250 Hz and 400 Hz Filter Measurements

[Joe Subich, W4TV-4]

 > Sure enough, with band noise as input to the K3, I see a similar
 > spectrum as Jim showed here
 >
 > http://audiosystemsgroup.com/K3FilterStudy-250HzRoof.pdf

Gentlemen, this is the typical noise response of a distributed
filter system.  The noise peak is front end noise through the
roofing filter and the broader "pedestal" is the noise generated
within the receiver following the roofing filter.

This greater the gain in the 2nd IF (or 3rd IF for upconversion
receivers like the FT-1000MP) the more pronounced the pedestal
and the less the "hump" produced by the roofing (or 2nd IF)
filter.  Jim/Chen, you can see this very effect in your Yaesu
rigs by forcing it to use the widest available 2nd IF filter
with a narrow 3rd IF filter (easy to do on the MP) and using
band noise to excite the radio.

This "hump" effect is one of the reasons for such approaches
as minimum gain IF stages following the last IF filter, post
IF filtering (e.g., a filter just before the product detector)
and active audio filtering following the product detector.
In all cases, these design approaches were intended to remove
the broadband noise generated in the last IF stage.

When operated as designed with the DSP providing the last
frequency selectivity, the K3 actually provides matched
filtering as the last stage in the IF and the first stage
of AF and shows no "pedestal."  The FT-1000 family, on the
other hand, still shows a noise pedestal with its cascaded
filters due to the excess gain in the 3rd IF unless one
follows W8JI's recommendations and reduces the IF gain or
enables DSP demodulation/audio DSP.

73,

    ... Joe, W4TV

On 7/18/2010 5:10 AM, Kok Chen wrote:

> I just did a quick experiment.
>
> First, I lied to my K3 that my 400 Hz filter is really a 2.8 kHz
> filter.  I then turned AGC off and set the K3 WIDTH setting to 2.4
> kHz.
>
> This way, I will be able to see the response of both the 2.4 kHz DSP
> filter and the 400 Hz roofing filter, as Dr. Grebenkemper KI6WX had
> described.
>
> Sure enough, with band noise as input to the K3, I see a similar
> spectrum as Jim showed here
>
> http://audiosystemsgroup.com/K3FilterStudy-250HzRoof.pdf
>
> In my case, it is a 400 Hz wide hump sitting on a wider 2.4 kHz noise
> pedestal.
>
> I switched to antennas of various gains and directions, and as
> expected, the peak of the crystal filter hump rises and falls, while
> the wider DSP filtered noise floor remains constant (lots of spectral
> averaging of my FFT output :-).  Remember that I have turned AGC
> off.
>
> I then changed to using the Elecraft N-gen as the noise source.  The
> noise became stronger now than using band noise -- the peak of the
> 400 Hz hump is now about 30 dB over the DSP noise pedestal.  But the
> wideband noise pedestal (presumably the internal K3 noise, filtered
> by the DSP filter) remained at about the same magnitude.
>
> When I changed the antenna input a dummy load, the 400 Hz hump
> disappears into the wider noise pedestal. But again, the 2.4 kHz
> noise pedestal did not change width nor amplitude. In my case, the
> DSP pedestal is a little over 20 dB higher than the noise floor
> outside of the DSP pedestal.
>
> Not finding a louder noise source, I resorted to the CW signal from
> the Elecraft XG2, set to 50 µV output.  As I tune across the carrier
> I can see it rise up to 70 dB above the DSP noise pedestal and then
> falling back to the DSP noise pedestal.  The shape looked very
> reasonable for a crystal filter.
>
> So I can definitely see at least 70 dB worth of decent 400 Hz
> filtering coming from the roofing filter.
>
> The lower amplitude noise humps (from 0 to 30 dB over the DSP noise
> pedestal, depending upon the strength of the noise source), like the
> plots by Jim, are in my case, the result of the band noise (and N-gen
> noise) not being strong enough for me to see the full dynamic range
> of the roofing filter.  The XG2 showed that the filter floor of the
> roofing filter is at least over 70 dB below the peak 50 µV signal.
>
> So, I would like to suggest that Jim try using a stronger noise
> source than band noise.  Even an N-gen (with K3 preamp on) was only
> giving me 30 dB over the DSP noise floor.  Perhaps a receiving preamp
> would be useful (no, I don't have one to try, otherwise I would :-).
>
> I think Jim will see what John was talking about, i.e., the plot with
> 25 dB to 30 dB hump over a wider pedestal is simply an artifact that
> the noise source to ping the roofing filter is only 25 dB to 30 dB
> louder than the internal noise of the K3 between the crystal filter
> and of the DSP filter.  Perhaps, the variation of the hump over the
> DSP pedestal as you vary the input noise might be persuading enough
> without resorting to using extremely strong noise sources.
>
> 73 Chen, W7AY
>
> P.S., now I need to go undo my filter settings, or I might wake up
> tomorrow thinking that my K3 had been bricked to only seeing a 400 Hz
> passband! :-)
>
>
>
> ______________________________________________________________
> Elecraft mailing list Home:
> http://mailman.qth.net/mailman/listinfo/elecraft Help:
> http://mailman.qth.net/mmfaq.htm Post:
> mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net Please help support this
> email list: http://www.qsl.net/donate.html
>

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[n7ws]

Nicely done.  One caveat: when using the audio out, have the equalizer flat.

