Comparing to Roland synths

[jpcima]

@SpotlightKid: If you want some refernce recordings, let me know.

Yes please, it would give me a idea of what we're targeting. A pair of recordings with chorus on and off, using the simplest input signal and the most dry configuration possible.

It wouldn't be bad also to measure for 1 low note, 1 mid and 1 high.

Is it possible to derive the delay time and modulation depth from the schematics?

By the knowledge of the chip and circuit, I would guess nominal delay approximately of 20ms.

It's difficult to know from schematics directly, although datasheets are able to help somewhat. Simulation is also complicated, as there don't exist SPICE models that I know of MN3101/MN3009 chips.

A decent possibility if you're able, it's to capture electronic signals directly on device.

But I think overall, the best way to go is based on current the gray-box method; ie. take recordings of the target chorus, and then adapt implementation as needed.

[jpcima]

I've found this demonstration sound: https://www.youtube.com/watch?v=HM-fIEEwaaA Around 00:25, there is a simple bass sawtooth, and from analyzing the sound region in audacity, it can be observed quite easily what is the variation range of delay.

Another thing, from collected information on the topic: the Juno chorus did not use the analog delay circuit with a compander for noise reduction. It's only the Dimension D which used a compander. Regardless, I guess it will be a good idea to implement it.

[SpotlightKid]

Hey, good find.

[jpcima]

It's a wave of this shape. From it, it can be seen the original, and then the two delays coming after.

The next thing is: I'll load this in numpy, and try estimate these delays with minima and maxima finding. Then, the variation of delay can be extracted in a plot, and this will be easier to reproduce it.

[jpcima]

A program is available in tools which produces data files of delays for gnuplot. But it's not accurate at all, and output is not so useful until fixing it.

It outputs tuples of (absolute time, left channel delay, right channel delay), with units in seconds.

[jpcima]

As imprecise as it is, this data still shows a visible tendency. If I make some gross extrapolations from these calculations, I can have some approximate characteristics.

Juno 60 (I): delay 1.5ms to 3.5ms, triangle 0.95 Hz

Juno 60 (II): delay 1.5ms to 3.5ms, triangle 1.6 Hz

[SpotlightKid]

The schematic says 0.5 Hz (Type I) and ~0.83 Hz (Type II), though.

[jpcima]

Well yeah, considering this LFO is input to a BBD clock generator, and BBD clock is proportional to inverse delay by a factor 2, that likely confirms it; because these frequencies which you cite are also proportional to measurement by a factor 2.

[jpcima]

Progress: this is a faust module emulating a delay with the lfo characteristics. (no bbd) You can see it compared to the juno60 sample in the video recording. chorus.dsp.gz chorus.ogv.gz

[jpcima]

Note: other resource of interest here https://github.com/pendragon-andyh/Juno60/tree/master/Chorus

[jpcima]

The sample files of youtube origin don't seem correct as a reference, they appear like mixed-down, and it's suspicious that they have two delay images, it should have 1 only.

I took the samples from the comment above's link, and did it over again. Try with the new file, which is expected to be nearly correct chorus2.dsp.gz

[jpcima]

I've pushed the chorus3.dsp prototype. This one adds I+II mode. It's a rewrite, it's simplified and fully functional, and it's matched with sawtooth sound samples found at the repository: https://github.com/pendragon-andyh/Juno60/tree/master/Chorus

I think the I+II frequency rating in that document's table is not matching, I had to reduce it to a lower value.

I think we can possibly go with this one, I'll like to hear your feedback about it.

[jpcima]

Next analysis, this is modulation of the I+II case.

It approximates to a sine of frequency 9.75Hz, delay variation 3.22ms to 3.56ms. Approximated in green on the below plot.

[jpcima]

I've rewritten the plugin with faust and the BBD delay hacked inside. The code is pushed on the master branch.

[jpcima]

Hi @b3ll, we are researching Juno chorus on a request of @SpotlightKid, and collect some information and research on this topic; so feel free to hop in if you like.

are you owner of Juno 106 by any chance?

[jpcima]

By experiment on sample juno60-1, I find following data lfoRate = 0.513 (no change) delaymin = 1.54e-3 (was 1.66e-3) delaymax = 5.15e-3 (was 5.35e-3)

Also the triangle wave is found asymmetric, by duty 49.3333%. (it rises faster than it declines)

[jpcima]

Jx-3p chorus: same as Juno 60 except lfoRate = 0.48 triangle duty = 46 %

[jpcima]

The JX-03 is particular, it used this curious sine-triangle hybrid waveform.

delaymin=1.43e-3 delaymax=4.33e-3 lfoRate=0.462

Note: as such, one must be careful about getting the inverse LFO by doing (1-x) of the original, as it will not work for this case.

channel1.dat.gz channel2.dat.gz

Faust code of JX-03 "sine-triangle" LFO

lf_sinetri(freq) = rdtable(tablesize, sinetriwaveform(tablesize), int(os.phasor(tablesize, freq))) with {
  tablesize = 1 << 16;
};

sinetriwaveform(tablesize) = ba.if(p<0.5, s(p), 0.5*(s(p)+t(2.*(p-0.5)))) with {
  p = float(ba.time)/float(tablesize);
  s(p) = sin(2*ma.PI*p);
  t(p) = ba.if(p<0.5, -2.*p, 2.*p-2.);
};

[b3ll]

@jpcima wow! this all sounds super cool. I don't own a 106, but I do own a System 8 which emulates a 106 (using hardware modeling), however I won't be home for another 2 weeks to test it :/

[jpcima]

I find the mixing part of direct signal vs chorus signal from schematic. It has an identical summing opamp part on all of Juno60/106/JX3P.

It's the following gains: g1=(-Rf/R1) and g2=(-Rf/R2) where Rf=100k R1=47k R2=39k

It establishes that direct signal is mixed in at -1.62dB compared to processed signal, which makes the delay a slight bit prominent over original. I have modified chorus.dsp source for this gain relation.

@b3ll it's a nice beast ! well feel free to contribute to our sample set anytime :)

[b3ll]

I should be able to in a week or so!