Fathom Synth Development Thread

[Frostline]

Which means you do not have the Visual Studio Redistributable installed on your system.
I did not think this was necessary on Windows 10, but apparently it is.
I will update the manual to explain this.

Please go here, download it, and install it:
https://www.microsoft.com/en-us/downloa ... x?id=48145

Let me know if that fixes it so I can update the manual for everyone else.

And from the error log

Code: Select all

[1BD4:23D0][2018-02-27T12:31:13]e000: Error 0x80070666: Cannot install a product when a newer version is installed.

Really dislike the idea of rolling it back and potentially breaking a product that uses the newer version.

[FathomSynth]

It's OK, it means you already have it installed.
I think it means that I will need to build for Windows 10 using VS 2015 instead of 2012.

I think I know what is happening for the pitch drop at the end of ADSR ramp.

For missing notes, does changing poly from 1 to 32 or 32 to 1 change it?

Also does low or high notes effect it?

[Tj Shredder]

TJ, Yes a video would be awesome, even better with sound.

Aliasing bugs are the highest priority so I'd like to have this fixed in the next release.

The Fathom code has the power to pull back partials per note frequency,
and also control the buffer size and interpolation,
so this should be easy to fix once I see it happen.

Ok, investigating again, it is triggered by setting a modulation (I choose Aftertouch) to the pulse width.

Here is a video with sound. (need some time to figure out how to do that...
https://www.dropbox.com/s/81g2om2z1fhfy ... g.mp4?dl=0

By the way, I did not find out how to delete a modulation...

And the puls in Fathom looks like that:

Fathom Pulse Width.png

But I think it should look like that:

Fathom Pulse Width Expected.jpg

[FathomSynth]

Tj, It is very important that I know the following.

I see the spectrum moving a lot, which dial are you moving, or is that just the after touch which is changing?

Does the aliasing occur if nothing is modulated?

Does the aliasing occur if you are not touching a dial?

If you remove all modulations and do a single mouse click on the partials dial does the aliasing go away?

Does it only happen on high notes which push the partials above 22,000 at a sample rate of 44100?

Also the waveform with three steps is caused by having the waveform dial set between square and pulse.

[Frostline]

For missing notes, does changing poly from 1 to 32 or 32 to 1 change it?

Perhaps a little, though the results are always random even in the mono. May go over a minute without skipping notes or may skip 5 notes, play 5 and then skip again.
Here is 32 poly playing original E3 notes.

Also does low or high notes effect it?

Tested with E6, seemed to do better and were only a couple of skipped notes and never more than one (or more accurately described as previous note hanging I suppose).

However using the 32 poly version and E1....

And the note is still playing now, transport stopped. Kill all MIDI message has no effect. Track is not armed. It's just very stuck.

[Tj Shredder]

Tj, It is very important that I know the following.

I see the spectrum moving a lot, which dial are you moving, or is that just the after touch which is changing?

I play a LinnStrument and bend the pitch

Does the aliasing occur if nothing is modulated?

Yes, as long there is a modulation assigned

Does the aliasing occur if you are not touching a dial?

Yes

If you remove all modulations and do a single mouse click on the partials dial does the aliasing go away?

As mentioned I don't know how to remove a modulation, but if I recreate it with no modulation, there is no aliasing.

Does it only happen on high notes which push the partials above 22,000 at a sample rate of 44100?

Yes, with lower partials and lower pitches its fine...

Also the waveform with three steps is caused by having the waveform dial set between square and pulse.

No, its only the pulse width I changed.

[Frostline]

By the way, I did not find out how to delete a modulation...

To use as example the modulator in your video:
Press the little note symbol on the modulator slat next to where it says pulse width.
Press the Note aftertouch button that will appear so you get the list of available modulators.
With that list open, press the delete button on the GUI and then press the red check mark at the top of the GUI to confirm the deletion.

It is an odd system, something about GUI needing focus in a particular area for it to know what is supposed to be deleted. I had to have help figuring it out too.

Also I notice that the amount for the modulator in your video is set @ zero and you are using the multiply type. If you change that to addition or bipolar do you still get the aliasing?

I mentioned in my video that the multiply option seems to have an effect even if the amount is at zero. You can hear it in the attacks of the notes where I was using the filter to show it following the envelope where pitch would not. Attack is sudden/sharp when using addition or subtraction and amount at zero, but softer (like a brass instrument attack) when using multiply with a zero amount.

[RPH]

Maybe someone can test the patch and midi item in another system.
https://1drv.ms/u/s!AuTU2-ocFX74hjstKk-1byHATSfu

Ran your project in Reaper 5.76.
Got the same as what you got, untill:
- I corrected the ADSR ( sustain on last point which is zero volume? ). More default and short worked fine.
- Removed all the changes you made to the waveform ( shape / detune / distortion / AM )
- And turned off the FX as well.

Then it was OK here, same in Renoise.
I see and hear some phasing, but that is probably normal.

[Frostline]

- I corrected the ADSR ( sustain on last point which is zero volume? ). More default and short worked fine.
- Removed all the changes you made to the waveform ( shape / detune / distortion / AM )
- And turned off the FX as well.

Then it was OK here, same in Renoise.

Thank you for testing.
Did just changing the ADSR eliminate the issue?
I'm thinking the issue is something in the AM or distortion because I would think the ADSR being the problem it would be a great deal more consistent in causing the issue and would be eliminated or greatly reduced at a slower BPM(which it was not in my testing).

