exploring tx upconverter designs

git-svn-id: https://svn.code.sf.net/p/freetel/code@2301 01035d8c-6547-0410-b346-abe4f91aad63

diff --git a/codec2-dev/octave/tximage.m b/codec2-dev/octave/tximage.m
new file mode 100644 (file)
index 0000000..f723e17
--- /dev/null
+++ b/codec2-dev/octave/tximage.m
@@ -0,0 +1,40 @@
+% tximage.m
+% David Rpowe Sep 2015
+
+% Say we have two signals from a mixer, at f_lo +/- f_if
+% We attenuate one by 20dB with a BPF
+% We then pass through a non linear amp
+% Note non linearities are tricky to simulate as sign() gives
+% us a lot of HF harmonics that aliased back down.
+%
+% However, it does appear that non-linear PA(sign) only supresses image
+% by another 6dB.  So mixing up does have it's problems compared
+% to direct generation.  A softer limiter (sqrt) seems to do a better job.
+
+fs = 1E4;
+f_rf = 146;
+f_if = 10.7;
+f_lo = f_rf - f_if;
+
+f_rf1 = f_if + f_lo;
+f_rf2 = f_if - f_lo;
+
+a_rf1 = 1;
+a_rf2 = 0.1;
+
+t = 1:fs;
+
+sig_rf = a_rf1*cos(2*pi*t*f_rf1/fs) + a_rf2*cos(2*pi*t*f_rf2/fs);
+sig_rf_clip = sign(sig_rf).* sig_rf .^ 1/2;
+b = fir1(100, f_rf/fs);
+sig_rf_clip_filter = filter(b,1,sig_rf_clip);
+
+figure(1);
+subplot(211)
+plot(sig_rf(1:1000))
+subplot(212)
+plot(sig_rf_clip_filter(1:1000))
+figure(2)
+plot(20*log10(abs(fft(sig_rf_clip_filter))))
+axis([1 1000 0 80])
+grid