Ideally you should tune the antenna for resonance and leave it there and put a tuner at the antenna feed point to match the antenna and feedline. That's assuming you're using a 50 ohm feed line to the antenna and that your xcvr expects a 50 ohm load.
From a radiation/reception perspective there is no
advantage in the antenna itself
being resonant if you are also going to add a tuner or some other type of matching. A
5/8 wave vertical or an EDZ, for example, is not resonant, but still works well as an antenna,
and changing the length to make it resonant prior to matching changes the gain and/or pattern
so you end up with a different antenna.
Here is an explanation:http://www.vk1od.net/antenna/iobr/
For simple antennas such as a dipole or quarter wave vertical that is suitable for direct coax
feed, there is very little difference between tuning for resonance or for low SWR (as long as
you are taking the measurements right at the antenna feedpoint.) That's because the reactance
changes faster than the resistance as you vary the length, and the point of minimum SWR will
be very close to the resonant frequency - within the accuracy that you can determine it with
most SWR meters. One reason for that is that you can't have a low SWR and still have a
significant amount of reactance at the feedpoint.
What many of us end up doing is trimming the antenna for a minimum SWR to keep the xcvr happy. The antenna can easily end up being non-resonant.
For an antenna such as a dipole or quarter wave vertical it is quite difficult to invent a situation
where the lengths for resonance and for minimum SWR are far enough apart to make a difference.
As an example, some of the simulations I've run had a shift of about 10 to 20 kHz on 20m, so
either setting gave about the same SWR across the band. That's because the R component of
the antenna feedpoint impedance changes much more slowly with frequency than the X value.
Where you do run into problems, however, is when you try to measure resonance at the shack
end of the feedline without correcting for the impedance transformation along the feedline. If
all one is interested in is X = 0 on your analyzer and there is any significant length of coax
between the analyzer and the antenna, then it isn't difficult to shift the antenna resonance
clear out of band in order to get the coveted X = 0 reading.
Consider the case where you have a 20m dipole with a radiation resistance around 72 ohms
and connect the analyzer to it using a 4' jumper of RG-213. At resonance (70+j0 ohms at
the antenna) the analyzer will read 56 - j19 ohms, so it looks like you need to lengthen the
antenna even though it is already resonant. So we'll lengthen it by 20ohms of reactance:
with 72 + j20 at the antenna, the analyzer now reads 74-j19 ohms. If we lengthen it to
add another 20 ohms of reactance (which in practice might move the R component up a
few ohms, but we'll ignore it for the moment) we end up with 96 - j26 ohms at the analyzer.
You can keep going until you are blue in the face, but you'll have to make it resonant
far out of band before achieving X = 0 on the analyzer through the 4' coax jumper.
The problem is that the impedance varies along the feedline when it isn't perfectly matched.
What will give X = 0 with one length won't do so if you add a feet, and for random lengths
of coax ther is no correlation between X = 0 at the shack end of the coax and at the
antenna feedpoint (where X = 0 means the antenna is resonant.)
On the other hand, adjusting the antenna so that the dip in SWR is in the middle of the
desired operating band pretty much assures you that the antenna is resonant in the
band regardless of the length of the coax or the reactance measured at the far end.
So if you are having the problem you describe, it may be due to adjusting it for X = 0 at
the shack end of the coax rather than at the antenna end. (VK1OD's calculator will allow
you to correct measurements at the shack end of the feedline to find the actual load
impedance at the antenna: http://www.vk1od.net/calc/tl/tllc.php
This is why I'm becoming more and more interested in doublets and similar antennas. They can be close enough to resonance on multiple bands and the antenna feedpoint is within arm's length of the operating position...
Hmmm... I've rarely used a doublet with the feedpoint that close to the operator: I'd rather
have it up high in the air to improve my signal, then connect it to the tuner with a length of
The same rules still apply, of course: it doesn't matter whether the antenna length is resonant or
not; the impedance at the antenna feedpoint is transformed by the feedline connecting it to the
shack; and the tuner (if used - it might not be needed for some combinations of antenna and
feedline lengths) matches the impedance at the end of the feedline to 50 ohms.