Personally, Jason confirmed what I was seeing with his initial comments on what he was hearing wherein he was "confusing" Depth with wet/dry, as his comment caused me to focus in on hearing what's going on such that something akin became apparent in addition to Cutoff Modulation.

What's fascinating is how the various modifiers in the MODX/Montage deal with extremes, and how the interplay between the various ways to modify Cutoff (and other aspects of the Filter) are summed/scaled by the different modifiers. This is where/how I first noticed (thought) that Filter Modulation, in this LFO, is doing something more than just swiping the Cutoff left and right.

In general, with regards all the modifiers that can be hooked up to all parameters, if they're not going to give us any kind of insight as to operator ordering and algorithmic handlings of transitions and limits, it'd be more than enough to have a resultant display, live, updating any resultant parameter we'd like to examine in its motions - so we can more easily "tune" out those areas where nothing's happening as we'd like it - enabling reversible insight into the unheard ranges (where things are crimped up against a zero or max value).

When and how things hit those edges would be invaluable insight, as after they hit that, it's no longer possible to easily tune by ear, as nothing happens until backing out of that edge. Still possible, but a lot of back and forth through menus to find where and how its hitting up against edges, which is tiresome detective work.

I wanted to put the picture to my thoughts that for Amplitude Modulation (LFO2) - the shape is not hypertriangle or an inverted quasi-square wave. I explained it by saying we hear amplitude in log scales. And, in fact, amplitude is measured in dB - not volts. I expressed the limitations of my software capture (wrong units of measure) - but I think previously using "volts" for a scale (which is how much we excite the loudspeakers) matched a bias and other realities were dismissed to some extent.

Now that I found a measurement tool to plot the audio in dB (a proper scale) - you can see for yourself my earlier assertions. Just regular triangles.

Throughout I was using the extended LFO. On the far left you see I was occasionally setting the speed to 0. And you can see it just holds the current modulation "amount". A true hold. When setting back to a non-zero value - the triangle continues where it left off. And following I demonstrate a few different speeds where my attempt was to change speeds when the triangle reached it minimum value.

I wish I could change the horizontal axis (time) to a different scale so I could show faster LFO speeds - but I think the point is well illustrated as-is.

I return back to - labeling this as a true triangle amplitude modulation is correct as we hear in dB not in volts going into a speaker. Volts to the speaker is a level removed from reality. I think if you perceive something other than a triangle it's important to look at the amplitude in dB as I've demonstrated here. This particular tool captures dB from a microphone and my source were my earphones. I don't see any slight of hand or anything other than what the user interface advertises as what should be happening for amplitude.

Tools used:

dB amplitude graph: Sonic Tools SVM by Rikki Systems Inc. (iOS app, version 2.1.4)
Oscilloscope: SignalScope X by Dynamic Signal Analysis Tools (iOS app, version 11.5.9 r2618)

The oscilloscope was used for the earlier blue/white plots and I use this to plot FM-X waveform outputs for individual carriers. Where in this case we are more concerned with level over time than "absolute" amplitude as in an envelope.

How much steam erupts if I point out (again) that Filter Modulation is the name of a system vs Amplitude and Pitch being the names of parameters?

Beginning to regret pointing out the Filter Modulation is not moving cutoff linearly in Triangle mode 😉

Another oddity:

Key On Reset in an OFF state only applies to Phase, not to Delay.

LFO Delay always operates as if Key On Reset is active, regardless of its state.

This could have been made visually instantly grokkable by putting the Key On Reset button immediately under the Phase parameter box, with a visual border as a boundary around the two of them.

OK... back on topic.

Filters. The question seems to be:-

If the LFO is Triangular, why don't I hear a Triangular change to Cutoff?

Equally, if LFO is a Sine, why don't I hear a sinusoidal change to Cutoff?

First, I would say, it is actually hard to "hear" a Triangle (and know it isn't a Triangle). Equally hard to perceive and recognise a Sine Wave (we are talking about LFO... Frequency lower than human hearing range).

Now if you listened to the same sound, first Modulated by a Triangle, and then by a Sine, you would very easily perceive a difference. However, to detect and verify the actual superimposed wave "shape" on the sound, you would likely need specific test and monitoring equipment.

Then as Jason pointed out, the actual wave "shape" you would see, would depend on the Calibration Scale of the equipment. This means whether you are measuring in absolute values, or some conveniently developed scale.... like Decibels.

The movement of a Frequency Response Curve is often measured in Log or Decibel scales. Take a look at Celestion's Website for their Speaker "Specs" in terms of frequency response (High and Low Cutoffs) and Resonance.

With those Frequency Log vs Decibel graphs, everything fits neatly onto a small graph, and still makes sense. If it were in absolute values, the graph would be 10m wide and 5m high. And not only that, it would look like a featureless hill... you wouldn't gain much "information" by looking at it. I exaggerate.

So, this is all getting a bit Monty Python-ish.

What really matters here is, it doesn't matter.

It's a case of "Do you like the sound?". If the answer is "No", choose another.

