How synthesizers, using the same chips, have different characteristics

Jump to navigation Jump to search

How synthesizers, using the same chips, have different characteristics.

For example, what is it that makes an Oberheim OB-XA an Oberheim and the Prophet 5 Rev. 3 a Prophet, while both are based on the CEM 3340 VCO and CEM 3320 VCF?

Think just the 3340 VCO drives the sonority of the instrument? Take the hard sync circuit. The coupling between master and slave determines how nasty the hard sync sound is.

Can a simple VCO mixer impact the sound? You betcha. A VCO mixer is usually an opamp summer and the summing inputs are not completely isolated. That means there's some loading between VCOs. In the Minimoog, the mixer is actually passive - no active devices, no opamp - and the VCOs are audibly loaded. You can hear this by turning down all VCOs except one - turn off mixer switches to all VCOs except the one under test and listen to it, then while you're listening to it turn on the mixer switch of any other VCO and you'll hear a difference, even though its mixer level is all the way down. The outputs of each VCO is loaded down. The same thing happens with opamp based mixers although less drastic.

Audio level to the VCF input, isolation between VCF gain blocks, and dielectric of the filtering caps are the reason why OB-Xa (in 4 pole mode) sounds different from the P5 rev3 despite both using the same 3320 VCF.

Audio level impacts two factors: overdrive and resonance quality. Some overdrive can be a good thing; unlike the P5, the VCF audio levels of the OB-Xa are fixed. Too much audio level can weaken the resonance quality, too little can make it sound anemic (I discovered this on my OB-X8 which has variable mix levels, many of the factory patches are too hot). There's a "sweet spot". To my ears, the 4 pole VCF mode on the OB-Xa has always sounded too... polite. The OB-Xa in 4 pole mode will not self oscillate but they will on the P5.

Each of the 3320 filter stages could have coupling in a way that weakens the isolation, which impacts the sound of the VCF. It depends on the design in the target synth.

The selection of dielectric on the VCF filter caps impacts the sonority of the 3320 (or any filter) between polysynths. Mica caps vs various other film caps are going to sound different.

The feedback architecture of a VCF provides the resonance, and it also impacts the resonant quality of the VCF. When the prototype model C Minimoogs were built, their input stages to the feedback architecture used JFET transistors. I have personally played one of those model C Minimoog and the resonant quality of the VCF was anemic and cold sounding. While it was a better circuit engineering-wise, it was not the best sound. When the model D was designed they changed from JFET to bipolar transistors which sounded much warmer.

VCAs. Too few people put enough credit in the VCA as the element that shapes the sound. The control element in a VCA can either be OTA based or differential amplifier. For any synth designed before 1980, there wasn't anything with high fidelity. The Minimoog uses EE textbook differential amplifiers which have the worst fidelity (OTAs first became available when the Minimoog was designed in 1969). But those differential amplifiers (the Minimoog has TWO) can impart a pleasant distortion, which was a big reason why the "Minimoog sound" was so elusive to later synth designs.

OTAs were an incremental improvement for VCAs. You could build an OTA-based VCA using fewer components. But the fidelity still wasn't great, and to stay in the linear range of an OTA the input levels had to stay within a miniscule +/-10mV (that's millivolts). This made the OTA a noisy little bugger, and even at maximum attenuation through Iabc there's still some level poking through.

But OTA imperfections can be a good thing. One reason why the OB-X and OB-Xa sound so big is the OTA element in their VCAs. Oberheim delivered a hot signal to the VCA which pushed the OTA into a nice pleasant overdrive. In the OB-X, the voice summer is built around an OTA where summing eight voicecards can make for a hot signal in a hurry! Oberheim used a predistorter on the input of the voice summing OTA, which improves the dynamic range and headroom that is vital for its function. Even with the predistorter, the OTA is still pushed into overdrive. All that overdrive made for a fat sounding synth.

Along comes the modern VCAs like the CEM3360 in 1980. Still OTA based but with advances to improve the performance. Greatly simplified design, much better dynamic range, much lower noise, much cleaner sound... too clean in certain stages. Oberheim used the 3360 as the OB-Xa voice summer and lost a little pleasant distortion. There were two versions of the OB-8; in the early version, the OTA-based VCA circuit on each voice was replaced by the 3360. Now all the OTA overdrive was gone. Tom noticed the OB-8 was "too clean". So in the later version of the OB-8 they reverted the voice VCA back to the OTA-based VCA. And I would note that the Memorymoog voicecard also uses the OTA-based VCA.

Then there's the PC board layout. There are certain design rules to make sure your circuits behave. Copper traces in parallel form a capacitor, copper traces of too little copper per square inch forms a resistance, long PC board traces form an inductor. All sorts of potential impedance reactions can muck up any CEM IC in a good or bad way.

Then there's the AC coupling between VCO/VCF/VCA stages. Every stage has an output impedance - the lower the better. The coupling technique between stages is going to load down the output of the preceding stage. The coupling circuit between stages does two things 1) it forms a voltage divider which is reactive 2) AC coupling forms a highpass filter thus some loss of bass is possible. This is inevitable. There have been improvements in OB-8 and Micromoog/Multimoog that make their bass response better.

Then there's active devices. Opamps are linear devices. Discrete transistors are non-linear and have to be threatened/coaxed into linear. Translates into different distortion effects. Both have their uses or poisons.

Then there's control law between modulators and destinations. LFO/EG modulation to VCF cutoff, VCO pitch, VCO PWM, VCA level will vary between polysynths. Why? Control law. The audible result of LFO modulation varies depending on linear or exponential or log control law. Not every audible result works with a single control law. When I say control law, I am referring to the relationship of the modulator signal (IE the shape of the EG transients - linear, exponential) to the input response of the receiving device (IE volts per hz or volts per octave). One of my sound design tricks for horns is applying a short EG transient to one VCO in a dual VCO patch, which makes a nice fat attack transient. But that trick always sounded better on my Oberheim than my Memorymoog. Why? Different control law between source and destination.

Another control law variation - even the transient shape of the CEM3310 EG can be altered. Moog did this in their Memorymoog. In most CEM3310-equipped polysynths, the EG shape is linear. The 3310 EG shape in the Memorymoog is exponential. This makes for punchier VCAs.

Then there's the opamps in the audio path. The early versions of ARP 2600 used Teledyne opamps with good fidelity. But those opamps failed in the field due to latch-ups - again, a product of early opamp designs. ARP changed the opamps to LM301s which eliminated the latch-up problems but resulted in a duller sound, losing high frequencies.

Yeah just about EVERY circuit contributes to the sound even with polysynths that share the same CEM ICs.

And BTW the Oberheim state variable SEM filter is not a Sallen-Key. They are different architectures with different transfer functions.[1]

See also


  1. ^ Chip Based Synths by Michael E Caloroso, Analogue Heaven mailing list, 28 Apr 2024

Further reading

External links

  • [], Wikipedia