Personally, when I wanted to measure my crystal filters, I pulled them out a looked at them with my N2PK network analyzer.

Wes  N7WS

--- On Sun, 7/18/10, Kok Chen <[hidden email]> wrote:


I just did a quick experiment.

First, I lied to my K3 that my 400 Hz filter is really a 2.8 kHz filter.  I then turned AGC off and set the K3 WIDTH setting to 2.4 kHz. 

This way, I will be able to see the response of both the 2.4 kHz DSP filter and the 400 Hz roofing filter, as Dr. Grebenkemper KI6WX had described.

Sure enough, with band noise as input to the K3, I see a similar spectrum as Jim showed here

http://audiosystemsgroup.com/K3FilterStudy-250HzRoof.pdf

In my case, it is a 400 Hz wide hump sitting on a wider 2.4 kHz noise pedestal.

I switched to antennas of various gains and directions, and as expected, the peak of the crystal filter hump rises and falls, while the wider DSP filtered noise floor remains constant (lots of spectral averaging of my FFT output :-).  Remember that I have turned AGC off.

I then changed to using the Elecraft N-gen as the noise source.  The noise became stronger now than using band noise -- the peak of the 400 Hz hump is now about 30 dB over the DSP noise pedestal.  But the wideband noise pedestal (presumably the internal K3 noise, filtered by the DSP filter) remained at about the same magnitude.

When I changed the antenna input a dummy load, the 400 Hz hump disappears into the wider noise pedestal. But again, the 2.4 kHz noise pedestal did not change width nor amplitude. In my case, the DSP pedestal is a little over 20 dB higher than the noise floor outside of the DSP pedestal.

Not finding a louder noise source, I resorted to the CW signal from the Elecraft XG2, set to 50 µV output.  As I tune across the carrier I can see it rise up to 70 dB above the DSP noise pedestal and then falling back to the DSP noise pedestal.  The shape looked very reasonable for a crystal filter.

So I can definitely see at least 70 dB worth of decent 400 Hz filtering coming from the roofing filter.

The lower amplitude noise humps (from 0 to 30 dB over the DSP noise pedestal, depending upon the strength of the noise source), like the plots by Jim, are in my case, the result of the band noise (and N-gen noise) not being strong enough for me to see the full dynamic range of the roofing filter.  The XG2 showed that the filter floor of the roofing filter is at least over 70 dB below the peak 50 µV signal.

So, I would like to suggest that Jim try using a stronger noise source than band noise.  Even an N-gen (with K3 preamp on) was only giving me 30 dB over the DSP noise floor.  Perhaps a receiving preamp would be useful (no, I don't have one to try, otherwise I would :-). 

I think Jim will see what John was talking about, i.e., the plot with 25 dB to 30 dB hump over a wider pedestal is simply an artifact that the noise source to ping the roofing filter is only 25 dB to 30 dB louder than the internal noise of the K3 between the crystal filter and of the DSP filter.  Perhaps, the variation of the hump over the DSP pedestal as you vary the input noise might be persuading enough without resorting to using extremely strong noise sources.

73
Chen, W7AY

P.S., now I need to go undo my filter settings, or I might wake up tomorrow thinking that my K3 had been bricked to only seeing a 400 Hz passband! :-)





     
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[Phil Salas]

"Is anyone else here familiar with notched noise IMD measurements?? I only
used them a few times in the 80's, so my recall is limited."

Tom - This is the way we used to measure our analg FM and SSB microwave
radios at Collins/Rockwell.  A broadband noise source with a very high-Q
notch filter was injected into the transmitter baseband input.  We had notch
filters at low, mid- and high baseband frequencies, though the highest
baseband frequency was always the worst.  However, we were using pre- and
de-emphasis to compensate the baseband signal.  We looked at total noise
(normal KTB plus IMD) in the received channen associated with the notches.
Since we knew the KTB noise floor, we could then determine the IMD noise.
Brings back memories.

Phil - AD5X

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[Bill W4ZV]

W8JI wrote

Is anyone else here familiar with notched noise IMD measurements?? I only
used them a few times in the 80's, so my recall is limited.

I suspect this guy might.  ;-)  

http://www.triquint.com/prodserv/tech_info/docs/WJ_classics/vol8_n6.pdf  (last page)

73,  Bill

[Jim Brown-10]

On Sun, 18 Jul 2010 02:10:38 -0700, Kok Chen wrote:

>So, I would like to suggest that Jim try using a stronger noise
source than
>band noise.  

You're exactly right, Kok. I wasn't pushing the K3 front end hard
enough. I found a much stronger noise source (a nasty switching
power supply that runs some low voltage lighting), fed it through
a DXE preamp and into the K3. The K3 was set for max RF gain, but
all the user gains were adjusted to minimize any obvious
overloading of the signal chain. I could, for example, hit the
audio chain harder and see significant harmonics and IM.

The new data is at the same link as the old data

http://audiosystemsgroup.com/K3FilterStudy-250HzRoof.pdf 

Executive Summary

The plateau effect WAS an artifact of my not driving the K3 hard
enough. My higher level noise source was enough to move that
plateau down to about -60dB (referenced to the peak of filter
response). Another point relative to the dynamic range of this
measurement -- this is a somewhat impulsive noise signal -
individual, un-averaged, measurements show peaks 6-10dB greater
than the averaged data, so the K3 is, indeed, being rather
robustly excited.

