TINA/Spice/LMH7322: Convergence issue with Comparator

Xavier,

A couple of minor edits and the circuit seems to converge now.
The initial sim after downloading the circuit gave an error. It looks like the THS3092 models in your schematic were corrupted.
Replacing those with a copy-and-paste from the online THS3092 reference circuit seemed to fix that problem.

After replacing the THS3092 models, the convergence problem appeared. 
Three changes seemed to help. They are included in the attached schematic 

1. TINA complained about a convergence error for a diode in the amp driving the diodes (U5 in the attached schematic).The diode mentioned in the error message was part of that model's output stage. On a hunch, a 100 ohm resistor was placed between the amp output pin and the diode D2. The rationale is that the simulator has trouble converging when the model drives a raw diode because of excessive current. The additional resistor will limit the device's output current. It may need to be tweaked for your application.
2. The THS3092 is a current-feedback part, so it needs to have some kind of feedback resistor.
   A 500 ohm resistor was added between the diode D3 and that device's inverting pin.
   I'm not sure if your design will need the gain resulting from this change, but its a start.
3. When starting the first transient sim, Zero Initial Values was chosen, That was changed to Find Calculate Operating Point.

Please let me know if you have any questions.

Regards,
John

Resonant Filter Board design_Revd.TSC

Xavier,

1. I suggested placing the resistor at the amp output because the convergence error had to do with the model's output stage. I think it also indicates that the real device would struggle to drive highly nonlinear loads like diodes. For high-speed amplifiers its good design practice to include some resistance between the output pin and a reactive or nonlinear load. This is because high-speed amps have high bandwidths - and sometimes lower compensation - and can become unstable because of phase shifts caused by tan unisolated load. A convergence error in a simulator can sometimes hint at possible instabilities in the real device.  I am pretty sure that is the case here.

2. CFAs are different from VFAs in that the CFB feedback resistor also helps provide compensation/stability. That's why CFBs are not generally used in voltage-follower configurations. If you want to preserve the topology of your circuit it might be best to consider switching that amplifier to a VFA. The other amp might be okay staying a CFB - it didn't seem to have any problems during simulation.

3. The need to start with zero initial values or calculate an initial operating point for a transient sim can depend on a lot of things: the models, the circuit, the types of devices, the sim analysis parameters, etc. If you get a convergence error at the beginning of the transient sim, its best to try changing that selection to see if it allows the sim to proceed. However, always check the behavior during the sim and the results to see if they make sense.
From the perspective of a real circuit, the devices will have to be at acceptable operating points before the circuit will work.

Hope this helps.

Regards,

John