Topics

Xtal filter capacitor values

g3oth <g3oth@...>
 

Hi,
I have been looking at the coupling capacitor values used to connect the Xtal Filter to the amplifier stages in the various popular Bitx design options (Bitx20 ver 3, Bitx20A, smdBITX and MKARS80) with which I am familiar and note that apart from the BITX20A design which employs 100pF, all the other variants employ 0.1 uF coupling capacitors for this function.

In a recent post Arv gave some general advice as follows:

"You can also play with the capacitors involved in the crystal filter. C-62 and C-53 affect transmit side impedance, and C-52 and C-89 affect receive side impedance (there is some interaction between receive and transmit side capacitors). Impedance mismatch causes ripple across the top of the passband. Capacitors to ground between the crystals affect bandwidth (increase C for less bandwidth). C-55 should always be larger than C-54 and C-56, but not always double their values."

In this context C-62, C-53 and C-52, C-89 are the values in point. Now, in anybody's book going down from 0.1 uF to 100 pF is a quantum leap so I am intrigued to learn more of what all this is about as regards mismatch, ripple etc. Qualitative rather than quantitative explanations sought at the moment please as the maths can be awesome.
Thanks es 73 de Charles G3OTH

Arv Evans
 

Charles, & others

I took a non-mathematical approach to evaluation of the BITX20A crystal
filter,
and similar filters for other IF frequencies in BITXxxA units for other
bands.

A BITXxxA testbed (a big section of PCB with the BITX20A reproduced
ugly-style)
was used for my tests of the crystal filters. Variable capacitors were
used for C54,
C55, and C56 (these caps have dials that are calibrated in pf). The
crystals are
socketed, for ease in changing IF frequency and to allow moving crystals
around
to get the best shape factor. Filter input and output coupling
capacitors are soldered
to PCB pads that allow quick changing of capacitor values.

From my experiments, using 0.1 mfd coupling caps for the crystal filter
will result
in a fairly good passband shape factor. There is some ripple (<3 db)
across the top
and bandwidth for 11 MHz crystals is usually around 2.8 to 3.4 KHz with
randomly
selected crystals. One observation that concerned me was a pop-up
frequency that
was approximately 6 KHz below the lower skirt of the filter, and about
25 db below
the filter passband level. This was present with some crystal sets, and
not there
with other sets. I am not entirely sure if this came from the crystal
filter, or was an
artifact of the BFO mixing process.

When I put 100 pf input & output coupling capacitors in place of the 0.1
mfd units,
this pop-up went away.

With the 100 pf coupling capacitors I was able to adjust the mid-filter
variable
capacitors so that ripple was reduced to less than 1 db and sideband
skirt steepness
was improved. It is possible to consistently get a passband of 2.5 to
2.7 KHz with
good shape factor by moving crystals around and optimizing the
mid-filter capacitance.

This is by no means a scientific approach to filter design. The BITX20A
design uses
the original capacitor values that Dan suggested. For BITX17A, BITX40A,
BITX30A,
and BITX80A experimental rigs I just adjusted filter components until it
looked right.
A good starting place is to calculate the impedance of each capacitor at
the BITX20A
IF frequency of 11 MHz, then determine what capacitance has the same
impedance
at your new IF frequency. This will result in a good filter, but you
can usually optimize
it a bit by making small changes to match the characteristics of your
particular crystals.

Now...before everybody decides to change their BITX crystal filter
designs, the original
BITX20 filter with 0.1 mfd coupling capacitors results in a good
sounding and efficient
SSB signal. Optimization of the crystal filter, and other parts of the
BITX20A, was driven
by need to allow kit builders to reliably build a transceiver that would
meet or exceed
the US FCC requirements for spurious emissions and signal purity. This
allows the
kit to be produced and sold in the US.

The Cohn or Ladder crystal filter design is quite forgiving and easy to
play with. It uses
same-frequency crystals and can be built with almost any reasonable
number of nodes.
If you know the crystal motional parameters, you can use AADE filter
design software
to design your filter.