IM would show up mostly as LF noise. The wide plots of response
with 250Hz DSP IF show LF noise to be more than 66dB down. The
small broad peaks at about 1.4kHz and 2.65kHz are also probably
IM, but they're at least 78dB down.

As to the use of broadband noise as a source -- the real world of
contesting and DX chasing does not consist of a few big sine
waves, rather, there are often several signals, plus noise, within
a few kHz of bandwidth, and for many hams, that noise can often be
nearly as strong as a strong signal. If you can figure out how to
use it as a measurement tool, noise is a FAR better representation
of the real world than even the world's best sine wave generators!  

The new data DOES show tha the Inrad filters, as integrated into
the K3, are well behaved at their skirts.

The curves showing the roofing filter response with a 1kHz wide IF
clearly show that the 250 Hz filter is about 22% narrower than the
400 Hz filter in the range where my data can be trusted (above
about -48dB).  That's 333 Hz vs. 464 Hz at -6dB, 501 Hz vs. 645 Hz
at -30dB, 620 Hz vs 771 Hz at -48dB.  As a roofing filter, it is
clearly a 22% improvement the 400 Hz filter. That does, however,
fall far short of the 38% improvement suggested by the ratio of
the nominal bandwidth of these filters, 400 Hz and 250 Hz. I think
many of us still want a real 250 Hz filter!

The curves showing the cascaded response of the two filters with
the 250 Hz DSP IF shows very little narrowing of the response by
the narrower filter. To see significant benefit from cascading,
one would need to set the switching point of these two filters to
wider bandwidths  perhaps 500 Hz and 350 Hz.  K2AV noted that
many users have chosen this path, and it does make sense. If you
have both filters, it might also make sense to set the 400 Hz
filter to 400 Hz, allowing you to hear a bit more bandwidth when
you're running, and set the 330 Hz filter to 350 Hz so that you
can quickly narrow it down when the going gets rougher.

Thanks to all those who have commented on my previous measurements
and shown me the error of my ways. Two things I learned long ago:

1) You learn a lot when you stick your neck out and say what you
think you know. When you're wrong, or when there are things you
haven't learned yet, someone will tell you. If you don't have an
ego problem, that's a good thing.

2) He who does nothing does nothing wrong.  

73, Jim Brown K9YC



______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[Guy, K2AV]

That measured 464 vs 333 is very close to what I have measured, and
why I set these two filters as 450 and 350 in the K3 utility setup.
In contesting lingo former is regular run width, the latter is tight
run width.   It is important to keep that guy up 350 Hz OUT of the IF
in a tight situation.  Opening up the pre-digital gain some to hear
someone at the noise level needs that down 70 skirt positioned
correctly to deal with the 70 dB wanted/unwanted differential and
keeping the unwanted from adding to IN-band crud before DSP can deal
with it.

Calling the 333 filter "250" is left over from Inrad supplying it for
a dual filter setup in the FT1000MP where the 250 was the COMBINED
result of TWO filters in successive IF's.

With the 50 Hz incremental DSP width, identifying a filter by a
combined bandpass number no longer makes sense.  If one sets the DSP
to 350 or 300 using the 333 filter, you will get a very decent 250-ish
overall passband.

With the flexibility of the K3, it is left to the operator to figure
out the result when combining 333 and 350.

I would like to have an 8 pole Inrad roofer that is designed
specifically for the optimum RTTY bandwidth presuming that the DSP was
set just a little wider for uncrowded and pulled in one step for
crowded.  What number do you call that?  I don't know.  I'd put it in
the narrow slot in the KRX3.

73, Guy.

On Sun, Jul 18, 2010 at 2:33 PM, Jim Brown <[hidden email]> wrote:

