Hi,
OK I have a 45MHz crystal filter for my first IF.
The first LO ranges from 45-75MHz.
To test I am using a 45.455MHz LO, which is a pll design.
When I connect up the LO to the mixer, with no input signal I get a -80dBm
level out of my IF stage!!
The PLL is clocked off 1MHz ref divided down to 5KHz. The 1MHz ref is a TTL
signal. Could this be the cause of the problem, and if so how do I get
around it?

Thanks,
John.

On Thu, 23 Mar 2006 07:50:41 +0000, John Wilkinson wrote:

Hi,
OK I have a 45MHz crystal filter for my first IF. The first LO ranges from
45-75MHz.
To test I am using a 45.455MHz LO, which is a pll design. When I connect
up the LO to the mixer, with no input signal I get a -80dBm level out of
my IF stage!!

-80 dBm at the IF output sounds pretty good. Are you using a
high gain, wide bandwidth IF amplifier? How good is the stopband
rejection of your IF filter? Two pole roofing filters are often
not much better than -20 to -25dB.

What kind of mixer are you using? A well balanced or double-balanced
mixer will help to keep the local oscillator signal out of the IF.

The PLL is clocked off 1MHz ref divided down to 5KHz. The 1MHz ref is a
TTL signal. Could this be the cause of the problem, and if so how do I get
around it?

Thanks,
John.

PLL oscillators can be very noisy, but it would take a very poor
design to produce strong spurious signals at +/- several hundred
kHz. Designing a PLL from scratch is not easy unless you have a
spectrum analyser or a good receiver that covers the VCO frequency
+/- a few MHz. Pay careful attention to grounding and screening,
especially around the VCO and loop filter. You must keep those
nasty TTL square waves away from the VCO. Make sure the DC supply
to the VCO and loop filter is very clean and stable.

73, Ed. EI9GQ.

--
Linux 2.6.15
Remove 'X' to reply by e-mail.
Yes, my username really is: nospam

I've been seeing your postings for a while now. How you thought of
writing a book?
Seriously.

The Eternal Squire

Unless you have to move the IF osc around, why not use a crystal?
With a 5 KHz PLL reference, you might have some sidebands from this.
What is the cutoff of the loop filter?

R

John Wilkinson wrote:
Hi,
OK I have a 45MHz crystal filter for my first IF.
The first LO ranges from 45-75MHz.
To test I am using a 45.455MHz LO, which is a pll design.
When I connect up the LO to the mixer, with no input signal I get a -80dBm
level out of my IF stage!!
The PLL is clocked off 1MHz ref divided down to 5KHz. The 1MHz ref is a TTL
signal. Could this be the cause of the problem, and if so how do I get
around it?

Thanks,
John.

On Thu, 23 Mar 2006 07:50:41 +0000, John Wilkinson
wrote:

Hi,
OK I have a 45MHz crystal filter for my first IF.
The first LO ranges from 45-75MHz.
To test I am using a 45.455MHz LO, which is a pll design.
When I connect up the LO to the mixer, with no input signal I get
a -80dBm
level out of my IF stage!!
The PLL is clocked off 1MHz ref divided down to 5KHz. The 1MHz ref is a
TTL
signal. Could this be the cause of the problem, and if so how do I get
around it?

Thanks,
John.




John,

When working toward the best ultimate attenuation (in the stop band) for
filters and high gain IF stages in general, you also need to pay attention
to ground currents for the input and output of the filter as well as
(applying the same principles to) the related stages-- And the stages well
before and after the filter.

I often say; "A grounds a ground the world around.", but this is a play on
words with the old saying "a pound's a pound the world around" as well as a
tongue-in-cheek (reverse) reference (more specifically irony) to the fact
that an RF ground can be elusive if not properly understood. [[in other
words, an RF ground is NOT necessarily an RF ground...or the concept of
"any ground in a storm" will NOT "do" for RF ]]



Make it so the ground current path for the input has as little in common
with the ground current path for the output. DO NOT just make a ground
plane and expect it to be better the bigger and more complete it is.
Consider ground current exactly as you do the "real" signal path. If the
filter has only one ground pin, make separate ground planes for the input
and output and have them connect __right at__ the pin...and ONLY _AT_ the
pin --- AND from opposite directions. If there are two pins, cut the
ground plane so the two pins have NO connection directly between, but only
via mass quantities of other routes, via far distant lands from the filter.
You must force the ground currents to go where the schematic says they
should go as thought you can not use the ground symbol. Any small amount of
conductor which has both the input and output ground return currents is a
source of coupling. Sometimes MAKING CUTS in a ground plane improves
spurious responses---been there--done that.



Hope that is clear. However, if the ultimate of the filter is only 50 dB,
you're screwed. You need to do what others do, use two or more filters
separated by stages, space and, of course, ground.



ALSO, random capacitive coupling (technically referred to as "spray" by
those in the profession) must be reduced by physical separation and shields.
AND don't allow a shields to provide a ground current path that is not
following the above ground recommendations.



Hope this helps

73, Steve, K9DCI