had to go outside to get work done with it.

But again, I'm not sure what getting it working "the way they wanted" necessarily means. (ie interfacing it in a cheaper manner without a dedicated MCU, etc)

 

I'm also not sure if the "go outside to get work done" comment refers to consulting some of the former Atari Inc staff, or outsourcing to Sight & Sound. (I'd gotten more the impression that they sold off the design to Sight & Sound, not outsourced engineering)

 

What docs I've seen of it seem to refer to interfacing with Intel processors, so maybe they used IBM machines in early testing.

 

As for the XEM - tragic that it never came out, but who knows - the CES demo unit may well have been a kludge using a hidden PC doing the grunt work. Wouldn't be the first time that sort of thing happened.

It wouldn't have anything to do with a PC though: the original AMY interface was intended to use a simple 8-bit MCU with am 8051 specifically referenced (successor to the older 8048 line -used in the Intellivision among many other embedded applications), but that doesn't mean they couldn't have gotten it working with a 6502... though getting it to work directly with a 68k might have been the bigger problem and they definitely wouldn't want to add the cost of a dedicated MCU to the ST. (and you'd probably still need GLUE logic to interface the dissimilar bus types if the MCU didn't mesh with the 68k arch -from Atari's PoV, if they could have used a 6502/6502C, that could at least have been more realistic, but still probably throwing it off the table for the ST's design points, though had they specifically designed the ST to include a 6502 coprocessor from the start, that could have meant some other things: short of including backwards compatibility -which would be more complex and costly- they could have re-used parts of the A8 chipset rather than off the shelf parts and saved some cost that way too -especially using POKEY for both audio and keyboard reading, maybe make some use of the SIO- but then again, with POKEY and a 2 MHz 6502 with a little bit of its own work RAM you'd have some nice options for up to 4 4-bit PCM channels at higher playback rates and no software mixing -and in parallel with the 68k)

But any of that would be a hack anyway, and designing more dedicated audio hardware would be a good deal cheaper. (let alone non PCM options, especially after the fact of selecting the YM2149 -YM2203 was a very interesting possibility)

 

I don't know if it even did sample playback as we know it. From what I gathered (looking some time back), you programmed all of the individual step characteristics for the waveform you wanted and the chip pieced them together and played them.

No, it's not a sample based chip, it's a synthesizer, sort of an advanced additive synthesizer accumulating sine waves (of which there are 64 oscillators) in various possible combinations (in steps of 2 oscillators) onto 8 channels (voices) with 64 independent amplitude ramps but only 8 frequency ramps (one per voice) with 1 master oscillator with frequency control and additional ones added to build up harmonics. So in some respects it's similar to Yamaha's sine wave fased FM synthesizers, but without the FM capabilities, only the additive synthesis. (all the FM chips had additive modes -among the various FM algortithms, with 2-op FM you only have 2 oscillators so only 1 FM mode and additive synth mode, 4-op would allow 4 sine waves added in that mode, etc, and of course you could pair channels on such chips to do more additive synthesis, but with some other flexibility but with many fewer oscillators to work with -Yamaha's arcade standard YM2151 used 32 oscillators with 8 voices -8 4-op FM channels)

 

However, due to the nature of the additive synthesis mechanism it could be used for voice playback as well as some other recorded sounds with varying results (voice being particularly good though) and with far, far lower bitrates than the lowest acceptable PCM quality could allow. (3 kHz 4-bit PCM is 12kb/s, but AMY could play voice at reasonable quality at only 2.4kb/s but it would be encoded differently, of course, and worked better for some sounds than others -but it did a wonderful job of speech compression, even 1-bit CVSD would have to drop to 2.4 kHz to match that bitrate -and it would be pretty worthless at that rate -note CVSD is what the williams arcade games used, including Sinistar)

Share this post


Link to post
Share on other sites

There's a Wiki article on it, but the best info is on the old Jindroush page.

 

http://en.wikipedia.org/wiki/Atari_AMY

 

http://www.atarimax.com/jindroush.atari.org/achamy.html

 

Yes, the same links I posted when I started the topic. ;)

 

There's also Curt's page on the XEM, but I'm not sure if that's up to date or not (some stuff has been recently updated on atariage, but I think a good chunk of other pages are still waiting for updates).

http://www.atarimuseum.com/computers/8bits/xe/xe_protos/65xem.html

Share this post


Link to post
Share on other sites

It's a shame they couldn't get that into the 7800 instead of the tired old TIA.

