Opamp oscillator with amplitude stabilization

[etf]

It is known R4=R5=1.2k. My task is to calculate amplitude of oscillations at v0:

Opamps and diodes are ideal.

I did analysis of circuit without amplitude stabilization and I got oscillation frequency, but I can't figure out how diodes stabilize amplitude in this circuit.
Any sugestion?
Thanks in advance.

Edit: I simulated this circuit in Multisim:

It looks like amplitude is 8.954*sqrt(2)V.
One strange thing: how can voltage on diode be 3.409V ??
I was thought that when diode is direct biased, voltage drop on it is about 0.7V.

 

[NascentOxygen]

In your simulation, V₄ and V₅ should have opposite polarities.

One diode conducts during part of v_o's positive half-cycle, and the other diode conducts during part of v_o's negative half-cycle. When conducting they supply additional heavy negative feedback (an additional path for signal v_o to feed back to the OP-AMP's inverting input).

A diode has 0.7V across it when conducting, but can have any voltage at all when reverse-biased.

 

[LvW]

Because the gain is only slightly larger than the critical value (3.03>3) we can assume that the diodes will start to conduct (and to reduce the gain) for values smaller than 0.7 volts (perhaps 0.4...0.5 V).
Without oscillation and for small oscillation amplituts both diodes are biased in reverse direction.
Only amplitudes above a certain value will slightly open the diodes - thereby reducing the gain (increasing feedback).

 

[rude man]

I think the diodes are shown in backwards. Your reading of U2 is impossible with the circuit as shown.
If the diodes are reversed then the negative clipping feedback thresholds make sense.
In your simulation did you run Vcc and Vee to the op amp, and what were the voltages if so?

Since the problem stated to assume ideal diodes, don't worry about 0.5V vs. 0.7V or whatever. Assume forward V = 0V.