We forgot the most important parameter. The frequency "jump". If the frequency change is large, i.e. from 10 Khz to 1 Ghz, then of course it will take longer for the tuner to equilibrate. In my scanner, when I approach a target I take 10 baby steps of 2 Khz each to zero in. Drift is negligible. I tried using the 5000 us in my scanner, and the results were the same as without it, except for the delay of an additional 50 s. I promptly removed it.

The way I see it, tuner needs some time to settle in a new frequency, but there is no fixed delay. It is situation specific and depends highly on the amount of frequency change.

HTH,

Nikos

On Tue, Jan 14, 2014 at 6:47 PM, Nikos Balkanas <nikos.balkanas@eyeonix.com> wrote:

On Tue, Jan 14, 2014 at 5:50 PM, <steve@steve-m.de> wrote:

Hi,

On 2014-01-14 15:23, Jiao Xianjun wrote:

Is there any information or hint on the settling time after new
frequency set to dongle?

That depends a bit on the tuner and the sample FIFO of the RTL2832,
but you need to drop a few buffers, as the PLL of the tuner takes
some time to lock and stabilize.

Sample FiFO should be full at all times in a DVB-T. I have never seen it empty due to interfernce, noise, etc.

Can I immediately read samples after new frequency is set by
rtlsdr_set_center_freq? If so, will the samples quality be degraded?

Kyle Keen did some experiments for his scanning application
(rtl_power), and he settled for waiting a few milliseconds and
flushing the buffer to get optimal results. See here:

http://cgit.osmocom.org/rtl-sdr/tree/src/rtl_power.c#n626

Regards,
Steve

That's a little too long, making large spectrum tuners inoperable. In my scanner I use a read immediately after the frequency
set without any problems. Of course I cannot vouch for correct frequency, but I have noticed that the frequency doesn't drift

appreciably when resetting frequency and locking to a band.

BR,
Nikos