Guest post by The Amiga has what is with no doubt in my mind, the absolute finest sound chip inside of any computer or console throughout the 1980’s as well as most, if not all of the 1990’s. Full disclosure; I have an MT-32… And the Amiga can actually do a piano. Yes, in a time when the vast majority of IBM and compatible PC owners were using a small speaker stuck deep inside of a metal tomb, Amiga users had a quality of sound nobody else could touch for that price.
[…]
To combat the story that has long been shaped that the Amiga was not popular to musicians because it did not have built in MIDI connectors I give you this quote given directly to me from the creator of the sequencing program Music-X, Talin:
“The story with MIDI is actually much more complex than most people realize. You see, the early Amiga models had a hardware bug which made the serial port unreliable at high data rates. Basically the problem was that the serial port hardware had only a one-byte buffer, and if you didn’t grab that byte before the next byte came in then data would be lost. Unfortunately, the Amiga’s four timer chips would generate a software interrupt at regular intervals, during which time the serial port could not be serviced. And while MIDI speed wasn’t super-high, it was high enough that you’d get a dropped byte every 10 minutes or so depending on how many notes you were sending over. Note that this did not affect the higher-end MIDI adapters which had their own dedicated serial point, but those were considerably more expensive.”
Interesting article about past MIDI challenges with the Amiga and how to hook up a modern synth to an Amiga to make music.
Interesting! I didn’t know about Amiga’s serial port problem… now I understand why nobody in the music industry used the Amiga!
Back in the day Atari ST was the gold standard among musicians, I remember people using it even in the middle/late 90s running Cubase… paired with a SM124 monochrome monitor It was pretty cool. Then everybody switched to the Mac or the PC, but the ST was huge.
Despite what the article implies, it wasn’t a hardware bug. There was nothing wrong with the hardware, it just didn’t favor high-speed serial without making some choices about how you structured your program. If having a one byte receive buffer was a bug, than the ST had the same “bug” as the 6850 in it also only had a one byte receive buffer. So what was the difference that made one better than the other? The interrupt priority. On the ST, the serial receive buffer full interrupt was nearly the highest in the system, which is good when you’re dealing with just over 3100 interrupts per second. On the Amiga, the serial receive buffer full interrupt was fairly low priority. That’s not a bug, it just means that the higher the serial rate, the more you have to pay attention to interrupt handler latency, something many programmers don’t even realize exists, much less know how to deal with. Throw in DMA contention if the programmer was trying to use a 16 color high res display and you can see where serial on the Amiga could occasionally drop a byte if you didn’t deal with it somehow. None of that were hardware BUGS, but limitations programmers would have to keep in mind when writing serial interface code for the Amiga. It was simply less hassle to deal with the ST.
By the way, the Z8530 on the Mac had a three byte receive buffer, making it much easier to do higher speed serial without losing any data. Even less hassle… just a much more expensive machine.
I get what you are saying about the interrupt, but I bet the various DMA channels that the Amiga had could also throw off the timing at lot more.
There was a lot of things going over the memory channel.
Edited 2017-03-29 01:22 UTC
That’s why I mentioned 16-color hires display – that used all the ram access slots on the chip ram for DMA, so if you tried to access chip ram during an active line, the cpu would be held until the line was done. Good high-speed serial was easier with fast ram, but you still had to be careful about chip ram accesses. You’d really want to make the display no more than 8-color so you’d have half the access cycles available for chip ram.
Four color would be even better, especially since we’re just talking about a music editor, not a game or graphics editor. But again, see what you have to keep in mind while doing serial on the Amiga? It’s not rocket science, but it’s still easy to do SOMETHING that makes missing bytes occur randomly with fast serial.
Indeed, and there are other minor mistakes in the article too – there aren’t 4 timing chips on an Amiga, there are two for example.
I did not know about that either. That could explain the rare but odd graphics I got on some BBSs that had a graphic interface.
The ANSI graphics would be wrong one time, but I never could repeat the same error on any single page.
It could be… or it could have been line noise between your modem and the BBS. It’s not like most BBSes used error correction on anything other than file transfers. Something like the display didn’t rate the overhead.
Not to disrespect the Amiga, because it did invent mod music after all, but I just can’t get past the guy saying it was the best chip throughout the 90’s.
Gravis Ultrasound was released in 1992. With 32 hardware channels versus 4 in the Amiga, its clear that by this point in time the Amiga chip had lost its lead. The GUS also had 16 bit channels, while the Amiga only had 8 bit.
In 1995 there was the Ultrasound PnP using the AMD InterWave chip. Once again 32 hardware channels, but this time with 16 MB of memory for the samples.
Both were somewhat disliked because they didn’t have the adlib synth chip, making adlib/soundblaster games sound sub-optimal, but then neither did the Amiga.
Yeah, the Amiga lost its lead in hardware in the EARLY 90s, not the mid or late 90s… it was dead by then. They actually lost it in the late 80s when they failed to produce an updated chipset for the A500/A2000. You have to be proactive in the hardware market or you get left behind. By the time they made the ECS for the A3000, they were now woefully behind in R&D, and AGA was a last gasp – too little, too late. They should have had something like AGA for the A500/A2000, and AAA for the A3000 to remain competitive on the hardware side of things.
