Mostro

I understand- but with respect, that's the sort of thing I had in mind covered by "niche uses". FWIW, I'm definitely aware of Spectrum 512, since I remember reading about it in the late '80s (long before I had a 16-bit machine myself). It looked like a nice trick, though from what I remember and what I've read since it was very CPU intensive, had some limitations (e.g. 40 colours per line) and probably wasn't suited to games. Still, I'll admit that some of the best examples I've seen of it were definitely comparable to HAM on the Amiga. Not that HAM itself was very suited to games either- the display wasn't CPU intensive, since it was hardware-based, but altering the image was still complicated; that's likely why even Amiga games stuck to the standard 32 colour mode (or 64 colour EHB on rare occasion).

Have to admit that's fairly impressive!

I thought the STE's increased palette was pointless even at the time- it still had the same 16 colour onscreen limit (in regular use) as my STFM, which was always the main limiting factor by far. So why bother? (IIRC even the ST magazines themselves thought this.) Pretty sure that if anyone had been offered the choice, they'd much rather have had 32+ colours even from 512, rather than 16 from 4096. (Yes, I know the former would likely have been harder, which probably explains why Atari went for the easy "enhancement"). I've no doubt if you wanted to create true 16-level greyscale images in Neochrome it might have offered a minor improvement- as well as marginally smoother gradients on games and possible other niche uses- but in general it appears to have been a "looks nice on paper until you think about it, useless in practice" improvement. The lack of support "vicious circle" I mentioned in my original post. The only way they could have made it worth commercially supporting- and extended the ST range's lifespan- would have been to sell it for £299 at launch in late 1989, replacing the STFM and ensuring a respectably-sized STE user base by default. No-one was going to buy a slightly improved ST with hobbled improvements and no support when they could get an already-supported Amiga 500 at the same price (£399). So the declining number of people still buying STs then (e.g. myself) did so- I assume- on price and got the rapidly-dating STFM. Atari reduced the STE to £299 in mid-1991... too late for anyone to care.

Yes, that's what I touched upon in my post- we're probably arguing more from an "end user" point of view, whereas Nebulon appears to be discussing it in the more demanding terms of being a production tool. I wouldn't have tried to use my no-name ISA soundcard in a professional recording studio, regardless.

Interestingly, even the BBC Micro conversion- almost certainly commissioned for the UK market (#)- used the "Mountain Dew" sign, despite the fact it wasn't sold here at the time. Playing Tapper on my friend's BBC B in the mid-80s was first and the only place I'd heard of it. IIRC I didn't even know what it actually was until they launched it in the UK in the mid-90s. (##) Anyway, I guess what I'm saying is that "Mountain Dew" will always be connected to Tapper for me, so I'd rather have that sign anyway. Sorry! (#) The BBC Micro's market share outside the UK was negligible; even here, it was pretty "second tier" in terms of game support, being primarily successful in the education field. (##) Then withdrew it again a few years later. They reintroduced the name for an energy drink a while back, but it's definitely not the same product. (That's assuming the 90s "Dew" was the same as the US version anyway.)

Er... I mentioned 1998 because you did!.... My point- in response to this- was that since even the lower-end PC I'd bought earlier that same year already included a generic soundcard with support for 16-bit sampled audio, I'd assume this sort of thing had been available on (at least) more state-of-the-art PC soundcards for years by this point. I don't know how many channels it had or what the limitations were. I doubt it was anything other than an anonymous manufacturer gluing together a bunch of chips designed for mass-market, playback-oriented use (ESS ES1869F, ES689F and ES981P- IIRC the first was the important one), but the fact that even a cheap, plain PC's, cheap generic ISA(!) card nominally supported CD-resolution sampled sound suggested that the "revolution" had already taken place. From what I can see on its Wikipedia page, the Soundblaster Live does appear to have been a step forward, and I suspect you're judging on different criteria than I am- I'm going to assume the SB Live's DSP made it more powerful and flexible- and may be what you're comparing to the Falcon. But, as I said, that's where 1998 came from.