There is zero point in getting upset because something doesn't sound "Exactly like a Triangle".

This is totally unproven, but my theory would be that the LFOs "model" the behaviour of "vintage analogue" LFOs. These were often dependent on whichever choice of electronic analogue components that were used. An obvious either/or case here would be the choice of potentiometer... Linear or Audio/Log. A Triangle (linear) turning an Audio Pot (Logarithmic) results in a "Hypertriangular" output. Used loosely here to describe a waveform that is round on one side and pointy on the other.

As an aside... I nearly drove myself insane, and bankrupt, trying to find a Flanger pedal that sounded "Close To" a 1970s Electroharmonix Electric Mistress. I now have about 15 Flangers!!

Aural analysis of the "vibrato" of the Electric Mistress led me all over the realm of LFO Waveshapes.

What I now know is that, the heard "Vibrato" is directly proportional to the Rate Of Change of Delay Time. The Rate of Change of Delay Time is dependent on the Slew Rate (reaction time) of the "BBD" Chip. The Slew Rate of the BBD, while Physical, is also responsive to the shape and amplitude of the Clock pulses (yet another IC Chip). The Clock is sensitive to the shape and amplitude of the LFO (which is a slow VCO).

Triangle Schmiangle.

The Electric Mistress sound, irrespective of LFO, was borne of an unusual (some would say amateurish) design, and the "Reticon (TM)" chips that were used (now Unicorn Tears). But, for all its "ham" engineering, it sounds fantastic, and that is all Mike Matthews (EHM CEO) had to guide him... a good set of ears.

Can you imagine if Mike Matthews, in 1974 had said "Screw this... I can't hear a Triangle LFO!!"

????

[quotePost id=116899]OK... back on topic.

Filters. The question seems to be:-

If the LFO is Triangular, why don't I hear a Triangular change to Cutoff?

Equally, if LFO is a Sine, why don't I hear a sinusoidal change to Cutoff?

[/quotePost]

This has nothing to do with the question, other than an interesting aside.

edit: ADD -

that interesting aside being that consideration was clearly given to the response of the Cutoff's behaviour in response to the LFO - it's not a literal or 1: 1 from the wave to the Cutoff's movement AND there is "pinching/easing" (normalisation to ranges) obviously occurring.

Both of these characteristics are interesting because they indicate designed intent implemented in the Filter Modulation, making it more likely/possible/probable that perhaps other aspects of the Filter are also being modulated, by design.

The original question was/is: what else (if anything) is being modulated, other than just Cutoff?

ADDITIONALLY:

Resonance, the filter envelope's depth and the duration of the envelope's segments are the most obvious candidates for other things happening.

Other than Cutoff's obvious activity:

My spider senses, when using the top-of-list Low Pass Filter give me the impression there's more Resonance at a Filter Modulation Depth of 50 than at 99, and that this is on purpose, to prevent peaking... or this could be just an increase in the envelope depth (scalar) of the Filter's influence that's strongest around the middle of this Filter Modulation Depth setting (again, to avoid peaking when Cutoff is being moved more by values greater than 50).

The Resonance really scrapes along, quite unpleasantly in my opinion, whilst Cutoff is being modulated by this LFO's Filter Modulation, and I think they have made some effort to minimise this. Which might be adjustments to the Resonance, as well as the rate of Cutoff change and/or the filter's envelope depth.

As Jason has pointed out, these choices (if I'm not merely imagining things) were probably made a long time ago.

For filter, my tools are less telling which is why I took at a look at amplitude first. Pitch would be as easily demonstrated.

Today I did some experiments with the filter and it's true that, looking at the spectrum analyzer, this is not a case of the depth affecting a wet/dry mix. It's easier to see with spectrum analysis. The behavior was well described before, but I'll more or less restate.

The speed determines how "fast" the cutoff is adjust from minimum to maximum values - whatever those values are. And the depth adjusts the spread. Therefore if you have a speed of "10" and a depth of "99" then the frequency spread will be the widest and therefore the sweep will travel "faster" through frequencies because it has further to go in terms of frequencies to sweep. But still, the timing of where the sweep reverses directions is the same. Depth is a matter of how many frequencies are involved.

If the depth is small then the frequency sweep will not be very wide and therefore the cutoff may not be reaching certain bands as it modulates which is why it "sounds" like a wet/dry mix. Really, there is a narrow band of frequencies wherever the center point of the sweep is for the cutoff modulation and the rest of the frequencies are not "reached". Of course this depends on the filter type what exact behavior you will see. This was previously astutely stated (not by me).

When I choose an "All" spectrum and wide skirt I can see discrete harmonics with a note played. Upper harmonics are closer together, in terms of frequency, as is inherit in the harmonic series.

With an "All" waveform has an LFO2 Filter applied and say a notch-type filter, you will see each individual harmonic get squelched as the LFO2 progresses back and forth through the triangle wave shape.

