Yoshimi extras

[Folderol]

Actually I forgot a bit
SubSynth harmonic 1 bandwidth multiplier needs to be set to maximum - sorry!

A personal preference is also to have reverb on

[Folderol]

Some fun with the AddSynth Waveform.

Work your way up the AddSynth window stack to the Waveform one, then select 'Gauss' as the base Function.

Next, MIDI-learn the base function Par. control to your keyboard pitch bend control.
That's it

As this is a real-time control, you get an interesting tonal change along with the bend. It sounds even better if you add one of the reverb effects.

If you set 'Block' on the learned line, then you just have the tone change with no pitch bend.

[Martin Walker]

Thanks again for all this info Will

Martin

[The Elf]

and based on what Elf has said about noise in the 2500 (see! I do read other stuff)...

And I remain Yoshimi-Pi-curious too, and read all your stuff!

[Folderol]

and based on what Elf has said about noise in the 2500 (see! I do read other stuff)...

And I remain Yoshimi-Pi-curious too, and read all your stuff!

[Folderol]

Something I completely forgot about

In January, one of Yoshimi's Little Band of Helpers produced this youtube vid of the then brand new LFO sync feature. One thing I particularly like about this is that he is clearly experimenting and finding out what happens when just twiddling knobs (in the time-honoured way). Also right at the end he proves that the lock is very responsive to changes in real-time.

[Martin Walker]

You didn't forget Will - I remember watching this video some time ago, but it's just as fascinating this time round

Martin

[Folderol]

Aaand another thing

A most under-used feature of Linux window managers is the ability to switch between up to 4 different screen views. This is especially useful where screen real estate is limited - such as with the 7 inch screens used with Yoshimi Pi, so here is a scenario where this is especially useful.

You want to run Yoshimi on a new project using Rosegarden as a sequencer and eventually recording with Audacity. You only have a single keyboard controller outputting on channel 1

Start up QjackCtl, then Yoshimi. Select an instrument from a bank, but leave the bank window open. In one corner of QjackCtl you'll see a small 'pin' icon. Set it.

Over a clear area of the desktop background scroll just one 'notch' (most wind managers accept this for screen change), and you'll now only see QjackCtl.

Start Rosegarden, and make sure it is connected to Yoshimi. Also open the 'Transport' window if it isn't already, and pin that as well.

Set Rosegarden track 1 for recording. Hit record and play. Once you have enough done, you can stop and then dodge backwards and forwards between the screens editing the track, and changing instruments, master volume, key shifts etc. for best results.

With Yoshimi showing, change part 1 to MIDI CH2, enable part 2 and set its channel to 1. Flick to Rosegarden and uncheck record for track 1 and set it's channel to 2. Set track 2 for recording and channel 1.

Record your next track to part 2.

Rinse and repeat.

Once you have a substantial amount there you might want to flick to Yoshimi and start tweaking parts effects etc. which you can do while Rosegarden is playing the piece. With it's Transport window being visible you can stop, start, rewind at any time.

With Yoshimi and Rosegarden project files saved, it's now time to record your masterpiece, so flick to an empty window (just the Jack and Transport windows visible) and start up Audacity.

Check it is set up to receive audio from Yoshimi. Click record, then Play on Transport. While it's recording you can flick between this window and the others to monitor their status.

You'll immediately see any Xruns in the Jack window even if they are inaudible, so can stop and take appropriate actions.

At one time with the kit I had at the time, this was the only way I was able to work.

[Folderol]

I've just listened to a track on soundcloud. I would never have guessed this was done with Yoshimi

https://soundcloud.com/diegosimak/2022a

[Drew Stephenson]

It certainly demonstrates quite a range. Secret submission for the next one-synth challenge!

[Folderol]

I've just been reminded of this:
https://www.youtube.com/watch?v=ifdOGFZkeT0

It's the work of one of Yoshimi's little band of helpers. See how he's using Pure Data to control Yoshimi in real-time. As well as switching various parts on and off and playing with internal settings while sending MIDI data, he's applying that low pass EQ to the entire output, and has MIDI-learned the cutoff point

[Folderol]

Layering sounds the Yoshimi way

Mostly people seem to go with using multiple instruments set to the same channel. Well you can do this if you want to, either by redirecting the normal 16 parts, and/or by switching to 64 parts, when you get 4 parts for every existing channel already set to the respective numbers.

However Yoshimi has several more efficient methods.

The first and most obvious is to use all three engine types, which immediately gives you quite different behaviors. As well as pitched sound these can all produce noise, but do so in completely different ways resulting in different spectrum.

Next , you can dive into AddSynth->voice where you then have 8 semi independent voices. Each can have a completely different waveform (or noise pattern) and set of envelopes filters etc. The start of a voice can also be delayed. They can have their own modulation types and waveforms, can be octave shifted and detuned from the group. There is also the unison feature which can give some nice chorus effects. Higher numbered voices and modulators can share the lower numbered waveform oscillators or indeed the entire voice/modulator.