If you're talking about internal audio, I think it had happened long before then. My first PC, which I bought in spring 1998, was a bland upper-entry-level model with a conservative (albeit not stingy) spec, and even that came with a generic 16-bit soundcard with no obvious limitations I remember. Was that common on almost all models, though, or just the higher-end ones at that time?

The basic principle- as far as I'm aware- is that the 16-bit input sample is re-rendered with "dithering" as 1-bit sample at a much higher frequency (somewhat akin to converting a low-res greyscale image to a much higher res two-colour one). This is then fed to a literal 1-bit (on/off) DAC, and the output from this is filtered to merge adjacent samples. (#) This should- as far as I can tell- give output equivalent to a regular 16-bit DAC. As far as I'm aware, the reason is that this makes the DAC much simpler, as well as having other advantages in design terms. I'll go no further than that- Wikipedia is probably a more reliable source than me. :-) (#) Which- since it takes the output of the DAC- is presumably in the analogue domain?, and uses a low-pass filter or something similar to "average" the samples (analogous to slightly blurring the high-res two colour image to restore the lower-resolution greyscale).

Oh, I've no doubt about that. I don't claim that what I said was the be all and end all. Tricks like- e.g.- adding together the "volume level" (i.e. 4-bit ersatz PCM position) for multiple channels to get better than 4-bit resolution are quite clever, and I'm sure there are other sneaky tricks and exploitation of the chip's characteristics. If I'm wrong that the "jagginess" one would expect from unsmoothable bit-bashing was at least partially the cause of the "tinniness" of ST sample playback Zip referred to (or other artifacts) then fair enough, but I'd be surprised if it didn't have an effect, even if there are other factors in play too. That said, I was starting to bog myself down in technical sidetracking in that last post, so I think I'll leave that for now.

With respect, I'm not convinced that I am- at least, not in the specific aspects you raise. I think you're either miscontruing what I said... or I explained myself badly, which is just as likely(!) I'll admit that from what I- vaguely- remember, and have seen on YouTube (assuming the playback is via a standard ST), the sound quality is still better than the harsh impression I gave above. It was just an attempt to explain where the "tinniness" might come from, which I assume is related to the sharp corners, and the spurious high frequency artifacts one would expect to be associated with such distortion of the original waveform. You've got that the wrong way round- I specifically mentioned oversampling in the footnote because I wanted to qualify that what I'd said didn't extend to general theoretical principle. (And yes, it was the 1-bit DACs in CD players I had in mind). At the same time, it was relegated to a footnote because I didn't believe it could be usefully applied to the situation under discussion (i.e. bit-bashed sample playback on a standard ST). I don't feel the analogy comparing 1-bit DACs with the ST's 4-bit audio is legitimate. As far as I'm aware, 1-bit DACs rely upon- as a fundamental and integral aspect of their design- post-processing and/or low-pass filters to accurately reconstruct the intended signal. The ST soundchip- of course- has nothing like this, because it was never designed with sample playback in mind, let alone oversampling! And I don't know exactly what the audio output of a 1-bit DAC (which I assume would effectively be a pulse width modulated signal?) would sound like if it was fed directly to an amplifier without these steps. I assume that much of the distortion would be way, way above the range of human hearing, and that the amplifier itself would effectively filter out frequencies much higher than it was designed for. Yet, while I'll guess that the signal unintentionally filtered in this way might more closely resemble the intended waveform than the purely rectangular stream of output from a 1-bit DAC, it would still have enough audible distortion to make it less then pleasant to listen to. Yes, I'm aware of the Nyquist-Shannon theorem; or at least its basic principle. FWIW, I had that it in mind when I said that "I'm guessing that moderate artifacts introduced in this manner were probably tolerated in preference to the muffling that'd be caused by completely low-pass filtering the source(?)". In other words, I knew that one could (theoretically) avoid artifacts by filtering out everything above half the sampling rate, but I suspected that in practice, that might have been intentionally ignored- assuming the people in question understood Nyquist!- if any moderate artifacts introduced were preferable to the loss of top end on the sample. Also, bear in mind that Nyquist only states what's theoretically (and mathematically) possible. In practice, it's not always easy- or even practical- to do things that perfectly. For example, it's very difficult- particularly with analogue filters- to perfectly remove all frequencies above a certain point while leaving those below untouched. The cutoff is generally more like a curve/slope than brick wall, so you have to either filter well below the desired frequency to avoid some higher ones getting through, or tolerate some above the limit- and the risk of artifacts- anyway. In addition, I've no idea what- if anything- the Nyquist-Shannon theorem says about the effect of quantisation on the individual point samples (i.e. the digitisation of the original theoretically "infinite" analogue value to quantised 4, 8 or 16-bit values). I'm guessing it says nothing at all- because that's not what it's supposed to be about!- but nonetheless, there must be some effect on the result if the resolution is low enough.