<http://www.aade.com/filter32/download.htm>

This runs well on Linux/Wine, even though it is written for MS-Windows
operating
systems. Results are very close to optimum (your motional parameters
must be
accurate), but a little tweaking of the actual build helps to optimize
the filter, and
makes you more confident that it is as good as it gets.

Crystal filter experimentation and design seems to be an aspect of our
hobby that
could be all-encompassing and become a whole hobby in it's own right.

Happy filtering,

Arv K7HKL
_._

On 05/26/2010 01:57 PM, g3oth wrote:

Hi,
I have been looking at the coupling capacitor values used to connect
the Xtal Filter to the amplifier stages in the various popular Bitx
design options (Bitx20 ver 3, Bitx20A, smdBITX and MKARS80) with which
I am familiar and note that apart from the BITX20A design which
employs 100pF, all the other variants employ 0.1 uF coupling
capacitors for this function.

In a recent post Arv gave some general advice as follows:

"You can also play with the capacitors involved in the crystal filter.
C-62 and C-53 affect transmit side impedance, and C-52 and C-89 affect
receive side impedance (there is some interaction between receive and
transmit side capacitors). Impedance mismatch causes ripple across the
top of the passband. Capacitors to ground between the crystals affect
bandwidth (increase C for less bandwidth). C-55 should always be
larger than C-54 and C-56, but not always double their values."

In this context C-62, C-53 and C-52, C-89 are the values in point.
Now, in anybody's book going down from 0.1 uF to 100 pF is a quantum
leap so I am intrigued to learn more of what all this is about as
regards mismatch, ripple etc. Qualitative rather than quantitative
explanations sought at the moment please as the maths can be awesome.
Thanks es 73 de Charles G3OTH

g3oth <g3oth@...>
 

Hi Arv,
Thanks for your detailed explanation of your experiments and practical findings.
I have always obtained sure fire results from all the four homebrew Ladder Filters using values as specified for the Bitx clones and the N3ZI GCRX which uses a similar topology which I have built to date so I previously had not given the value of the coupling capacitors much thought until you mentioned them recently.
I am now currently battling with a similar ladder filter in my smdBITX build using very well matched smd xtals, 0.1 uF coupling capacitors but with signicantly lower value shunt capacitors.
However I am finding it to be very much wider and more peaky with a very deep trough than what I was previously accustomed to and what the theory and design programs would suggest from their Co and motional parameters.
I was therefore looking for any practical indicators and ways as to how best improve the response when I came across your earlier remarks regarding their influence on ripple, matching and bandwidth.
Elia in the meantime is currently working on a sweeper which should enable him to further optimise the cap values, also Inge has the benefit of a VNA so hopefully with help from Len and the other builders we will eventually crack this one.
73 de Charles G3OTH

--- In BITX20@..., Arv Evans <arvid.evans@...> wrote:

Charles, & others

I took a non-mathematical approach to evaluation of the BITX20A crystal
filter,
and similar filters for other IF frequencies in BITXxxA units for other
bands.

A BITXxxA testbed (a big section of PCB with the BITX20A reproduced
ugly-style)
was used for my tests of the crystal filters. Variable capacitors were
used for C54,
C55, and C56 (these caps have dials that are calibrated in pf). The
crystals are
socketed, for ease in changing IF frequency and to allow moving crystals
around
to get the best shape factor. Filter input and output coupling
capacitors are soldered
to PCB pads that allow quick changing of capacitor values.

From my experiments, using 0.1 mfd coupling caps for the crystal filter
will result
in a fairly good passband shape factor. There is some ripple (<3 db)
across the top
and bandwidth for 11 MHz crystals is usually around 2.8 to 3.4 KHz with
randomly
selected crystals. One observation that concerned me was a pop-up
frequency that
was approximately 6 KHz below the lower skirt of the filter, and about
25 db below
the filter passband level. This was present with some crystal sets, and
not there
with other sets. I am not entirely sure if this came from the crystal
filter, or was an
artifact of the BFO mixing process.

