Here’s a business plan for you, should you ever run into an old silicon fab sitting in a dumpster: build Commodore SID chips. The MOS 6581 and 8580 are synthesizers on a chip, famously used in the demoscene, and even today command prices of up to $40 USD per chip. There’s a market for this, and with the right process, this could conceivably be a viable business plan.
Finding a silicon fab in a dumpster is a longshot, but here’s the next best thing: an FPGASID project. The FPGASID is a project to re-create the now-unobtanium MOS 6581 found in the Commodore 64.
The Commodore SID chip has been out of production for a while now, and nearly every available SID chip has already been snapped up by people building MIDIbox SIDs, or by Elektron for their SidStation, which has been out of production for nearly a decade. There is a demand for SID chips, one that has been filled by “clones” or recreations using ATmegas, Propellers, and nearly every other microcontroller architecture available. While these clones can get the four voices of the SID right, there’s one universal problem: the SID had analog filters, and no two SIDs ever sounded alike.
From the audio samples available on the project page for the FPGASID, the filters might be a solved problem. The output from the FPGASID sounds a lot like the output from a vintage SID. Whether or not this is what everyone agrees a SID should sound like is another matter entirely, but this is the best attempt so far to drag the synth on a chip found in the Commodore 64 into modern times.
The files, firmware, and FPGA special sauce aren’t available yet, but the FPGASID is in alpha testing, with a proper release tentatively scheduled for early 2017. Maybe now it’s time to dig out those plans for the Uber MIDIbox, with octophonic SID goodness.
It continually amazes me how far we’ve come in chip fabrication, and yet despite the amazing tools we have available antiquated chip fabrication remains(nearly) impossible – instead we throw ever more advanced chips at the problem to emulate things
Part of this seems to be the inability of patents to adequately describe processes enough to be replicated; the other the lack of incentive for expired intellectual property to be released
So although the patent expired in 2004, it seems amazing that we instead use semiconductors of orders of magnitude more complex fabrication
More complex, sure, but often far cheaper through technology improvements and economies of scale. A raspberry pi zero is orders of magnitude more complex than a C64, but it’s also orders of magnitude cheaper than that machine ever was.
At this point, it’s possible to actually make a multiphoton lithography system at home with the resolution needed to make a MOS 6581. Also, yes there are die shots that have enough information so that you can recreate it using free VLSI layout software and preserve it’s digital and analog hybrid goodness.
This is why I love HaD, the comments can point people at things that they never knew existed. Thanks for the link to “Magic VLSI layout tool”
+1
+2
+4
+8
Ooh I’ve been looking into what you’d need to fab chips DIY… Always saw that you’d need to outsource or use an FPGA…
“While these clones can get the four voices of the SID right, there’s one universal problem: the SID had analog filters, and no two SIDs ever sounded alike.”
Recreate using…tubes. ;-)
Tubes of what?
(I just HaD to go there.)
Best dig them sid chips out. I got over 20 of them packed away with other parts for a project I never got round too. Keep meaning to do something with them but never did. Going to waste.
Would you be interested in selling any of those chips? :)
I’m half of a team in a computer club in Sweden working on converting an old electric organ into a retro chip synthesiser. We have 6 chips as it is, but a few more wouldn’t hurt.
I would like to purchase 4 SIDs for my planned ultimate 8-bit computer*. Name the price, and I’d think about it. Would be better if they are of the same model to simplify power supply things. Of course I need chips with working filters.
* (ATMega 2560 as CPU, 16MB of RAM, 400*300 (max 256 colors) or 800*600 (hi-res, max 4 colors) resolution via VGA, 2 OPL3’s and 4 SIDs for sound, and some more features).
I think this is an unsolvable problem, because it’s not a technical problem, it’s a psychological problem. You could recreate the original hardware perfectly, and people would still complain that it’s not as good. See also: Stradivari’s violins. We’ve had much better violins for a while now. People still crave the expensive old ones, and think they sound better, even if they fail every blind test.
Amen.
Audiophiles… Or is it Audiophools?
I know someone who got an extra line (with meter), by the power company (who were laughing at the idea as well), to their house and gold kettle leads for their audio equipment. I would agree with Audiophools.
But on the other hand, some of the really high end Audiophools sound cards 24-bit 384kHz can be damn fine when repurposed as an ADC card for Radio reception e.g. listening to the Schumann Resonances (global electromagnetic resonances created by lightning discharges in the cavity formed by the Earth’s surface and the ionosphere). Or any other natural or man made ELF/SLF/SLF/VLF transmissions.
https://en.wikipedia.org/wiki/Communication_with_submarines
Because of the massive spike at 50Hz/60Hz globally from AC power you actually do need such a large dynamic range to be able to see anything else at the very bottom of the RF spectrum.
