Let's build a "perfect" MOS SID (6581)

[aciddose]

The linked thread is about a full spice schematic/netlist that models the SID filter + amplifier analog stages.

I've decided to "spam" this thread linking to a past thread in the instruments forum in case it was missed back then. At the time I couldn't decide the best forum to post it in. I should also make sure the thread is linked in SID-related threads related to the sidplay open source library and similar locations as well.

viewtopic.php?f=1&t=526480

[aciddose]

I've wondered whether it might be possible to match the timbre using NJFETs for the input buffer and VCRs and NPNs for the other transistors, or whether the NMOS curve is too unique and an NPN model would end up far too linear.

Variations on approximated discrete SID-filters would be a really cool project. I wonder if the modular crowd might be interested in PCBs + parts kits?

I suspect the 6581 SID filter might be one of if not THE most distorted filters ever

[justin3am]

No need for three of the same topic in different sub-forums. If you would like this thread moved, we can do that but otherwise lets try to keep the discussion in one place.

[aciddose]

Well, the question is whether the SID can be considered an instrument (it can't really) and where the content of the thread would be most accessible to those most interested in it.

DSP / development would make more sense if it were focused on a software re-implementation. So posting there was probably not too useful. DIY here makes sense from the perspective of electronics projects since the main content is a ltspice schematic used to simulate the circuit. Hardware would make sense due to any SID-based hardware being most likely the sole application of this stuff.

The DIY aspect may make sense... if there is any interest in the idea. I'd like to make an attempt myself although I can't see it happening any time in the next year or so. Possibly next year... if I do make any modified schematics/models I'll post them here though.

[justin3am]

Cool, I would certainly be interested in your progress.
If you decide that you would like to have the thread moved later, we can do that.

I don't know about the modular crowd in general but I'm definitely interested in a PCB (I can source my own parts). If you want to aim a product specifically at DIY Eurorack, you might think about designing a front panel for it too; there is a white paper which contains panel specs on the Doepfer website.

I find that there is demand for DIYsynths but it's mostly coming from folks who will try to find someone to build the module for them. There are a number of professional kit assemblers who advertise their services on the Muffwiggler forum. I think there could be demand for the x0xbox equivalent of a SID based synth. ALM Busy Circuits has their SID Guts module but I think that's a BYOIC affair.

[Skupje]

Think the DIY application makes sense since it has an analog filter section.

[aciddose]

I've wondered whether it might be possible to match the timbre using NJFETs for the input buffer and VCRs and NPNs for the other transistors, or whether the NMOS curve is too unique and an NPN model would end up far too linear.

It's far too linear



At this point you might as well just use jfet opamps (tl07x or 8x) and OTAs with a lot of biasing to approximate the timbre. JFETs are horrible to use for VCRs because they're way worse than NMOS transistors for VGS (gate-source voltage threshold) variability. It's "possible", but almost completely impractical and mostly pointless.

I created a ltspice schematic/netlist to test this but the results while they sounded great just didn't have any special quality to them at all. You're still stuck tracking down some NMOS transistors for the VCRs anyway, rendering the whole circuit mostly useless compared to opamps + OTAs. You'd get the same or better results using an LM13700 + TL082 or NMOS matched pair + TL082 in a much smaller footprint.

When it comes to accurately modelling the whole NMOS circuit you end up with the problem that there simply are no cheap and readily available discrete NMOS transistors... they're not used for analog audio (high distortion!) and there is not really any other practical application for the sort of discrete transistors we're looking for.

There are an absolute ton of ICs packaging sets of N/PMOS transistors configured for dedicated tasks like analog power switching, but one you get into anything resembling digital it makes more sense to jump directly to the 40xx series ICs. In most cases the price is through the roof on these things, although I have found some possible solutions using SMDs for a dedicated SID emulation circuit.

[Jim Y]

Although they are not conveniently arranged or separated, can the N & P channels in the CD4007 array chip be useful?

[aciddose]

Possibly, it depends upon the particular properties of the N-channel devices in the 4007. 14 pins is a lot of footprint for NMOS transistors... also since the body (B) is linked to the first in each chain... you only get a single independent N out of the whole device. Pretty worthless for anything other than its intended purpose.