VC Phaser using the LM13600

[Magnus Danielson]

>>>>> "jb" == jorgen bergfors <jorgen.bergfors at idg.se> writes:

 jb> Hi all.
 jb> I designed and built a test circuit inspired by the EH Small Stone. 
 jb> Instead of the CA3094, I wanted to use the LM13600. The test
 jb> circuit consists of an exponential voltage to current converter
 jb> and two all-pass stages. In an actual phaser more than two stages
 jb> of course would be used. The Small Stone has four, some other
 jb> phasers have six. Phase shift is around zero at 0V CV and
 jb> increases with increasing control voltage. 

 jb> The following was measured om my test circuit:

 jb> Frequency Phase shift (0 to 10V CV)
 jb> 18 Hz 45 - 350 degrees
 jb> 37 Hz 30 - 340 degrees
 jb> 75 Hz 20 - 330 degrees
 jb> 150 Hz 15 - 300 degrees
 jb> 300 Hz 10 - 250 degrees
 jb> 600 Hz 0 - 250 degrees
 jb> 1200 Hz 0 - 110 degrees
 jb> 2400 Hz 0 - 55 degrees

 jb> As can be seen, there is some phase shift at lower frequencies
 jb> even when the control voltage is 0 volts. The frequency response
 jb> falls to -3 dB at 24 Hz. This is the worst case, at 5V CV. At 10V
 jb> CV it seems to be flat down to DC. At 0V CV it goes down to 18 Hz.
 jb> There is no HF damping (below 20 kHz). 

 jb> Noise and distortion seems to be low. I haven't measured them,
 jb> but they aren't noticable on the scope or at listening tests. 
 jb> Note that phase shift is higher att low frequencies. Is this
 jb> common to all phasers or are there phasers that are the other way
 jb> round (higher phase shifts at high frequencies)? 

Well, the phase distorsion from a single pole or zero becomes greater
the nearer you come to them. In the s-plane your above analysis would
appear as a number of test-points on the jw-axis. The classical
allpass filters in phasers (such as the EH SmallStone) are built in a
1-pole/1-zero structure, where all the poles and zeros will be real,
and thus placed on the sigma-axis (which is real). I don't recall
exactly how the zero's moved when you made the feed-forward, but they
move. Now, when you place your analysis point on 18 Hz (113 rad) you
place yourself very close to origo (and thus very close to the sigma
axis) and this will be the closest point to the poles and zeros and
thus see the greatest phase distorsion as compared to the higher
frequencies (which are inherently farther away). As you modulate you
move the poles and zero on the sigma axis, closer (or further away) to
the jw-axis, and then will the higher frequency points become closer
and thus become more affected by the preccens of each of the poles and
zeros. In the allpass filters will the phase contribution of a pole
and its corrensponding zero be additative, so whatever phase
distorsion that the single pole would contribute with, the zero will
add exactly (well, in theory) the same.

With all this in mind... yes, if you allowed for complex poles and
zeros in your allpass filter function, you could move the effect. For
a full bank of allpass you could optimize them to acheive flattness
and all of a sudden you are making a Hilbert transform approximation,
oups!

For a Phaser, what really intests you is where zeros ends up (when
including the feed forward part), since the effect when these move all
over the place is at least in my mind much of the phaser sound, not
the phase itself.

Cheers,
Magnus