[TowerTalk] amplifiers for phased arrays

[Jim Lux]

>
> 2). Scanned phased arrays do suffer from problems related
> to mutual coupling (scan blindness). This I think is an area that
> has received a lot of research attention.

Yes.. scan blindness is a problem where the elements are evenly spaced in a
uniform environment.  As a practical matter, it's not likely to occur in a
ham installation.  As you say, a lot has to do with the mutual impedance
things, and the fact that in a typical radar array, everything is designed
to work with 50 ohms, which is not the case here.

>
> 3). At very high frequencies ferrite isolators can be used to
> solve the IMD the problem Tom describes.  At HF frequencies
> I could see how this might be a big problem. Thinking out
> of the box a little, however, if one sampled the output of each
> amplifier it might be possible to use the DSPs to predistort
> the drive signals in order to minimize IMD in a manner similar
> to what is done in cellular base station amplifiers. Even in
> single amplifier situations, this is an area where software
> defined radios using DSPs in the TX chain might be able to
> really push the state of the art (as Tom loves to point out - the
> state of the art in amateur amplifier IMD performance for
> the most part sucks big time).

Another couple interesting aspects to the IMD issue are:
1) While commercially available amplifiers may have relatively bad IMD
performance, obviously they are "good enough", since they sell (legally)
lots of them, and people use them.  While I (and probably others) would love
to have perfect amplifiers, I (and others) are not amplifier designers and
builders, nor can I afford to pay someone to do it for me on a one-off basis
for a hobby application. What I have been hoping for (for many years) is
that some ham, who IS an amplifier designer, would take that on.  There's a
lot of new amplifier design approaches out there that are efficient, have
better performance, etc., but the ham market (being relatively tiny) seems
to be willing to buy the same old triode, bipolar, and fet designs that have
been around for decades.
Prompting my original post:  If I'm constrained to buying modular amps off
the shelf, what's my best bet?

2) There is some amount of variability between amplifiers, so the precise
nature of the spurious signals created by IMD might not be correlated
between amplifiers (for instance, the phases of the spurs might be
different), in which case, the radiated field might have less IMD than
created by the component amplifiers.  I'll have to go look this up...
Normally you just worry about the level of IMD in a general way, not the
exact phases and amplitudes of the spurious signals.  You might have the
happy circumstance, like you do when combining multiple oscillators, that
the undesirable signals are relatively smaller when 'averaged out'.

>
> 4) It isn't very constructive to compare ideal installations (100ft
> tower with a steppIR) to the kinds of installations Jim is aiming
> his concept at (e.g. CC&RVille). What would be useful to the
> discussion is to focus in on a non-ideal installation and then try to
> find out if distributed amplifiers would have any advantage in that
> scenario. Say for instance, I can put three 20' whips along a chain
> link fence (no additional ground beyond that provided by the fence).
> A typical urban CC&RVille lot might be 50 x 100' with a chain link
> fence running around the backyard perimeter (say a 50' x 30'
> rectangle with the back of the house comprising one of the 4 sides
> of the rectangle). Looking at that kind of environment (and assuming
> the IMD problem could be solved with some kind of DSP based
> linearizer), what, if any, advantage would there be to using the
> distributed solid state amps versus those same amps combined
> into a single output and then split back with passive couplers to
> drive the 3 array elements?

I was thinking about creating just such a model environment.  50x100 lot,
30x50 ft house in the middle, maximum height of 30 feet for radiators,
radiate legal limit power, rf power density must be below legal limit inside
the house.


>
> 5) I am a little confused about the advantages of the distributed
> array. I can see how it provides you with fine amplitude and
> phase control vis-a-vis the individual software defined radio
> exciters for each element, but at the same time doesn't the
> requirement to provide equal power to each element diminish
> the relative advantage of that amplitude and phase control?
> In other words does having to be mindful of mutual coupling
> put a straight jacket on you that offsets the benefit of having
> the flexibility of DSP controlled amplitude and phase?

Perhaps..  you have a trade between
a) being able to do the phasing at low levels (in a DSP) and the higher
electrical efficiency of spatial combining
b) having to do the phasing at high power, and having more flexibility in
what phases and amplitudes you can feed to the radiators (since you don't
have the "equal power" constraint).

I think that for small numbers of elements(4-6), the traditional approach
(one big amp, high power phasing) is probably the one. For large numbers of
elements (10-15), the fully distributed approach is optimal.  Just like in
big radar arrays, when you start having lots of elements, the constraints
imposed by mutual coupling start to be become less significant.  Ttaking the
"negative power element" for an example (assuming you want that particular
phasing).. if you have a 4 element array, and one is negative, you have to
sacrifice 1/4 of your total power, and connect the element to a load.  If
you have 15 elements in the array, and one is negative, you only have to
sacrifice 1/15th of the total power.

>
> 6) Can you really ignore the mutual coupling problem? If I
> am driving two mutually coupled elements with 500 watts
> each from two independent amplifiers each one connected
> to one of the two elements respectively, I would have to
> provide a matching network so that the driving point impedance
> of each element is transformed to 50 ohms. In that case,
> each amplifier would see 50 ohms resistive and it would
> therefore deliver 500 watts into each element. However,
> wouldn't some of that  power going into each element mutually
> couple into the adjacent element and then get dissipated in
> the ouput of that amplifier?

Not if you choose the phases correctly.  One can, in general (but not
necessarily in specific cases), find relative phasings that radiate all the
power (500 W to each element from the amp, one element radiates 700W, the
other radiates 300W).   The tricky thing is that finding those phasings is
not trivial (as in, I don't think you can do it analytically with an
equation, like you can with other phasing approaches).  It's really a
non-linear optimization problem with N variables (the phases) and a
constraint (no reflected power) and a goal (maximize sum of fields in
desired direction).  The solution to the problem is very, very dependent on
the exact mutual impedances and phases, so self calibration is essential.

There IS something else to think about.. On transmit, where sidelobe
performance isn't hugely important, you can tolerate fairly large phasing
errors (as in tens of degrees) in the array, and the forward gain doesn't
drop very much (yes, F/B goes to heck, but for Tx, you don't care).  Say I
have 4 elements all phased for a given direction, and all radiating the same
power.  If one of those elements is now misphased by 90 degrees (and stays
at the same power), the field in the desired direction is only down a dB or
so.

>
> Okay my head hurts sufficiently now. I look forward to hearing
> from each camp.