Umm, I think the first step up would be POKEY before they'd even consider that. :P (and supposedly the only reason they didn't go with POKEY was board space and they apparently didn't want to use a riser board to address that on early models -or another/larger riser board rather as I believe the early revisions already had one riser board)

 

In that same line of thought, it would have made a lot more sense to enhance the A8's audio with a 2nd POKEY rather than something as elaborate as AMY. (and no MCU or separate DAC needed) Hell, if they were going to use POKEY dedicated to audio in the 7800, doubled in the A8, or on-cart with the 7800, they probably could have easily used a cost-cut version with the POT/IO/key lines removed and cut to 24 pins -and you could leave all the timer/irq lines still. (going below 24 pins would mean a narrow DIP and that might not work without shrinking the die used -ie more engineering cost and new masks rather than simply changing the package/pinout)

I know they dropped GCC's GUMBY/MINI chip, but that was due to GCC owning the IP with the related R&D costs and such, Atari Corp owned POKEY's IP and it was still in production. (granted they had an existing stockpile of standard POKEYs as well, but they'd still use those as long as they were making the A8XEGS in any case)

 

Dual POKEYs would be nice, especially with the greater flexibility to use paired (16 bit res) channels... and hell, with a 24-pin DIP it could ahev made more sense to add to the ST alongside the YM chip (still needed for I/O) unless they had trouble interfacing it with the 68k bus -running it at 2 MHz rather than 1.79 shouldn't have been a big issue -ie the convienient division from the 8 MHz master clock, or 1.6 MHz if that was preferable -unless the ST actually uses faster than an 8 MHz master clock -in which case you'd have more options: 16 MHz would divide to 1.78 MHz. (1 YM plus 1 POKEY could have been really interesting and make PCM playback a bit easier as well -in some cases you might still want the pseudo 8-bit YM playback, but for others plain linear 4-bit playback would be preferable and with 4 hardware channels avoiding mixing and giving the ability to control pitch by changing playback rate alone -plus timers tied to 3 of those channels, and of course for in-game stuff you still likely wouldn't be pushing 4 channel PCM playback, but still more flexible in general on top of the POKEY's chip sounds alongside the YM)

Share this post


Link to post
Share on other sites

The YM chip in the ST does more than sound, that is not the only reason it is there. Having YM AND Amy would be crazy from board complexity and cost point of view I guess.

 

And once the first ST is out the design is locked probably, it's likely to be a time and resource issue. A simple DAC is what they went for and I bet part of that was for compatibility as well as cost issues.

 

Wasn't the ST selling well anyway until A500 dropped to £399? The ST was £299 at one point too before Amiga was discounted down from £499+sales tax. Mac and PC had no real genius solution to sound so I guess they wanted to copy Paula as cheaply as possible...and hence we have the way it turned out with STE too.

Share this post


Link to post
Share on other sites

It wouldn't have anything to do with a PC though: the original AMY interface was intended to use a simple 8-bit MCU with am 8051 specifically referenced (successor to the older 8048 line -used in the Intellivision among many other embedded applications)

 

Nope! The Intellivision uses the CP1610 which is a 16bit CPU and nothing like an 8051.

Share this post


Link to post
Share on other sites

The YM chip in the ST does more than sound, that is not the only reason it is there. Having YM AND Amy would be crazy from board complexity and cost point of view I guess.

 

And once the first ST is out the design is locked probably, it's likely to be a time and resource issue. A simple DAC is what they went for and I bet part of that was for compatibility as well as cost issues.

 

Wasn't the ST selling well anyway until A500 dropped to £399? The ST was £299 at one point too before Amiga was discounted down from £499+sales tax. Mac and PC had no real genius solution to sound so I guess they wanted to copy Paula as cheaply as possible...and hence we have the way it turned out with STE too.

 

 

The way I heard it, the ST had to be made quickly using all off the shelf parts. No custom IC's were allowed in order to speed development and help contain costs. JT wanted the ST out the door as far ahead of the Amiga as possible to try to blunt the impact of it's release somewhat. That's why the Yamaha sound chip was chosen. There was no time or budget to make or buy anything more exotic.