> On Sun, 18 Jul 2010 02:10:38 -0700, Kok Chen wrote:
>
>>So, I would like to suggest that Jim try using a stronger noise
> source than
>>band noise.
>
> You're exactly right, Kok. I wasn't pushing the K3 front end hard
> enough. I found a much stronger noise source (a nasty switching
> power supply that runs some low voltage lighting), fed it through
> a DXE preamp and into the K3. The K3 was set for max RF gain, but
> all the user gains were adjusted to minimize any obvious
> overloading of the signal chain. I could, for example, hit the
> audio chain harder and see significant harmonics and IM.
>
> The new data is at the same link as the old data
>
> http://audiosystemsgroup.com/K3FilterStudy-250HzRoof.pdf
>
> Executive Summary
>
> The plateau effect WAS an artifact of my not driving the K3 hard
> enough. My higher level noise source was enough to move that
> plateau down to about -60dB (referenced to the peak of filter
> response). Another point relative to the dynamic range of this
> measurement -- this is a somewhat impulsive noise signal -
> individual, un-averaged, measurements show peaks 6-10dB greater
> than the averaged data, so the K3 is, indeed, being rather
> robustly excited.
>
> IM would show up mostly as LF noise. The wide plots of response
> with 250Hz DSP IF show LF noise to be more than 66dB down. The
> small broad peaks at about 1.4kHz and 2.65kHz are also probably
> IM, but they're at least 78dB down.
>
> As to the use of broadband noise as a source -- the real world of
> contesting and DX chasing does not consist of a few big sine
> waves, rather, there are often several signals, plus noise, within
> a few kHz of bandwidth, and for many hams, that noise can often be
> nearly as strong as a strong signal. If you can figure out how to
> use it as a measurement tool, noise is a FAR better representation
> of the real world than even the world's best sine wave generators!
>
> The new data DOES show tha the Inrad filters, as integrated into
> the K3, are well behaved at their skirts.
>
> The curves showing the roofing filter response with a 1kHz wide IF
> clearly show that the 250 Hz filter is about 22% narrower than the
> 400 Hz filter in the range where my data can be trusted (above
> about -48dB).  That's 333 Hz vs. 464 Hz at -6dB, 501 Hz vs. 645 Hz
> at -30dB, 620 Hz vs 771 Hz at -48dB.  As a roofing filter, it is
> clearly a 22% improvement the 400 Hz filter. That does, however,
> fall far short of the 38% improvement suggested by the ratio of
> the nominal bandwidth of these filters, 400 Hz and 250 Hz. I think
> many of us still want a real 250 Hz filter!
>
> The curves showing the cascaded response of the two filters with
> the 250 Hz DSP IF shows very little narrowing of the response by
> the narrower filter. To see significant benefit from cascading,
> one would need to set the switching point of these two filters to
> wider bandwidths   perhaps 500 Hz and 350 Hz.  K2AV noted that
> many users have chosen this path, and it does make sense. If you
> have both filters, it might also make sense to set the 400 Hz
> filter to 400 Hz, allowing you to hear a bit more bandwidth when
> you're running, and set the 330 Hz filter to 350 Hz so that you
> can quickly narrow it down when the going gets rougher.
>
> Thanks to all those who have commented on my previous measurements
> and shown me the error of my ways. Two things I learned long ago:
>
> 1) You learn a lot when you stick your neck out and say what you
> think you know. When you're wrong, or when there are things you
> haven't learned yet, someone will tell you. If you don't have an
> ego problem, that's a good thing.
>
> 2) He who does nothing does nothing wrong.
>
> 73, Jim Brown K9YC
>
>
>
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
>

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[Dunc Carter - W5DC]

I write software for vibration data collection and the test case with a
narrow hardware filter followed by a wider dsp filter really looks like
the case that I often see where there are not of bits in the analog data
driving the A/D converter.  In our case, we use hardware from vendors
such as National Insturments.  When NI first introduced their 24 bit USB
A/D converters, they seemed to think this would give adequate dynamic
range but eventually realized that additional hardware amplification was
needed.   24 bit A/D converter resolution may seem like a lot, it's not
.  We still need adjustable gain and attenuation front ends.  I don't
know of any way to look at the K3 A/D output in a way that would allow
assessment of this.  If you could calculate a spectrum of the A/D
converter output, you could compare the peak and average spectrum output
to the noise floor of the spectrum.  Back in the 12 and 16 bit A/D
converter days, we found that 12 to 14 bit A/D converter output was
necessary for good performance and our transducer signals levels
typically range from 5 to 20volts peak down to fractional microvolt
levels.  You might look at the audio output but I'm not at all sure that
there wouln't be other issues that would spoil the measurement.

Dunc, W5DC.
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[Dunc Carter - W5DC]

Duncan Carter wrote:

> I write software for vibration data collection and the test case with a
> narrow hardware filter followed by a wider dsp filter really looks like
> the case that I often see where there are not ENOUGH of bits in the analog data
> driving the A/D converter.  In our case, we use hardware from vendors
> such as National Insturments.  When NI first introduced their 24 bit USB
> A/D converters, they seemed to think this would give adequate dynamic
> range but eventually realized that additional hardware amplification was
> needed.   24 bit A/D converter resolution may seem like a lot, it's not.  We still need adjustable gain and attenuation front ends.  I don't know of any way to look at the K3 A/D output in a way that would allow assessment of this.  If you could calculate a spectrum of the A/D
> converter output, you could compare the peak and average spectrum output
> to the noise floor of the spectrum.  Back in the 12 and 16 bit A/D
> converter days, we found that 12 to 14 bit A/D converter output was
> necessary for good performance and our transducer signals levels
> typically range from 5 to 20volts peak down to fractional microvolt
> levels.  You might look at the audio output but I'm not at all sure that
> there wouln't be other issues that would spoil the measurement.
>
> Dunc, W5DC.
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
>
>  

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[Joe Subich, W4TV-4]

Jim,

 > The curves showing the cascaded response of the two filters with the
 > 250 Hz DSP IF shows very little narrowing of the response by the
 > narrower filter.

This is because your choice of a 250 Hz DSP bandwidth makes the DSP
the dominant filter.  The DSP filters have a skirt selectivity (slope)
of about .6dB/Hz which can be seen in the upper skirt of the 400 Hz
filter (nearly the entire 60dB range is within the -6dB passband of
the 450 Hz wide filter).

 > The new data DOES show tha the Inrad filters, as integrated into the
 > K3, are well behaved at their skirts.

No, your data - particularly the first two graphs show only the skirt
selectivity of the DSP filtering.  The third graph begins to show the
skirt selectivity of the "250 Hz" (300+ Hz) crystal filter but the
DSP bandwidth would need to be considerably wider - its "corners" must
be wider than the expected -80 dB points of the filter being measured
to avoid coloring the results.

