Diagrams imply that each reverb or variation send is not combined before entering the respective system effects and each return is only a return based on the siloed send of each Part.  If they were combined before entering into the system effects then you would hear another Part's system effected signal on the Part returned signal.  So if you say had 100% send to variation on Parts 1 and 2 then 0% return on Parts 1 and 100% return on Parts 2 then if they were combined before reaching variation you would still hear Part 1's variation signal fed back on Parts 2's return.

That's one piece.

Some detail on missing dry level information:

https://yamahasynth.com/community/montage-series-synthesizers/preset-sound-routing-to-var-rock-hard-organ/#post-104183

So elements are combined AFTER Insert effects?

Diagrams imply that each reverb or variation send is not combined before entering the respective system effects and each return is only a return based on the siloed send of each Part.  

That is what I thought also. But I am having trouble testing it.

The diagrams also seem to imply that for a single part each element takes its own path through the 3-band EQ and the insert effect(s).

That is, the elements of a part aren't combined until at least after the insert effects and perhaps not until after the 2-band EQ. But they are combined before Var, Rev and System effect into a single left and single right signal.

Then any 'pan' occurs after all effects just before final output. As far as I know panning does NOT use simple scaling but uses a SIN/COS form of panning rule to try to maintain a constant power output of the signal. This was presented here: https://yamahasynth.com/community/postid/102932/

But combining Dry, Var and Rev doesn't need to maintain any such constant power output so can use basic proportional scaling.

Jason' old thread and doc link has valuable info re topic.

Thanks for the link to the old thread and the MotifFSRack doc. The older docs generally have much more detailed info but I had been unable to find any about the routing.

My last remaining Q on routing is just WHEN the elements of a part are combined into a single signal. I'm assuming it is before the 2-band equalizer and thus before the Var, Rev and Master effects.

But is each element signal sent through the insert effects separate from signals from other elements of the part? Or have all elements been combined into one BEFORE the insert effects?

Correction on var/rev return:  the sends are per Part but returns are single globals.  So there's not a way really to tease out where the mixing happens necessarily.  The sends could all be mixed then sent to the effect.  This would make more sense from a resource perspective.

Clarification here: 

https://yamahasynth.com/community/montage-series-synthesizers/what-is-%e2%80%98var-send%e2%80%99/#post-30799

Correction on var/rev return:  the sends are per Part but returns are single globals.  So there's not a way really to tease out where the mixing happens necessarily.  The sends could all be mixed then sent to the effect.  This would make more sense from a resource perspective.

Clarification here: 

https://yamahasynth.com/community/montage-series-synthesizers/what-is-%e2%80%98var-send%e2%80%99/#post-30799

Any way to tell if effect handles parts separately?

 So there's not a way really to tease out where the mixing happens necessarily.  The sends could all be mixed then sent to the effect.  This would make more sense from a resource perspective.

That is also what I thought. I agree that each part has its own 'send' to Var/Rev and that 'send' is based on the parts 'level' value - the parts 'dry level' isn't involved.

What I am trying to tease out is whether the Var effect acts separately on each parts var send and THEN combines the 16 results into one 'var return'? Or if the 16 'var send' values are combined into one signal and that one signal is acted on by the var effect.

Take filtering as as example. If you have two signals, one high Hz and one low Hz you can get a different result if you apply some filters separately to each signal and then combine the results than if you combine the two freqs first and then apply the filter. But that can be because of the way filters work given their cutoff ramps, resonance impacts and sideband reflections.

So, with filters, it is possible in some cases to create tests with two extreme signals and determine if filtering is done before, or after, combining the signals because there will be detectable artifacts in the results that indicate it.

Not worth beating it to death. I was just wondering there is any way to do a similar test with either var or rev. Would either of those produce a different result if two signals were combined first before the effect acted than if the effect was applied separately.

Just FYI - the original question arose from trying to determine the extent to which the sound of a single element (AWM2) could be controlled as opposed to the part as a whole.

A single element might be a violin that should sound as if it was at far 'stage left' against a partition. It will sound differently than the same raw sound position center or right - especially in a small space.

So the ideal would be to be able to apply effects like reverb to that violin/element sound but differently than a second violin/element in the same part. The part might represent the entire violin section but you want to treat each violin separately.

I don't think that is possible until/unless you can apply effects at the element level rather than the part level.

The complexity of the signal doesn't really affect how the effect works. So having to process elements separately would really require twice the resources of process a single signal no matter how complex that signal was.

Nuf said.

Issue is that 100% dry signal path is NOT shown.

I'll post an IdeaScale for it but none of the graphics I've seen, including the one in the link you provide, actually show the path that a the full, 100%, dry signal takes to get to the Var/Rev/MX block (see my #1 and #2 in the original post).

1. Init Normal (AWM2)

2. notice that part volume = 100, dry level = 127, var send = 0, rev send = 0

3. go to the effect routing screen - notice both insert effects are set to 'Thru'

4. play a note - hear the sound - the sound is ONLY raw element with no insert effects 

5. set dry level = 0, var send = 127

6. play a note - hear the sound - the sound is ONLY the var effect (G Chorus) even though dry level is zero

My tests seem to show that:

1. The variation effect receives a signal level based on part 'Volume' regardless of what the part 'Dry Level' is set to.

2. It receives that same signal, and level even if there are no insert effects.

3. It receives that same signal, and level even WITH insert effects. But with insert effects it also receives a 'Var Send' portion of the signal produced by the insert effects. And the source signal for the insert effects is NOT the part 'Volume' level signal but a scaled signal value based on both Dry Level and Part Volume.