Share this post


Link to post
Share on other sites

In the interview linked, the architect of the chip was asked (indirectly) why it was never produced, and he said it best:

"Amy did not result from mainstream Atari product requirements, unfortunately... Like almost everything at Atari Research..."

 

I don't have any inside information, but I am an engineer and I know how research projects go.

 

Research projects are not products. Research projects usually do cool and amazing things, often BECAUSE they fail to meet most or all requirements of a successful product.

 

Many a hapless investor or silicon valley entrepreneur has seen a research project and tried to ship it. After some killer demos and press releases, it's never heard of again.

 

These are the sorts of 'product requirements' not found in research projects:

Easy to understand (fits the industry mold)

Easy to program, even for beginners

Easy to design into a system

Not a big paradigm shift (see above)

Easy to manufacture, with common/cheap equipment

Reliable in hostile environments: buggy code, blocked vents, static discharge

Requires no tuning, calibration, or exotic components

Generates little/no heat or electrical interference

Well-documented

Inexpensive

 

Hints abound that AMY had problems with the above list. It definitely seems that Atari wanted to ship it, but you only have to miss one of the above requirements and you're doomed to the annals of 'What if...' threads.

 

- KS

  • Like 2

Share this post


Link to post
Share on other sites
The way I heard it, the ST had to be made quickly using all off the shelf parts. No custom IC's were allowed in order to speed development and help contain costs. JT wanted the ST out the door as far ahead of the Amiga as possible to try to blunt the impact of it's release somewhat. That's why the Yamaha sound chip was chosen. There was no time or budget to make or buy anything more exotic.

 

Sounds like you need your ears testing ;). The ST contains four custom chips GLUE, MMU, DMA and SHIFTER. Five if you count BLITTER. Without those custom chips the ST motherboard would have been enormous and too costly to produce in mass production. The ST's major off the shelf parts are a 68K (CPU), WD1772 (FDC), MFP68901 (RS232, OS timers), 6850 x 2 (MIDI, mouse/keyboard) and a YM2149 (sound/Centronics).

Share this post


Link to post
Share on other sites

In the interview linked, the architect of the chip was asked (indirectly) why it was never produced, and he said it best:

"Amy did not result from mainstream Atari product requirements, unfortunately... Like almost everything at Atari Research..."

 

I don't have any inside information, but I am an engineer and I know how research projects go.

 

Research projects are not products. Research projects usually do cool and amazing things, often BECAUSE they fail to meet most or all requirements of a successful product.

 

Many a hapless investor or silicon valley entrepreneur has seen a research project and tried to ship it. After some killer demos and press releases, it's never heard of again.

 

These are the sorts of 'product requirements' not found in research projects:

Easy to understand (fits the industry mold)

Easy to program, even for beginners

Easy to design into a system

Not a big paradigm shift (see above)

Easy to manufacture, with common/cheap equipment

Reliable in hostile environments: buggy code, blocked vents, static discharge

Requires no tuning, calibration, or exotic components

Generates little/no heat or electrical interference

Well-documented

Inexpensive

 

Hints abound that AMY had problems with the above list. It definitely seems that Atari wanted to ship it, but you only have to miss one of the above requirements and you're doomed to the annals of 'What if...' threads.

 

- KS

 

 

 

 

 

Pointless then in investing in R&D projects then if v. little of them ever see the light of day (seems though Warner were very good at splashing money on atari R&D things when they knew full well they really were'nt committed to that tech. reaching the marketplace)

Share this post


Link to post
Share on other sites

The only reason the ST was rushed is because JT inherited all of Atari's debts and the sale cost him most of his personal fortune. Money was going out, nothing was coming in as there was little interest in the A8 without major updates to various aspects of it, all Atari did was change the case of the 800XL and do a 128k version of the same machine and the public were having none of it without loads of cheap games (as opposed to expensive £30 imports from the USA) ;)

 

The reason the ST was decided as their first new major machine was because 68000 was the way to go and 16bit was clearly the future looking at Mac and Amiga really I guess.8086 was DODO as far as technicians were concerned and in 1985 the 80286 CPU probably cost as much as a complete Amiga 1000! So there is certainly an element of truth in that, but everything was rush rush faster faster including CP/M 68k conversion and GEM 68k conversion. We don't complain about GEM we do complain about the Spectrum/Amstrad sound chip. Can't win em all and it was significantly cheaper than Amiga 1000 PC XT and Mac and wasn't the worst of the four so good value all in all for 1985.