 > To see significant benefit from cascading, one would need to set
 > the switching point of these two filters to wider bandwidths 
 > perhaps 500 Hz and 350 Hz.

I don't know that "benefit" is necessarily the correct word to use
ton describe cascading.  The better term would be "effect."  The K3
still exhibits benefits from the distributed filter design - the
roofing filter protects the 2nd IF and DSP from overload, AGC
pumping and IMD from strong signals outside the passband of the
DSP filter.

However, the roofing filter is not intended to contribute significant
selectivity within the DSP passbad.  That's not the function of a
roofing filter - just like the VHF filter in an upconversion based
receiver is not intended to provide significant baseband selectivity.
In an upconversion receiver the baseband (ultimate) selectivity is
provided by the 2nd/3rd IF filters (FT-1000MP/MKV, etc) or DSP (IC-
756Pro/ProII/ProIII, FT-2000/5000/9000).  In either case the sole
design function of the "roofing" filter is to protect the IF chain
from strong out of band signals.

73,

    ... Joe, W4TV


On 7/18/2010 2:33 PM, Jim Brown wrote:

> On Sun, 18 Jul 2010 02:10:38 -0700, Kok Chen wrote:
>
>> So, I would like to suggest that Jim try using a stronger noise
> source than
>> band noise.
>
> You're exactly right, Kok. I wasn't pushing the K3 front end hard
> enough. I found a much stronger noise source (a nasty switching power
> supply that runs some low voltage lighting), fed it through a DXE
> preamp and into the K3. The K3 was set for max RF gain, but all the
> user gains were adjusted to minimize any obvious overloading of the
> signal chain. I could, for example, hit the audio chain harder and
> see significant harmonics and IM.
>
> The new data is at the same link as the old data
>
> http://audiosystemsgroup.com/K3FilterStudy-250HzRoof.pdf
>
> Executive Summary
>
> The plateau effect WAS an artifact of my not driving the K3 hard
> enough. My higher level noise source was enough to move that plateau
> down to about -60dB (referenced to the peak of filter response).
> Another point relative to the dynamic range of this measurement --
> this is a somewhat impulsive noise signal - individual, un-averaged,
> measurements show peaks 6-10dB greater than the averaged data, so the
> K3 is, indeed, being rather robustly excited.
>
> IM would show up mostly as LF noise. The wide plots of response with
> 250Hz DSP IF show LF noise to be more than 66dB down. The small broad
> peaks at about 1.4kHz and 2.65kHz are also probably IM, but they're
> at least 78dB down.
>
> As to the use of broadband noise as a source -- the real world of
> contesting and DX chasing does not consist of a few big sine waves,
> rather, there are often several signals, plus noise, within a few kHz
> of bandwidth, and for many hams, that noise can often be nearly as
> strong as a strong signal. If you can figure out how to use it as a
> measurement tool, noise is a FAR better representation of the real
> world than even the world's best sine wave generators!
>
> The new data DOES show tha the Inrad filters, as integrated into the
> K3, are well behaved at their skirts.
>
> The curves showing the roofing filter response with a 1kHz wide IF
> clearly show that the 250 Hz filter is about 22% narrower than the
> 400 Hz filter in the range where my data can be trusted (above about
> -48dB).  That's 333 Hz vs. 464 Hz at -6dB, 501 Hz vs. 645 Hz at
> -30dB, 620 Hz vs 771 Hz at -48dB.  As a roofing filter, it is clearly
> a 22% improvement the 400 Hz filter. That does, however, fall far
> short of the 38% improvement suggested by the ratio of the nominal
> bandwidth of these filters, 400 Hz and 250 Hz. I think many of us
> still want a real 250 Hz filter!
>
> The curves showing the cascaded response of the two filters with the
> 250 Hz DSP IF shows very little narrowing of the response by the
> narrower filter. To see significant benefit from cascading, one would
> need to set the switching point of these two filters to wider
> bandwidths  perhaps 500 Hz and 350 Hz.  K2AV noted that many users
> have chosen this path, and it does make sense. If you have both
> filters, it might also make sense to set the 400 Hz filter to 400 Hz,
> allowing you to hear a bit more bandwidth when you're running, and
> set the 330 Hz filter to 350 Hz so that you can quickly narrow it
> down when the going gets rougher.
>
> Thanks to all those who have commented on my previous measurements
> and shown me the error of my ways. Two things I learned long ago:
>
> 1) You learn a lot when you stick your neck out and say what you
> think you know. When you're wrong, or when there are things you
> haven't learned yet, someone will tell you. If you don't have an ego
> problem, that's a good thing.
>
> 2) He who does nothing does nothing wrong.
>
> 73, Jim Brown K9YC
>
>
>
> ______________________________________________________________
> Elecraft mailing list Home:
> http://mailman.qth.net/mailman/listinfo/elecraft Help:
> http://mailman.qth.net/mmfaq.htm Post:
> mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net Please help support this
> email list: http://www.qsl.net/donate.html
>

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[n7ws]

Agreed with one exception; in the K3 the "roofing" (I hate that term) filter also rejects the 2nd i-f image.

