First take a look at the simple case below:
*************************
.param FREQ    = 2Meg
.param T       = '1/FREQ'

 Vctrl1 ctrl1 0 pulse(0   1 0 0 0 'T/2'  T)
 Rctrl1 ctrl1 0 1k
 .hb    TONES = 'FREQ'  nharms = 50

 .option tranforhb=1 post     = 1        

 .probe hbtran     v(*)
 .probe hb     v(*)
.end
*********************************
With this case, Spectre RF will give ideal square waveform and HSPICE RF gives square waves with some High Frequency effect taking ( viewing the attached picture. )

Spectre says their result is correct since the source is a square waveform; while HSPICE mentions that HSPICE's one is correct since Gibbs effect will take effect at the case.

Which is one on earth correct? Why?

Thanks

BangBang

Please note that I have checked both HB results. For spectreRF, use Flexible Balance, For HSPICE RF, use HB analysis.  I don't use Shooting Newton.

I realize that continuous and discrete time fourier transforms are different in the explanation. For continuous one, it'said that f(t) ~ sin(t) + sin(3t)/3 + sin(5t)/5 + ... while for discrete one, read Eq 1 at http://en.wikipedia.org/wiki/Discrete_Fourier_transform .
Is it the reason to explain the difference between spectre and HSPICE RF? Which one is the correct one in fact?

Thanks

-BangBang

If HSPICE RF gives what is the expected, then two things comes out:
1. The result from SpectreRF HB may be not accpected to us. Or is square source adjusted by Cadence? Do we need trust Cadence's HB for square source?  Or need file CCR against SpectreRF HB for square source?
2. For square source, then we have to use Shooting. Since under HB, much error comes out due to FFT of square, so does it for small signal analysis (for example, for phase noise analysis, will it affect phase noise curve? ). Am I right?


Thanks

-BangBang

That's indeed probably the best way of putting it: For arbitrary signals, HB only gives you accurate results for a limited number of time points that depends on the number of harmonics chosen.

If you are applying square waves to your circuit, shooting pss might really be the better alternative (see also http://www.cadence.com/community/blogs/rf/archive/2008/09/03/tip-of-the-week-whe...).