First analog design -- knee function generator

[twilightsentry]

First, let me apologize if this is the wrong forum for posts relating to analog processing; my search for similar results found things here, so...

I'm starting my first project in the strange, nonquantized (at the level I care about) world of analog: designing a compressor. Before I design the whole thing, I want to present what I have and see if I'm on the right track and understanding analog properly. So far, I've sketched out a rough outline of the compressor's structure. Only the function generator for the compressor's knee is designed down to the circuit level.

For the parts of the circuit that need a VCA/multiplier, I used an op-amp driving a pair of mosfets. Both mosfets are in voltage dividers; one generates the op-amp's feedback, the other divides the input, forming the output. The second input is a control voltage applied to the op-amp's positive input. I didn't use an OTA-based circuit, as my grasp of current sources is not nearly as good as my understanding of voltage sources.

The circuit implements either a soft knee or a hard knee. The function for the hard knee is: (assuming that 'i' is the circuit's input, 't' is the threshold setting, and 'r' is the ratio setting)



It's implemented by the upper signal path in the diagram, which generates (t-i), multiplies by the ratio, and adds i.

The soft knee is defined as a quadratic bézier curve with endpoints at 0 volts and twice the threshold and a control point at the threshold. This generates a nice, smooth curve. (Well, I assume that it's a nice curve. I won't know how any of this sounds until I actually build it.)



The soft knee circuit starts with a log-sum-antilog section, whose intermediate results are labeled in the diagram. It then multiplies by the ratio and subtracts from the input.

And that (along with the actual circuit design, of course) is what I've got so far. I'd really appreciate any commentary/flaming/[telling me that I should become a luddite farmer and never touch technology again] you'd like to give!

[GK]

Those BJT based log and antilog generators are great for textbook analysis but not so practical in real life due to huge junction temperature dependence.

Cheer,
Glen

[twilightsentry]

Many thanks for the reply, Glen.

I think that I've gotten the thermal voltage and saturation current to cancel out (are there any other temperature-dependent factors?), assuming that the three transistors are matched and on the same die. I'm leaning toward a MAT04 for that.