Questions about ARRL review

[Douglas G. Bonett]

Table 2 of the ARRL review finally answered my questions about the relative
performance of the 2700 and 2800 Hz filters at 1 Hz and 2 Hz spacing. The
results are very impressive. Can someone explain to me why the above/below
results for these two filters are so disparate at these narrow spacings?  The
review suggests that the observed asymmetry is quite unusual. Is the performance
of a radio most accurately described by the larger or the smaller of the
above/below values? Is the asymmetry due to a design problem in the roofing
filters or something in the K3 circuitry? If the asymmetry could be corrected by
a change in roofing filter design, would the above/below values both become
similar to the average of the two values? For instance, would the 94/81 IMD
values at 1 kH spacing for a redesigned 2800 Hz filter approach 87.5/87.5, which
would be a truly spectacular result for a radio with a single 2800 Hz roofing
filter.

Doug N0HH  (K3/10 #1213)


_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Don Wilhelm-4]

Doug,

I have never seen a crystal filter that is truly symmetrical.  That is
just the "facts of life".  The real question is, just how much asymmetry
is the user willing to tolerate.  Since this is a roofing filter, and
not the ultimate filter for the K3, I tend to believe that the 5 pole
filter is adequate for my purposes.  YMMV.

73,
Don W3FPR

Douglas G. Bonett wrote:

> Table 2 of the ARRL review finally answered my questions about the relative
> performance of the 2700 and 2800 Hz filters at 1 Hz and 2 Hz spacing. The
> results are very impressive. Can someone explain to me why the above/below
> results for these two filters are so disparate at these narrow spacings?  The
> review suggests that the observed asymmetry is quite unusual. Is the performance
> of a radio most accurately described by the larger or the smaller of the
> above/below values? Is the asymmetry due to a design problem in the roofing
> filters or something in the K3 circuitry? If the asymmetry could be corrected by
> a change in roofing filter design, would the above/below values both become
> similar to the average of the two values? For instance, would the 94/81 IMD
> values at 1 kH spacing for a redesigned 2800 Hz filter approach 87.5/87.5, which
> would be a truly spectacular result for a radio with a single 2800 Hz roofing
> filter.
>  
>

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[dj7mgq]

Hi Doug,

 > Can someone explain to me why the above/below results
 > for these two filters are so disparate at these narrow
 > spacings?

Let me speculate...

you might want to read the description by PA3AKE, especially the
chapters on his development of his roofers:

<http://www.xs4all.nl/~martein/pa3ake/hmode/index.html>

He talks about, among other things, how important clean surfaces are for
good IMD performance in crystals filters.

---------------------------------------------------------------------
Quartz and IMD

The surface finish plays a big role in the linearity of the
quartz. Particles (micro dust) polluting the surface of the
quartz are known to cause IMD. To make things considerably
worse, the IMD they cause is not following 3rd order law by
any means.

(from the PA3AKE website)
---------------------------------------------------------------------

Sherwood also mentions this in the information (in one of the audio
files, if I remember correctly) available at his web site.

If one of the crystals in the filter is "dusty" then you ought to see an
immediate difference.

My 2700Hz filter, if our numbers (DD5FZ & DK4YJ) are really correct, is
better than the ARRL filter but the difference is not major. I assume
this is due to normal variance which occurs during the production
process (and minor differences in the calibration of the measurement
equipment).

Also the 8 pole 2800Hz filter is likely to have a bit more attenuation
in the passband than a 5 pole filter, which may have a minor affect on
the amount of IMD in the post filter IF amps and second mixer. All of
this could combine to explain the differences at 1kHz and 2kHz spacing.

On the other hand, at 20kHz both filters are almost identical, so a
"dusty" quartz crystal may not be part of the reason after all.


 > review suggests that the observed asymmetry is quite unusual

For many filter topologies the asymmetry is perfectly normal. Not to
forget that in a design such as the K3, it's the first 20dB to 30dB of
stop band attenuation which are really important. The final total
selectivity comes from the DSP filter and not from the roofing filter.

This said, I am playing with idea of rolling my own roofing filters for
my K3 for three reasons. Firstly, can a couple dBs be tickled out by
using extremely good crystals (expensive!)? Secondly, it should be
educational. And most importantly, HAMs just wanna have fun.


One thing the ARRL report does show, and I agree with this after having
a 2700Hz filter in my K3 for about a year now, the five pole 2700Hz
filter is plenty good. An 8 pole filter (i.e. the 2800Hz filter) does
not appear to be necessary.

vy 73 de toby

--
DD5FZ (ex 4n6fz, dj7mgq, dg5mgq, dd5fz)
K2 #885, K2/100 #3248, K3/100 #67

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Geoffrey Mackenzie-Kennedy]

Toby Deinhardt wrote on Thursday, November 27, 2008 at 9:17 AM:

> you might want to read the description by PA3AKE, especially the chapters
> on his development of his roofers:
>
> <http://www.xs4all.nl/~martein/pa3ake/hmode/index.html>
>
> He talks about, among other things, how important clean surfaces are for
> good IMD performance in crystals filters.

