LM3409: output behavior driver

The falling curve is limited by the slew rate of the inductor: L* di/dt= Vo 

You may consider lower dimming frequency, reducing inductance or consider shunt FET dimming. 

Regards 

XG 

Thank you for the information.
But I have set the same dimming frequency as in the data sheet. Therefore, I can not understand the behavior. The inductance has already been reduced.

The Vo in your application is 2-3V when inductor current decreases, 

It means di/dt is much lower than that in datasheet (VO=42V) with same inductor. 

BTW, when Vf is low, ILED is also very low. Thus, the current shown in the waveform may also includes other current like discharge current of capacitor. ILED is smaller than the tested current. 

I have simulated the shunt FET dimming. However, I probably still have an error there.
Can you please check where my error could be?

I have tried 2 variants.

Once as on the evaluation board:

and once as in the datasheet:

The load is not correct. 

With 0.075 ohm sense resistor, output current is set to be 3.2A. With load = 10k, the target regulation voltage is 32kV. You may use LED or diodes in series as load, or use resister like 1- 5 ohm as load. With 1ohm load, output voltage is around 3.2V

Thank you. I have changed the load. But I can not understand my output voltage in both cases. But the EN input is also wired correctly? It must be permanently high, right?

R64 isn't needed. Make sure that PWM shunt path impedance much lower than output load that output current can flow through shunt FET. 

EN input always be high. 

You may also monitor L6 current, Coff, and switch node voltage the same time. 

The current flows exclusively through R51.
I have used the standard mosfets from Pspice. As soon as I use models from ti, the simulation brings an error message?

Different Mosfet parameter like Rdson, Cgate and so on might lead to different performance. May consider these facts when simulating or in real application. 

I used the standard mosfets from Pspice in the last simulation. As soon as I replace them with Mosfet models from TI, an error message appears. The simulation is aborted during the calculation.

It is ok to use standard EFTs from PSPICE. I guess you met converge error as usually happens. 

How does the convergence error occur in the calculation?

However, the output voltage in the last picture is not according to the PWM. R51 was already set to 5Ohm. How can I make sure that the PWM shunt path impedance is much lower than output load?

From the simulation waveform, it seems M11 didn't turn on. Either the device properties are not set or not set correctly for this application. You may just select POWER_NMOS_P which is parameterized already.  This is 0.0059ohm device which is much lower than 5ohm. 

I have replaced the MOSFET. At least there is no error message anymore.
What is the difference to the normal NMOS that was used before?

The result still doesn't look like it should be:

You may also try changing M9. The below circuit works at my side.