On Sat, Dec 03, 2005 at 08:16:53PM +0100, Ralph Metzler wrote:
> Can you tell me which part of the calculations in the old code is
> wrong? I have a card with such a tuner here and with the old code
> there are some tuning problems. Sometimes it tunes right away,
> sometimes it takes a very long time.
The original code in tua6100_set_tv_freq (this was in stv0299.c before all
the rearrangement on the tuner/PLL drivers started, don't know where it is
now):
R = 64;
A = 64;
P = 64; //32
M = (freq * R) / 4; /* in Mhz */
N = (M - A * 1000) / (P * 1000);
This one is much better:
int nfreq=freq/125;
R=64;
P=64;
A=(2*nfreq)&(R-1); // 1/16MHz Steps
N=nfreq/(4*8); // 64*1/16=4 MHz Steps
Additionally, I looked at the output of the switchable bandwidth filter with
a spectrum analyzer. It looks like the filter should be better used for
symbolrates >8Ms/s instead of the 4Ms/s mentioned in the SU1278 data sheet.
--
Georg Acher, acher@xxxxxxxxx
http://www.lrr.in.tum.de/~acher
"Oh no, not again !" The bowl of petunias
