DSP Robotics Support

Hello Flowmates!

Continuing my theme of using FlowStone to simulate sounds similar to real instruments, here’s my attempt at making Harpsichord-like sounds without using samples.
The sound will not be as authentic as a sample-based plugin, but it has a lot of flexibility and a small hard drive footprint.

Here’s my YouTube video:

https://youtu.be/_ii9qZWwFxw

Download it here:

https://www.dropbox.com/s/dh103k1oum8ov ... 2.zip?dl=0

The schematic was made with FS 3.06 and the 32 bit and 64 bit plugins were made with the current FS4 alpha. Lots of background info is included, for those who wish to explore it, and my User Guide.

Have some Baroque fun!

Cheers

Spogg

Nicely done!

Looking back on your work, I'm thinking maybe the only two methods of synthesis you haven't delved into are pulsar train synthesis (like Hamburg Audio's Nuklear, for a theoretical basis there's https://asa.scitation.org/doi/abs/10.1121/1.4744478), and scanned synthesis, (as implemented in Enzyme, an interesting system of "hammers", "nodes" and "weights", here's another academic paper http://www.billverplank.com/ScannedSynthesis.PDF)

Spogg wrote:Continuing my theme of using FlowStone to simulate sounds similar to real instruments, here’s my attempt at making Harpsichord-like sounds without using samples.
The sound will not be as authentic as a sample-based plugin, but it has a lot of flexibility and a small hard drive footprint.

Little is known that Mozart composed a lot of his work on his little home harpsichord, for example the opera "The Magic Flute". The harpsichord still exists, because it was given to the Mozarteum by his son early on. Here you can hear it being played just a few years ago:
https://www.dropbox.com/s/llcqafwkmbbtej9/CLAVICHORD_P59306_Andante_KV616.wav?dl=0

Imagine, that is exactly the sound, Mozart heard when composing "The Magic Flute"! Fascinating.

Duckett wrote:Nicely done!
Looking back on your work, I'm thinking maybe the only two methods of synthesis you haven't delved into are pulsar train synthesis (like Hamburg Audio's Nuklear, for a theoretical basis there's https://asa.scitation.org/doi/abs/10.1121/1.4744478), and scanned synthesis, (as implemented in Enzyme, an interesting system of "hammers", "nodes" and "weights", here's another academic paper http://www.billverplank.com/ScannedSynthesis.PDF)

Thanks Duckett!

The first link takes me to just an abstract. Reading that it seems like they’re describing granular synthesis, but maybe there’s more to it. Curtis Roads has a free “Pulsar generator” synthesiser but it’s Mac only.

Scanned synthesis reminds me of my Quilcom NINE:
viewtopic.php?f=3&t=4419&start=0&hilit=quilcom+nine
The interesting aspect in the paper you linked to was the method of generating the cross-fade (scan) control signal. Not just an envelope, LFO or modwheel, but a whole method of influencing a physical model in real time. Food for thought…

Cheers

Spogg

Apologies for the first link only being the abstract, lack of due diligence on my part

Here's an actual lecture in pdf from 2006 from Roads, discussing PS- I realize that this is 14 years old and therefore not "cutting-edge" for granular synthesis in general, and I'm not sure if he's discussing aspects you haven't looked into, but I'd hazard to guess that from the bottom of page 2 to page 4 is the most useful: https://pdfs.semanticscholar.org/36ea/94a82ce58abee7102e6a6cf45ac86c41c888.pdf

Also, just for completeness' sake, here's the Nuklear manual pdf- relevant material starts with page 24 (pulsaret waveforms and envelopes shown starting page 21): http://hamburg-audio.com/NUKLEAR_GB.pdf

I think I had NINE in mind when posting my first reply and mentioning scanned synthesis.
Food for thought indeed- AFAIK, Humanoid Sound Systems has been the only dev to really dive into that PM-based method.