Share this post


Link to post
Share on other sites

Pointless then in investing in R&D projects then if v. little of them ever see the light of day

It's not pointless for big/healthy companies. Sometimes you invent important things, that you can put into real products later. Other times you get patents out of it. And it attracts really smart people, that you just couldn't hire otherwise.

 

But it's true that research groups running amok can waste both resources and opportunities. There are enough famous stories of that in the valley, all the way back to Xerox PARC. If you're taking products from research groups and showing them at CES, you're probably about to miss the boat.

 

- KS

  • Like 1

Share this post


Link to post
Share on other sites

The YM chip in the ST does more than sound, that is not the only reason it is there. Having YM AND Amy would be crazy from board complexity and cost point of view I guess.

 

And once the first ST is out the design is locked probably, it's likely to be a time and resource issue. A simple DAC is what they went for and I bet part of that was for compatibility as well as cost issues.

 

Wasn't the ST selling well anyway until A500 dropped to £399? The ST was £299 at one point too before Amiga was discounted down from £499+sales tax. Mac and PC had no real genius solution to sound so I guess they wanted to copy Paula as cheaply as possible...and hence we have the way it turned out with STE too.

 

 

The way I heard it, the ST had to be made quickly using all off the shelf parts. No custom IC's were allowed in order to speed development and help contain costs. JT wanted the ST out the door as far ahead of the Amiga as possible to try to blunt the impact of it's release somewhat. That's why the Yamaha sound chip was chosen. There was no time or budget to make or buy anything more exotic.

 

Last time I looked even the basic ST had some custom silicon. The DMA controller sure as heck isn't an off the shelf part. Neither is the GLUE/Shifter etc. ;)

Edited by frank1974

Share this post


Link to post
Share on other sites

The only reason the ST was rushed is because JT inherited all of Atari's debts and the sale cost him most of his personal fortune. Money was going out, nothing was coming in as there was little interest in the A8 without major updates to various aspects of it, all Atari did was change the case of the 800XL and do a 128k version of the same machine and the public were having none of it without loads of cheap games (as opposed to expensive £30 imports from the USA) ;)

 

The reason the ST was decided as their first new major machine was because 68000 was the way to go and 16bit was clearly the future looking at Mac and Amiga really I guess.8086 was DODO as far as technicians were concerned and in 1985 the 80286 CPU probably cost as much as a complete Amiga 1000! So there is certainly an element of truth in that, but everything was rush rush faster faster including CP/M 68k conversion and GEM 68k conversion. We don't complain about GEM we do complain about the Spectrum/Amstrad sound chip. Can't win em all and it was significantly cheaper than Amiga 1000 PC XT and Mac and wasn't the worst of the four so good value all in all for 1985.

 

Yet the spectrum didn't use that soundchip till about 3 years later. That's another thing people keep repeating. The ST didn't have a spectrum soundchip. It was the other way around.

Technically they're not the same sound chip either. The Spectrum used the AY variant. The off the shelf nonsense Amiga owners keep saying is getting tiresome too.

Share this post


Link to post
Share on other sites

Not at all. The Spectrum 128k+ was launched in September 1985 so the same year as the ST ;)

 

No idea where you got the 3 years later from. Alot of games that were released for the Spectrum and ST had exactly the same music.

Share this post


Link to post
Share on other sites

Not at all. The Spectrum 128k+ was launched in September 1985 so the same year as the ST ;)

 

No idea where you got the 3 years later from. Alot of games that were released for the Spectrum and ST had exactly the same music.

 

Heh. I didn't think the AY was used till the +2 in 87. ;) Looks like you're right. I forgot about the Sinclair model

Plenty of custom silicon in the ST though. It sure as heck wasn't built from off the shelf parts.

