Vocoder design

[Scott Bernardi]

jorgen.bergfors at idg.se wrote:

> I looked a little more at the schematics.
> I think the text on the website is a little misleading. If the filters are 12 dB/oct, the response will be 24 dB/oct far away from the passband, eventhough the centre frequencies are staggered.
>

You're right - the text is misleading, and I've updated it.
On the bandpass channels, each of the filters made of one opamp plus associated R's and C's contribute one pole and one zero, with 6db/octave slopes. Since there are two cascaded filters, the
ultimate slope is 12db/octave.
One the lowpass and highpass channels however, they are all poles are all zeros, so the ultimate slope is 24db/octave.


>
> On to the schematics:
> I guess that you are going to connect all OTA outputs together and buffer that. Right? I suppose that will work with current outputs like those on the 3080.

Yes, I'll be summing the output currents of the OTA's in an output stage which I didn't show.


>
> The input buffers (U3A,B) should not be necessary on every channel. One or two buffers for all channels should be enough, I think.

I just calculated the combined input resistance of all 18 channels at 3.2k at DC, which should not require an input buffer. However, I'm not sure what the AC input impedence would look like on one
of these filters - does is drop near the corner frequency?  Also, I read on PAIA's site about a mod to their vocoder, which was adding input buffers to each stage.



>
> I would go for the LM13600 or NE5517 OTA instead of the 3080.
>
>

I had originally laid out a PC board with one channel per board, so I figured a single OTA.... Using an LM13600 or LM13700 with the linearizing diodes would improve S/N ratio.
I'm not familiar with the NE5517 - is that a single OTA with linearizing diodes?


> Do you have a value for R19?

I've got a resistor missing on my schematic. R19, running to -Vee, should be a high value like 1M or so. There should also be another resistor from pin 9 to ground, on the order of 4.7k or so. I'll
correct my schematic.

> /Jorgen

--
Scott Bernardi
sbernardi at home.net