revisiting that Bode Phaser

[Haible Juergen]

	>Yes, I see now. The signal is split into two branches,
	>so only phase difference matters.

... and the all pass function common to both PDN branches
actually adds to (or subtracts from) the external all pass network
suggestet by Bode. In fact, there are 3 notches without an external
all pass filter, just from the "average" phase shift of the two PDN
paths. So you'd get a phaser even without anything extra, just FS
(now I call it that way again (;->) ) mixed with straight signal. Only
that the notches are so widely spaced in the frequency range that this
is boring - thus an external all pass network.

	>Suggestions for multipliers?
	>Do you think that an external predistortion (antidistortion)
	>stage for a 1496 will be a good idea? Recently a parcel
	>from Reichelt arrived with some nice japanese dual transistors.
	>I think it could be worth a try to use them to upgrade the
	>1496 to the linear 1495.

My first choice would be a couple of RC4200's - probably no trimming
required (?) - bleedthru is hardly an issue at sub audio range !

Even the PDN might (?) give decent results without component selection
or trimming. We're not aiming for a flat frequency response after all.
Somebody should check the theory if any of such errors result in
unsatisfactory notches. Mabe we get away with one single trimpot
to adjust for optimal cancellation in the notches ??
(Math experts, your turn !)

One thing I am after is a generalized CV input. You know it works fine
with quadrature LFO and with a joystick moved in circles. And with
Bode's proposed method of sync'ing two LFOs (I'm paraphrasing)
you can emulate linear shifts over several cycles. But there are two many 
constraints, such as reversal while zero crossing.
What I want is a linear input CV that is converted to sine and cosine
functions _over_several_periods_. Is there something like that ?
An EPROM / DAC even ?

JH.