Share this post


Link to post
Share on other sites

seeming as though the St wouldn't get AMY, perhaps Atari coin op (atari games) should have sold Atari corp the rights to Quad pokey and got Atari corp to adapt it for A8 and ST use (after all, the coin op, gauntlet uses a 68k and pokey)

Edited by carmel_andrews

Share this post


Link to post
Share on other sites

It wouldn't have anything to do with a PC though: the original AMY interface was intended to use a simple 8-bit MCU with am 8051 specifically referenced (successor to the older 8048 line -used in the Intellivision among many other embedded applications)

 

Nope! The Intellivision uses the CP1610 which is a 16bit CPU and nothing like an 8051.

Sorry, I meant the Odyssey2 ;)

 

 

 

 

The way I heard it, the ST had to be made quickly using all off the shelf parts. No custom IC's were allowed in order to speed development and help contain costs. JT wanted the ST out the door as far ahead of the Amiga as possible to try to blunt the impact of it's release somewhat. That's why the Yamaha sound chip was chosen. There was no time or budget to make or buy anything more exotic.

Time mattered, but whether it used custom chips is tangent to that. Using all off the shelf parts was impractical in a cost/performance respect and on top of that it would be very tough to interface some dissimilar components: thus the ST had the custom SHIFTER, GLU, and MMU in the system alongside the off the shelf parts.

 

However, my question was pertaining to why they didn't add it to later STs, not the launch models. Though for that matter I wonder if they ever went back and considered re-using some components from the Sierra/Gaza/etc chipsets going forward with the ST. (ie cutting out R&D costs)

And repurposing some of the existing 8-bit chips would be another matter as well, but might have been out the window due to interfacing difficulties (and not wanting to add even more to the GLU), plus POKEY's key scanning was limited to 64 keys, not enough for the ST (unless they used dual POKEYs), but lots of cost trade-offs. (including things like relegating a 6502 to I/O and sound coprocessor and used POKEY or 2 POKEYs plus PIA to use in place of the AY chip and hitachi keyboard scanners -I'm not sure the SIO interface would be useful in a standard RS232 serial port either and the POT lines would be wasted though you could cut the pin count of POKEY back to 28-32 pins to save on that, but still that's going to be more board space used than the AY+hitachi chips in any case but would have had the advantage of Atari owning the POKEY IP and already having 6502s and PIAs in stock)

I wonder if they ever even considered that as a possible cost saving option or if the ST was too far along by mid 1984 to even consider shifting the GLU to that purpose. (OTOH they could cut even more of the other off the shelf parts out as well, more so if Atari already had a license for PIA rather than buying it off the shelf -I think they may have and they'd have to for the 6502 core and RIOT given the consolidated that into a custom single chip VCS, and RIOT could be a better choice anyway with the timer capabilities on top of I/O -might find the scratchpad RAM useful too- though POKEY itself had 3 interval timers to use as well -so 6 with dual POKEYs, plus if they owned the license/IP to the chips used they could more reasonably later integrate them into a single ASIC as well as remove the unused features like POKEY's SIO if not used: a 6502+RIOT+dual POKEY IC could have been extremely useful)

And remember in the context of POKEY+6502 you have 4-bit DACs with timers and a CPU for sample playback.

 

But I'm way off of the main topic now. ;)

 

 

I was really just wondering if there was more definitive information on why AMY was never used commercially.

Edited by kool kitty89

Share this post


Link to post
Share on other sites

The YM chip in the ST does more than sound, that is not the only reason it is there. Having YM AND Amy would be crazy from board complexity and cost point of view I guess.

Yes, though that's exactly why I was thinking in the context of a later addition or for higher-end models. (supposedly it added $30-50 to the suggested retail price of the XEM over the 65XE) And again, it would have cut board space if they integrated a DAC with the chip. (ie 1 24 pin DIP rather than 1 40 PIN DIP plus 1 18-pin DIP for the 16-bit DAC, or 12 pin if they dropped to 8-bit output) If they couldn't get the interface working without the need for a separate MCU that could have been a significant problem though. (even using a 6507 would add significant board space... and integrating a 6502 core with AMY would be a lot more engineering than just adding a DAC assuming it would be too much silicon to even be practical without shrinking AMY's die)

 

And once the first ST is out the design is locked probably, it's likely to be a time and resource issue. A simple DAC is what they went for and I bet part of that was for compatibility as well as cost issues.

