Ok, i had a class in cfd many years ago, and i'm trying to refresh some code i wrote back in the day. It was for euler equations, using the MacCormack method for supersonic flow over a supersonic airfoil. See a contour plot of the rho*u field here:

http://home.mn.rr.com/curlydave/cfdsample.jpg

I am using forward differencing in the predictor, and backward differencing in the corrector. I have tried rotating the differencing direction every other time step but it seems to have convergence problems. The data shown is with 2nd order artificial viscosity of 0.01 or so, which i think is already pretty large. Does the trend in the figure suggest any particular problem with my code?

Be gentle, i'm an experimentalist.

I added the solution for a subsonic case, and it looks nice and smooth.

http://home.mn.rr.com/curlydave/cfdsample.jpg

These results are pretty good, for both subsonic and supersonic, considering that you are using a McCormack scheme, which uses no 'upwinding'. What is your complaint?

If you want to get a 'second opinion' on your Euler solutions, try downloading my MicroTunnel shareware and run both cases (free 3-day trial period). MicroTunnel is based on Van Leer's flux vector splitting, which is a very 'robust' upwind scheme (robust meaning numerical viscosity is already built in to the scheme).

www.microcfd.com

It would help to see the airfoil in your contour plots as well. I assume it is one of those very thin supersonic airfoils!?!

Thanks for the reply axel. Yeah it is a thin airfoil assumption, the boundary is purely rectangular but you impose a v component such to match the slope of the airfoil. It is hokey but that is what we did in the class.

Again i'm not a cfd guy so if this result is ok with MacCormick that is news to me. Thanks for the input!