If I remember correctly, the Falcon was mentioned in one of the Amiga magazines the same month they announced the A4000 (i.e. the first AGA Amiga). I'll admit that I do remember thinking some of the specifications looked pretty nice. (Bear in mind that- IIRC- the affordably-priced A1200 hadn't been announced yet). Unfortunately, I also remember knowing- even at the time, and without the benefit of hindsight- that it didn't stand a cat's chance in hell. Firstly, it was clear that Atari couldn't market their way out of a paper bag. Secondly, the ST line- of which the Falcon was still effectively a descendant- had been in decline in the face of the Amiga since the end of the 1980s, even on the UK market. By this point was practically yesterday's news. At a time when it was starting to become clear that the tide was turning against even the Amiga and in favour of the PC.... Well, if there was any company that could achieve the miracle of persuading the mass market to move back to the ST, it sure as heck wasn't going to be Atari. So, interesting looking machine. Not remotely surprised that it didn't go anywhere, or that they (apparently) ditched it a year later.

I might have been thinking along the right lines then- that "tinniness" could well be the square-wave flavoured "harshness" I had in mind. Almost forgot one thing I'd read which is going to make this problem much worse (along with explaining why sample playback on the vanilla ST is mediocre). From what I've read, the basic bit-bashing technique on the original ST only supports 4-bit (i.e. 2^4 = 16 level) audio resolution (#) rather than the 8-bit (256 level) audio of the Amiga and STE(?) or Compact Disc's 16-bit (65,536 level). This image shows just how horribly coarse 4-bit is. It means that beyond a certain point, high sample rates (i.e. the "horizontal" resolution) aren't going to buy you higher quality if you still only have sixteen output levels (i.e. the "vertical" resolution) to jump between. (##) So, in addition to the poxy resolution, there are going to be major, unsmoothed jagged edges (###) with sample playback on a regular ST, and that's going to have a major negative effect on the quality. I've heard of people combining the output level of multiple channels (and other aspects of the chip) to go above 4-bit quality, but I'm pretty sure it's never going to match 8-bit, hardware-based sample playback. (Besides which, wouldn't it effectively reduce you to one audio channel?) (BTW, if anything I've said here is incorrect or misleading, feel free to point that out). --- (#) Assuming the process is similar to sample playback on my 800XL's (also square-wave based) POKEY soundchip, this'll be because the output level per channel can only be set to one of sixteen positions. AFAICT these corresponded to the 16 volumes of square wave it could generate in normal usage (presumably by toggling abruptly between zero and the desired position at the correct frequency). (##) This assumes that oversampling isn't possible. (From my limited understanding of the subject, this wouldn't be possible on the ST since it would require hardware-based filtering of the output on top of much higher sample rates.) (###) Smoothing them off effectively being what a low-pass filter would do here- something which the standard ST soundchip doesn't have, since it was never designed for sample playback.

Ah, good point- I should have thought of that. Though if that was the intent, was it really good design to have required it to use bank switching for more than that minimal 128KB in the first place?

