Technical Discussion Topic
"DDS-60
Signal Purity"
I was intrigued with the "ratty-looking
signal" being produced by the DDS-60 when running the wrong (DDS-30)
software for it, either on the Micro908 or on the PIC-EL platforms. Others
had commented similarly, and the general thought was "gee, the only
difference between these two chips from a programming sense is that upper bit of
the control words that enables the PLL on the AD9851 to give a 6x reference
clock. So why aren't we see a nice-looking signal when keeping the
fundamental frequency below 10 MHz, which is 1/3 the refclock of 30 MHz??"
Well, to accommodate an easy comparison, I added an experimental mode to the
AA-908 configuration menu that selects the DDS chip that one has installed in
his system. The experimental mode allowed selection of the AD9851
*without* the PLL turned on so as to efectively use the 30 MHz clock oscillator
as the main refclock for the DDS chip. In fact, this experiment was also
designed to let me see how "clean" the direct oscillator could be as
the refclock, without the dreaded phase noise that is so often talked about with
the use of PLLs. (Note that this same experiment of turning off the PLL
control bit could also be done on the PEGen software or the PIC-EL Test Suite
program on the PIC-EL platform ... I just happened to be in the bowels of the
AA-908 code and decided to do it there.)
Okay, so here's the simple observations as seen on a spectrum analyzer ...
1) With a fundamental of (say) 1 MHz, a pair of sampling artifacts are also seen
at 29 MHz and 31 MHz, and another pair are seen at 59 MHz and 61 MHz. The
artifact pairs "converge" to 30 MHz and 60 MHz, respectively, when the
frequency dial on the AA-908 reduces the signal to 0. (Craig and I saw
this "mirroring" while debugging the PIC-EL platform trying to control
the DDS-60 card.) These artifact pairs would been seen out farther and
farther at 30 MHz spacings if it were not for the LPF on the DDS card rolling
off at about 70 MHz, per design. THIS is the reason for the
"ratty" signals being seen on an oscilloscope in the time domain ...
multiple frequencies contribute to distorting the nice sine wave we like to see
coming out of the DDS cards.
2) The part that has me still stumped is that I do NOT see a clock signal (feedthru,
or whatever) at 30 MHz, but I DO see a (small) constant one at 60 MHz. In
other words, the artifact pairs converge to an empty (signal-less) point at 30
MHz, yet the next pair converges to a small signal at 60 MHz. I'm sure the
answer for this lies in the mathematics for the sampling ... but seeing the
clock "feedthru" at 60 Mhz but not at 30 MHz is not intuitive.
3) The signal quality of the fundamental signal all by itself is very nice and
clean ... there was no discernable difference between that "direct
reference clock-generated" signal and a signal created by the 6x PLL. But
the sampling artifacts when using the 30 MHz clock with a 70 MHz LPF will start
becoming visible above 10 MHz (i.e., 1/3 of 30 MHz refclock), so this operating
mode is really quite unusable. One would need to move the LPF down to 10
MHz in order to use the 30 MHz oscillator as the 1x refclock for the AD9851.
(I commented out the selection of this 1x mode in the production release of the
AA908 v5.0i code, but it is easy to re-enable by just uncommenting the two lines
of code in the DDS Settings routine in Config_m.asm and then recompiling.)
4) As a final observation, I explored just *what* signal(s) I was seeing while
running AD9850 code on the AD9851 DDS chip. The 10 MHz signal commanded by
the AA-908 upon initial power-up is actually interpretted as a 9 MHz signal by a
DDS-60 daughtercard. The differences are caused by two reasons: (1) the
AD9850 software computes a control word based on using a 100 MHz refclk and the
AD9851 is using a 30 MHz refclk; and (2) the different weighting/scaling of the
displayed frequency digits when computing the actual control word (plus the PLL
control bit being turned off, of course). When turning the dial lower to
approach 0 Hz, the signal (and its artifacts on the SA screen) reacted similarly
to what I described above.
Anyway, I thought all this was interesting and perhaps others would find it so
too.
73, George N2APB
Back
to Micro908 Project Page
Page last updated: January
2, 2006
.JPG){kind=logo}
