On RF PCBs to bend a trace 90° you have many choices but among them Curved and Mitered bend considered as a good choice from performance POV (Both shown below).
For many years I thought that if you have enough space on your board, curved bend is a better choice over mitered bend but lately I hear an opposite recommendation from one of my colleague.
My question is in case of having enough room which one of these options is a better choice? (simulated results or real world measurement is appreciated)
Neither a mitred nor curved bend is as 'good' as the equivalent length of straight track.
There are two main aspects to goodness, S11 and S21.
S11. Other things being equal, width and thickness of trace, dielectric performance, the curved bend can be designed to have good S11 to a higher frequency than the mitred. That's because the mitre is effectively a lowpass filter. The two 135 degree corners produce a slight extra capacitive loading, the thinner region in the elbow of the bend a slight series inductance. With a properly designed mitred bend (that mitre you illustrate is not properly designed, more should be taken off the corner, see below) the result is a matched 3rd order filter with good S11 up to a certain frequency. As the curved bend has smaller parasitic loadings, the good S11 passband is wider.
S21. A mitred bend can be made more compact than a curved bend, and will therefore have lower track loss, and allow tighter component packing. Either might be critical for your application, but I think there's quite a tendency to prefer to make things small these days.
The performance differences are small, and the design would have to be very marginal if they made the difference between working or not.
This is more like the proportions I'd expect to see from a correctly designed mitre, from microwaves101.com