For an IF filter, I'm planning on using a an off-the-DigiKey-shelf crystal
filter. I need (well, want) to match the filter to 50 ohms.

Could someone verify I'm interpreting the data sheets & terminology correctly?

Say I'm getting an ECS-10.7-7.5A filter, data sheet he
http://www.ecsxtal.com/pdf/mono.pdf ... I'm reading the "termination
impedance" as 1800 ohms in parallel with 5pF or, at 10.7MHz, 1800 || -j2975 =
1318 - j797.2 ohms. From reading the application notes on ECS's site, the
idea here is that -- for the filter to work as designed -- it wants to "see"
1318 - j797.2 ohms on both the input and output, correct?

I'm also thinking that just because the filter wants to see a certain
impedance doesn't necessarily imply that the filter's own internal impedance
is the complex conjugate.

Thanks,
---Joel

Joel Kolstad wrote:
For an IF filter, I'm planning on using a an off-the-DigiKey-shelf crystal
filter. I need (well, want) to match the filter to 50 ohms.

Could someone verify I'm interpreting the data sheets & terminology correctly?

Say I'm getting an ECS-10.7-7.5A filter, data sheet he
http://www.ecsxtal.com/pdf/mono.pdf ... I'm reading the "termination
impedance" as 1800 ohms in parallel with 5pF or, at 10.7MHz, 1800 || -j2975 =
1318 - j797.2 ohms. From reading the application notes on ECS's site, the
idea here is that -- for the filter to work as designed -- it wants to "see"
1318 - j797.2 ohms on both the input and output, correct?

I'm also thinking that just because the filter wants to see a certain
impedance doesn't necessarily imply that the filter's own internal impedance
is the complex conjugate.

Thanks,
---Joel

Andy writes:
Joel, whenever I put in a Xtal filter, I make a parallel tank on the
input
and output so I can tune the reactance of the tank to minimize the
ripple in the filter. This is usually part of the source amp, anyway,
and having a high Q tank on the output is only a couple more parts.

You are correct in that the filter "wants" an 1800 ohm source and
load resistance. This can be achieved by impedance matching, or
by termination resistors ( whilch will lose a lot of your power ). It
is
good for the entire termination to be the input of the next device,
transformed up or down.

The same can be done with the source impedance. However,
the source may not like the Xtal filter as a load, since, even tho you
may have the transformation exactly right, the load presented to
the source is very complex, and may be very high ( or low) as soon
as you leave the xtal filter bandpass frequency..... Consequently ,
I have always used a load resistor in the output of the source
amplifier, either as part of the coil Q, or as a fixed value resistor
....... just to
keep the load from going very very high outside of the bandpass
and making the source amplifier go unstable......Sure, I lose a bit
of gain, but a stable amp is more important than an extra few db
of gain....

Finally, the last adjustment is made by sweeping the bandpass
while diddling the input and output tuning until the flattest response
is noted, or at least the response the manufacturer specifies......
I peak the response at the center of the banpass first, then
sweep back and forth to get a flatter response. Usually the
peaking at the center is very close, tho..and if you have no
way to sweep, it will probly work well for voice type signals...
With data signals, well, it just depends.....

It may be useful to build a little jig to test the filter in, using
fixed resistors for source and load, and parallel tanks on both
sides, just to see if the filter does what you think it does. There is

a lot of loss here, but you would only be looking for the response.
If you don't have a sweeper, maybe you can just rock the
generator back and forth with one hand while diddling the
tanks with the other......

Good luck..

Andy W4OAH

"AndyS" wrote in message
ups.com...

Joel Kolstad wrote:
For an IF filter, I'm planning on using a an off-the-DigiKey-shelf
crystal
filter. I need (well, want) to match the filter to 50 ohms.

Could someone verify I'm interpreting the data sheets & terminology
correctly?


The same can be done with the source impedance. However,
the source may not like the Xtal filter as a load, since, even tho you
may have the transformation exactly right, the load presented to
the source is very complex, and may be very high ( or low) as soon
as you leave the xtal filter bandpass frequency..... Consequently ,
I have always used a load resistor in the output of the source
amplifier, either as part of the coil Q, or as a fixed value resistor
...... just to
keep the load from going very very high outside of the bandpass
and making the source amplifier go unstable......Sure, I lose a bit
of gain, but a stable amp is more important than an extra few db
of gain....


A large increase in a receiver amplifier load impedance just outside the
passband of a crystal filter (or other types of filter) can cause adjacent
channel susceptibility. The network between the amplifier and the filter
should be designed to prevent this occurrence. There are networks that can
reverse this effect, so the impedance drops instead of increasing.
Simulations and impedance measurements can verify.

Bill W0IYH

William E. Sabin wrote:
"AndyS" andysharpe@
A large increase in a receiver amplifier load impedance just outside the
passband of a crystal filter (or other types of filter) can cause adjacent
channel susceptibility. The network between the amplifier and the filter
should be designed to prevent this occurrence. There are networks that can
reverse this effect, so the impedance drops instead of increasing.
Simulations and impedance measurements can verify.

Bill W0IYH

Andy comments:

A good point. But I am not aware of any passive networks than
can roll the impedance off ( or bring it up) fast enough to handle
the adjacent channel problem. Adjacent channels are usually
(in my experience) only a few Khz away while the xtal filter is
usually many Mhz, and a passive network just doesn't change
the impedance that quickly. For those problems, I have always
had to go to a passive resistor source, such as a collector resistor,
or use a coil with a Q low enough to provide the same effect.

