SynthMakerDigitalWhiteByte ·
I speak of course the famous Twin-T Resonator, which are located in the legendary drum machine TR-808 and TR-606. 😉
Twin-T Resonator, no way to achieve it?
From the original SynthMaker forum (2004–2014), with its members’ names and dates.
trogluddite ·
Should be possible in theory.
The twin T is basically a filter with just enough feedback (resonance) for the resulting oscillations to decay rather than become full-on self-resonance - as far as I remember the TR-machines just kick this into life with a 1ms square pulse. Some slight 'deliberate imperfections' in the Roland design mean that the resonant frequency decays slightly as the amplitude drops. So i guess that a good place to start would be a nice resonant filter (not sure what slope a T would be) with the resonance set 'just so' to get the right decay - with maybe an envelope generator to alter the filter frequency to simulate the 'imperfections' A 'perfect' T-oscillator should generate a pure sine wave that decays exponentially in amplitude, so maybe even a sine oscillator with the right envelope shape and a bit of waveshaping (for added interest) might work too.
tor ·
http://www.discovercircuits.com/Andy/Tw ... llator.pdf
For those with the skills to translat from a analog schematic to DSP 🙂
But as trogluddite mention. In practice it can possibly be archived with a moddified biquad hp and lp in a paralell impementaion. Then just test and tweak until you get the right resonance range. The oscillation could the be adjusted with frequency and resonance.
If you try my equalizer then move the "timetravel" slider all the way to 1950, then take two or three peakfilters on the same frequency and crank them up you get this oscillating effect.
So I guess it is very possible to do this 😉
tor ·
Ill see if I can put together something in the spirit of the twin T osc.. But it will not be a correct or modelled thing... just give me some time..
trogluddite ·
Yes, having seen the schematic, filters are definitely the right track - scrap my sine osc suggestion, that just wouldn't have the right transient response for the thump at the start of the note.
From the schematic you have a little unfiltered +ve feedback, and a bandstop filter in the -ve feedback path. That shouldn't be too different to a bandpass filter (signal minus bandstop) with a little feedback from output back to input - a bit like the old Moog trick of feeding the audio output back into the external input for extra oomph. Mind your speakers though - the feedback values will be VERY sensitive to the tiniest of changes.
Just given this a bit of a try - it's working OK in principle, but needs trying with different filters to find just the right one - it's just a bit too pure sine with the stock SVF - just the right decay shape though.
tor ·
Here is a example on how it could be done. Actually used element from my EQ and tweaked them a little. It is far from perfect but it is kind of neat I think. I do not know what the actuallt twin-T sound like. But in this thing you can make many different sounds. One thing is certain... you do not get a clean sine what so ever 😉
The impulse I use is a one sample @ 0.5 per 44100 samples looped.
BE CAREFUL WITH YOUR SPEAKERS AND EARS!
Enjoy:
tor ·
If you want a cleaner sine'ish sound it might work out better to use peak filters instead of the HP and LP as I used in my example here.
tor ·
Here is a demo sound made with the twin-t example but also run through my eq. This sounds more like a bass drum though:
tor ·
I found this a interesting project. So will keep on develop this. First is to go back to code in the filters with regular floatingpoint to reduce cpu load. I see alot of exciting thing that could be done here.
trogluddite ·
Yes, keeping doing - the example .osm sounded very different to anything else I've heard lately - nicely organic; no two hits sound quite identical. Reminds me a little of a freebie I downloaded a while back (name escapes me, I'll have to look it up) - all the oscillators were just white noise, so all the pitched/harmonic elements to the notes were made by very finely tweaked filtering - I guess part of the charm is that the phase relationships aren't so set-in-stone as when using pitched oscillators.
DigitalWhiteByte ·
thanks guys, after having studied a little the case, I can say this: the Twin-T oscillator is an RC circuit, in this case by a RCR / CRC, or a notch filter, with other components and the feedback system.
I am trying various ways to emulate the circuit, one would think that you could do before using a bandpass filter with a lot of resonance, but does not sound the same.
