One-Synth-Challenge 57: Tunefish - Chat & Gossip

[RichardSemper]

2. The Spline and Additive OSC seem to pitched differently and are quite discordant when played together

I raised this in earlier posts - at 48khz the additive synth plays 1.5 semitones below the spline - changing setup to 44.1khz remedied it.

hope this helps

[H-man]

additive synth is playing 3/4 semitone below the note played.

Could be a bug, does this happen at 44.1 khz or another frequency?

EDIT: I just tried it, I get the described effect at 48khz. Not at 44.1khz.
Apparently that does not work right yet

Thanks. Changing to 44.1khz fixed this.

Hi again, thx Richard. My bad for not checking the older posts

@Payne, is this a difficult thing to rectify or should I just let go of 96kHz?

[payne_bc]

additive synth is playing 3/4 semitone below the note played.

Could be a bug, does this happen at 44.1 khz or another frequency?

EDIT: I just tried it, I get the described effect at 48khz. Not at 44.1khz.
Apparently that does not work right yet

Thanks. Changing to 44.1khz fixed this.

Hi again, thx Richard. My bad for not checking the older posts

@Payne, is this a difficult thing to rectify or should I just let go of 96kHz?

I'll look into this again and hope to resolve it in time before the end of month, so you can all make high quality versions of your songs. But for now stick to 44.1khz until this is resolved.

[payne_bc]

I got clicks at the start and end of each note play by turning everything off,
so just using the single additive synth, no fx,filters etc.
I rendered out some single note play and stuck the wave in a sample editor. It looks like as if your not starting/ending your waveform on a zero crossover so you get clicks, but playing with it a bit more, e.g. add adsr 1 to the volume via the matrix. When setting the attack to zero then a single click is heard on the note start, setting a longer attack you hear multiple clicks at the beginning of the attack, same happens on release, short release single click, longer release more clicks towards the end of the release.
Now I maybe wrong (and I don't want to come across as someone telling their "grandma how to suck eggs") as I don't know what number representation you are using (integer/single or double floating point numbers, but guessing one of the floating points.
This honestly sounds like you are possibly using floating point numbers and the calculations with small numbers are causing floating point numbers that are too small to be represent a number and end up as NaNs.

I am using single presicion float with SIMD instructions in "flush to zero" mode, so the CPU actually takes care that too small numbers are automatically zeroed. If I did not do that, it would not cause clicks either, cause I would just get denormalized floats which would cause performance slowdown.
But thanks for the input, I'm currently looking for the cause of this. I get clicks mainly in release so I am doing those first and hoping the cause (whatever it is) is the reason for ALL clicks

[payne_bc]

1. Negative Mod values seem to have exactly the same effect rather than inverse

I'll have a look at those.

EDIT: found the issue... but testing it now, I am wondering why and where does a negative value make sense? Maybe I should get rid of this feature?

[payne_bc]

I got clicks at the start and end of each note play by turning everything off,
so just using the single additive synth, no fx,filters etc.

Ok, found the reason for the clicks in the addsynth and fixed that. There seems to be a second source though and that is the highpass which causes the click in the release of basedrum for example. That one I'm still looking for.

http://www.tunefish-synth.com/beta6/vst24_tf3_x64.zip
http://www.tunefish-synth.com/beta6/vst24_tf3_x86.zip

[payne_bc]

Ok, thanks to RichardSemper, I noticed there is another big issue with tunefish. Depending on the size of the data the DAW requests from Tunefish, the programs can sound very different or just plain broken. Tunefish executes the modulation matrix once per frame and that results in a very poor modmatrix resolution when the DAW requests large chunks of data.
So I implemented a little piece of code that will sit in between the DAW and the Tunefish synth engine and splits up big requests into smaller ones that should result in better sound quality for some of you.
Please test it and let me know if you experience any problems with it.

http://www.tunefish-synth.com/beta7/vst24_tf3_x64.zip
http://www.tunefish-synth.com/beta7/vst24_tf3_x86.zip

[payne_bc]

Ok, this is becoming ridiculous now. I should try and limit myself to 1 release a day max
I tested now 44.1khz, 48khz, 88.2khz and 96khz and they all seem to work properly, no detuned addsynth anymore and also no broken percussions in higher samplerates.
Anybody noticing other issues in non 44.1khz?