When I put 100 pf input & output coupling capacitors in place of the 0.1
mfd units,
this pop-up went away.

With the 100 pf coupling capacitors I was able to adjust the mid-filter
variable
capacitors so that ripple was reduced to less than 1 db and sideband
skirt steepness
was improved. It is possible to consistently get a passband of 2.5 to
2.7 KHz with
good shape factor by moving crystals around and optimizing the
mid-filter capacitance.

This is by no means a scientific approach to filter design. The BITX20A
design uses
the original capacitor values that Dan suggested. For BITX17A, BITX40A,
BITX30A,
and BITX80A experimental rigs I just adjusted filter components until it
looked right.
A good starting place is to calculate the impedance of each capacitor at
the BITX20A
IF frequency of 11 MHz, then determine what capacitance has the same
impedance
at your new IF frequency. This will result in a good filter, but you
can usually optimize
it a bit by making small changes to match the characteristics of your
particular crystals.

Now...before everybody decides to change their BITX crystal filter
designs, the original
BITX20 filter with 0.1 mfd coupling capacitors results in a good
sounding and efficient
SSB signal. Optimization of the crystal filter, and other parts of the
BITX20A, was driven
by need to allow kit builders to reliably build a transceiver that would
meet or exceed
the US FCC requirements for spurious emissions and signal purity. This
allows the
kit to be produced and sold in the US.

The Cohn or Ladder crystal filter design is quite forgiving and easy to
play with. It uses
same-frequency crystals and can be built with almost any reasonable
number of nodes.
If you know the crystal motional parameters, you can use AADE filter
design software
to design your filter.

<http://www.aade.com/filter32/download.htm>

This runs well on Linux/Wine, even though it is written for MS-Windows
operating
systems. Results are very close to optimum (your motional parameters
must be
accurate), but a little tweaking of the actual build helps to optimize
the filter, and
makes you more confident that it is as good as it gets.

Crystal filter experimentation and design seems to be an aspect of our
hobby that
could be all-encompassing and become a whole hobby in it's own right.

Happy filtering,

Arv K7HKL

Charles Cook <g3oth@...>
 

Hi agn Arv
Just in case you are maybe thinking I am not using very well matched xtals, then you will see from an extract of an email I sent to Elia and Len earlier that the xtals I used were all quite as well matched as any others I have used in the past and previously had no problem with.

QUOTE
I have now finished measuring the seven Xtals you sent up to me Elia.
I built the test rig on perf board using the actual G3UUR circuit and values, see attached photo.
I measured the total switched capacitance of my jig including strays to be 34.3 pF
Elia, I don't know if the value of 32.5 pF that you quoted was the measured value of the capacitor you used before you soldered it in circuit or whether it was like mine the actual total switched capacitance including jig strays that you measured ?
My measurements taken tonight at about 20 degrees C using a 10.2 volts supply compared with yours are as follows:

Elia's Measurements: Charles's Measurements:

No. Short (Hz) Open (Hz) No. Short (Hz) Open (Hz)
222 9999060 10000070 222 9999059 10000158
113 9999060 10000080 113 9999050 10000160
226 9999060 10000080 226 9999072 10000171
127 9999060 10000090 127 9999083 10000184
172 9999060 10000090 172 9999052 10000164
256 9999060 10000100 256 9999073 10000189
267 9999060 10000100 267 9999087 10000186

Although I used a fairly crude jig and relied on an elastic band pressure contact to a pair of wire pads, I found the measurements were repeatable to one or two Hertz each time I placed the Xtals in situ.
I allowed about 2 minutes for each Xtal to stabilise and the settling drift during this time was always less than about 5 Hz.
Each measurement taken was averaged using a Gate Time of 10 seconds.
I have no idea about the absolute accuracy of my Frequency Counter, but I think it is adequate for comparison purposes.