If you find that too limiting, you can set the complete part to kit mode as well. This means you can have up to 16 of these three engine kit items. these can all sound together across the entire key range, or set such that they sound one at a time across the keys, or finally in pairs fade from one to the other for a given key overlap. In kit mode you can change where each kit item links to the chain of the three available part effects.

[Folderol]

Been a while - just look at all the dust round here

Yoshimi's Envelopes have some interesting behaviour.

Looking at the Amplitude one, the bit that can cause some puzzlement is 'Stretch'. By default changing this control will have no effect at 440Hz (normally A4), at lower frequencies the envelope gets progressively longer, and conversely at higher frequencies it gets shorter. This is to emulate the behaviour of real instruments. However, if you don't want that, just set it to zero, or for an even greater range, increase it.

Al the other envelope types, frequency, filter, bandwidth behave in the same way, but have different default settings depending on exactly where they are.

Amplitude envelope is the big daddy. It is the only thing that affects the overall time, but that in itself is quite complex. The three engines can have quite different envelopes but of these, it's whichever is the longest that sets the end of a note - so you may hear different engines stop before others.

The AddSynth engine takes this further, as when you have multiple voices active, the longest one sets the time. Well not quite! Of the Voice group and the global envelope it's the shorter that sets the AddSynth time.

Each Voice can have a delay so logically its time becomes the envelope time plus the delay time. If the delay time is longer than the other voices (or global) then it won't sound at all.

Finally, LFOs have the same stretch behaviour, but in this case the default is always for zero stretch.

[Martin Walker]

Thanks for all the extra info!

[Eddy Deegan]

Interesting information indeed, thank you Will. This is particularly relevant to me at the moment as I've been doing a little digging into how to create my own virtual sound-chip for a side project I've got on the go.

Originally (Will and I spoke of this previously in person) I was considering utilising some of the Yoshimi code for the purpose but I do like making my own wheels and as such I found some great resources online which help with the process.

For the non-programmers an envelope may seem a simple thing. It's anything but. An envelope function is complex because it needs to be called at sample-rate and return a value, usually between 0 and 1, which is multiplied by the amplitude of a single sample at the correct time-delta from the note-on event to accurately reflect the correct point of the (usually, but not always, ADSR) envelope at that moment in time.

The maths gets quite heavy (at least for a non-mathematician like me) quickly. It's essentially a time-sensitive state machine that heavily leans on formula.

Buffers of the kind found in ASIO drivers sort-of simplify it, but not much. It's beautiful wizardry.

For anyone interested there is a rather good developer-orientated video on implementing envelopes here. It's the second in a short series on how to construct the basics of your own standalone soft-synth.

[Folderol]

That guy has a very good Video style - rare!
Incidentally I've just realised I've been wittering on about Yoshimi since 2013

[Folderol]

I've just realised I've been somewhat remiss
For those into ambient drone-style music, here's Lorenzo driving Yoshimi with Pd.. again

https://www.youtube.com/watch?v=h6kiqR5tg80

[Eddy Deegan]

That's very impressive indeed Will

[Martin Walker]

[Folderol]

Lorenzo is an extremely accomplished musician, and Yoshimi's best possible advert - we're lucky to have him!

[Folderol]

Like most of Yoshimi, the way it's formant filter works is slightly odd.

The display looks as though vowels are inside the formants, when it is actually the other way of course. As an aside, the actual mathematical hoops it jumps through are quite different again! Unlike the other filters, when a formant one is created for the first time all continuously variable controls are set to reasonable, but to some degree randomised values.

However, while working on the (much neglected) Copy/Paste presets code I discovered some further things things about the filter I had been unaware of.

First, all six vowels are always active, and from the main filter window a preset stores all of them, their formants and the sequence information, so even if you think you're not using the vowels you are! Probably the best way to effectively cancel the effect of a vowel would be to wind the Q down to zero (rather than the gain)

Secondly, when actually in the formant window itself, using the C/P buttons you only manage the formants of the currently selected vowel as a preset. No other data at all. The parent filter will store and recover everything.

Third, even though you may only have 3-4 formants per vowel enabled and active, Yoshimi will store (and recover) all 12 possible formants per vowel - which may or may not be useful

I don't yet understand how the sequence size operates, but can confirm that all vowels get seen regardless of the sequence size.

[Folderol]

A correction
The number of active Vowels is limited to the sequence size - this is if you're not using a buggy development branch
However, I'm considering making a change so you can't even select a vowel outside the sequence size.

[Folderol]

Good Heavens! Look at the dust round here. Don't the cleaners ever come in?
Anyway, here's something that may be of interest.

Yoshimi can produce some extremely spiky waveforms, especially the percussion sounds - such as Natural Drum Kit, but there is a way you can tame these without having to use external means. This is by applying the default distortion effect - Overdrive 1.
The only controls you need to adjust are the Drive and Level ones to get the degree of limiting you want while maintaining output level.
This will act as a fairly 'soft' limiter, and applied to just that part will not disturb any other playing parts.

Wind the drive up a bit more and you can 'thicken' the sound.

[Martin Walker]

Thanks for the hint Will