Wes  N7WS

--- On Sun, 7/18/10, Joe Subich, W4TV <[hidden email]> wrote:
In either case the sole
design function of the "roofing" filter is to
 protect the IF chain
from strong out of band signals.


     
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[hf4me]

I chose the 250 and 500 filters for my narrow ones.  I set the 250(370) at
400 and the 500(565) at 600.  Works great for me.  Just played in the RTTY
contest last night and with the bandwidth down to 400 or 350(bringing on the
select for the "250") if needed after tuning to the signal (S&P) I was able
to work them all even though some were closer to the next station than their
own mark and space traces were to each other (one exception, one had the
space of his directly on top of the next guys mark signal or he was running
RTTTY).

Made me wonder though if one could decode two signals intertwined but
individual mark and spaces traces in the clear form each other?.  I would
think it might be possible.  Anybody done this?

73, de Jim KG0KP

----- Original Message -----
From: "Joe Subich, W4TV" <[hidden email]>
To: <[hidden email]>; "Jim Brown" <[hidden email]>
Sent: Sunday, July 18, 2010 3:33 PM
Subject: Re: [Elecraft] 250 Hz and 400 Hz Filter Measurements

>
> Jim,
>
> > The curves showing the cascaded response of the two filters with the
> > 250 Hz DSP IF shows very little narrowing of the response by the
> > narrower filter.
>
> This is because your choice of a 250 Hz DSP bandwidth makes the DSP
> the dominant filter.  The DSP filters have a skirt selectivity (slope)
> of about .6dB/Hz which can be seen in the upper skirt of the 400 Hz
> filter (nearly the entire 60dB range is within the -6dB passband of
> the 450 Hz wide filter).
>
> > The new data DOES show tha the Inrad filters, as integrated into the
> > K3, are well behaved at their skirts.
>
> No, your data - particularly the first two graphs show only the skirt
> selectivity of the DSP filtering.  The third graph begins to show the
> skirt selectivity of the "250 Hz" (300+ Hz) crystal filter but the
> DSP bandwidth would need to be considerably wider - its "corners" must
> be wider than the expected -80 dB points of the filter being measured
> to avoid coloring the results.
>
> > To see significant benefit from cascading, one would need to set
> > the switching point of these two filters to wider bandwidths 
> > perhaps 500 Hz and 350 Hz.
>
> I don't know that "benefit" is necessarily the correct word to use
> ton describe cascading.  The better term would be "effect."  The K3
> still exhibits benefits from the distributed filter design - the
> roofing filter protects the 2nd IF and DSP from overload, AGC
> pumping and IMD from strong signals outside the passband of the
> DSP filter.
>
> However, the roofing filter is not intended to contribute significant
> selectivity within the DSP passbad.  That's not the function of a
> roofing filter - just like the VHF filter in an upconversion based
> receiver is not intended to provide significant baseband selectivity.
> In an upconversion receiver the baseband (ultimate) selectivity is
> provided by the 2nd/3rd IF filters (FT-1000MP/MKV, etc) or DSP (IC-
> 756Pro/ProII/ProIII, FT-2000/5000/9000).  In either case the sole
> design function of the "roofing" filter is to protect the IF chain
> from strong out of band signals.
>
> 73,
>
>    ... Joe, W4TV
>
>
> On 7/18/2010 2:33 PM, Jim Brown wrote:
>> On Sun, 18 Jul 2010 02:10:38 -0700, Kok Chen wrote:
>>
>>> So, I would like to suggest that Jim try using a stronger noise
>> source than
>>> band noise.
>>
>> You're exactly right, Kok. I wasn't pushing the K3 front end hard
>> enough. I found a much stronger noise source (a nasty switching power
>> supply that runs some low voltage lighting), fed it through a DXE
>> preamp and into the K3. The K3 was set for max RF gain, but all the
>> user gains were adjusted to minimize any obvious overloading of the
>> signal chain. I could, for example, hit the audio chain harder and
>> see significant harmonics and IM.
>>
>> The new data is at the same link as the old data
>>
>> http://audiosystemsgroup.com/K3FilterStudy-250HzRoof.pdf
>>
>> Executive Summary
>>
>> The plateau effect WAS an artifact of my not driving the K3 hard
>> enough. My higher level noise source was enough to move that plateau
>> down to about -60dB (referenced to the peak of filter response).
>> Another point relative to the dynamic range of this measurement --
>> this is a somewhat impulsive noise signal - individual, un-averaged,
>> measurements show peaks 6-10dB greater than the averaged data, so the
>> K3 is, indeed, being rather robustly excited.
>>
>> IM would show up mostly as LF noise. The wide plots of response with
>> 250Hz DSP IF show LF noise to be more than 66dB down. The small broad
>> peaks at about 1.4kHz and 2.65kHz are also probably IM, but they're
>> at least 78dB down.
>>
>> As to the use of broadband noise as a source -- the real world of
>> contesting and DX chasing does not consist of a few big sine waves,
>> rather, there are often several signals, plus noise, within a few kHz
>> of bandwidth, and for many hams, that noise can often be nearly as
>> strong as a strong signal. If you can figure out how to use it as a
>> measurement tool, noise is a FAR better representation of the real
>> world than even the world's best sine wave generators!
>>
>> The new data DOES show tha the Inrad filters, as integrated into the
>> K3, are well behaved at their skirts.
>>
>> The curves showing the roofing filter response with a 1kHz wide IF
>> clearly show that the 250 Hz filter is about 22% narrower than the
>> 400 Hz filter in the range where my data can be trusted (above about
>> -48dB).  That's 333 Hz vs. 464 Hz at -6dB, 501 Hz vs. 645 Hz at
>> -30dB, 620 Hz vs 771 Hz at -48dB.  As a roofing filter, it is clearly
>> a 22% improvement the 400 Hz filter. That does, however, fall far
>> short of the 38% improvement suggested by the ratio of the nominal
>> bandwidth of these filters, 400 Hz and 250 Hz. I think many of us
>> still want a real 250 Hz filter!
>>
>> The curves showing the cascaded response of the two filters with the
>> 250 Hz DSP IF shows very little narrowing of the response by the
>> narrower filter. To see significant benefit from cascading, one would
>> need to set the switching point of these two filters to wider
>> bandwidths  perhaps 500 Hz and 350 Hz.  K2AV noted that many users
>> have chosen this path, and it does make sense. If you have both
>> filters, it might also make sense to set the 400 Hz filter to 400 Hz,
>> allowing you to hear a bit more bandwidth when you're running, and
>> set the 330 Hz filter to 350 Hz so that you can quickly narrow it
>> down when the going gets rougher.
>>
>> Thanks to all those who have commented on my previous measurements
>> and shown me the error of my ways. Two things I learned long ago:
>>
>> 1) You learn a lot when you stick your neck out and say what you
>> think you know. When you're wrong, or when there are things you
>> haven't learned yet, someone will tell you. If you don't have an ego
>> problem, that's a good thing.
>>
>> 2) He who does nothing does nothing wrong.
>>
>> 73, Jim Brown K9YC
>>
>>
>>
>> ______________________________________________________________
>> Elecraft mailing list Home:
>> http://mailman.qth.net/mailman/listinfo/elecraft Help:
>> http://mailman.qth.net/mmfaq.htm Post:
>> mailto:[hidden email]
>>
>> This list hosted by: http://www.qsl.net Please help support this
>> email list: http://www.qsl.net/donate.html
>>
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
>

