Strange comment in wikipedia about TB303's square waves

[Kriminal]

Wiki is hardly a reliable source for any kind of information...

[Urs]

On a side note, a highpass filtered squarewave looks exactly like the description on wp: It's a squarewave with with a falling slope on the +1 and a rising slope on the -1 bit. Which one might easily confuse with a sawtoothish appearance or even, uhm, even harmonics.

The threshold/comparator theory makes sense. A square wave that oscillates between 0 and 1 has to be highpassed to get rid of the dc-offset (easier to use an RC-thing than wasting another opamp to subtract 0.5), which would explain the highpassed appearance...

[antto]

Urs: probably, btw, i find a lot of other web pages about "The Roland TB-303" (generaly, small ones) using the same text from that article on Wikipedia..

and i've seen other music-related topics from wikipedia, being discused at other forums that are incorrect.. guess they are right too..

well, wikipedia can give you a simple idea about something, but if you want to really know the "truth" then you will want to dig deeper, and not blindly trust wikipedia
that's what i think i'll do from now on

[uvacom]

I don't really know the method by which the TB-303 generates its square wave (although whatever it is, it has a unique sound). But I think the idea behind which this is suggested is more like the negative cycle is "reversed" rather than flipped. Basically odd harmonics are produced by symmetry and even harmonics are produced by asymmetry. A square, a triangle, and a sine wave are all symmetric about the X axis and so only produce even harmonics. A sawtooth is perfectly asymmetric about the x axis, so it produces odd harmonics. But flip the negative cycle and it becomes completely symmetric.


So this:

Code: Select all

|\ |\ |\ |\ |\ |\ |\ |\ |\ |\ |\ |\ 
  \| \| \| \| \| \| \| \| \| \| \| \|

Becomes this:

Code: Select all

|\ |\ |\ |\ |\ |\ |\ |\ |\ |\ |\ |\ 
  |/ |/ |/ |/ |/ |/ |/ |/ |/ |/ |/ |/

This was somewhat useful in hardware design a long, long time ago, because triangle, sawtooth, and something resembling a "square" wave could be generated all from an up/down counter and some basic logic circuitry without any processing or RAM necessary. I don't know how or if there is an efficient implementation of the same concept in the analog domain, if there is I've never heard of it.

[Urs]

well, wikipedia can give you a simple idea about something, but if you want to really know the "truth" then you will want to dig deeper, and not blindly trust wikipedia
that's what i think i'll do from now on

Absolutely! I spent another enlightning day reading Hal Chamberlin's book (errm bible?) to figure out where wp is just a tad too, hmmm, lacking. I love wp though for all the things I don't know about.

[antto]

i agree about the Transistor/Comparator version
it was a cheap synth, i don't believe there are 2 saw oscs in there really

but i guess the PW could be adjustable, if it compares the Saw waveform and "splits it in half" then some DC offset could be added with the Saw signal, and it'll split it at a different place (the amount of DC offset will adjust the PW)

the "strange" look of the Square is due to the HighPass + Resonant LowPass filters after the Osc (as mystran and Urs said)

it's not so difficult to achieve the "look" of it
look at this.mp3 in a sound editor, it looks damn close to the audio materials i got of the real thing..

[Gamma-UT]

i agree about the Transistor/Comparator version
it was a cheap synth, i don't believe there are 2 saw oscs in there really

but i guess the PW could be adjustable, if it compares the Saw waveform and "splits it in half" then some DC offset could be added with the Saw signal, and it'll split it at a different place (the amount of DC offset will adjust the PW)

In practice, it's easier to alter the point at which the waveshaper switches rather than mess with the sawtooth generator. It's a matter of adjusting the resistors used to bias (or control) the behaviour of the transistor. Apparently R45 is the one: http://www.ladyada.net/forums/viewtopic.php?f=13&t=6389.

I've edited the wiki entry BTW: just a matter of waiting to see if an over-enthusiastic admin rolls back the change.

[antto]

yeah, i meant adding DC offset to the comparator (on input, combined with the saw), not messing with the sawtooth osc in any way..

over-enthusiastic admin? hm..
tell me if he does..

[Urs]

yeah, i meant adding DC offset to the comparator (on input, combined with the saw), not messing with the sawtooth osc in any way..

That's the usual way, yes.

[mystran]

Anyone for any interesting ideas of where the asymmetry in the resonance comes? It must be something clipping somewhere where there's DC offset (?) present because at least the samples on xox site seem to show that it stabilizes back to symmetry further into a long note.

edit: nevermind, probably a later stage... turns out the resonant frequency stays constant, which couldn't possibly happen (as far as I can tell) if it was different resonance.. so somewhere after the filter, but before the hipass it gets clipped asymmetrically..

[mistertoast]

I highpass my squares and they still look very squarish. What do other people set their cutoff to? (And my highpass is only 12db per octave, should I be going for something a bit more severe to go after any bias that might creep in?)

I assume the point you guys make is that the filters (low-, high-, or all-pass) will affect the phases of the various partials, and that's why the waveforms from the oscillator can look so different from the waveforms coming out of the filters.

I completely understand that.

[mystran]

I highpass my squares and they still look very squarish. What do other people set their cutoff to? (And my highpass is only 12db per octave, should I be going for something a bit more severe to go after any bias that might creep in?)

Try something like 100Hz.

[mistertoast]

I highpass my squares and they still look very squarish. What do other people set their cutoff to? (And my highpass is only 12db per octave, should I be going for something a bit more severe to go after any bias that might creep in?)

Try something like 100Hz.

I figured 30 hertz. A bass guitar in standard tuning goes as low as 41 hertz and a bass with extra strings goes as low as 31 hertz.

[nollock]

nvm. Should read whole thread before replying.

[mistertoast]

I don't really know the method by which the TB-303 generates its square wave (although whatever it is, it has a unique sound). But I think the idea behind which this is suggested is more like the negative cycle is "reversed" rather than flipped. Basically odd harmonics are produced by symmetry and even harmonics are produced by asymmetry. A square, a triangle, and a sine wave are all symmetric about the X axis and so only produce even harmonics. A sawtooth is perfectly asymmetric about the x axis, so it produces odd harmonics. But flip the negative cycle and it becomes completely symmetric...

That's interesting. It makes a lot more sense than what was in the wikipedia article (since changed, of course).