It depends how big the difference is. If there’s a gross difference it may be that the reverse tool has missed something, but in general a system’s performance will be slightly different the other way round unless it’s a unity magnification system. The image and object f/#s won’t be the same, for example. 

Hi Kaleb,

I have used this technique on a cell phone lens with about 900um of aspheric departure and got the same performance.  Something to consider is the object side in sequential mode always has a “perfect” ray bundle and the image side has all the aberrations, so only a single (Chief) ray can provide true forward-reverse comparison. 

The easiest way to ensure the reversed system is geometrically correct, I would suggest tracing a single ray in the forward configuration, save the XYZ and LMN values in image space (either Merit Function operands or the ZPL), then using the ZPL (or API), perform a direct ray trace.  You would simply need to change the signs for the LMN direction cosines to reverse the ray.  If the forward and reverse ray has the same XYZ coordinates at the conjugate surfaces, then the system is geometrically correct in the reversed configuration.

Depending on what analysis you need to do on the reversed system, you can also try to convert to non-sequential mode and setup your source anywhere you want; you don’t have to worry about “reversing” a non-sequential system.