By 94, I had a GUS Max… awesome card for music. Made those 32 track XM songs really shine. Without needing to use all the CPU time, of course. You could play XM on a SB16, but used most of the CPU time mixing the sound channels.
Edited 2017-03-29 15:26 UTC
Yeah, a little strange that one. Perhaps it was meant in the context of standard supplied hardware rather than expensive, specialist add-on cards, in which case it was competing with such cards as the original 8-bit Soundblasters from the ’80s, with which the Amiga hardware could easily compete and outperform in many aspects. Once the 16-bit cards started becoming standard in the ’90s and the Amiga’s chips weren’t updated, they got left behind.
That’s probably what the author was thinking about. The SoundBlaster cards (until the 16) had the FM synth chip for music, and one 8-bit stereo PCM channel that could be driven by one of the onboard DMA channels. The Amiga sound was better than that… not by much, but still better. The SB16 had one 16-bit stereo channel – you could argue that since it was only ONE channel, it was still worse than the Amiga, or at least not much better. It wasn’t until the next gen of sound cards that finally made clear that PC sound cards were much better. You could argue they were still plug-in cards, and built-in audio didn’t get better until motherboard manufacturers started putting cheap AC97 codecs onto the motherboard that PC sound was better without a card, but that’s just splitting hairs.
The Atari STE had better sound than the Amiga, then the Atari Falcon was a blast in the early 90s, probably the most underrated hardware ever. So after 89, I think the Amiga was behind the competition regarding sound.
Meh, bit of a toss up. The STe added four 8-bit PCM channels. What made them better was they were STEREO channels while the four channels in the Amiga were directed to left or right only. However, they were worse in that they could only be set to a few different frequencies, and said frequency applied to all channels. The Amiga allowed a very wide range of rates that were set per channel. Even worse for the STe, all the channels had to be interleaved in memory, making playing different samples on different channels much harder. The Amiga had pointers for each channel. So I’d argue that OVERALL, the Amiga still had better sound than the STe.
I accept your argument, I did not know about the limitations. Yet I think you might agree that the Falcon showed what the next gen Amiga sound and video should have been(even if on a 16-bit bus)
Yeah, the Falcon was a clear step up from the rest. No arguing – it was better. It had 16-bit audio as well as a DSP. I’ve seen some killer stuff done with it, like helping speed up the math for Quake. Too bad Atari kinda imploded about that time, throwing everything behind the Jaguar in the game console market rather than taking the lead in the home computer market.
I love my Falcon, it is the little computer that could’ve, if only Atari hadn’t felt the need to try to get back into the video game market.
I have read things that suggested the Amiga 3000+ that was never released was supposed to also have a built in DSP, at least I am pretty sure that was the one. As others have said, if either of them had released the upgraded standard systems a year or two earlier than they had, they could have lasted much longer.
Falcon in ’91 and AGA in ’90 would have been much better, but both systems always hurt because developers would target the lowest common denomination, which meant things like the Shadow of the Beast port for the ST only using 512kb of ram. My Mega STe with 2mb and then upgraded to 4mb never had anything that would really take advantage of the 16mhz mode outside of desktop applications.
Back on topic a bit, Paula did have a fairly unique flavor of sound from it, so that even to this day you can tell it apart. I think after this article, I should see if I could get some mods and play them on the falcon vs my A4000 to see what one sounds better.
One of the more distinctive aspects of the sound was the rather low frequency low-pass filter on the audio. This was part of the original A1000 from back when 8kHz samples were the norm. Luckily, when they made the A500 and A2000, they also added a way to turn off the filter so that higher sample rates wouldn’t be as muddied: you turn off the power indicator LED to turn off the filter. Better games would do that, and music players often had an option for doing so. There were also a few small apps you could run from Workbench that would turn off the filter that you could run before older music apps and games that launched from Workbench. A few of the hacker groups released cracks for some old games that would turn off the filter as well. Of course, you’d get some aliasing in the sound for some of those games, to which they’d tell folks to turn down the treble knob on your audio amp for those games.
Yeah, the TSR sb/adlib emulator was….underwhelming but as I recall it did get better with later software updates.
I had both a soundblaster and an ultrasound so I could work around it (by dealing with IRQ madness. Ah, the gold ole days…) but it was still a bit awkward.
I personally didn’t find the emulator that bad. I preferred the emulated sound in half the titles over the FM synth chip.
Still, a lot of people seemed to hate the GUS for the fact their games didn’t just work. I guess I can’t blame them if they weren’t that tech savvy.
I’d recommend listening to the music album the author links to as well. He is not a professional musician, but hooked a keyboard up to his Amiga and created 13 tracks from the heart all by hand and they sound amazing. No sequencers, and all Amiga chip. Pretty impressive IMO.
Edited 2017-03-29 18:55 UTC