[hpsdr] "static" predistortion

[Hans-Jürgen Hartfuß]

Hi,
I am the person Warren is mentioning below in his response to Jim.
About three years ago I tried this static approach with my Atlas based 
system by carefully measuring the transfer characteristic of a Mosfet PA 
with four MRF 150 (EB104, CCI), fitted the transfer function with a 
fifth order polynomial and thought by multiplying the input with the 
inverse function suitable linearization could be obtained.
Warren provided me with a modified software version which allowed me to 
implement the determined parameters into the code.
Basically it worked. In two-tone tests I obtained an improvement in IMD 
by up to about 10 dB after careful playing with the various parameters 
of the polynomial, which is not bad.
However, at the end I was  not really happy with it, since no coherent 
picture resulted. The best correction was not obtained with the 
coefficients as derived from the measured transfer characteristic, 
that's what most irritated me.
By trial and error, by just playing with the parameters, I could improve 
either the IMD3 or IMD5 by up to about 10 dB, however, the situation is 
not really reproducible as one might expect, depending on operation 
history, band, power, temperature, drive etc. In addition the 
improvement could not really be obtained in SSB QSOs.
Thus I gave up and ordered a Hermes board.
73, Hans DL2MDQ



>
> Hi Jim,
>
> I am the author of PureSignal.  Please see below for answers and 
> comments.
>
> 73,
>
> Warren  NR0V
>
>
>> How dynamic is the Pure Signal correction data, once the correction 
>> transfer function is determined?  If not very dynamic, would it be 
>> possible to start with a default transfer function, and then stations 
>> suitably equipped to measure distortion or linearity could "tweak" 
>> constants to reduce IMD?  Presumably the correction function would be 
>> several segments in a piecewise linear approximation.  This would 
>> probably have to be a set of constatnts per band.
>>
>> Would this be at all possible?  Or am I totally out to lunch?
> It is possible to do this; however, it is perhaps not very practical 
> to use.  Any change in Drive level or in antenna/tuner load impedance 
> changes the shape of the curves and requires some re-adjustment.  At 
> one point I did a "manual adjust" version, using polynomials, for 
> another person who was interested in this approach.  After 
> experimentation, I believe that it was abandoned.
>>
>>

-- 
---------------------
Hans-Jürgen Hartfuß
Am Schäperhoff 30
17498 Weitenhagen, MV