Thursday, February 27, 2014

Signal Generator Spectra

One of the intended uses for the boards is as a low cost and flexible signal generator that I can program with simple shell/perl/python scripts. In this scenario the B-C board pair is used in the SDR configuration with the IF output used as the signal source.  This allows better sinusoids than just using the aux outputs.  It also allows frequency synthesis below the ADF4351 lower limit of ~34MHz.  To that end, I wanted to make sure I understood the spectral purity of the output across a broad range of frequencies.  The following are some notes, observations and measurements revolving around using a B-C pair as a signal generator.

The second mixer output is the sum and difference of its inputs including all of the harmonics and their pair wise sum/differences.  The following table provides an example of the difference frequencies present at the second mixer output.

F1 / F0
F0
1
2
3
4
5
1575
3150
4725
6300
7875
F1
1
2075
500
-1075
-2650
-4225
-5800
2
4150
2575
1000
-575
-2150
-3725
3
6225
4650
3075
1500
-75
-1650
4
8300
6725
5150
3575
2000
425
5
10375
8800
7225
5650
4075
2500

The input is F0 while the LO is F1.  In our case the B board output at 1575MHz is centered within the C board SAW filter (This is actually the first stage LO setting leveraging the singly balanced mixer).  F1 is chosen to generate a difference output of 500MHz.  Said differently, the frequency of interest is |F1-F0|.  The harmonics of F0 and F1 mix to produce multiples of the original difference (e.g. 2F0 – 2F1 = 2F0 – 2*(F0 + 500) = 2*500).  These are the diagonal elements of the table in blue.  The items immediately off diagonal in orange are mixing harmonics MF0 + NF1 where N=M-1.Those in green are the mixing harmonics where N=M+1. The ones of particular interest are the -1’s on the upper diagonal.  Here we have F1 - 2F0 which results in a negative frequency but is observed as a positive frequency (F1-F0) below F0.  When both the orange and green elements are taken together they form a mirror of the diagonal elements centered on F0.  As the frequency is increased, the frequencies centered on F0/F1 overlap with the difference signal of interest.  By the time 600MHz is reached the first negative -1 product is between the carrier and its first harmonic.


The following figures capture the spectra at 100MHz steps through 1GHz.  For 100MHz through 300MHz the third harmonic is visible (roughly -25dB) while at 400MHz and above the third harmonic is beyond the scale used.  If you look very close, the second harmonic is sometimes visible (roughly -31dBc or more).  This makes sense since the second harmonic of F0 and F1 is significantly smaller than the odd harmonics (F0 due to the SAW filter and F1 due to the balanced square wave synthesized by the ADF4351).  The |3F1-4F0| is visible starting at 200MHz and above and is just above the noise floor of the 7L12.

The following figure captures the spectra from 600MHz to 1000MHz, again in steps of 100MHz.  All of the second and third harmonics are beyond the horizontal scale as monitored.  At 600 MHz the |1F1-2F0| is visible at -13dB below the fundamental at 975MHz while the |3F1-4F0| is barely visible at 225MHz.  At 700MHz the |1F0-2F0| is visible at 875MHz at -16dB below the fundamental with the |3F1-4F0| barely visible at 525MHz.  At 800MHz the |1F1-2F0| at 775MHz and |3F1-4F0| at 825MHz cluster around the fundamental and are -20dB or more lower – this is one of the more problematic cases with only 25MHz separating the primary from undesired harmonics.  At 900MHz and 1000MHz only |1F1-2F0| products are visible at 675MHz and 575MHz respectively, both of which are approximately 18dB below the fundamental.

The amplitude can be adjusted from -25dBm to at least -58dBm by moving the IF mixing frequency along the SAW filter response.

While the harmonics are a little higher than I had hoped, having an easily programmable signal source available across this range is very helpful (when you don't have real test equipment).