http://www.tunefish-synth.com/beta8/vst24_tf3_x64.zip
http://www.tunefish-synth.com/beta8/vst24_tf3_x86.zip

[RichardSemper]

I implemented a little piece of code that will sit in between the DAW and the Tunefish synth engine and splits up big requests into smaller ones that should result in better sound quality for some of you.
Please test it and let me know if you experience any problems with it.

http://www.tunefish-synth.com/beta7/vst24_tf3_x64.zip
http://www.tunefish-synth.com/beta7/vst24_tf3_x86.zip

wow! another really prompt update
quick tested on my setup and sounds are now consistent whether at 50ms or 7ms latency, though still timing issues on the higher latency, so will stick to 7ms.
Unfortunately it also seems to have introduced a squeak on some sounds - notably kicks I was working on - where ADSR1 used on the frequency set to 0.7 or below in mod matrix. I'm still checking to see if its something I have done wrong.

[RichardSemper]

hang on, I'm trying to keep up here

[H-man]

1. Negative Mod values seem to have exactly the same effect rather than inverse

I'll have a look at those.

EDIT: found the issue... but testing it now, I am wondering why and where does a negative value make sense? Maybe I should get rid of this feature?

Hey Payne, first, I would like to say thanks for the ongoing dedication to this Synth and the OSC.



96kHz is not essential, certainly not for the time-being.

One example of possible use for negative mod values would be when using the ADSR to mod the frequency. Positive values would have the sound come down to the set note/frequency and with negative values the frequency would rise to that value instead.

This could of course be applied to filter cutoff, spread etc.

Given the up/down ramps included with the LFOs (and DAW automation), maybe it's not so essential however it would extend the sound sculpting capabilities of TuneFish IMHO.

Off to test beta 7

[payne_bc]

96kHz is not essential, certainly not for the time-being.

too late

One example of possible use for negative mod values would be when using the ADSR to mod the frequency. Positive values would have the sound come down to the set note/frequency and with negative values the frequency would rise to that value instead.

This could of course be applied to filter cutoff, spread etc.

Given the up/down ramps included with the LFOs (and DAW automation), maybe it's not so essential however it would extend the sound sculpting capabilities of TuneFish IMHO.

Good point. I'll try to implement it that way, could be interesting

[H-man]

hang on, I'm trying to keep up here

Holy crapcakes! Me too ....make that Beta 8

Payne_bc aka "Flash".

[chilledpanda]

I got clicks at the start and end of each note play by turning everything off,
so just using the single additive synth, no fx,filters etc.
I rendered out some single note play and stuck the wave in a sample editor. It looks like as if your not starting/ending your waveform on a zero crossover so you get clicks, but playing with it a bit more, e.g. add adsr 1 to the volume via the matrix. When setting the attack to zero then a single click is heard on the note start, setting a longer attack you hear multiple clicks at the beginning of the attack, same happens on release, short release single click, longer release more clicks towards the end of the release.
Now I maybe wrong (and I don't want to come across as someone telling their "grandma how to suck eggs") as I don't know what number representation you are using (integer/single or double floating point numbers, but guessing one of the floating points.
This honestly sounds like you are possibly using floating point numbers and the calculations with small numbers are causing floating point numbers that are too small to be represent a number and end up as NaNs.

I am using single presicion float with SIMD instructions in "flush to zero" mode, so the CPU actually takes care that too small numbers are automatically zeroed. If I did not do that, it would not cause clicks either, cause I would just get denormalized floats which would cause performance slowdown.
But thanks for the input, I'm currently looking for the cause of this. I get clicks mainly in release so I am doing those first and hoping the cause (whatever it is) is the reason for ALL clicks

No worries, the simple test i did kinda pointed in the direction of possible Input source data or scalar issues, glad your well versed enough with SIMD, so I will shutup;)

[payne_bc]

No worries, the simple test i did kinda pointed in the direction of possible Input source data or scalar issues, glad your well versed enough with SIMD, so I will shutup;)

Actually found one more reason for clicks. Polyphony. When you use a low one and an instance of a note gets kicked, it is not properly ended and will click. So watch out of the polyphony setting if you get clicking. It *might* be the reason.

Yes, I know what you think... no beta9?! no, not today. I want something to do tomorrow.