If the analogue source contains any frequencies too high for the (low) sampling rate, that would cause artifacts, but that'd be an issue for the sampling hardware- which might well not be an ST anyway- and the sample itself. Then again, such sampling would probably be done with playback limitations in mind. And I'm guessing that moderate artifacts introduced in this manner were probably tolerated in preference to the muffling that'd be caused by completely low-pass filtering the source(?) I'm still pretty sure that the "jaggies" caused by unsmoothed bit-bashing would cause some unpleasant high-end harshness, though I'm not sure whether it would translate into hiss? I could waffle further on this, but I don't intend boring everyone with (potentially misinformed) guesswork based on my half-baked- well, quarter-baked!- understanding of sampling theory.

It's not really "funny", it's quite sensible to borrow good techniques, and not surprising that they might have originated on the ST. The limitations of its inbuilt sound hardware would- I assume- have provided a greater impetus to overcome them, whereas the Amiga's audio limits (which undeniably existed) weren't that much of an issue until later on, because the early software didn't really fulfil its potential. (Now that you mention it, it would have been nice if latterly they'd upgraded the soundchip to support hardware-based panning, since I assume that wouldn't have required a major redesign.) Ah, now you are arguing semantics. Technically, that's correct. However, when I talked about doing something "in software", I meant purely using the machine's existing processing functionality to accomplish a task- rather than relying upon additional *hardware* intended to aid it in that purpose. The DSP is a piece of additional *hardware*, and while it might- in theory- be turned to other, more general uses, it's quite clearly designed with digital signal processing in mind. (The clue's in the name ). My point was that the Falcon is somewhat different to the STE, because it has additional hardware for that sort of processing. (But not forgetting that the STE *does* have hardware-based sample playback- it's just that we're discussing doing things "in software" because of the playback rate limitations).

I know it probably pales next to some childhood traumas, but that's still pretty sad- if only because I can imagine how I would have felt in that position. I mean, I appreciate it was probably an oversight and not intentional sadism on their part (since they bought you an ST!), but wasn't it possible to get even a dirt-cheap joystick from somewhere?

As far as I'm aware, the cartridge slot was limited to 128K. Granted, that would have left the RAM free, but it would still have been a serious limitation for games.

Badly phrased in hindsight, as that's not what I meant to imply... particularly as I had- and remember- one of those MOD trackers myself when I owned a 520STFM in the early 90s! (I think I was trying to say that it *was* technically possible- but not by design and with limitations.) I don't recall hearing that much use of samples during actual in-game play, though. I'll admit that I only had my ST for a year or so (and not that many games). However, from what I've seen on YouTube, most of them still went with the horribly dated square wave sounds the chip was designed around. I don't remember it that clearly, but I'm guessing this may have been because- since the STFM's bit-bashed samples wouldn't have had the advantage of a hardware cutoff filter for high frequencies- the jagged edges of the unsmoothed PCM would have resulted in aliasing and noise.

I assume you mean rendering sample fragments in real time? I'd argue that's still "multiple samples"- it's just that you're creating them "on the fly" rather than holding them in memory. That said, I'm nitpicking semantics , so let's skip that and get onto the real issue... I'm sorry, but this is trying to make a virtue out a limitation. If it had been so great, there was nothing stopping you from doing the same (very CPU-intensive) software mixing on the Amiga. Indeed, I was actually going to use OctaMED as a counter-example until I saw you'd already mentioned it! It was clever, but I remember it slowed down the computer quite significantly in 8 channel mode. In theory, even the regular STFM could play back sampled sound to not-too-far-off-the-Amiga quality by crude "bit bashing" of the 8-bit era soundchip with PCM values. (#) But you didn't see that often in games, presumably because the CPU load- and strict timing required- make it impractical. And the "real time re-sampling" solution to the STE's limitations has the same problem. The need- as opposed to the choice- of carrying out CPU-intensive software-based mixing to get around the limitations of the hardware in the STE doesn't indicate an advantage. Quite the opposite- it's a serious limitation, pure and simple. (#) Technically, most of the "square wave" 8-bits could manage sample playback in a similar manner; I wrote a routine to do it on my Atari 800XL, and I was never an expert programmer. If a machine is using a DSP like the Falcon did, I wouldn't consider that "software mixing"- it's dedicated hardware designed to take the load off the CPU.