I wouldn't call a 2-channel DMA chip accepting PCM "a simple DAC" as such... a simple DAC would not be very useful but at least better for PCM than the YM chip, a small array of DACs might be more useful (which is what POKEY is, albeit only 4-bit), let alone other options between a bare DAC and an actual DMA chip closer to Paula. (ie a FIFO buffer attached to a DAC, maybe with multiple hardware channels, or a simple DMA circuit like the Mac) You could use a dedicated CPU/MCU+DAC to drive PCM, but that's not very cost effective unless it can also be used to displace other hardware on the board. (so mainly in the context of the original design -using the existing 8-bit computer and VCS chipset would be interesting possibilities assuming the GLU could be modified to address that -even if using more board space initially they could cut pin count to only use the desired features followed by actual consolidation of the chips -given they owned the IP or licenses to all the chips already with the 6502/RIOT/PIA/POKEY in various combinations you could displace the dual hitachi keyboard chips, YM sounc chip, interval timers -POKEY and RIOT- and potentially more -especially depending on how SIO might be used... or whether POKEY's ADC capabilities were fast/precise enough to be useful for more than polling pots -like digitizing analog audio)

 

Does the ST even use both of the 8-bit I/O ports of the YM2149 or is one wasted? (which would make sense if the YM2149 was cheaper than the 28 pin AY-3-8912)

 

 

Wasn't the ST selling well anyway until A500 dropped to £399? The ST was £299 at one point too before Amiga was discounted down from £499+sales tax. Mac and PC had no real genius solution to sound so I guess they wanted to copy Paula as cheaply as possible...and hence we have the way it turned out with STE too.

After the fact, the best option was the YM2203 with it being fully YM2149 compatible and adding 3 FM synth channels: that would be parallel to any PCM chip (of whatever mechanism) being developed. There were other attractive FM synth chips too, but none that would save as much board space. (YM2612 would probably be the next best to add on top of the YM2149... you could use the OPL2 of adlib but that wasn't nearly as nice and didn't integrate the DAC while the even lower end OPLL/YM2413 was small at 18 pins it was REALLY limited with only 1 user programmable instrument on top of 15 presets -and 3 4-op FM channels integral to the 2203 would be great while keeping board space to a minimum) By 1989 there were some pretty nice off the shelf PCM chips (most notably one by Ricoh used in the FM Towns, Sega System 18/32, and Mega CD) but a simpler in-house solution would be much better.

 

PCs indeed had pretty much nothing (other than Tandy/PCJr and later 48 kHz Tandy DAC) beyond the PC speaker until Adlib came in 1987 (not even AY/YM/SN PSG cards) and others quickly followed. Of course you did have DACs appearing earlier still with Covox selling their parallel port Speech Thing in 1986 and a lot of DIY adaptations following (and the Disney Sound Source and some ISA cards including backwards compatibility), but those were all bare 8-bit DACs requiring CPU resource (albeit better than using a PSG and far better than PWM via the PC speaker -so much so that Adlib probably should have included a bare 8-bit DAC/resistor ladder to the Adlib card back in '87 mapped for parallel port DAC compatibility).

 

But in 1989 (same year as the STe) you had actual 8-bit DMA sound appearing with the Sound Blaster and Covox's Sound Master, albeit only mono and with the SB 1.0/1.5 only up to 23 kHz, but still a very nice addition and including ADPCM decoding. (incidentally using a Intel MCS-51 MCU -branded as Creative Labs "DSP") It took a little while for the DAC to get common use, but by 1991 it was pretty significant (and you also had the Sound Blaster 2.0 with 44.1 kHz playback, Pro adding 22 kHz stereo and finally the 1992 SB16 with 44.1 kHz 16-bit stereo -with the pro also adding a 2nd OPL2 and Pro 2.0/SB16 adding the more powerful OPL3 -though it was never fully taken advantage of and more often treated as a 18 channel OPL2 with stereo panning if supported specifically)

