[hpsdr] Question about understanding Digital Predistortion

[John Laur]

I have been reading a lot about predistortion, and I have noticed a few
conflicting opinions in the literature that I am trying to resolve in my
own mind.

Here is a chart to visualize during this discussion:
http://www.mathworks.com/cmsimages/63380_wl_91990v00_commscope_fig7_wl.jpgwhere
the red is signal before DPD and green is signal after DPD.

Regarding PA efficiency, there are claims that DPD increases efficiency
because the PA's output power only goes to the transmitted signal. Yet on
comparison plots from a spectrum analyzer, there seems to be no additional
power in the signal when predistortion is applied. It seems to me that
digital predistortion would technically make the PA marginally less
efficient (at least insofar as ratio of total input to total output power
is concerned) as the cancelled power would be converted to heat instead of
RF.

However, the flip side of course is that the signal to noise ratio appears
to be increased. In the spectrum linked above, there is intermod on the
sidebands -10dB from the signal peaks, so we might assume that that signal
has about a 10dB SnR.  The signal with DPD applied brings the sidebands
down about 25dB is it also cancelling intermod that would be present within
the signal itself and thus increasing the SnR?

I would be very eager to see the results of testing digital predistortion
with a mode like FreeDV's 16-carrier QPSK/FDM modem which is very sensitive
to amplifier distortion and performs well under low SnR conditions. I
suspect this mode would see extreme benefits from this technique at both
high and low power levels.

Thanks for your opinions,
John KF5SAB
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openhpsdr.org/pipermail/hpsdr-openhpsdr.org/attachments/20131213/96462ca8/attachment-0003.htm>