[sdiy] SAW core VCO flyback time

[Ian Fritz]

Nice post.  A couple of additional comments inline.

On 9/7/2016 8:25 AM, Neil Johnson wrote:
> The problem with the original Michaels (sp?) saw core VCO (really a
> CCO with a voltage-to-current expo converter front end) is that the
> reset time is outside of the control loop, by which I mean that while
> you can derive an expression relating the input current to output
> frequency that includes a term for the reset time, you don't know what
> that time is, so the output frequency will always be an approximation.
> Franco compensation addresses this, but it is not perfect (the output
> level varies with frequency)

True enough, but it is simple to measure the reset time.  I have always 
had trouble with the Franco compensation distorting the waveform -- 
never figured out why.  I typically put in a bit of compensation here 
and get the rest from the compensation in the expo converter.  Of 
course, this compensation has the wrong functional form, but, since the 
errors are small with a fast reset time, a first-order correction is 
good enough.  (Yes, even for me :-))

> The venerable Pease-designed LM331 does this so eloquently, using what
> is termed "charge balancing".  If you look at the datasheet:
>
> http://www.ti.com/lit/ds/snosbi2c/snosbi2c.pdf
>
> you can see how it works.  Note that the critical part is the
> stability of the reset pulse timing and gain - once that is set the
> output frequency is very linear.
>
> If you consider figure 16 then that _IS_ a current-controlled
> oscillator (Rin does the V-to-I conversion for the integrator loop
> around A1, Cf).
>
> Granted, as you approach the high end of the frequency range the shape
> slowly moves towards a triangular shape, but I would argue that by the
> time you're above 20kHz you don't really care about triangle or saw or
> whatnot as any harmonics will be well above hearing range.

I just had another quick look at my LM331 design.  It is  basically the 
same as the Pease circuit.  The retrace time is 2.5 us, and the waveform 
runs from 4V to 6V.  So at 10kHz the reset time is 2.5% of the period. 
I agree that this is pretty much inaudible.  But of course a problem 
arises if you try to do a Saw-Tri waveshaping. There will always be some 
even-order harmonics mixed in.  I looked up the Chroma manuals and I see 
there are only Saw and Pulse waveforms -- no attempt at a Tri.

Ian

-- 
ijfritz.byethost4.com