Re: Upsampling by 4 in DSPs
- From: "markt" <takatz@xxxxxxxxxxx>
- Date: Thu, 01 May 2008 11:00:33 -0500
1- For fs=4IF: Let us say that IF=30 MHz and fs=4IF=120 MHz. In this casefs=(M/N)IF.
if the last mixing stage before ADC (which is from RF to IF) produced
harmonics then we will have at the input of ADC the IF singal and second
harmonics (2IF) and maybe 3rd Harmonics (3IF). Now 3IF=90MHz and it will
allias with the original IF at 30 MHz (since fs=120MHz). That is why I
found some people go for what is called "non-IQ sampling" where
So by choosing for example fs=125MHz. This problem will not exist.the
To tell you the truth. May be my understanding is wrong since I found
majority goes with fs=4IF but here is the link from which I got the ideaof
non_IQ sampling (slide 24):
http://cas.web.cern.ch/cas/Sweden-2007/Lectures/Web-versions/Schilcher-1.pdf
This is what a bandpass filter (or lowpass if your IF is low enough)
around your IF is for.
But the first case is subsampling and it is more prone to jittereffects. > the second one is oversampling which is less sensitive to
jitter. My
question is, if we want to oversample, should we choose fs=4IF or the
other one. How can I decide?
I disagree. The effect of jitter on SNR is a function of your analog
input frequency, not your sample rate. Plus, in the 4IF case, you have to
run your ADC at a much higher frequency. Generally, bandpass sampling is
done at higher a higher IF, which will result in increased noise due to
jitter and maybe what you're getting at. However, given equal IF, both
will suffer the same (though I could also argue that the increased Fs in
the 4IF case makes things tougher, too).
2-for upsampling at pulse shaping filter by 4. Does this number have anygo
relation with phase and frequency tracking capabilities? To my
understanding I know high upsampling ratio for pulse shaping relaxes the
synchronizer circuit. But I am not sure. If not, then it makes sense to
with low upsampling ratio to relax RRC filter order. What do you think?
No relation to phase and frequency tracking that I know of, at least not
directly. What matters most is the bandwidth of the data itself. For a
given data rate, a higher interpolation rate translates to a higher sample
rate at the DAC.
A lot of the pulse shaping is done to relax your transmitter
amplifier/filter requirements (reduces the peak to average power ratio by
limiting bandwidth). I worked at a company once that spec'd a Tx
amp/filter combo for the rms power (DS-CDMA system, just under 45 dBm rms
power). Since the crest-factor (essentially PAPR) was on the order of
12-15 dB, our peak power was in the neighborhood of 57-60 dBm, or nearly 1
kW. The cavity filter we were using had burn marks in it. Oddly, the
amplifier hung tough, but its performance did begin to degrade after a
while.
Your receiver can operate on any oversample rate you desire (with a
matched filter for best performance). Using a different rate on the
receive end simply works out to a different RRC filter order for matching
(assuming you transmit with an RRC).
Mark
.
- Follow-Ups:
- Re: Upsampling by 4 in DSPs
- From: John
- Re: Upsampling by 4 in DSPs
- References:
- Re: Upsampling by 4 in DSPs
- From: ytach
- Re: Upsampling by 4 in DSPs
- Prev by Date: Re: IFFT of custom frequency array - confusion with the time domain sampling frequency
- Next by Date: Re: Demodulating AM pulse width coding (long post)
- Previous by thread: Re: Upsampling by 4 in DSPs
- Next by thread: Re: Upsampling by 4 in DSPs
- Index(es):
Relevant Pages
|
Loading