Ooh, I like that. I’ve been using Audiots myself ;)
I’ve been using “audiopiles” for a while now :)
Good point…
… Ima sell 1980s quarters for $25 a roll, for the coinop freaks…. because obviously 1990s and later quarters don’t have the same heft and jingle and clunk when inserted.
Change 1980 to 1964 and you very much do have a difference, thanks to 90% silver content.
This completely misses out on the point a Stradivarius violin is like a piece of art. It’s not necessarily the music it makes but the person who created it or the craftsmanship and history in it. It’s like saying people have painted mona lisa better many times. The mona lisa will always be priceless because of who and when and why it was made.
Wrong. The Mona Lisa is priceless because it was stolen once, and became a mass media phenomenon. That’s how this stuff works.
He is not talking about the value of an instrument as art to collectors, or historical value. His point is you can make a new violin that is In many ways better than a Stradivarius, and regardless of the fact that in blind tests not even experts and world class musicians can tell them apart, or will like as not choose the sound of the newer instrument as “better” and it does nothing to diminish the perception that Stradivarius instruments are somehow better in a way that can not be reproduced. One aspect of anything dealing with the senses is that there ultimately is no completely objective way to quantify perceptions. What sounds better? It depends on who is listening.
It will never be as good to the original sid, it was a special chip in its own rights.
It’s like the Roland Jupiter, they make update versions of it but still the original shines. But we can only try to make them sound like the original with today’s tech.
Your never best the original ones.
Yeah, that’s the sort of rubbish that audiophiles tend to spout. You could replicate an audio signal perfectly but someone would claim it’s not ‘authentic’ and therefore inferior, although it’s identical. Humans are irrational.
But that does not apply to musical instruments, electronic or not. The sound of a vintage Roland Jupiter-8 is unique and while Roland’s digital re-creations come very close they do not sound exactly like the real thing.
Does blind test confirm that? Or is it just another fallacy of perception caused by having wrong expectations? The same fallacy happens when audiophile believes that power cord for 10k USD improves the sound quality of his amplifier. It has no impact whatsoever, but he still will be hearing the difference, with his wallet…
Technically there WILL be a difference. But only a 30k Rigol scope will be able to tell the difference…
I know from a reliable source (maker of some of the “highendiest” gear avail) that after about € 200,- / m the difference in sound and even “normal” measurement will be irrelevant. Of cause absurd wattages in amp and speakers have to be accounted for in thickness of the wires – you won’t attach a 5kW heater to a phonewire either…
First you create stress in an audiophile that his existing equipment may be imperfect. He is then unable to relax and enjoy listening, straining his ears in a hyperattentive state to catch the problem. Then you offer a solution a kind of purgative sacrificial ritual where vast sums of cash are offered up to expunge his sins against perfect fidelity, whereupon he may relax and enjoy listening again… For a while… Until slowly creeping anxiety that his system is getting out of date forces him to commune again with the scene…. Then he’s ripe for shearing once more.
@Stefan_Z
Problem with replacing ordinary power cord with something extremely expensive is the fact that wiring between wall socket and nearest power plant is rather very low quality compared to that 10k USD/m cord. So if someone believes in magic of power cords, he should replace with them entire wiring of his home, and probably the line to local distribution substation/transformer too. When it comes to power cords the only important thing is power rating. Everything else is irrelevant…
You’re missing the fact that if you’re explicitly looking for miniscule differences between an anolog synth and a good digital recreation then you’ll certainly find them. But as with all analog gear, if you compared two “identical” analog synths side by side, you’d also detect just as many differences between them..
Well to be fair you won’t get the same results out of the same synth two days running if you write all the settings down super precisely on a patch sheet, reset it all, then set it again off the sheet. You’ll get “somewhere near” the same sound but not exactly the same.
Sorry beat not best on my last comment lol. Always in a rush lol.
The Problem here is the completely unique digital/analog design of the SID… Anything purely digital is an emulation.
And anything analog that isn’t the exact same chip design is similar at best.
The summary states that no two SID chips sound the same. So a real SID is similar at best.
Does this mean no real SID exists?
The SID chip that was released wasnt hte SID chip yannes wanted to release – commodore cheaped out so it has less features that initially intended.
Everything in the SID apart from the filter is digital. I have got a HardSID Quattro that contains one of each revision of the SID chip. If you playback a SID file with a player that supports the reSID engine (I recommend jsidplay) you can not tell the difference between the emulation and the real hardware. The filter emulation is unbelievably accurate thanks to analysis by a few people of photographs of the SID die, and the digital implementation, complete with flaws, is absolutely identical.
Isn’t the problem that the SID’s unique sound comes not from its design, but from unknown flaws in manufacturing?
The flaws are very well known by now and can be recreated. See https://bel.fi/alankila/c64-sw/index-cpp.html for information about the filter errors. The filter on the 6581 is main culprit.
