Good work. I read your comments in the file about frequency modulation. I really don't think you need FM if you’re already doing additive synthesis. As far as I understand, those 80s synths only used FM/PM to create complex sidebands because the hardware back then only had a few oscillators and couldn't handle even 7 or 8 sine waves for real additive synthesis. I think going the additive route is the way to go with DSP and the CPUs we have now. Just watch out because too many harmonics can still overload slower CPUs. I found that out the hard way in my own experiments with Xpressive performing STFT on speech ‘samples’ where I was forced to prioritise harmonics. Posted by: ewanpettigrew  on 2026-04-04 00:26:31

I've been enjoying using (some version) of your piano preset for some time now. I don't think I've published anything with it yet though... Well, thanks for the hard work you've put in it! And I love the demo song too. Posted by: NiftySardine  on 2026-04-06 17:58:45

@ewanpettigrew I want to try out FM synthesis, because there are many other instrument samples that I have tried to synthesize with my Python script, but none of them work nearly as well as this one upright piano sample that I used. It might have to do with the sample having a higher quality timbre than the other samples, but even just trying another piano, like a grand piano, a lot of those overtones sound like they're missing. I hope that using frequency modulation might capture the more complex samples that I am going for, and the synth even allows me to control the strength of the frequency modulation with envelopes and LFOs, which might be used to change the timbre across time. But, I don't know how any of this works yet, let alone how I could make my script automatically determine what the best parameters are for the frequency modulation, to match the original sample as closely as possible. Posted by: sampletext  on 2026-04-06 23:23:51

@NiftySardine This project has been a year and a half in the making, yet there is still so much I can do to make the piano better. I also adopted a technique that I saw in one of your projects, where you assigned a peak controller to its own channel, but the channel is not sent to any other channels. I found the frequency that the peak controller samples at, so I made a synth preset to create a wave for the peak controller at that frequency, which gets rid of the jitter that sometimes happens when the peak controller samples lower amplitude portions of the wave. Posted by: sampletext  on 2026-04-06 23:34:02

@Sampletext I've had a think about this. My idea was that the process would be to perform an FFT to find your target frequencies and amplitudes, and then find the ratio between the carrier and modulator. The next step would likely be to match a modulation index between Bessel amplitudes and the FFT amplitudes. I thought since you would already be performing an FFT, why then perform the next steps when you can go straight to plugging in the frequency, phase, and amplitude? However, I can now see the benefit in your idea. Maybe your patch was not sounding as good as you wanted due to the number of sine waves you were using, and to add more would increase CPU load causing jumping in playing. FM would be able to let you use say, 6 sine waves to generate complex overtones, whereas increasing the resolution in additive synthesis would require plugging in the frequency, phase, and amplitude of dozens of individual sine waves per note. Therefore, looking forward to hearing the results of your experiments with FM. Posted by: ewanpettigrew  on 2026-04-09 11:01:48