On Tue, 03 Jun 2008 06:50:47 -0000 "cocokcocok" <[EMAIL PROTECTED]> wrote:
> Thanks Leif, for the explanation. It does make sense. > > Is it fair to say that 12KHz IF version will not suffer any I/Q > imbalance (phase, amplitude and DC offset)as it is created and > manipulated in digital domain. It is not necessary to go to an I/Q pair in the digital domain, but I would think most SDR softwares would prefer to do that. In the digital domain, "component tolerances" correspond to the accuracy of digital information and it can be made high enough for balance and DC offset to become perfect. > while the I/Q complex base band produced > by analog circuit (component tolerances. has problems. Yes. > It is still difficult for me to visualise why "direct conversion" > would have twice the bandwidth. You feed two different signals of the same bandwidth into the computer. A signal that is above the LO will have a certain phase relation between the channels and come out as a positive frequency from a FFT. A signal that is below the LO by the same amount will also be present in both channels, but with another phase relation and it would ideally not give any output at all on the positive side. It would instead present itself as a negative frequency. In othe words, you may send two channels of 20 kHz bandwidth each into the soundcard. Both channels have the same amplitude for the signal and the mirror image. When processed as a complex pair, I and Q together have twice the bandwidth going from -20kHz to +20kHz. 73 Leif / SM5BSZ > > Regards, > > > --- In [email protected], Leif Asbrink <[EMAIL PROTECTED]> wrote: > > > > On Sun, 01 Jun 2008 07:10:56 -0000 > > "cocokcocok" <[EMAIL PROTECTED]> wrote: > > > > > What would be the advantage of having baseband complex I/Q signal feed > > > to sound card instead of having direct 12KHz IF signals which can be > > > digitally splitted to I/Q signals for further processing. > > > TenTec and Winradio with their old models using the later approach as > > > far as I know. > > > > The problem is the mirror image. One way is to use a filter > > that passes a bandwidth of something like 10 kHz at i.e. 455 kHz. > > > > That would mean that 455 is converted to 8 kHz while the entire > > passband would be from 3 to 13 kHz. The mirror frequency that > > creates a signal at the end point of the passband would be 6 kHz > > away and a reasonable filter could attenuate it by 80 dB. > > > > The filter would guarantee freedom from spurs caused by the mirror > > image. > > > > With direct conversion there would be no high Q filter at RF > frequencies. > > One would feed two wideband audio signals into the soundcard and > > this would increase the useful bandwidth quite a lot. There would > > be a false signal at the image frequency. With 1% components one > > can expect the image level somewhere around -30 to -40 dB. Careful > > balancing with trimmers or by special software procedures is required > > to bring the image spurs to a reasonable level. The advantage with > > "direct conversion" giving I and Q is that the useful bandwidth > > becomes at least two times larger. Typically more depending on > > the shape factor of the filter in use. > > > > Here is a description of a system with the filter method that uses > > an extreme filter to get nearly the maximum possible bandwidth: > > http://www.sm5bsz.com/pcdsp/hware.htm and particularly the filter > > here: http://www.sm5bsz.com/pcdsp/pcif.htm > > As you can see it takes many crystals to get a usable bandwidth > > of 20 kHz when sampling at 44.1 kHz - but it can be done. > > The links show my first two-channel receiver in a PC computer. > > Nowadays I use I and Q with a four-channel soundcard. > > > > 73 > > > > Leif / SM5BSZ > > > > > > ------------------------------------ > > Yahoo! Groups Links > > >