Oops-almost forgot: I've always enjoyed the Moppelsynths family of freebies (Sonitarium probably being the best-known),
and their developer was kind enough to write and post an article about the concept behind them: https://www.verklagekasper.de/synths/dsfsynthesis/dsfsynthesis.html

Here's to having interests that can be indulged quite safely in 2020.. Cheers

Great work as always! I have a lot of free time lately so here are some GUI touches (and finally my 3.0.6 works ok ). Your last 2 instruments are on fire, Spogg. Great sounding, perfectly built. I love them very much. And by the way, only after finishing building the sidechaining system I've posted lately, someone told me that you already posted something way better. On 2016! Can't believe I missed that... So sorry for not paying attention and thanks for lovely the contribution!

Cheers.

https://www.imageupload.net/image/jE23g

I downloaded the original and will download the new by Kortezz, thanks and I will comment soon.

Duckett wrote:Here's an actual lecture in pdf from 2006 from Roads, discussing PS- I realize that this is 14 years old and therefore not "cutting-edge" for granular synthesis in general, and I'm not sure if he's discussing aspects you haven't looked into, but I'd hazard to guess that from the bottom of page 2 to page 4 is the most useful: https://pdfs.semanticscholar.org/36ea/94a82ce58abee7102e6a6cf45ac86c41c888.pdf

Also, just for completeness' sake, here's the Nuklear manual pdf- relevant material starts with page 24 (pulsaret waveforms and envelopes shown starting page 21): http://hamburg-audio.com/NUKLEAR_GB.pdf

Oops-almost forgot: I've always enjoyed the Moppelsynths family of freebies (Sonitarium probably being the best-known),
and their developer was kind enough to write and post an article about the concept behind them: https://www.verklagekasper.de/synths/dsfsynthesis/dsfsynthesis.html

Well I finally got round to looking into this stuff and I must thank you for the links.

I looked at the paper on DSF and nearly passed out with all that maths. Something for Martin maybe, but not me.

But I’ve had a very interesting afternoon playing with the NUKLEAR demo, having read and re-read the Roads paper first. I think I get it, having analysed the waveforms with a scope while adjusting parameters with knobs and settings. They’ve done a really good job and once again, I’m so envious of developers than can do so much with such relatively low CPU. The aliasing isn’t too bad either (but present for some settings and high notes).

Essentially they take a single cycle waveform as a grain and modulate the speed of playing the cycle. The repetition rate is determined by the midi note played. So unlike a more conventional granular synth or effect, it’s not based on scanning a loaded wav sound clip. This can lead to a type of formant and/or filter sweep sound.

I explored something rather similar in my Quilcom Informant, part of which was based on a patent by Yamaha and used in their FS1R synth. This made grains based on scanning a sinewave and applying an envelope (”window”) running at fundamental pitch rate. The sinewave was scanned at a higher frequency than the fundamental pitch and was not necessarily an integer ratio to it. This emulated a static formant without having to use a filter (it was digital hardware synthesis in those days).

viewtopic.php?f=3&t=8405&p=34330&hilit=quilcom+informant#p33866


So, the pulsar train synthesis is indeed something I haven’t done yet, and may well tackle at some point. So many thanks for the idea.

Cheers

Spogg

Spogg wrote:I looked at the paper on DSF and nearly passed out with all that maths. Something for Martin maybe, but not me.

Yes, the typesetting is really terrible, I agree. I remember having posted a DSF based oscillator once (was it on the old SM forum?) and then never used it again. Dunno.

martinvicanek wrote:

Spogg wrote:I looked at the paper on DSF and nearly passed out with all that maths. Something for Martin maybe, but not me.

Yes, the typesetting is really terrible, I agree. I remember having posted a DSF based oscillator once (was it on the old SM forum?) and then never used it again. Dunno.

It wasn’t the typesetting that caused my anguish!

I’d love to see what you posted if you can find it and be bothered. Or, a simplified (very simplified) description of how it works.

Cheers

Spogg