开云体育

On Sun, Sep 2, 2018 at 12:45 PM, Henning Weddig wrote:

The result is intersting: als Allison already stated the gain of the TX stage after the first mixer is 48 dB under these assumptions. My guess is that "normally" a lower level (-10 dBm) will do better in respect of IMD resulting in nearly 60 dB of gain!

Doesn't sound at all correct.? Since the mixer output is in the near -20dbm you need 60db of gain to
achieve +40DBm (10W).? ?I believe I've been saying at least 60db is the target for some time.? What we get is
far less but that an amplifier design issue. FYI most that get maybe 1.7w at 10M are getting far less maybe 48-50db
at 10M.

Most level 7 mixers for "clean" (quoted as they are never pure or clean) output need an input 10db less than
the LO drive minus the insertion loss of the mixer.? So we start at 7dbm (or up to 10 dbM) of LO and that means
input max is -3dbm -another 7db for the mixer loss or minus 10dbm (-10).? Than another 2-4db loss for a
bandpass filter if done right so you at -14.? So you can if lucky get away with 54 db of gain but you need a
tiny bit more for losses and errors.? That is idealized sometimes? you can push the mixer harder, but you
get more unwanted products.

uBitx case is more complicated as you have ground currents inducing signals where they should not be
and a lack of filters to scrub the output.? Now add to that a output low pas filter system that is compromised
by layout.? Now you have a mix of signals that should not be there at all mixing with those that should be
plus excess gain in the IF to overdrive the RF starved ( I measured not more than 3dbm at the mixer) mixer,
output contains a lot of not mathematically predicted outputs because there are inputs not recognized like
a sample of what the power amp is putting out ( coupling though incidental and current loops).

So for a simple 3.5mhz output the mixer may have 45mhx, 48mhz, 3.5mhz, 7mhz, 14mhz, and on
with unpredictable levels and those are the likely inputs the output products are .[...]? many!

Now I did try a 45mhz low pass and it helped a little, less than 6db.? Reason was simple you
have unpredictable paths (ground currents, DC supply lines with RF...) you cannot stop.? But
some are also the nature of the?DBM and cannot be filtered with a low pass filter.? An example
of that is the 2IF (90mhz) mixing with the LO (we will use 28Mhz) of 73mhz to get both 28mhz
output and 90-73 or 17mhz.? A low pass cannot prevent the 17mhz if it is to pass 28mhz.? Also
the existing 33mhz low pass reflects sum products back the mixer so figure all the possible
sums returning to the mixer.

Remember a DBM is both a 4 quadrant multiplier and a chopped sampling system the products
out suggest both as its linear and no linear.? With clean 45mhz you still get diode limited and
squared 45mhz with harmonics circulating internally.

In short you have to map all the possible signals that may exist and their unwanted return
paths and then do the matrix of sums and differences for the fundamentals and their harmonics.

IT made my head hurt and spurtune went nuts listing all of them though the 11th order.

By then levels are no longer an issue.

Allison