And you also saw Sound Blaster based MOD players appearing in the early/mid 90s. (Aladdin, Jazz Jackrabbit, Quik the Thunder Rabbit, James Pond II, Turrican II, etc, etc) Technically speaking, with a DMA sound channel and a decent CPU (386 or maybe even a fast 286) it should have been very reasonable to have software mixed MOD players sooner too, not sure why they waited unless most developers felt FM+SFX were good enough alone and the added CPU resource was better spent on graphics. (but that still wouldn't explain the lack of simple interleaved or even single channel music samples alongside sfx and in addition to FM -no reason to waste FM, though all MOD player using games seemed to drop FM entirely -percussion was one of the weakest points of FM in particular, especially with 2-op only, and offloading that would give more options for additive synth/harmonizing using paired FM channels)

 

 

 

 

 

The way I heard it, the ST had to be made quickly using all off the shelf parts. No custom IC's were allowed in order to speed development and help contain costs. JT wanted the ST out the door as far ahead of the Amiga as possible to try to blunt the impact of it's release somewhat. That's why the Yamaha sound chip was chosen. There was no time or budget to make or buy anything more exotic.

 

Sounds like you need your ears testing ;). The ST contains four custom chips GLUE, MMU, DMA and SHIFTER. Five if you count BLITTER. Without those custom chips the ST motherboard would have been enormous and too costly to produce in mass production. The ST's major off the shelf parts are a 68K (CPU), WD1772 (FDC), MFP68901 (RS232, OS timers), 6850 x 2 (MIDI, mouse/keyboard) and a YM2149 (sound/Centronics).

Ah, yeah forgot about DMA... But was the YM2149 really only used for the parallel port and floppy signaling? The only way I could see it being used in such a case is if Yamaha undercut even the AY8912 (28 pin single 8-bit parallel port).

 

And it wouldn't have been enormous, but they'd have been limited to buying other off the shelf video ASICs and such... other than the GLU which was needed to mate the components using incombatible bus designs, but they saved bus space and improved cost/performance certainly. (though it also implies they could have used it for the mouse/joy ports at least for the first 4 I/O lines on each port -ie use the keyboard scanner only for button presses)

There's also the point where custom chips are off the shelf parts too: like in '79 the TMS9918 and SN76489 were custom chips designed for the TI99/4A but by a couple years later they were being used in 3rd party designs. (Colecovision, SG-1000, MSX, etc) Though I think the AY8910 was a commercial product from the start. (better than the SN76489 anyway -that's right, the Intellivision has a better sound chip than the CV ;) -and it's got the I/O functionality like that of RIOT/PIA)

 

The Apple II and CoCo were pretty much off the shelf, at least I think early Apple IIs were all TTL stuff... the CoCo used a bunch of LSI ICs but all were off the shelf (SAM, 6821 PIA, etc) from Motorola and a bare 6-bit resistor DAC for audio. (granted better than the beeper of the Speccy and Apple II -and not forcing PWM to do more than square ways... a shame there weren't interval timers to drive square waves like the PC speaker -and with a 6-bit DAC allowing some simple software volume and mixing -if you had multiple timers)

 

 

 

 

 

The only reason the ST was rushed is because JT inherited all of Atari's debts and the sale cost him most of his personal fortune. Money was going out, nothing was coming in as there was little interest in the A8 without major updates to various aspects of it, all Atari did was change the case of the 800XL and do a 128k version of the same machine and the public were having none of it without loads of cheap games (as opposed to expensive £30 imports from the USA) ;)

I thought it was bought with loans (promisary notes) to be paid back if/when the company became profitable again but TTL (Atari Corp) taking on the liability of the debt in any case.

 

The ST had to be rushed to get out for the time desired, in the mean time they were banking on Atari Inc's existing products supporting them alongside the ST and especially before it was in full swing. (2600 and A8 mainly -7800 was delayed by contention with Warner, 5200 wasn't seen as viable)

 

The RBP was designed to fill a hole in the Market Tramiel saw and was worried the Japanese would fill: a low cost high performance computer... Commodore's management of the Amiga certainly didn't fill that role. Tramiel probably would have bought Amiga if they'd had what he needed, but they only had the chipset, not the name/distribution/manufacturing network that Atari provided. (had he still been with CBM, he'd probably have bought Amiga outright under the same circumstances in 1984)

