Designer's Notebook: How NOT to do it
I wasn't ready for the unbelievable design problems in the supposedly mature analog music synthesizer kit I bought a while ago. But damned if I was going to return my slaved-over modules.
I'd decided to augment my small synthesizer collection with an old-school modular analog system. A small, low-cost starter system to be sure, but from a grand old name in the biz. Yes – modular analog systems (think "Moog"), like LPs and bell-bottoms, have made a bit of a comeback.
The system I purchased was a full kit version, and the sub-rack contains four modules: a MIDI-to-CV converter (CV, Control Voltage, is needed to control the synthesis modules; MIDI is a practical nod to modern digital interfacing needs), a dual VCO, a dual VCF, and a triple VCA/multiplier. Each of the three synthesis modules also contains an auto-retriggerable envelope generator. All in all, a compact and versatile little system.
I'm only going to tell you about the MIDI-CV problems here – there are so many.
Conceptual problems: The DAC
The goal of the MIDI-CV module is to convert MIDI data from a keyboard or computer into analog voltages that can be used to control various module parameters, such as oscillator and filter frequency. The signalling level between modules is 1V/octave (modules need to respond exponentially to the CV). MIDI specifies 128 notes – about 10.5 octaves – thus, a perfect 7-bit DAC could be used. Given that DACs aren't perfect, and that we'd like at least 0.01 LSB DNL & INL for good tuning accuracy, adding seven unused bits, for a total of 14, is a reasonable minimum. Personally, I would use a 16-bit DAC with a good INL spec. What do they use? An el cheapo old warhorse 8-bit "DAC08" chip. The horror.
If that isn't bad enough, the "reference" voltage is taken from the 5V supply (from a standard three-terminal regulator), and all the resistors in the signal path are cheap carbon types. One could hardly design a less stable and accurate system if one tried.
The left half of the sub-rack. Notice the accurately applied silk-screening on the MIDI module (look around the jacks). I've pimped out the synth a bit. The crappy original pitch controls are replaced with sexy 10-turn "clock-face" pots, the power LED, with a vintage, ~35 year-old yellow one, and the envelope LED with a pink one!
Details: That poor, poor chip
When I first powered up my completed system, everything basically worked, though one of the first things I noticed was a lot of grunge on the oscillator outputs. After some basic troubleshooting, it became clear that the oscillators were clean enough, but the CV outputs themselves were noisy. That's when I started digging into the design.
One of the first, and most shocking, things I found was a 4051 CMOS analog demultiplexer chip, being powered by approximately 33V! The chip was quite warm. I pulled the power. What egregious wiring error had I made? None, it turned out. The design was running a chip with an 18V-20V absolute maximum rating at 33V! How it survived is anyone's guess. This gross overvoltage also caused some supply noise to leak into the signals – the first of several grunge sources.
I added zener regulators to get the voltages to a safe level (and reduce the ripple, too).
The power circuitry on this module, such that it is, also supplies juice for the other three modules. So while my zeners reduced the local ripple, the outgoing DC was still pretty bad – about 1V P-P. I tacked in a couple of electrolytics, reducing ripple by a factor of 10.
The 5V supply powers both analog and digital circuitry, and the poor PCB layout exacerbated the amount of digital grunge making its way to the DAC. I sliced and diced the power traces, so that the analog circuits were powered by a direct connection to the regulator, not through a shared impedance with the digital load. Decoupling was also added to the digital parts, and a bigger filter cap to the DAC reference.
By now, the CV outputs were much improved, but still exhibited some noise. Given the poor power layout, I started to wonder about the ground, too, which wasn't much better. Sure enough, I found about 10mV P-P of line-frequency ground differential where I really didn't want it – the relatively high filter cap currents were contaminating analog signal paths. Some more slicing and dicing cured this.
To do, maybe
Finally, my CVs were cleaned up. My sine waves no longer sounded like grunge-modulated FM (unless I wanted them to).
I may replace the DAC and associated circuitry in order to achieve excellent stability and accuracy, but that's for another day.
What horrendously bad designs have you seen in the wild, and have you troubled yourself to fix any?
- Modular synthesizers
- MIDI Manufacturers Association
- PAiA (the synth manufacturer)
- Example 4051 datasheet
- Moog synthesizer inventor is born, May 23, 1934
- PCB design course & checklist
- Successful PCB grounding with mixed-signal chips - Part 1: Principles of current flow
- Successful PCB grounding with mixed-signal chips - Part 3: Power currents and multiple mixed-signal ICs