I've heard people have decapped these and even recreated (in some fashion) versions of the masks, or circuit layouts. I'd love to figure out how to start a project making actual chips.
Here is someone's heroic effort in reverse engineering the SID chip from decapped die shots, containing almost complete schematics of the entire chip [0], both the digital parts (oscillators, control logic) and the analog parts (filter, envelope, mixer).
Unfortunately, re-creating a SID chip is almost impossible given that we don't have 1970s era NMOS semiconductor manufacturing facilities available. It is very difficult to build analog circuitry with similar characteristics from discrete components or modern IC technology.
For example the filter section (see page ~9 of the forum post) is a dead simple state variable filter with 4 operational amplifier (opamp) stages and just a handful of resistors and capacitors. But the opamps in the original 6581 were not real opamps, but something hacked together from 3 MOSFET transistors due to silicon area constraints. The later revision SID (8581) had something more like an opamp, but SID enthusiasts claim it "doesn't sound as good" (and it definitely sounds different).
For a hobby synth project, I re-created the filter section with a standard quad opamp chip (TL084 or similar), and sure enough it behaves like a filter, but it doesn't sound like the original (the resonant peak is way too sharp). In particular, the original filter changes frequency response with the amplitude of the incoming signal.
If you look around the web, you can see several attempts at re-creating the circuitry (the filter in particular), but it's darn difficult to re-create it faithfully. And the original design is not a great musical filter design, its character is almost entirely due to the characteristics of the manufacturing process and the "bad" ompamp circuitry. For creating analog filters, the Moog transistor ladder and other classic synth circuits easier to get good sounds out of.
> Unfortunately, re-creating a SID chip is almost impossible given that we don't have 1970s era NMOS semiconductor manufacturing facilities available.
Do you have any idea how hard that would be to recreate? Would it be in the realm of possibilty for an ambitious amateur to fabricate a chip? (Maybe even if not something as relatively new as a SID, possibly something simpler?)
Jeri Ellsworth (who is far beyond "ambitious amateur") managed the first home CMOS process a few years ago, and I think the state of the art has advanced to about 1975 tech: http://sam.zeloof.xyz/first-ic/
It's a lot of work, obviously, and requires a big set of interdisciplinary tools.
An ambitious amateur could probably figure out how to fab a chip of the same complexity in small quantities (using e.g. electron beam techniques), but I think recreating the process faithfully, with the same parameters and variation as the original process, would be almost impossible for anything short of a major corporate R&D project.
Because the SID deals with analog as well as digital signals, it's not that easy to clone, however you can get fpga versions such as the SwinSID - https://www.c64-wiki.com/wiki/SwinSID
Unfortunately, the SwinSID (and similar) do not sound anything like the original chips when playing music.
Consequently SID chips are quite expensive to buy (relative to what they are) - any effort to accurately clone a SID chip would be welcomed, as it will drive the cost down of these chips on ebay etc.
Sorry to be picking a nit here but the SwinSID is not FPGA based, it's a software emulator running on an AVR mcu.
There is an FPGA based version of the SID called (surprise) FGPASID (http://www.fpgasid.de/) which is pretty much indistinguishable from an original SID. The only real downside to it is the cost, it's not a cheap one.
The FPGASID was created by decapping and reverse engineering original SID chips and the result is pretty spectacular.
And, it emulates two SID chips and has multiple modes (6581 and 8580), so it seems like a pretty darned good deal, honestly.
Real SIDs are pretty pricey these days, and they aren't making any more of them, so they will keep climbing in price for as long as people find them interesting. So, it's good to have an alternative.
There are other, newer alternatives, and they are getting difficult to distinguish from a real SID, especially given no two SIDs sound the same - e.g. http://dzi.n.cz/8bit/arm2sid/index_en.php
Original SID chips are often semi-broken as well. I had a few when they were cheaper and I was building a midibox SID. Most had weird VCA issues or oscillators leaking to the output always. Stuff like that. If you end up with a half decent one the oddities just become part of the charm.
There were two variants of the SID. The early one ran on 12v and the latter on 5v. As I understand it, the reduction in voltage greatly improved their reliability.
More than just the voltage, the 8581 had significantly revised circuitry and better opamps due because they had much more silicon area and a better process available.