I don't recall reports of a shortage in the UK, only of the price going up. According to this post, Tramiel gave preference to European markets because they were more profitable:- So I'm guessing that if he was having trouble sourcing affordable RAM, the machines he could produce were more likely to end up here.

I'd assume it would use the blitter- or any other hardware support- if it's available. (IIRC, didn't TOS on the ST also automatically use the blitter when executing standard OS calls to the graphical subsystem if installed?) My point was that the OS didn't seem to be *reliant* upon these advantages; i.e. while graphical updates might run a little slower without the blitter, I'd still expect the OS as a whole to be workable. Indeed, I might be wrong, but I'd assume that it should be able to be ported to the ST hardware without any fundamental changes to its design.

I vaguely remember that the STFM- which had come down to £299 in late 1987- went up to £399 the following year due to increases in the cost of RAM. (Atari tried to mitigate this by bundling it with a lot of games.) This article confirms that RAM prices went up massively in 1988; ditto this contemporary report. Perhaps you're thinking of that? I've no idea if you're right about the STE or not- but unless it used completely different memory chips to the STFM (and couldn't be easily redesigned to accept different ones), surely any increase in RAM prices for the STE would have affected the STFM as well? The Wikipedia article doesn't indicate a "chip famine"(!) circa 1990 or 91. Also, by 1990-91, wouldn't prices have fallen enough in general so that even a (proportionately) similar increase to the 1988 one wouldn't increase the cost of 512KB by as much as it had then?

There's actually more AGA-specific software than you (or I) might have realised. (Bear in mind that the CD32 console was essentially an A1200 plus CD drive and "planar" chip, sans keyboard.) Some of it came out surprisingly late- I was surprised to find out recently that there had been an AGA-specific version of Street Fighter II Turbo released in 1995. (#) But- like the A1200 itself- it came out just a bit too late for most people to shift their attention back to the Amiga. The A600 on the other hand? An indefensible debacle. Most accounts make clear it was intended as the budget "A300"- which might have worked- but for some reason they positioned it as the successor to the A500 at the same price, despite the core spec being unimproved and inferior in many ways. At just about the point (it's clear in hindsight) that the Amiga was starting to prove vulnerable, this utterly pointless muddying of the waters would have given the impression that the format was stale and dated. The "new" Amiga? Ha ha ha. The A1200- more obviously the true successor to the A500- came out six months later, but it never attained the same status and looking back, the A600 was probably the Amiga's "jump the shark" moment at which it lost its former desirability and never really recovered. (#) To put this in perspective, C= were already bankrupt by this point- they went under in 1994, and after more than a year of inactivity, Escom's plan in late 1995 to relaunch the (by now three year old and unimproved) A1200 at £100 more than it was selling for before the bankruptcy(!!!) was so obviously destined for failure that even I knew it was dead.