After several attempts I managed to get a decent result. The difficulty is that the trig, which serves to excite the system, consisting of a square wave of 1 ms at 56Hz, is actually after trial by a capacitor, its performance would be in effect with an attack at the highest level with an exponential release, then type an envelope, and the other problem would be to have a sort of automatic gain control to balance the feedback.
I'm working with a sampling of 96000KHz, then using 96 samples of the envelope and hold for his release even 96 samples. The result has nothing to do with a trig generated in analog ...
No one has the ability to record an analog square wave trig 1ms? I think what makes the difference are the harmonics generated in analog.
MegaHurtz ·
It`s an active bandpass filter design.. if im not mistaking. 😉 To propperly implement this you would need the values of the resistors and caps.
DigitalWhiteByte ·
It was taken from the manual of the original TR-808, in truth is far more complex, is based with a bridged T-filter.
I attach a document , Chapter 2 shows the bridged T-filter, and it seems to behave as a notch filter.
MegaHurtz ·
It`s actually a very simple filter design, only here it seems to be a passive one.? The first cap and resistor to ground are working as a highpass, The overflow, phase shift, and resistance to sum, makes it a 6db notch at 50% of the output level. (correct me if im wrong) As you may have noticed varying the frequency is no option here. And the frequency will depend on the values for C (mostly). The actual response of these is a very simple butterworth ish one. very smooth transfer curve. If you want to go outside synthmaker a little also check out RLC cirquits for resonators. The benefit is that these can be manually tuned to perfection, because the L can be bent. ( I used to test and repair these cirquits for mri machines 🙂 Too bad i dont have those power supply and frequency analysers at home..
DigitalWhiteByte ·
In my schematic I was able to simulate something like that, however, attach the link of the original pattern of the TR-808, so just for funny 🙂
TR-808 Service Manual
DigitalWhiteByte ·
I found this very good site that explains some steps to have a good resonator. Now I just try to study how to achieve it. Its complex resonator looks great.
trogluddite ·
Great post! Wish my schools maths teacher had been able to explain things as clearly as that. I like the idea of the complex integrator having frequency independent decay - sounds like a potentially nice route into some psuedo-physical modellng type stuff.
trogluddite ·
And here it is (no optimisation or anything yet) - just had to post it coz it looks so cool on my goniometer!
sunsynth ·
Awesome - me likey 🙂
Keep on doing - dude 😉
DigitalWhiteByte ·
Great Trog. This seems a good start to undertake some experiments to give a bit of analog sound. I try to optimize some code, and eliminate some of subnormal numbers
stw ·
trogluddite wrote: And here it is (no optimisation or anything yet) - just had to post it coz it looks so cool on my goniometer!
As usual a great inspiration 😉 . Thanks trog! I built a quick filter demo for the resonator to listen to its sound. I guess that thing can be tweaked into many directions and different uses.
infuzion ·
Thanks! BTW, what does the OutR & OutI do please?
DigitalWhiteByte ·
trogluddite wrote: Great post! sounds like a potentially nice route into some psuedo-physical modellng type stuff.
In the post there are many other interesting things, At this point it would take a serious FFT in Blue Stream so we can start some serious synthesis.
We hope that Malcolm examines the request.
infuzion wrote: Thanks! BTW, what does the OutR & OutI do please?
Are the outputs of the computation of complex integration in the real and imaginary part.
trogluddite ·
infuzion wrote: BTW, what does the OutR & OutI do please?
Yes as DWB says it's because of the comlex number maths having to have two parts. For audio purposes, just using one of them would be fine, it only requires both values internally - they are both exponentially decaying sine waves, just with different phases. The article suggests feeding the impulse into the R (real) input, and taking the audio from the I (imaginary) output, as it gives a cleaner 'zero crossing' start to the waveform - using the other output might be cool though if you wanted a big DC 'thump' at the start of the note for kicks etc.
infuzion ·
Ah, OK, thanks!
MegaHurtz ·
Cool stuff really neat filters in the process, thanks people!
Wrote something here? To have a post removed or your name hidden, write to [email protected].