[sdiy] Roland Juno 106

[Tom Wiltshire]

> On 11 Feb 2024, at 09:50, Gordonjcp <gordonjcp at gjcp.net> wrote:
> 
> On Sat, Feb 10, 2024 at 04:42:08PM -0800, brianw wrote:
>> What's "the digital part"?
>> 
>> I assume you're not referring to the 'Digital' part of the DCO, which you referred to as boring, although the digital master clock could have a significant affect on the sound.
>> 
> 
> No, the DCOs and filters are pretty boring and ordinary. I'm talking about the microprocessor that runs the voice card.
> 
>> There's no need to 'clone' the firmware, since that digital aspect doesn't really affect the sound.
> 
> Okay, so you're saying that the way the control ranges interact, the way the envelopes work over particular ranges, and the way the LFO works - and the way that is all mixed together to control the oscillators and envelopes - you're saying that this "doesn't really affect the sound"?
> 
> That's an interesting take on it.
> 
>> I'm just curious what you're saying has yet to be reproduced with modern technology...
> 
> Nobody seems to really understand what the voice CPU is doing, so they try to "improve" it. This is nothing short of disastrous. If you compare the JU-06 to a real Juno 106 for example, it doesn't sound anything like the "real deal" because it just doesn't get the envelopes even close to correct.
> 
> That's leaving out all the "higher order effects" like the inherent wobblyness that the LFO has, the variation in the envelope times, and so on. There's a lot going on in that 2kB of code and 1.5kB of lookup tables.

While this might not have been done, it would be easy enough to do. Especially if Roland have any notes kept from the original design. Fundamentally, there's nothing in that firmware that couldn't be *exactly* reproduced on a more modern chip.

Most of the details are in the service manual. It includes the envelope times, LFO range, DAC resolution (12-bit) and update rate (238Hz). We'd still need to know whether envelopes are curved or linear, for example, and any other quirks they have.

The interaction between the LFO and the DCOs is one quirk in the 106. Since the timers only update when they finish a cycle, pitch modulation is only applied at the end of each waveform. This means it gets more and more steppy as the frequency goes lower. For a 1KHz note, the waveform cycle is finishing much more rapidly than the LFO is being updated, although it's highly unlikely that the two cycles coincide, so there's still an effect there. For a 100Hz note, that effect is going to be much more notable, since the LFO's 238Hz update rate is essentially being "down sampled" at the DCO's 100Hz pitch.

Tom