If I am incorrect in this, could you please elaborate on what
passive matching circuit you are reccommending. ?

Andy W4OAH

"AndyS" wrote in message
oups.com...

William E. Sabin wrote:
"AndyS" andysharpe@
A large increase in a receiver amplifier load impedance just outside the
passband of a crystal filter (or other types of filter) can cause
adjacent
channel susceptibility. The network between the amplifier and the filter
should be designed to prevent this occurrence. There are networks that
can
reverse this effect, so the impedance drops instead of increasing.
Simulations and impedance measurements can verify.

Bill W0IYH

Andy comments:

A good point. But I am not aware of any passive networks than
can roll the impedance off ( or bring it up) fast enough to handle
the adjacent channel problem. Adjacent channels are usually
(in my experience) only a few Khz away while the xtal filter is
usually many Mhz, and a passive network just doesn't change
the impedance that quickly. For those problems, I have always
had to go to a passive resistor source, such as a collector resistor,
or use a coil with a Q low enough to provide the same effect.

If I am incorrect in this, could you please elaborate on what
passive matching circuit you are reccommending. ?

Andy W4OAH

If you have the July 1970 QST there is an article starting on page 35 that
discusses this situation. The load impedance at the drain of the JFET
amplifier increased very rapidly just outside the passband of the xtal
filter as shown in Fig 8. The network in Fig 9A inverts the impedance change
as shown in Fig 9B. This approach was tested for this article. Pages 39 and
40 show the text that applies to this situation. I have confirmed this
approach a few times since then. However, if a fixed resistor is used to
load the drain of the JFET this effect is pretty well mitigated. A large
increase in the load impedance that the amplifier transistor sees can cause
the JFET drain to become overloaded just outside the passband, and I
confirmed this in a homebrew receiver that used a particular xtal filter.
Different filter design approaches (xtal filter black magic) might produce
results somewhat different than mine, so the problem and the solution always
need to be verified experimentally. In general the gain of the JFET etc.
must be restricted so that drain overload is avoided at all frequencies, as
the article explains. Also, if you have the 2004 ARRL Handbook, page 17.5
column one mentions a similar problem that can occur.

Bill W0IYH

Thanks for the help, Andy... I muddled through it (it's a little odd the first
time you've normalized your Smith chart to 1800 ohms and the termination is 50
ohms!) and the filters do work as their data sheets say.

I did find that xtal filters can easily drive a network analyzer nuts... even
for looking at a single frequency, I had to significantly increase the
sampling time well above the default to get a stable display. I'm thinking
this is due to the absurdly high Q of the crystals causing a very long
transient response, thereby necessitating a long sampling time to finally
measure the steady state response.

---Joel

Joel Kolstad wrote:
Thanks for the help, Andy... I muddled through it (it's a little odd the first
time you've normalized your Smith chart to 1800 ohms and the termination is 50
ohms!) and the filters do work as their data sheets say.

I did find that xtal filters can easily drive a network analyzer nuts... even
for looking at a single frequency, I had to significantly increase the
sampling time well above the default to get a stable display. I'm thinking
this is due to the absurdly high Q of the crystals causing a very long
transient response, thereby necessitating a long sampling time to finally
measure the steady state response.

---Joel

Andy writes:

I liked using the old Telonic sweep generators, with the two big
dials, and a diode detector/scope to scan the filter bandpass.
I don't have one, or access to one anymore, and have to do
it by hand.....
But, what the hell. Retirement gives me a lot of time to
twiddle knobs...

Andy W4OAH

Joel Kolstad wrote:
Thanks for the help, Andy... I muddled through it (it's a little odd the first
time you've normalized your Smith chart to 1800 ohms and the termination is 50
ohms!) and the filters do work as their data sheets say.

I did find that xtal filters can easily drive a network analyzer nuts... even
for looking at a single frequency, I had to significantly increase the
sampling time well above the default to get a stable display. I'm thinking
this is due to the absurdly high Q of the crystals causing a very long
transient response, thereby necessitating a long sampling time to finally
measure the steady state response.

---Joel

Andy writes:

Joel, one more thing comes to mind.

I like to match impedances using a "double Smith chart". On this
chart,
a normal smith chart is in one color while the mirror image is in
another
color.....
It is called an "impedance-admittance" chart.

One can follow the series reactances along one color, and then
the parallel reactances along the other, to get from the starting
point to the ending part....

Once a person is comfortable using this method, it gives a LOT
of gut feel insight to the various combinations of matching networks
you can use.....

I had to design an automated antenna tuner for Texas Instruments
once, for the HF SSB marine system, and I used the double Smith
Chart to figure out what the combinations needed to be, and then
wrote a computer program that would step reactive elements in series
or parallel . The automated system therefore just stepped along the
curves exactly as I had done by hand..... That was about 30 years
ago.

Anyway, the Smith Chart, and various derivative versions, is great
because you can see a much larger 'big picture" rather than just
discrete values at discrete frequencies...... It's a good and very
useful skill to have... among others..

Andy in Eureka, Texas W4OAH

"AndyS" wrote in message
ups.com...
I liked using the old Telonic sweep generators, with the two big
dials, and a diode detector/scope to scan the filter bandpass.
I don't have one, or access to one anymore, and have to do
it by hand.....

Perhaps I'm spoiled by having access to VNAs and spectrum analyzers with
tracking generators. :-)

It's sad that *good* spectrum analyzers are still quite spendy, even on the
used market. There are several "homebew" spec. ans. out there that are quite
impressive, but none that come close to even something like a 15 year old HP
8594.

---Joel