Yeah, replacing the SID by an FPGA isn’t going to cut it; you can easily put reSID-FP on some SoC and obtain the same result without even so much as touching a soldering iron. The unique sound comes from the hybrid digital/analog design and then ten [?] cheap D/As that run at the full 1MHz (easiest way to get rid of aliasing ever :))
Basically what Gravis said – take the original die shots and go from there, and don’t be to keen on “fixing” things – while I personally like the 8580’s improved oscillators and filter response; whoever thought “oh, let’s get rid of that DC offset in the output signal” killed the volume register sample playback right along with it. Same for the weird, pretty useless register latches that come from a former, more complicated design – without those a lot of the current tricks like 8bit sample playback etc wouldn’t work.
What of course would work is hardening or EM proofing the analog parts more. While it has its charm that you can faintly hear about everything that’s going on in the whole machine (eg. the badline video DMAs being responsible for that characteristic chirp) most people could probably live without. :)
you forgot to add that from 1982 to 2016, tecnology gave a speed up of MANY degree of magnitude to power (up) and to prices (down).
Where’s my 50Ghz Z80? :-P
No market for it now that relatively more modern designs have matured enough for organisations like NASA. :-P
Right, they given up on 386es yet?
Nope. Still mostly 80386es and rad hardened MIPS/PowerPC chips since the supply of 8086es ran out.
18 comments and no one has noticed that the SID has THREE voices NOT FOUR!
Picard agrees with you. ;-)
They might’ve been referring to the sounds – noise, triangle, pulse, sawtooth.
Should give it backing vocals too, so it’s like SID and the SIDettes.
SIDnees might be more appropriate for some musical styles.
I was going to take on this challenge but after quite considerable research I dropped because of the lack of specifications for the analogue filters.
Quote: [Benchoff] “no two SIDs ever sounded alike”
There are specific reasons for variations but I think the above statement goes a little ahead of the reality.
I would say that SIDs sound very similar but not perfectly the same.
Here are the reasons as I understand them –
1) A lot of fake SIDs are coming from China now as the filter specification was not complete there are huge variations and that is even if it *was* filters at all.
2) The SID design changed a number of times.
3) One of the changes to the SID was made after ‘tape out’ so only the designer and the fab know what was changed.
Thanks [Benchoff] for this great article and I hope they do well.
PS: aparts from the filter, the SID is more CPLD size than FPGA.
Its not about how synths “sound” its about how they play…. You can get soft versions of all the old moogs, and arps and rolands etc. The thing they all lack is the user interface of the originals. It’s the feel, the attack, the release. things people who don’t know how to play an instrument don’t know anything about but love to attack other people because they claim they know how to measure something… I bet none of you who are down on audiophiles are instrumentation experts either… Just internet trolls with opinions that are as worthless as one is to the other… Meaningless, mindless drivel. When you can listen with MY ears, call yourself and let you know that you are an idiot. But please don’t tell me, I already know…
Of course soft synths aren’t equivalent to the actual instruments they emulate. Anyone arguing that is a dope, or is missing the obvious. This isn’t about that.
This is about a chip that had no user interface other than a bunch of programmable registers.
As for what you hear, it isn’t hard to design listening tests that eliminate differences in peoples ears and brains, by using the same ear and the same brain to do the listening. What they often show is that people can’t actually identify differences they hear with their own ears a split-second apart.
It’s not that other people can’t know what you hear, its that you don’t know what you hear, not as well as you think you do.
Is this on?
thing is… reproduction is absolutely no problem. we can just kick in the order right now. just that we do have a ‘working prototype’ (brings down engineeringcosts at the factory by some 20k usd to just 80k usd) in the form of plenty of original chips still existing ;) but we don’t have the drawings of what should go -in- the chip. and that is the whole problem right there. the original plans are missing. now if someone would find -those- in a dumpster the matter would be resolved with a single email and some 80k usd to place the order and get production going again. adding some logic to make it cmos bus compatible would be practical tho. as despite commodore always having failed to address any markets other than their own computers actually the sid would sell FAR more for ‘other applications’ ranging from sid box midi expanders to doorbells to washingmachines than it ever did for the c64. (in fact commodore could have kept on running their entire factory on just selling sid’s alone – they never did ;) as commodore also only produced these themselves in their own factory there also are now plans for this chip at any still existing external contractor.
not that the sid really ever was that special compared to let’s say the yamaha opl-3 or the atari sound chips but yeah there still would be a market for it today. also for things that never were part of cbm their original concept. while we’re at it we’d also like to see reproduction runs of the motorola 6854 econet controller. probably the best local loop network bus ever. kinda wonder why the entire 8 bit microcomputer industry didn’t jump on that either, but then again most didn’t even bother to implement rs232 properly. lol. “networks? who could possibly need that” :P