For some reason I see there is a spread of 37 Hz in my "Short" values compared with 0 Hz spread with yours
However there is a spread of 31 Hz in my "Open" values compared with a spread of 30 Hz with yours
The least change between my Short and Open values was1099 Hz and the most change was 1116 Hz, only a spread of 17 Hz
Whereas the least change between your Short and Open values was 1010 Hz and the most change was 1040 Hz, a spread of 30 Hz

Bearing in mind that we used two different frequency counters, jigs and switched capacitor values, the results look as if they will compare quite well.

END OF QUOTE

Regards Charles G3OTH

Arv Evans
 

Charles

You seem to be seeing the same thing I did with 0.1 mfd coupling caps
(deep trough across the top of the passband).
I think this is indicative of over-coupling (input Z is too low...?).
Try changing to I/O coupling caps with Z that matches
the filter impedance (probably around 200 ohms). Lacking a sweeper, you
can use the manual sweep method that
was outlined in an earlier email. It is slow and laborious, but does
seem to work.

Arv
_._

On 05/26/2010 05:55 PM, g3oth wrote:

Hi Arv,
Thanks for your detailed explanation of your experiments and practical
findings.
I have always obtained sure fire results from all the four homebrew
Ladder Filters using values as specified for the Bitx clones and the
N3ZI GCRX which uses a similar topology which I have built to date so
I previously had not given the value of the coupling capacitors much
thought until you mentioned them recently.
I am now currently battling with a similar ladder filter in my smdBITX
build using very well matched smd xtals, 0.1 uF coupling capacitors
but with signicantly lower value shunt capacitors.
However I am finding it to be very much wider and more peaky with a
very deep trough than what I was previously accustomed to and what the
theory and design programs would suggest from their Co and motional
parameters.
I was therefore looking for any practical indicators and ways as to
how best improve the response when I came across your earlier remarks
regarding their influence on ripple, matching and bandwidth.
Elia in the meantime is currently working on a sweeper which should
enable him to further optimise the cap values, also Inge has the
benefit of a VNA so hopefully with help from Len and the other
builders we will eventually crack this one.
73 de Charles G3OTH

--- In BITX20@... <mailto:BITX20%40yahoogroups.com>, Arv
Evans <arvid.evans@...> wrote:

Charles, & others

I took a non-mathematical approach to evaluation of the BITX20A crystal
filter,
and similar filters for other IF frequencies in BITXxxA units for other
bands.

A BITXxxA testbed (a big section of PCB with the BITX20A reproduced
ugly-style)
was used for my tests of the crystal filters. Variable capacitors were
used for C54,
C55, and C56 (these caps have dials that are calibrated in pf). The
crystals are
socketed, for ease in changing IF frequency and to allow moving
crystals
around
to get the best shape factor. Filter input and output coupling
capacitors are soldered
to PCB pads that allow quick changing of capacitor values.

From my experiments, using 0.1 mfd coupling caps for the crystal filter
will result
in a fairly good passband shape factor. There is some ripple (<3 db)
across the top
and bandwidth for 11 MHz crystals is usually around 2.8 to 3.4 KHz with
randomly
selected crystals. One observation that concerned me was a pop-up
frequency that
was approximately 6 KHz below the lower skirt of the filter, and about
25 db below
the filter passband level. This was present with some crystal sets, and
not there
with other sets. I am not entirely sure if this came from the crystal
filter, or was an
artifact of the BFO mixing process.

When I put 100 pf input & output coupling capacitors in place of the
0.1
mfd units,
this pop-up went away.

With the 100 pf coupling capacitors I was able to adjust the mid-filter
variable
capacitors so that ripple was reduced to less than 1 db and sideband
skirt steepness
was improved. It is possible to consistently get a passband of 2.5 to
2.7 KHz with
good shape factor by moving crystals around and optimizing the
mid-filter capacitance.

