I was about to make a simple A440 tune patch to be able to tune to the analouges to the NM. While I was at it I couldn't resist adding an octave switch. Sometimes it's easier to tune to something an octave higher/lower. Then I stumbled onto a something I don't fully understand.

You have one oscillator running at A440 with the keyboard turned off. Dead simple. There's a second oscillator also A440 with a Control Constant going thru a Notescaler so it can only raise the second oscillator by one octave. Or at least that was what is was meant to be. I was suprised to find that when the Constant hits +64 the second oscillator is NOT one octave up but only almost. I had to add the cable going from the notescaler to a second pitch modulation input and set that gain knob to 1 to get the desired exactly one octave raise. Try it yourselves. The error, if it at all is one, also occurs when you use a KeyQuantizer instead of a Notescaler and also with different types of oscillators. See the patch file.

What's going on here? Are the oscillator modulation inputs not scaled exactly so that you get precisly one octave by adding 64 thru a notescaler/quantizer? But shouldn't they be? Or am I doing some backwards thinking?

There are three attenuation types in the Nord Mod. The oscillator pitch mod is a type II exponential and will therefore not produce a 1:1 relationship to it's input. It's all on page 32 of the text under "modulation".

> There are three attenuation types in the Nord Mod.

Yuppers! So does analog synths/modules. Log for "musical" use and lin for example when you modulate a couple of OSCs to get constant frequency beating (between the OSCs) across the keyboard. In my attempt to do an octave switch I had to use the pitch, or log, inputs since my aim were to have several OSCs running at different octaves. However now I simply morph the Coarse setting of the OSCs but I'm still bugged by the OSC not responding the way I expected them to.

> The oscillator pitch mod is a type II exponential and will therefore not produce a 1:1 relationship to it's input.

The log response should not be on the way of a perfect octave raise when modulated via a Notescaler set to modulate one octave should it? I mean, what's the point having a Notescaler or a KeyQuantizer if it cannot accurately modulate the OSC?

You sollution by using two PitchControls looks a little complicated to me

Oh yes - it was just how I ended up getting a one octave raise using the Notescaler set at one octave. In the final tuner patch I modulated the OSCs directly with a morph but that doesn't adress the problem with the OSCs responding odd.

This works in a way, but... When set in the same pitch, there is a little bit of indifference again: about every 10 sec the sound will fade a little.

Ah - you're using the master/slave cables - that's a different approach. Perhaps they are more precise than the modulation inputs. But you're right - these two OSCs drifts against each other just a little bit aswell.

The maximum frequency resolution of the Nord is 1/2^23 * 96000, which is exactly 0,011444091796875 Hz, the 'binary' frequency resolution of the Coarse/Fine knobs and the grey in/outputs. Its 2^23 and not 2^24 as the frequencyrange is actually from -48kHz to 48kHz, the 'negative' frequencies being 180 degree phase shifted instances of the 'positive' frequencies, giving 2^24 /2 possible frequencies on an oscillator. All this on a lin scale.

Now when modulating the pitch there are accumulating integer math roundoff errors, the -64 to 64 unit value has to be recalculated to a 24 bit integer value, which is multiplied by a 127 value from the pitch knob which is also recalculated to a 24 bit integer value and then converted to an integer log value before this multiplication, the result is then converted to an integerlog value again and finally scaled to the internal lin scale of theoscillator. I remember Magnus Kjellander say that they took a long time to figure out how to scale the 1 unit/semitone scale to the internal 24 bit lin scale of the oscillators and they didn't compensate for the roundoff errors on the pitch inputs to keep that analog dual oscillator synth characteristic and save some DSP time as well. The grey master/slave connections are more precise as they represent the actual 24 bit lin value internally used for the frequency parameter of an oscillator. This way of working seems unavoidable with the 'modular' approach, most other digital synths and samplers actually use a fixed tuning table with the closest integer values that the oscillators need for every distinct semitone/key stored in memory. Pitch modulation is then done by calculating the integer offset value and _adding_ it to the table value, instead of using the multiplications needed with a modular (read: freely and incrementally modulatable) approach. I agree with Magnus, Clavia's way makes the NM more lively. Remember that on a sampler the liveliness of a sound is already recorded in the sample. Being too accurate on a modular VA would prolly make it a sterile and boring machine. So a little grey input mangling kb patch to have some fun, as: all work and no play makes jack a dull boy...;-) (quote 'One flew over the CooCoo nest')