4. set dry level to 64, leave part volume at 100 and set Var Send to 64

5. insert effects will receive a signal level of 50 (1/2 of 100) since dry level 64 is 1/2 of 128

6. var effect will receive a signal level of 100 based on part volume of 100 (not based on dry level of 50)

7. var effect will also receive a signal level of 25 from the effect output because the effect source signal level is 50 and and Var Send of 64 (1/2 of 128) cuts it in half.

Net result is the var effect RECEIVES a totally dry signal of 100 and a 'wet' signal of 25 from the insert effect path.

That 'totally dry' signal path isn't shown on the graphics. The graphics would have you believe that the insert block only receives the 'dry' portion of the signal but my tests show that isn't the case.

This is how I see it:

https://ibb.co/H4hLBhW

(Looks like we can't do inline images anymore -- or I lost the mojo).

The red is global and the blue is part-level.  The red pluses are where I see each part mixing together.  All of this seems consistent with a few experiments using one Part and two Parts.   I don't see that Rev vs Var vs Dry are proportional.  Increasing/Decreasing one doesn't appear to impact the level of the others.

One thought on how to do this is to use the pitch shift effect for Var.  Then crank up the delay to maximum so that the pitch shifted output happens I believe it's 400mS later.  You can play a short note whereby the effect's output doesn't sound until you've left off a key and the dry signal sounds alone.  Then the Var effect will sound alone without any dry signal (because it's gone).   You can turn things up and down and see how each portion of the sound is impacted since they are very isolated this way without overlap.

Note that in general I have limited time and resources generally answering on a cell phone which is cumbersome.  Took me a while to get the keyboard out and actually look at this.  There may still be unanswered portions of your original ask - just a lot of things going on.  Hopefully the tip on looking at effect types that can help you in your quest can help you self-serve if there's anything outstanding.

Also, I arrived at the relationship between the Part volume slider and the various trims (dry/rev_send/var_send) by listening to what's happening with the touch wah effect.  As you increase or decrease the level going into touch wah (I used Init Normal AWM2 unmodified except for effect settings) you will notice that the frequency range of wah's peak goes brighter for higher level inputs and darker for lower levels.  I had the Var Send set to 127 (max) and if the slider was not trimming the var send then I would not hear a difference in the wah return brightness.  It would just get softer but not harmonically different.  However, it did change brightness so I inferred the slider is inserted into the signal chain as pictured.

.. spoke too soon about inline graphics.  No lost mojo.  Just used the wrong link.

Thanks for the test suggestions.

I will spend some time for the next several days trying your suggestion about working up a more suitable test case. For now my focus is on trying to understand how the effects and signals are combined just for one part - multiple parts can come later.

1. Not sure what 'Trims Each Equally' means

2. Three comments on the graphic you provided

    A. Thanks - I appreciate the time it took to create that

    B. It doesn't show the 'Raw Dry' (unscaled dry - essentially part volume) that is an input to the Var/Rev effects

    C. I don't see how the slider can be in the position you show it since part volume is FIRST in the chain rather than near the end.

3. This is my current thinking

    A. The slider is equivalent to the part volume setting

    B. The volume setting is then scaled to obtain a 'Dry Level' value

    C. The Dry Level value is the INPUT to the insert effects

    D. The output of an effect can be either higher or lower than the dry level input

    E. The 'Send' values are scaled to some portion of the effect output level

    F. The output of a var/rev effect can be either higher or lower than the Send input level

    G. The Var/Rev return then scales the output of the var/rev effects

There has to be a raw (dry of all insert effects) signal being fed to the Var/Rev effects since they can produce full output even if the 'Dry Level' parameter value is zero.

So, forgive the terminology, this is how it appears to me:

A. Part Volume sets the upper bound for part signal level fed to the insert effects. I call this 'Raw Dry' 

B. Part Dry Level (the parameter) is used to scale part volume. If part volume is 64 and dry level is 64 then that dry level represents a signal level of 32 (1/2 of the part volume level)

C. The insert effects receive the scaled signal level - 32 for the example values in B above

D. The var/rev send values then scale the output of the insert effects to produced a signal level for the var/rev effects.

E. The Var/Rev effects receive the 'Raw Dry' signal from A above as well as as scaled var/rev send values from D above. The Var/Rev effects can produce full output even if the 'Dry Level' parameter is zero or any other value. So It seems to me that var/rev must be receiving both a 'pre insert effect' AND a 'post insert effect' signal.

F. By 'scaled' I mean proportional based on a max range of 0-128 to bring all signals into the same range.

G. Somewhere the element signals are combined but I don't know where - could be before the insert effects or after.

H. Somewhere the part signals are combined but I don't know where - could be early in the stream or at the very end after ALL effects (insert, var, rev, master)

If you turn dry down all the way the insertion effects still come out of var.  So I believe as pictured that rev and var get the output of the insertsion chain.   

Dry is kind of a misnomer.  It still carries the Ins effects.  Dry is in the sense that it doesn't have any system effects.

The slider I think trims the dry level and the var send and the rev send by the same amount.   

Since two trims in series add up it doesn't matter if the placement is before or after the dry level/rev_send/var_send knobs pictured since you'll get the same result no matter the order.  So you can think of the slider on the other side of the knobs.  That's still consistent with what's pictured.

Another way to explain the slider comment is to think of the slider as 3 trim knobs (separate from the sends and dry level knobs) that have a rubber band around them whereby turning one knob (any of the 3) turns the other two to the same position and the slider sets this position.  Low slider and all 3 are full CCW (counter clockwise) and all the up slider all 3 are full CW.  

Due to the nature of electronics, schematic representation, etc - there's more than one way to express the same thing which turns out to be equivalent.