This is by no means a scientific approach to filter design. The BITX20A
design uses
the original capacitor values that Dan suggested. For BITX17A, BITX40A,
BITX30A,
and BITX80A experimental rigs I just adjusted filter components
until it
looked right.
A good starting place is to calculate the impedance of each
capacitor at
the BITX20A
IF frequency of 11 MHz, then determine what capacitance has the same
impedance
at your new IF frequency. This will result in a good filter, but you
can usually optimize
it a bit by making small changes to match the characteristics of your
particular crystals.

Now...before everybody decides to change their BITX crystal filter
designs, the original
BITX20 filter with 0.1 mfd coupling capacitors results in a good
sounding and efficient
SSB signal. Optimization of the crystal filter, and other parts of the
BITX20A, was driven
by need to allow kit builders to reliably build a transceiver that
would
meet or exceed
the US FCC requirements for spurious emissions and signal purity. This
allows the
kit to be produced and sold in the US.

The Cohn or Ladder crystal filter design is quite forgiving and easy to
play with. It uses
same-frequency crystals and can be built with almost any reasonable
number of nodes.
If you know the crystal motional parameters, you can use AADE filter
design software
to design your filter.

<http://www.aade.com/filter32/download.htm>

This runs well on Linux/Wine, even though it is written for MS-Windows
operating
systems. Results are very close to optimum (your motional parameters
must be
accurate), but a little tweaking of the actual build helps to optimize
the filter, and
makes you more confident that it is as good as it gets.

Crystal filter experimentation and design seems to be an aspect of our
hobby that
could be all-encompassing and become a whole hobby in it's own right.

Happy filtering,

Arv K7HKL

Arv Evans
 

Charles

I have to admit that I am a fan of "empirical design" (try it to see if
it works).
For simple ladder filters, this seems to be adequate, even though it may
not
be an elegant approach to the problem. I have an oscillator with
shorting switch
so I can perform motional parameter tests, but do not find a lot of use
for it.

From what I can see, your crystals should be perfect for filter use.
But admittedly,
I am not an expert in this area.

I would probably set up the test jig with 100 pf in & out, 100 pf to
ground except for
the middle crystal junction which would have 180 pf to ground. Then
play with the
values until it looked like a good sideband filter response.

see attached
Filter test jig suggested by John VK6JY, and one that I use frequently.
It is just a 40-pin
IC socket with some jumpers soldered on the bottom side. Input & output
caps are on
the ends, and caps-to-ground are between the crystals.

This 13.5 MHz filter with 47 pf coupling from/to 50 ohms and 68-136-68
pf to ground is
6 KHz wide and has a deep trough on the lower frequency side.

Changing the I/O caps to 100 pf narrowed the bandwidth to 4.2 KHz and
widened the
trough while decreasing it's depth by almost 1/2.

I changed the I/O caps to 220 pf, now the trough is nearly gone and
bandwidth is 5 KHz.

Next the caps to ground were changed to 68-100-68. Now the bandwidth is
up to 5.5 KHz
and the top is nearly flat with very minimal ripple.

Changing the caps-to-ground to 68-220-68 brought back a 6 db trough in
the center but
narrowed the bandwidth to 1.4 KHz.

OK...I'm beginning to see a pattern here.

Changing the caps-to-ground to 100-220-100 pf, still with 220 pf I/O and
the bandwidth is
13.495,600 to 13.493,800 (1.8 KHz). The top is flat as a board. It is
not often that I get a
really flat top, but this one seemed to fall into place.


see attached
That variable capacitor is a 3-section unit of 15 to 500 pf per
section. I'm only using 2
sections here for the outside caps-to-ground.

Now, a bit of very rough analysis...