Toby,

I'll second that, I am using the same 9MHz crystals that Martein uses.

> Not to forget that in a design such as the K3, it's the first 20dB to 30dB
> of stop band attenuation which are really important.

Maybe, assuming that the IMD performance of the roofer and ALL stages which
follow the roofer is good enough at those spacings.

> This said, I am playing with idea of rolling my own roofing filters for my
> K3 for three reasons. Firstly, can a couple dBs be tickled out by using
> extremely good crystals (expensive!)? Secondly, it should be educational.
> And most importantly, HAMs just wanna have fun.

Something to bear in mind is that the passband insertion loss of a
quadrature type roofer using good crystals is less than that of most roofing
filters of the same bandwidth, which will increase signal levels at the IF's
input all other things being equal. The difference can be as much as 6 - 8db
in the case of a 500 Hz roofer. Also if you do build these filters, include
a shield so that the input - output hybrids do not see one another and
forget about using disc ceramic capacitors - bad for IMD!

73,
Geoff
GM4ESD



_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[dj7mgq]

Hi Geoff,


 >> Not to forget that in a design such as the K3, it's the first 20dB to
 >> 30dB of stop band attenuation which are really important.
 >
 > Maybe, assuming that the IMD performance of the roofer and ALL stages
 > which follow the roofer is good enough at those spacings.

It also depends on whether you are talking about signals within the pass
band of the roofing filter or signals in the stop band. I would argue
that within the roofing pass band the K3 does have weaknesses, but if
the roofing filter bandwidth is close to the DSP bandwidth, then these
become secondary as other aspects such as TX and RX phase noise will
start to become limiting factors, especially at narrow bandwidths.


 > Something to bear in mind is that the passband insertion loss of a
 > quadrature type roofer using good crystals is less than that of most
 > roofing filters of the same bandwidth, which will increase signal
 > levels at the IF's input all other things being equal. The difference
 > can be as much as 6 - 8db in the case of a 500 Hz roofer.

Which could be taken care of by using a 6dB resistive pad, should the
lower loss become a problem. I rather doubt that a quadrature type will
fit into the space which Elecraft has "alloted" per filter. It could be
the largest challenge if I decide to go the path which you and Martein have.


 > disc ceramic capacitors - bad for IMD!

Full Ack.


very 73 de toby


_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[KK7P]

> Table 2 of the ARRL review finally answered my questions about the relative
> performance of the 2700 and 2800 Hz filters at 1 Hz and 2 Hz spacing. The
> results are very impressive. Can someone explain to me why the above/below
> results for these two filters are so disparate at these narrow spacings?

There are really two questions here.

The first question has to do with the skirt selectivity of the roofing
filter.

If you look at the larger values of attenuation (134 vs 115, or 113 vs
93) the result is mostly due to the difference in shape factor of the
roofing filter.  Expressed differently, an 8-pole filter is likely to
have a steeper slope in its selectivity curve than a 5-pole filter.

The second question has to do with the difference between the above and
below values for a given filter when the signal spacing is less than the
nominal width of the filter.

This asymmetry is due to the way in which the K3 aligns the roofing and
DSP filter passbands.  The signal is not necessarily centered in the
roofing filter passband; rather, the signal is shifted towards an edge
to maximize the use of the roofing filter's skirt selectivity.

Why is this important?

Consider the case in which you are using a wider filter in a crowded
band while operating CW.  A huge signal appears very close by, perhaps 1
or 2 kHz away.  The use of CW REV or CW may make a considerable
difference on the impact of that signal on the receiver.  If the
passbands were centered, this tool would be less effective.

And if huge signal are on both sides?  Time to get a narrower roofing
filter!

73 and Happy Thanksgiving,

Lyle KK7P

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[dj7mgq]

Hi Lyle,

> If you look at the larger values of attenuation (134 vs 115, or 113 vs
> 93) the result is mostly due to the difference in shape factor of the
> roofing filter.  Expressed differently, an 8-pole filter is likely to
> have a steeper slope in its selectivity curve than a 5-pole filter.

But on the other hand, AFAIK, an 8 pole filter will tend to be more
likely to create higher distortion of the Nutzsignal due to group and
phase delay, ringing, etc. at each pole. So more poles is not always
better than less poles.

vy 73 de toby
_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[dj7mgq]

Hi lyle & the List,

... higher distortion of the Nutzsignal due ...

das Nutzsignal = the wanted signal

Sorry about the German which kinda snuck in.

vy 73 de toby

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Gary, W7TEA]

Thank you Lyle.  It never occurred to me that this was purposeful. I adjusted
the filter offsets with the 500 and 200hz filters to more closely "center" them.

Guess I'll believe the factory settings from now on!

tnx,  Gary W7TEA

<quote author="Lyle Johnson">

This asymmetry is due to the way in which the K3 aligns the roofing and
DSP filter passbands.  The signal is not necessarily centered in the
roofing filter passband; rather, the signal is shifted towards an edge
to maximize the use of the roofing filter's skirt selectivity.

