Wanted: distortion free soft clipping circuit

I tested both variants of the clipper and like to share the results.
Here is THD vs. level for the variant with Zener diodes:
overvoltage_protection_caps_THD_vs_level.png

The THD measurement setup wasn't spectacular with a lot of mains frequency and other disturbance.
Not sure whether this counts as "soft" clipping.

Clipping at 1kHz looks unspectacular:
overvoltage_protection_caps_1kHz.jpg
Looks still good at 20kHz:
overvoltage_protection_caps_20kHz.jpg
Apparently the fast recovery diodes do a good job.
 
The variant with the relative clipping threshold has issues and the root cause is that I set the threshold to ~70% of the power supply voltage.
The TL072 has limited swing towards the negative rail and it shows as some kind of oscillation:
overvoltage_protection_oscillation.jpg
I bet this goes away with the clipping threshold set lower, like 50% of the rail voltage.
 
Not sure whether this counts as "soft" clipping.
Good question - the distortion jumps more than 60dB up with just one dB more input.
Not really soft I would think.

One cause is that the breakdown of a zener is rather abrubt beween 0V and Vz - that abruptness (sp?) is reflected in the distortion curve.
In my experience, a circuit that uses the curve of a regular diode will be softer because that curve is much softer between 0V and Vd.
As I showed somewhere before in this thread.

Jan