Wayne's comment on Filter Selection

[DOUGLAS ZWIEBEL]

Wayne:

Thanks for your multiple posts.  Your responsiveness is beyond
impressive.  I would not doubt the CTO's point of view on his own
design.  But now, I do have other comments and I would greatly
appreciate learning how you "see" these from your point of view.

Let's talk CW.

If a triple conversion radio has a 15kc wide roofing filter (1st IF),
there is a genuine need for a narrower filter in later stages (let's
say 2nd IF).  Since we are letting in a 15kc swath of the band
(ignoring IMD and AGC pumping for the moment), we might find that the
signal we want to copy is very close (say 500hz) to an interfering
SINGLE signal.  We can use a narrower (multiple choice of selected or
variable passband type) filter to isolate the desired signal from the
interfering signal.  These filters work well at this.

It is my presumption that the K3 DSP at 15kc (some call audio), is the
equivalent of the 2nd IF "narrow" passband filter you referenced in
other radios.  All the time, other radios still have a 15kc wide (not
to be confused with the DSP operating freq) roofing filter, still
letting other signals into the front end.

If I now insert a 1kc wide roofing filter, THAT has become a limiting
factor.  I should discern no difference (other than, perhaps some
audio) between using a 2kc wide 2nd IF filter and a 1kc wide 2nd IF
filter in a qrm-free environment.  If I encounter  "single signal" QRM
nearby to where I am listening, I can still narrow the 2nd IF (say
again to 500hz) and achieve some improvement.

So...if I get a K3 and (for cw) I use a 500hz roofing filter, I can
still use the DSP to get "narrower" for any single signal, nearby
interference.  If I use a 200hz roofing filter (assuming all other
factors could remain unchanged), there is no reason to use a DSP (2nd
IF filter) any WIDER than 200 because it is already limited at the 1st
IF.  I can use the DSP to get narrower than 200 and then I will HEAR a
difference.

All of the above basically is dealing with single signal QRM.  On a
band without IMD, a 15kc wide 1st IF is fine.  I use my "narrow" 2nd
IF (or DSP) to remove local qrm.

For me, the real purpose of the "narrow" roofing filter is kill the
potential for IMD, not "nearby, single signal qrm."  There has been a
lot of discussion here already about HOW the roofing does that.
So...assuming that IMD is the culprit that we are trying to solve, I
would simply dump in the narrowest filter possible which met my
operating style.

For example, during a DX contest, I have to "tune the band" beside
just sitting on one frequency and running.  For me, I simply can NOT
tune the band with a 200hz filter; I like to (need to) hear what is
coming up as I approach it.  It gives me an "advanced notice" on
whether to keep tuning or to slow down and stop as I hit the guys
frequency.  With 200hz, signals only "pop" into the passband.  If you
"tune too fast" you'll miss them.

So I want to pick a roofing filter that is narrow enough to remove
most (if not all) IMD while still allowing me to tune the band in the
style I am comfortable with.  And I don't think that tuning the band
in 200hz "steps" is an efficient way to win a contest (or even to
enjoy tuning the band).

The same is analogous on SSB.  Using a 200hz (or 400hz) roofing filter
on SSB just won't work because the first IF is now narrower than the
single signal.  So I need something at least as wide as the minimum
for intelligible SSB reception.  This is why you recommend/offer a 6kc
filter for AM and an "FM" filter for FM (whatever width that is).

If the above is true, I want to use the narrowest roofing filter by
mode that works for my operating style only when I am having IMD
issues (some have said AGC issues also).  My use of the word
"narrowest" is key.  What possible advantage is there to "trying" a
filter a which is a little bit narrower?  If it doesn't work, I still
get IMD.  If it works "this time," will it work next time under
harsher IMD conditions?  During a contest, I don't have time or
motivation to "play around."  I want works all the time (or 99%).

So for me, it just seems way more efficient to go the narrowEST
roofing possible by mode "all at once."  Meaning that I want a single
roofing filter by mode.  And with that single roofing filter, I can
still narrow the 2nd IF via the DSP (to 200, 100 or whatever hz)
depending on the local (by frequency) qrm.

It's fine with me that you are able to "offer for sale" so many
options, but I agree 100.0% with Bill (ZV) as to why you would do
that.  It would appear to be huge overkill and is an apparent source
of confusion to the subscribers.  And saying that the competition also
offers an array of filters, but in their 2nd if is not quite apple to
apples.  I found that to be a bit misleading.