How the Amiga/C64/VIC would have been manged from '85 onward under Tramiel would be another story entirely though. ;) (and whether the 128/Plus/4/C16/C64GS ect would ever have existed as such at all -integrating parts of the Amiga chipset with the C64 back in '85 would have been interesting -ie something more like the C65 rather than the C128, let alone if he actually took interest in backwards compatibility -ie include/integrate the C64 chipset with the Amiga with compatibility with the C65/etc as well, though the VIC to C64 didn't point to such an interest in backwards compatibility even when the design facilitated it fairly reasonably, though the C64 was far more popular than the VIC... in any case the C16/Plus4 didn't make that much sense and the timing of the C64GS was all wrong and the design was mismanaged in any case -any such game console shouldn't have been directly compatible with the C64 but by '90/91 the Amiga was a FAR better choice for a console)

 

 

The reason the ST was decided as their first new major machine was because 68000 was the way to go and 16bit was clearly the future looking at Mac and Amiga really I guess.8086 was DODO as far as technicians were concerned and in 1985 the 80286 CPU probably cost as much as a complete Amiga 1000! So there is certainly an element of truth in that, but everything was rush rush faster faster including CP/M 68k conversion and GEM 68k conversion. We don't complain about GEM we do complain about the Spectrum/Amstrad sound chip. Can't win em all and it was significantly cheaper than Amiga 1000 PC XT and Mac and wasn't the worst of the four so good value all in all for 1985.

Not so much... the ST was in development (RBP) before Atari Corp existed (or rather under TTL before it was renamed Atari Corp). Tramiel acquired Atari Inc's consumer holdings specifically because of the ST/RBP design and needing the resources/name/network to bring it to market.

 

Atari Inc was already set on the 68000 and had been since ~1983 with several advanced chipsets being developed (AMY eventually becoming part of that) along with a Unix based OS and "Snowcap" GUI. Warner initially took strong interest but pushed back when they decided they really wanted to focus in the consumer market more than the high-end. (the 68k machines had predominantly been configured as high-end workstations, albeit in 1983 even something like the ST would have been high-end and there was a range of how high-end the Atari machines were, but that was also up to implementation and NOT the chipsets themselves -like including 2 or 3

 

As for the x86 architecture, aside from annoying design from a proramming standpoint (albeit, technically, the 64k segmentation could have been avoided entirely if external bank switching had been used -ie using friendlier 4 or 8 kB banks rather than 64k -though INTEL could have designed it to have such modes in the first place), but in terms of lower-end cost/performance, while the 286 would have been very expesive initially (from a manufacturing standpoint not nearly as much as you point to, but what Intel priced it at would be a different matter, though I highly doubt anywhere close to what you're implying... and not more than a 68020 obviously -less silicon, older release, and much lower pin count similar to the 68k), but aside from that you had both the 80186 (with some improved performance and higher max clock rate) but that wasted some silicon on integrating peripherals which didn't match IBM's lay out (so it was OK, but not as cheap as it could have been if omitting the integration), but for non PCs or only semi-compatibles that was not a problem in any case.

Then you had NEC's x86 compatibles with similar ISA and speed enhancement to the 186 (hardware multiplier) and almost certainly favorable cost over Intel while also offering up to 16 MHz 8088/8086 compatible CPUs. (and interesting bonus was the inclusion of an 8080/Z80 compatible mode, so attractive as a successor for Z80 based systems -NEC did that in their own PC8801) NEC later introduced the 80286 comparable V33 though that dropped the Z80 mode. So what the x86 chips lacked in per clock performance, they could have addressed (to some extent) with higher clock speeds plus the ~30% performance boost the 186/NECV20 added over the original 8088/86. (plus you have the inherent cost advantages of the x86 chips using less silicon and having a low pin count -though the latter changed with the 186 but not the NEC chips) NEC later introduced PC compatible embedded versions of their V20 family. (integrating PC compatible components/peripherals unlike the 80186)

 

The fundamental architecture of the 68000 from a programming standpoint if far nicer though and allowed a cleaner path to successors. (though Motorola kind of screwed themselves with the closed architecture policies with later 68k CPUs -ie no licencing/2nd sourcing like the 68k making it tough to compete with Intel, let alone due to the PC's success driving x86 -probably one reason the ST and Amiga took so long to get 32-bit variants, though not explaining the absence of 12 or 16 MHz 68ks)

Edited by kool kitty89

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
Sign in to follow this  

  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...