What I see is that higher harmonics are squelched "faster" than lower harmonics. That is because frequency is more spaced out for the lower harmonics. And higher harmonics are closer together making these appear to get squelched "faster".

Really what's going on here is that, by way of the cutoff, frequency is modulated. And due to the harmonic series if you have sounds with discrete harmonics such as all, odd, etc - then you may "hear" what sounds like speeding up at higher frequencies and slowing down at lower frequencies. Frequency is the modulation -- or the consequence of modulating cutoff is a "travel" through frequency.

There's FS1R documentation (3rd party) that mentions the filter LFO (FS1R had filter modulation with LFO1 and LFO2) travels through essentially the entire frequency range. I haven't put the Control/Assign cutoff offset version of sweeping through the analysis to see if that "reach" is just not as wide which would explain the difference. The FS1R also allowed for various sources to control the filter modulation such as mod wheel, expression pedal, etc. Presumably all of these could create the same "wah" effect as the LFO2 we have today. Therefore, I'm a bit surprised I wasn't able to quickly make Control/Assign (destination cutoff) translate to the same sound as LFO2.

@Jason

I'm testing with a fast playing, wide array of notes, being sprayed at the MODX by an onboard Arp.

This has a couple of unique qualities, different from holding infinite notes, and one huge difference from an externally MIDI driven sequence of notes, or a Pattern Sequence.

Try with an Arp, and you'll see the curve I've drawn, at extremities - the curve is a time on x, amount of cutoff change on left.

There's an easing, seemingly, going in and out of midpoints, and in and out of the extremities.

This might be harder to check on a scope, as each note refreshes, and is a different pitch. I'm using an arp I'm super familiar with, particularly with regards filter sweeps, which (combined) are about the extent of my musical prowess. I'm an arpist!

Notes all over the place are nearly impossible to "see" on my software scope because even without the filter the frequency peaks are dancing around all over the place. Even if I start holding multiple notes together without adding or subtracting notes much less movement with the notes themselves.

This is just a matter of analysis with tools. Setting up an observable controlled environment where conclusions can be made.

I think we have some reasonable swings at the answer to the question backed by stimulus/observe experiments and some past documentation. We don't have the DSP code (if that's where LFO lives) or more official sources. As an aside, it's interesting to know that the FS1R would reduce its polyphony if all FS1R Parts had the filter turned on.

[quotePost id=116920]As an aside, it's interesting to know that the FS1R would reduce its polyphony if all FS1R Parts had the filter turned on. [/quotePost]

WOW!

That is an interesting "aside".

Thank you!

[quotePost id=116896]Beginning to regret pointing out the Filter Modulation is not moving cutoff linearly in Triangle mode ;)[/quotePost]

[quotePost id=116900][quotePost id=116899]OK... back on topic.

Filters. The question seems to be:-

If the LFO is Triangular, why don't I hear a Triangular change to Cutoff?

Equally, if LFO is a Sine, why don't I hear a sinusoidal change to Cutoff?

[/quotePost]

This has nothing to do with the question, other than an interesting aside.

[/quotePost]

Whatever.

I think your questions have been well and truly answered.

By all means continue your nonsensical destructive testing, if that is what floats your boat. Assuming you successfully mine more dirt on Yamaha's design, please be sure to not bother anyone with it.

Thank you.

@Antony: this remains a simple question, and there's still things other than men can be made with straw.

https://www.swheatbottle.com/en-au/blogs/news/7-things-to-know-about-wheat-straw

@Jason, some of your posts in this thread really resonated. Thank you!

Particularly the one about thoughts and feelings on results and what's happening, here: https://www.yamahasynth.com/forum/2nd-lfo-of-fm-x-filter-modulation?start=20#reply-116754

It's a great comfort to know I'm not alone in my muted rage at the lack of consideration of the works involved in programming sounds.

This is especially true of FM-X, wherein a lot of tedium and busy work is interspersed with tight range experimentation and (for lack of a better word) calibration, just for a static sound that's got agreeable qualities.

When it comes to adding Motion and Playing Dynamics (the whole point of the Montage and MODX) it feels like I'm fighting the OS almost every step of the way, and that there are even times when it's downright hostile towards the processes of discernment of relationships between those things that can be animated easily and knowingly, and those things we might like to animate in response/coordination that require a bit more 'calibration' and pointed framing/containing.

Melas' tools don't get near this kind of thing.

I can do it by ear. I can even work backwards from intentions and objectives. But given how much is numbers, it'd be nice to see some of them, in their resultant forms, at the places where we're scaling/ratio-ing or otherwise animating/ranging them.

This is especially so when we need to rig up Assignable Knobs to get at modifying values more easily across the dozens of parameters involved in shaping and testing an FM sound's ideal boundaries for animation within itself and throughout its effects... it'd all be vastly easier and more productive if we could see resultant good boundaries/ranges, so we could jot them down and then reassign animation/Motion features to these in properties and be certain, quickly, that we're transferring them from our own shorthand knobs to the ideal discovered ranges of animation/Motion via whatever means is used to create these dynamics.