The Amiga was also a victim of lack of development. It was so utterly state-of-the-art (and expensive) when it launched in 1985 that the barely-improved (and still 7.14 MHz 68000 based) architecture was still desirable five years later. But ultimately I think this led Commodore to get used to resting on its laurels and the need to plan ahead- so that when everyone else *did* start to catch up, there wasn't really enough inertia. The AGA-based A1200 and A4000 were a nice improvement, but evolutionary rather than revolutionary, and came out just a bit too late to reverse the trend. Had they come out 18 months earlier, the Amiga might have survived longer. Might. From what I understand there were quite a few talented designers working at C=, who were quite proud of the high-end Amiga 3000 from 1990. That seems to have been the closest it came to a significant improvement in its heyday. But nice though its design may have been, I'm not sure it had much that could have led the way forward in mass market terms. Perhaps Commodore (like Atari) was just too small- and too cynically run- to take on the changing marketplace and the threat from commodity PCs and better-equipped console manufacturers. Indeed, that's why I thought the STE *had* to entirely replace the STFM when it first came out- to ensure that there were enough people with the STE (by default) to make it worth supporting and overcome the "anti-network effect" (i.e. no software -> no buyers -> no software -> no buyers...), whereas continuing to sell the STFM *increased* the proportion of lowest-common denominator STs. Early Amiga games had a reputation for being straight ports of the ST version not taking advantage of the enhanced sound and graphics. However, from the late 80s onwards, this applied less and less, and most games took advantage of it. You couldn't have been sure that was going to be the case even if the STE had been as good as the Amiga, so which machine would most people go for at the same price point? Enough said. That sounds about right. I'll also note that the thing Tramiel got right with Commodore in the 8-bit days was the vertical integration that gave them a huge advantage in both price terms and ability to design and manufacture (e.g.) the C64's custom chips. When he fell out with the C= board and formed Atari Corp, he didn't have the same advantage. Perhaps that was also a factor? I've heard it argue that Tramiel's approach didn't shift with the changing market. In the late 70s and early 80s, Commodore practically had people coming to *them* to find out about- and buy- their computers. I've also heard it said that Tramiel was a hardware- not software- guy. (#) This worked for him first time round, and he didn't feel- or understand- the need to market the computers or that- by then- software sold hardware. At the time- though I might not have admitted it as an Atari fanboy- and more obviously in hindsight, Atari's marketing and software development felt utterly shoestring and third-tier compared to anything their "rivals" were doing. Even in the early 90s, the technically-brilliant Lynx was marred by crappy marketing and "knock off unlicensed soundalike" game titles reminiscent of small companies in the early 80s (e.g. "Dirty Larry" (cough), "Blue Lightning" (Afterburner clone, name obviously nicked from Blue Thunder, etc.) The XEGS was originally meant to be released in Europe instead of the 7800, but we eventually got the latter here anyway. Why? They were competing against each other and going nowhere even compared to Nintendo or Sega. (Despite criticism of the XEGS, however, there's a very interesting Usenet post from the late 1980s that suggests it was never quite the "sad attempt to compete with the NES" it might have appeared, but a slightly more knowing attempt to "trojan horse" the 65XE in disguise to retailers who were more interested in consoles). Contrary to what some US-centric viewpoints claim, the ST *was* pretty successful for a while in Europe during the mid-to-late 80s, but overall Atari Corp gives the impression of a damp squib. (#) Which might explain the lousy BASIC and OS in the C64- that, and the fact they wanted to keep taking advantage of the dirt-cheap royalty deal they'd got on the original, but dated, Microsoft BASIC!

That may have been the case in the US. However, AFAIK the ST and Amiga never really took off there in the first place- you can't lose what you never really had!- and since they enjoyed their greatest success in Europe, the US-centric perspective may be unintentionally misleading. The threat from Nintendo? Not in Europe, so much. Here's the thing. During the late 80s, the NES was never remotely close to being as popular in the UK as it was in the US. It wasn't a flop by normal standards, but it wasn't that big either- it even was outsold by the Sega Master System- and neither console came close to unseating the home computer formats as the NES had in the US. (#) It wasn't until the Game Boy came out here circa 1990 that Nintendo really took off. As far as I'm aware, there was a similar pattern elsewhere in Western Europe, not helped by Nintendo's well-known indifference to European markets in general compared to the US and Japan. So, no. Nintendo wasn't the cause of the ST's decline in its largest market around the turn of the decade; that was more obviously due to the Amiga. The PC? Yes, they started to take off here when Amstrad launched affordable PC clones from 1986 onwards. But those were never really popular among gamers, who still tended to go for the ST initially, and then the Amiga later in the 80s and during the early 90s. Eventually the Amiga itself did start to decline in the face of the Mega Drive- and later, SNES- on one side and increasingly cheap (and higher-specced) commodity PCs on the other. But that was from 1992-93 onwards, long after the late 80s US market you describe. I'd say that the European market appears to be far more similar to the US one nowadays, but please don't assume this was always the case. (#) This article makes essentially the same point that I've been railing about for years regarding Nintendo-centric rewriting of history.