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[KK7P]

> Made me wonder though if one could decode two signals intertwined but
> individual mark and spaces traces in the clear form each other?.  I would
> think it might be possible.  Anybody done this?

RTTY tones have sidebands, and since the Tx logic is not usually shaped
to move optimally between the tones, but instead tends to switch
abruptly, each tone has sidebands beyond the minimum necessary.

I've often thought it might be fun to create an FSK demodulator for
systems that use more spacing than is needed (for 45.45 baud RTTY, that
is about 30 Hz spacing) that would examine and attempt to demodulate
using three parallel algorithms:

1) The typical differential amplitude algorithm using dual filters
feeding a data rate filter and "slicing" the output.

2) A single tone amplitude algorithm using just the space tone filter
and a reference derived from hi/low values in that filter (or some
more-sophisticated sliding reference), feeding a data rate filter and
slicer...

3) The same as 2, but using the mark tone filter only.

In most cases, the first algorithm will give best results of the three.  
But in cases where signals are overlapping, or partially overlapping, or
there is selective fading or QRM on one tone, the demodulator looking at
the other tone is likely to give useful copy if the S/N is sufficient.

Implemented in a PC with the three text streams aligned, it could make
the difference between a contact or not, or a repeat or not.  Stream it
with the output from your DXP-38, also with text aligned...

73,

Lyle KK7P
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[Kok Chen]

On Jul 18, 2010, at 6:11 PM, Jim Miller KG0KP wrote:

> Made me wonder though if one could decode two signals intertwined but
> individual mark and spaces traces in the clear form each other?.

Yes, you can, and the method is regularly used by some RTTY folks.

It is called Mark-only or Space-only copy.  

You basically treat the corrupted FSK signal as an OOK (on-off keyed) signal by only listening to one of the two FSK tones.  If you go back to your communications theory textbooks, you will notice that under quiet band conditions -- the so called "Additive White Gaussian Noise" or AWGN case, you just need to raise the peak transmit power by about 3 dB to attain the same error rate as binary FSK

OOK radioteletype was actually the original mode that is used by hams.  Hams eventually migrated to binary (two tone) FSK to do better under selective fading.

With Mark-only or Space-only, you basically compare the single demodulated output against the noise floor (setting the decision threshold at 1/2 of the long term peak voltage) instead of comparing the matched filtered output of the Mark tone against the Space tone in the case of binary FSK.

Binary FSK got another boost after the discovery of the automatic threshold correction (ATC) circuit (see US Patent 2,999,925 that was issued in 1961; the same patent also has claims on a method of copying RTTY using diversity reception).

With good demodulators today, when you switch from FSK to OOK, you also lose the additional robustness that the ATC circuit provides under selective fading, in addition to the aforementioned 3 dB.

Many modems support Mark-only and Space-only copy.  The legendary HAL ST-8000 has "MO" and "SO" as two of the settings in the "Detector Mode" switch.  Many big gun RTTY contesters and DX'ers still have an ST-8000 or two sitting next to their radios (when new, they sold for about $3,000 :-).  (The CEO of HAL is a ham, by the way.)