[Geoffrey Mackenzie-Kennedy]

Toby Deinhardt wrote on Thursday, November 27, 2008 3:33 PM:

Hi Toby,

> It also depends on whether you are talking about signals within the pass
> band of the roofing filter or signals in the stop band. I would argue that
> within the roofing pass band the K3 does have weaknesses, but if the
> roofing filter bandwidth is close to the DSP bandwidth, then these become
> secondary as other aspects such as TX and RX phase noise will start to
> become limiting factors, especially at narrow bandwidths.

Putting aside the K3 and thinking in the context of receivers in general, if
you plot the Input IP3 of a ladder crystal filter vs frequency you will find
that in most cases the value of its IIP3 has not reached a maximum untill
the frequency is well into the upper and lower stopbands. For example the
IIP3 of the 500Hz quadrature ladder filter that I use starts to decrease
from + 55dbm at delta 6kHz from centre passband frequency, to a minimum of +
34 dbm at delta 500Hz, stays between +34dbm and +38dbm until reaching the
opposite delta 600 Hz, then increases back to near +54 dbm at delta 10kHz
before settling.

Now if the plot of a filter's IIP3 is superimposed onto the plot of its
frequency response it then becomes possible to determine relatively quickly,
in the context of Gain Distribution, what is the level of two or more
signals appearing in the skirt and stopband regions and their positions that
will result in noticeable IMD products generated by the filter appearing in
the passband. This is why I said "maybe" with ladder filters in mind, and is
one of the reasons why three and four tone IMD tests are used by some.

I agree with your comment about RX LO phase noise which must be suitably low
if one objective is good in-passband performance, which also requires a
bullet proof IF. Surprisingly TX phase noise has not been an issue here,
even when digging out a SSB DX station sitting between very strong 40m BC
stations.

 > > Something to bear in mind is that the passband insertion loss of a
> > quadrature type roofer using good crystals is less than that of most
> > roofing filters of the same bandwidth, which will increase signal
> > levels at the IF's input all other things being equal. The difference
> > can be as much as 6 - 8db in the case of a 500 Hz roofer.
>
> Which could be taken care of by using a 6dB resistive pad, should the
> lower loss become a problem.

A pity to lose any potential improvement in overall noise figure by doing
that, especially on the higher bands!

73,
Geoff
GM4ESD




_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[dj7mgq]

Hi Geoff,

 >> Which could be taken care of by using a 6dB resistive pad, should the
 >> lower loss become a problem.
 >
 > A pity to lose any potential improvement in overall noise figure by
 > doing that, especially on the higher bands!

Yep. It would be nice to avoid the resistive pad but any "gain" on the
high bands could be detrimental on the lower. A problem if your two
favorite shortwave bands are 160m and 10m.

Of course front end attenuation is only a button push away...


 > Putting aside the K3 and thinking in the context of receivers in...

Thanks for pointing this out.


vy 73 de toby

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Ignacy]

 > disc ceramic capacitors - bad for IMD!

If there is a problem with those capacitors at 30dbm, I am wondering about the power amps especially SS where ceramic caps are used extensively. Seems nearly all amp's IMD cam generated in those caps. Am I missing something?

Ignacy

[Jack Smith-6]

Not all disk caps are the same - it depends on the dielectric. NP0 types
are pretty good but the high K formulations, Z5U and Y5U are not so good.

I've written about capacitance versus voltage and harmonic distortion in
various dielectrics at
http://www.cliftonlaboratories.com/capacitor_voltage_change.htm

Will revisit this in the future with some RF data.

Jack K8ZOA


Ignacy wrote:
>  > disc ceramic capacitors - bad for IMD!
>
> If there is a problem with those capacitors at 30dbm, I am wondering about
> the power amps especially SS where ceramic caps are used extensively. Seems
> nearly all amp's IMD cam generated in those caps. Am I missing something?
>
> Ignacy
>  
_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Geoffrey Mackenzie-Kennedy]

Ignacy wrote:

> If there is a problem with those capacitors at 30dbm, I am wondering about
> the power amps especially SS where ceramic caps are used extensively.
> Seems
> nearly all amp's IMD cam generated in those caps. Am I missing something?

----------------------------------------------------------------

I suspect that when used in power amps any IMD products produced by ceramic
capacitors will be masked by the IMD products generated by the active
devices, be they tube or solid state. In a receiver the difference between
the level of a given IMD product generated by an active device and the level
of one of the 'parent' signals involved is much greater than it is in the
majority if not all power amps, which means that any IMD contributed by
passive devices can in some cases affect performance.

It is the NPO disc ceramics that I have found to be troublesome when used in
crystal filters and / or RF tuned circuits.  Instead I use AVX's AQ12
multilayer ceramic chip (SMD) capacitors which are rated for use at
microwave.

Jack, I look forward to seeing your RF data.

73,
Geoff
GM4ESD












_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com