[BC] "Cusp Rotation"

[Phil Alexander]

On 16 Mar 2007 at 9:49, Jeff Welton wrote:

> Phil;
> 
> I have to respectfully disagree on one point you raise below and 
> offer an adversarial perspective on another.  Other points have been 
> duly taken under advisement.... Once the thread dies, I'll condense 
> it and pass it over to Engineering, so they can use the input from the field.
> 
> You suggest phase shifting to put the PA load so that it appears 
> "horns up" (oriented around a 12:00 axis).  This is counter to what I 
> have been taught and seen where the horns should be oriented to the 
> left or right (3:00 or 9:00 axis... what Gray calls drunken smiley 
> faces).  Either 3:00 or 9:00 rotation will work for IBOC, although a 
> 3:00 orientation provides a bit of pre-emphasis that can be 
> beneficial to the analog.  Incidentally, among other literary 
> endeavours that are about to be, or have already been, published, 
> Bobby Cox at Kintronic Labs wrote a very detailed and well researched 
> article about this topic, available on their website:
> http://www.kintronic.com/site/techpapers/TP-AM_Case_Studies.pdf

Yes, I'm familiar with the article although I haven't 
read it for a couple of years. As I recall, that article 
was based on a Nautel wasn't it?

However, that is oriented to the Tx output, thus not 
relevant to what I was trying to say.

I'm not sure we have a disagreement at all. In fact, 
methinks we are speaking the same things differently. <g> 

If I understand you correctly, you are referring to the 
Smith at the ***output*** of the box, in which case I 
fully agree.

I was speaking of the Smith at the output of the ***PA*** 
before matching or combining.

> I don't disagree with the concept of having a measuring 
> point for the PA outputs... the trick is to make it 
> physically close enough to the actual PA output point(s) 
> so that it doesn't induce any phase delay of its own, 

Exactly, which is why it would be a really neat Tx addition
IMHO.

> or to provide a correction chart.  In addition, due to 
> safety concerns, it would need to be either interlocked 
> or shielded to prevent direct human contact with the hot 
> conductor.  All of this is going to affect equipment cost.

No, a sample is what I said. That shouldn't be much hotter
than the BNC for driving a mod. mon.

> Our position historically has been that since the antenna 
> system needs to be optimized for HD Radio anyway, the cost 
> to the station is less if we provide the phase delay 
> information publicly so the consultant can factor that 
> into the optimization process, as opposed to adding cost 
> to the transmitter.  

Yes, BUT, the actual phase delay and the design phase
delay depend on what happens in tuning during test at
the factory, and they may not be quite the same. As I
understand it, the only way Nautel can provide the exact
rotation is by customer request for factory measurement
during final test. (And I understand you do that upon 
request, so I do appreciate that point.)

> Correct me if I'm wrong here, but if you have a Smith 
> chart plot with specific starting, ending and center 
> points, how is it harder to set them up with a network 
> analyzer so that they are oriented around, for example 
> 11:00 as opposed to 3:00?  Obviously antenna work is not 
> my strong suit, so this is a chance for me to learn 
> something from the perspective of those who are doing 
> it regularly.

First, consider that the ONLY important point is that the
load on the ACTUAL PA is a j0 load that is essentially
symmetrical (IOW something near Hermitian symmetry). What
has to be done after the combining/matching network is
intended to simply achieve that objective and the specifics
are not really important as long as the job is done. 

Second, consider that shutting down a array for measurement
is not a trivial event because it means taking the station
off the air and getting someone with the right tools to do
the measurement. Once upon a time this was the case with
common point impedance measurements, but that brought 
about the common point version of the OIB and eliminated
the problem. In the case of "rotation", the true problem is 
the match at the PA output, not the box. Thus it seems to
me the best place to address it is in the transmitter.
That is where the information is always available.

Third, consider that transmitter frequencies are changed 
in the field, and the best way of assuring the best match
to the PA(s) would be measurement at their output.

> Regarding the anto phase correction and/or manual rotators in the 
> transmitter network, there are various arguments on both sides of the 
> fence.  One argument against would be the need for measuring a 
> complex impedance at carrier and at both sidebands, then either 
> displaying it, or using it to drive the auto circuit.  However.... 
> what happens if (oh my beating heart) the load is near perfect and 
> the levels are too close to be measured?  

Dreamer. <g>

> What happens if the load is 
> really bad and the correction circuit goes off the rails trying to 
> fix something that won't pass HD, no matter how much 
> tweaking?  Designing for all these possibilities again impacts cost.

No, I agree limits would be needed, and these might need to be 
determined and set at commissioning.

> The problem as I see it - and this is probably a topic for a 
> different thread - is that we have to design the units to keep the 
> guys who have to install them and work on them happy... but we also 
> have to keep the price low enough to make the guys who pay for them 
> happy as well.  Unfortunately, in the majority of cases, those guys 
> are not the same person and frequently they don't share the same 
> objectives.  That's probably one of the biggest challenges any 
> manufacturer has to face.  There's no question that we could design a 
> box that would perform virtually any task any engineer could ever ask 
> for - but if it's twice as much as another box that makes the same 
> output power, regardless of the added features, it's not going to 
> sell well enough to keep it in the stable.

Understood and agreed. However, and this is the point to consider, as
matching to the actual PA(s) go(es) so goes the bit error rate of
digital transmission. In the digital world, as Bobby Cox's paper shows,
audience is essentially an inverse function of bit error rate. This
is a fact the guys who pay for the hardware will "get" in the future.

Keeping in mind that ANY necessary phasor adjustment needed for 
monitor point maintenance may upset rotation, it seems to my poor
befuddled brain that dynamic measurement of PA match becomes essential
just as dynamic measurement of CP Z became essential with the advent
of 24/7 programming and reduced staffing. It further seems that the
best place to do that would be at the point that must be satisfied,
namely the PA itself. There must be a way, and the first Tx 
manufacturer who does it economically would gain what the guys who
pay for the iron call a "market advantage."

Thanks for the reply Jeff, I always enjoy your insightful viewpoint.
BTW congrats on the new job. It is a great move on Nautel's part
to put knowledgeable people in the field when others are making all
the wrong moves.

All the best of luck to you.

---------------------------------------------
Phil Alexander, CSRE, AMD
Broadcast Engineering Services and Technology 
(a Div. of Advanced Parts Corporation) 
Ph. (317) 335-2065   FAX (317) 335-9037