I guess I will try eliminating things one at a time until the issue goes away and then put everything back except that last removed one and see if it works. Would like to have the patch more like it is than just an analog saw playing 1/16 notes.

[RPH]

- I corrected the ADSR ( sustain on last point which is zero volume? ). More default and short worked fine.
- Removed all the changes you made to the waveform ( shape / detune / distortion / AM )
- And turned off the FX as well.

Then it was OK here, same in Renoise.

Thank you for testing.
Did just changing the ADSR eliminate the issue?
I'm thinking the issue is something in the AM or distortion because I would think the ADSR being the problem it would be a great deal more consistent in causing the issue and would be eliminated or greatly reduced at a slower BPM(which it was not in my testing).

I did a quick test, as always strip down to check where the error is.
My bet is the ADSR, when i replaced it with a default one and shortend it the notes were fine. Didn’t put the rest back after, time was up.
So just replace the ADSR and keep your edits and fx.

Btw, did you also setup Reaper with the performance tips from Everett?

[FathomSynth]

Tj Shredder, OK, I have good news for you.

I know now exactly what is causing the aliasing problem.

Normally when oscillators play from the partial buffers, if the note frequency pushes the partials over the limit (half the same rate) it adjusts the partials in real time and chooses the buffer with the right number of partials which go up to but not over the limit for that note frequency. Let's call this the "partial limiting mechanism."

Modulating the Pulse Width changes the waveform and is causing the audio engine to switch to a mode in which the oscillator output is built from the partial sin waves for a pulse in real time, that is by adding all the partials in real time, and playing from the buffers is temporarily turned off, so far so good. But ...

The bug is that I forgot to add the partial limiting mechanism to the real time partials mode!

Doh!!!

Good news is, that this is a super easy fix.

It's easy because building the partials in real time is perfectly suited to only selecting partials up to the limit. So I will change this so that modulating a waveform parameter does not turn off the "smart" partials limit.

It's actually a really good thing you found this problem, since it's a really big bug, and easily fixed. A lot of people would think "Wow, this sounds horrible!" but not know why.

So this fix will go in 2.8, which will be a special bug fix release with all these problems fixed.

TJ, also to delete a modulation please see the manual, 5.4.4 Deleting a Modulation. I would recommend reading section 5 on modulation, it's not very long and basically tells you everything you need to know about modulation.

On the note skipping thing, sorry to say, I'm super confused reading the last few posts. If anyone feels they have tested it to the point where they can say what the isolated problem is, please post a conclusion below to help me a bit recreating this.

Missing notes should be easy to catch on my system if I know how to set it up, and will then be easy to fix also, so I will include that in the bug fix release.

But please, a synopsis and the skipped notes, so I know what to look for.

[Frostline]

My bet is the ADSR, when i replaced it with a default one and shortend it the notes were fine. Didn’t put the rest back after, time was up.
So just replace the ADSR and keep your edits and fx.

So I relaunched Reaper, put in a Fathom poly MT 2.71 and my midi item. Moved the notes in the midi item back down to E1 since previous tests seemed to indicate the lower pitch makes the unwanted effect happen more frequently.
Then I completely removed the ADSR from the patch.

I was really hoping it was simply the ADSR (because that means less alteration to my patch).
But elimination of the ADSR had no effect.

Btw, did you also setup Reaper with the performance tips from Everett?

I believe so. I followed the instructions as presented when I first started using Fathom.
To check I went back to the webpage today and looked at the settings recommended to be changed there.
From what I can tell my settings seem to match the recommended.

[Frostline]

Missing notes should be easy to catch on my system if I know how to set it up, and will then be easy to fix also, so I will include that in the bug fix release.

But please, a synopsis and the skipped notes, so I know what to look for.

To set it up most efficiently load the Reaper project file I provided, drop the midi notes down 2 octaves and then play the midi item with the patch I provided. It may take almost an entire run of the midi item to hear/see the effect or it may start from the beginning. Or with your much more powerful system it may work perfectly.

It is not really skipping notes, but rather notes are hanging on and preventing the next note from playing. At least that is what it seems to me by the long tails on the notes and the fact that it can get stuck playing the note even after transport has stopped.

I have not yet completely removed every modification I did to the initial analog saw, but bypassing the filters and phasor, turning off AM, distortion and detune(in various combinations) and the effect still occurs quite often with the midi item on E1.

[FathomSynth]

OK, I think I might know what is happening here.

Fathom has a feature where it will never stop a note in mid cycle to prevent audio clicks.

It always finishes the current waveform cycle when it stops an oscillator at the end of a note.

It's possible that if the note is so low, and the midi notes so close together, then finishing the cycle might extend over the next note.

I will try to run this on Reaper myself.

Correct me if I'm wrong, but if this is true, then it should be happening even if there are no modulations, effects or detune of any kind, just with a single mono oscillator.

[Frostline]

It's possible that if the note is so low, and the midi notes so close together, then finishing the cycle might extend over the next note.

If that was the case wouldn't it happen more frequently and regularly? Like the 1st note in each measure always hangs and blocks the 2nd, or the 5th note in every 3rd measure or some type of pattern to it like that?
And it doesn't just extend over the next note, as visible in the images it extends over many notes sometimes.
Could be some odd confluence of settings in the patch that no (rational) user would ever do.
When trying to remake it with a different oscillator I noticed it almost started doing it again when adjusting the 1st ratio knob in the distortion, but then it stopped.