The 220 pf I/O capacitors have an impedance of 53.6 ohms at 13.5 MHz, so
that gives me
a fairly good impedance match with the 50 ohm signal generator and 51
ohm terminating
resistor.
The 220 pf to ground between a pair of 100 pf caps to ground is probably
higher than I
would have guessed, but it works, so I'm happy.

1.8 KHz is probably the minimum one would want for a SSB filter, but it
is in the general
range of being adequate for CW, so I think I have a filter design for
the BITX15A with a
13.5 MHz IF. The BITX amplifier impedance is a bit higher, but I can
juggle the I/O caps
to get a match.
13.5 crystals are plentiful and cheap because they are used in RFID
systems, new digital
TV circuits, and in Radon brand computer video cards.

There you see my ugly empirical method for designing crystal filters.
This time I got lucky
and reached a good design in only a few steps (about an hour worth of
tinkering). There
are probably much better ways, but I'm mathematically lazy and this
works for me.

Arv
_._

On 05/26/2010 06:56 PM, Charles Cook wrote:

Hi agn Arv
Just in case you are maybe thinking I am not using very well matched
xtals, then you will see from an extract of an email I sent to Elia
and Len earlier that the xtals I used were all quite as well matched
as any others I have used in the past and previously had no problem with.

QUOTE
I have now finished measuring the seven Xtals you sent up to me Elia.
I built the test rig on perf board using the actual G3UUR circuit and
values, see attached photo.
I measured the total switched capacitance of my jig including strays
to be 34.3 pF
Elia, I don't know if the value of 32.5 pF that you quoted was the
measured value of the capacitor you used before you soldered it in
circuit or whether it was like mine the actual total switched
capacitance including jig strays that you measured ?
My measurements taken tonight at about 20 degrees C using a 10.2 volts
supply compared with yours are as follows:

Elia's Measurements: Charles's Measurements:

No. Short (Hz) Open (Hz) No. Short (Hz) Open (Hz)
222 9999060 10000070 222 9999059 10000158
113 9999060 10000080 113 9999050 10000160
226 9999060 10000080 226 9999072 10000171
127 9999060 10000090 127 9999083 10000184
172 9999060 10000090 172 9999052 10000164
256 9999060 10000100 256 9999073 10000189
267 9999060 10000100 267 9999087 10000186

Although I used a fairly crude jig and relied on an elastic band
pressure contact to a pair of wire pads, I found the measurements were
repeatable to one or two Hertz each time I placed the Xtals in situ.
I allowed about 2 minutes for each Xtal to stabilise and the settling
drift during this time was always less than about 5 Hz.
Each measurement taken was averaged using a Gate Time of 10 seconds.
I have no idea about the absolute accuracy of my Frequency Counter,
but I think it is adequate for comparison purposes.

For some reason I see there is a spread of 37 Hz in my "Short" values
compared with 0 Hz spread with yours
However there is a spread of 31 Hz in my "Open" values compared with a
spread of 30 Hz with yours
The least change between my Short and Open values was1099 Hz and the
most change was 1116 Hz, only a spread of 17 Hz
Whereas the least change between your Short and Open values was 1010
Hz and the most change was 1040 Hz, a spread of 30 Hz

Bearing in mind that we used two different frequency counters, jigs
and switched capacitor values, the results look as if they will
compare quite well.

END OF QUOTE

Regards Charles G3OTH



kc0wox <bitx@...>
 

Here's a link to a web page I made a few month ago testing a smd crystal filter.

http://golddredgervideo.com/kc0wox/smdcrystalfilter/smdfilter.htm

I don't know if I have posted it before but it fits into this discussion.

The board I was using was from http://kitsandparts.com It is very versatile and only costs $4.00.

I'll be ready to test my version 4 crystal filter soon and I'll post the results.

Just a reminder, My construction manual for the version 4 board is at http://golddredgervideo.com/kc0wox/bitxsmd/ver4/

Much of it is untested at this point. I am writing it as I build and the only circuit I have actually tested is the RF amplifier.
Leonard
kc0wox.com