I guess the price of the variable passband roofing filter will add a
lot to the story.  If some guys operate "all modes" then they
POTENTIALLY could use a roofing filter for each: FM, AM, SSB, CW,
digital, if, and only if, they encounter bad IMD conditions (except FM
and AM which I believe need it just to transmit).

One last thought.  I have not heard the DSP output.  Perhaps there are
so many artifacts that using selectable "roofing" filters is a better
option?????  I kinda doubt it.

de Doug KR2Q
_______________________________________________
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

[wayne burdick]

[Note to all: We're planning to put any further information about
roofing filters (whether philosophical or practical) into the FAQ, not
on the reflector. Thanks for all of your great input and questions!]


Hi Doug,

DOUGLAS ZWIEBEL wrote:

> I want to use the narrowest roofing filter by
> mode that works for my operating style only when I am having IMD
> issues (some have said AGC issues also).  My use of the word
> "narrowest" is key.  What possible advantage is there to "trying" a
> filter a which is a little bit narrower?  If it doesn't work, I still
> get IMD.  If it works "this time," will it work next time under
> harsher IMD conditions?  During a contest, I don't have time or
> motivation to "play around."  I want works all the time (or 99%).

This is a multi-part answer -- bear with me  :)

As you adjust the DSP's WIDTH control (or hicut/locut), firmware
automatically selects that crystal filter which is closest to your
selected DSP passband. This means you don't have to worry about manual
crystal filter selection at all, normally -- let the K3 do the work.
The five filter icons (FL1-FL5) will indicate which crystal filter it
has selected. You can continue to narrow the DSP bandwidth all the way
down to 50 Hz in every mode, and the K3 will dutifully select the best
installed crystal filter. Slight differences in filter passband loss
are taken into account as well.

If you have a number of roofing filters, you'll have a three-point
filtering system with ultimate rejection that's basically infinite
(roofing filter, DSP I.F., DSP AF). Anyone who has suffered AGC
pumping, desense or other problems when using wide roofing filters will
notice the different the first time they use the K3 in crowded band
conditions.

You can select crystal filters manually if you wish using the XFIL
button. When you do this, the DSP passband is set to match the selected
crystal filter's passband. You can then alter it with WIDTH or
HICUT/LOCUT from there.

There are two other controls of interest that really simplify operation:

   NORM  --  This normalizes the overall system passband (crystal filter
and DSP filtering) to a per-mode value, e.g.
             2.7 or 2.8 kHz on SSB. This is a quick way to get back to
initial conditions; it also removes any SHIFT.

   I/II  --  This toggles between two "presets", i.e. two
previously-established combinations of DSP control settings
             and crystal filter selection. If you have two settings you
typically use (say wide/narrow for CW tuning),
             toggling between the two presets makes it very fast and
easy. In CW mode, I usually have preset I set
             for 400 Hz, and II set for 100 Hz.

All of the DSP/filter settings are per-mode and per-VFO. So if you have
the subreceiver installed, you can for example listen to CW at two
separate bandwidths in the two receivers. The audio can be mixed, or
one receiver's audio routed to the left earpiece/speaker and the other
to the right. This is why we wanted two high-performance receivers with
identical filter complements: versatility.


> I guess the price of the variable passband roofing filter will add a
> lot to the story.

Yes, these will in effect give you 4 to 8 filters in a single slot.
Since crystal filters are selected automatically as you adjust the DSP
controls, you can think of the variable-passband filters as
DSP-tracking crystal filters -- unique to the K3.


> One last thought.  I have not heard the DSP output.  Perhaps there are
> so many artifacts that using selectable "roofing" filters is a better
> option?????  I kinda doubt it.

The DSP subsystem has extremely high fidelity and low IMD. Until
signals reach a fairly high level, the DSP's own filtering could handle
the entire job. The roofing filters, if installed, will remove very
strong signals outside the communications bandwidth being used. There's
also pre-DSP hardware AGC that comes in at a high level to protect the
A to D converter when very strong signals are present in the crystal
filter passband.

73,
Wayne
N6KR


---

http://www.elecraft.com

_______________________________________________
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

[Jan Erik Holm]

"fairly high level"....can you present any numbers?
What is the dynamic range for the DSP?

73 Jim SM2EKM
-----------------------------
wayne burdick wrote:

>
> The DSP subsystem has extremely high fidelity and low IMD. Until signals
> reach a fairly high level, the DSP's own filtering could handle the
> entire job.
>
> 73,
> Wayne
> N6KR
>
>

_______________________________________________
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