Some software modems also directly support Mark-only and Space-only copy.  For modems that don't support it directly, you can still get a good approximation of Mark-only or Space-only copy by using a very narrow filter that passes only the uncorrupted tone as long as the modem also has a decent ATC circuit/algorithm.

By using Mark-only copy on one signal and Space-only copy on a second signal, I have successfully copied both RTTY signals that are overlapped to the point were a tone of one signal is located at the center of the second signal.

73
Chen, W7AY

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[KK7P]

> RTTY tones have sidebands, and since the Tx logic is not usually shaped
> to move optimally between the tones, but instead tends to switch
> abruptly, each tone has sidebands beyond the minimum necessary.

I should add that most systems, including the K3, use phase-continuous
tone switching to reduce the switching transient.

73,

Lyle KK7P
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[Jim Brown-10]

On Sun, 18 Jul 2010 16:33:40 -0400, Joe Subich, W4TV wrote:

>I don't know that "benefit" is necessarily the correct word to use
>ton describe cascading.  The better term would be "effect."  The K3
>still exhibits benefits from the distributed filter design - the
>roofing filter protects the 2nd IF and DSP from overload, AGC
>pumping and IMD from strong signals outside the passband of the
>DSP filter.

Of course. But as contesters and DXers, we always look for more, and
the cascading of filter responses can certainly be beneficial.

>However, the roofing filter is not intended to contribute significant
>selectivity within the DSP passband.  That's not the function of a
>roofing filter

Yes, but the laws of physics does result in cascaded response.   Just
because a tool is designed for one purpose doesn't necessarily mean
that it can't be used for others. :) Repeating what I stated earlier
-- when 100 big signals like W8JI, W4ZV, N6RO, and K1TTT are crammed
into 15-25 kHz to work EU or JA, you can never have enough front end!
And thankfully, I think all those stations are running K3s (I'm not
sure about TTT).

73, Jim K9YC


______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

[Joe Subich, W4TV-4]

 > Of course. But as contesters and DXers, we always look for more, and
 > the cascading of filter responses can certainly be beneficial.

Yes, cascading can be beneficial but the DSP filter set for 200 Hz
at the corners is narrower that the "cascaded" response of the
"250 Hz, 8-pole" filter and the DSP set to align with the corners
of the "250 Hz" filter.

 > Yes, but the laws of physics does result in cascaded response.   Just
 > because a tool is designed for one purpose doesn't necessarily mean
 > that it can't be used for others.

Yes, the laws of physics result in a cascaded response.  However, when
the DSP filter is significantly more narrow than the roofing filter,
the roofing filter provides very little in the way of "cascade gain."
Consider that the DSP filter set for 250 Hz is about 350 Hz wide at
-50 dB and the "250 Hz, 8-pole filter" response is only about -6dB
at 350 Hz bandwidth, the roofing filter is not providing much if any
selectivity ("cascade effect").

 > Repeating what I stated earlier -- when 100 big signals like W8JI,
 > W4ZV, N6RO, and K1TTT are crammed into 15-25 kHz to work EU or JA,
 > you can never have enough front end!

That's true but even the "250 Hz, 8-pole filter" is not doing much
except keeping them away from the DSP (and the hardware AGC) to the
extent that they are more than +/- 300 Hz off frequency.

The difference is one of receive (base band) selectivity vs. front
end (or 1st IF) selectivity.  The front end (IF selectivity) is
only important in its effects on blocking (HAGC "pumping") and
"wide" spaced IMD (where "wide" is defined as greater than the
DSP bandwidth).  In a multiple strong signal situation it is
convenient to have "front end" selectivity as closely matched to
the DSP bandwidth as possible to protect the IF, 2nd Mixer and
DSP.

*IF* strong signals exist within the IF bandwidth but not the DSP
bandwidth they can cause blocking and IMD effects.  This is most
apparent with narrow bandwidth modes like CW and RTTY where the
interfering bandwidth can be quite small and the interfering power
concentrated on a single frequency.  The issue is not nearly as
acute on SSB where the "channels" are wider and the power density
is lower.

73,

    ... Joe, W4TV

On 7/19/2010 12:39 AM, Jim Brown wrote:

> On Sun, 18 Jul 2010 16:33:40 -0400, Joe Subich, W4TV wrote:
>
>> I don't know that "benefit" is necessarily the correct word to use
>> ton describe cascading.  The better term would be "effect."  The K3
>> still exhibits benefits from the distributed filter design - the
>> roofing filter protects the 2nd IF and DSP from overload, AGC
>> pumping and IMD from strong signals outside the passband of the
>> DSP filter.
>
> Of course. But as contesters and DXers, we always look for more, and
> the cascading of filter responses can certainly be beneficial.
>
>> However, the roofing filter is not intended to contribute significant
>> selectivity within the DSP passband.  That's not the function of a
>> roofing filter
>
> Yes, but the laws of physics does result in cascaded response.   Just
> because a tool is designed for one purpose doesn't necessarily mean
> that it can't be used for others. :) Repeating what I stated earlier
> -- when 100 big signals like W8JI, W4ZV, N6RO, and K1TTT are crammed
> into 15-25 kHz to work EU or JA, you can never have enough front end!
> And thankfully, I think all those stations are running K3s (I'm not
> sure about TTT).
>
> 73, Jim K9YC
>
>
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
>

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html