K3 APF (audio peaking filter) progress report #2

[wayne burdick]

We now have our basic APF function in place, and I'm amazed at how  
well it works. Lyle (KK7P, our DSP guy) surprised even himself :)  It  
improves signal to noise ratio significantly, as evaluated both by  
measurement (S+N/N increase of ~3 to 5 dB using the K3's built-in dBV  
meter) and by ear (makes many weak signals buried in noise easier to  
copy). Because we're using DSP, we've been able to improve on analog  
implementations by providing constant gain over the full pitch range  
and ideal passband characteristics.

One explanation for why APF improves copy is that it is *not* a brick-
wall filter: it has a narrow peak, but the skirts are very broad,  
leading to a more "natural" sound. It has fewer of the artifacts you  
get with a really narrow filter.

To borrow terminology from our existing DUAL PB function, the K3's  
first and second I.F. filters provide the "context" bandwidth (along  
with a very high level of ultimate attenuation), while APF provides  
the "focus." The combination of the two sometimes works better than  
brick-wall filters at an equivalent narrow bandwidth, at least in some  
QRN/QRM scenarios. For example, I was listening to 40 meters last  
night with S5 background noise. Copy on many signals was better with  
WIDTH=400 and APF on, than with WIDTH set to 100 or 50 Hz and APF OFF.

Many of you sent suggestions for how APF should be turned on and  
controlled (thanks). We're weighing various alternatives.

Wayne
N6KR



______________________________________________________________
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

[wayne burdick]

I wrote:

> [the APF] has fewer of the artifacts that you get with a really  
> narrow filter.

Instead of "really narrow" I should have said "higher-order," i.e. a  
filter with the equivalent of many cascaded sections. The APF is a  
very low-order filter in comparison to the K3's normal audio DSP  
filters.

Wayne
N6KR


______________________________________________________________
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]

Goes without saying that a lot of us would love to have this for SS CW
and especially CQWWDX CW, even if it's only a field test.

73, Guy.

On Wed, Oct 27, 2010 at 1:47 PM, Wayne Burdick <[hidden email]> wrote:

> We now have our basic APF function in place, and I'm amazed at how
> well it works. Lyle (KK7P, our DSP guy) surprised even himself :)  It
> improves signal to noise ratio significantly, as evaluated both by
> measurement (S+N/N increase of ~3 to 5 dB using the K3's built-in dBV
> meter) and by ear (makes many weak signals buried in noise easier to
> copy). Because we're using DSP, we've been able to improve on analog
> implementations by providing constant gain over the full pitch range
> and ideal passband characteristics.
>
> One explanation for why APF improves copy is that it is *not* a brick-
> wall filter: it has a narrow peak, but the skirts are very broad,
> leading to a more "natural" sound. It has fewer of the artifacts you
> get with a really narrow filter.
>
> To borrow terminology from our existing DUAL PB function, the K3's
> first and second I.F. filters provide the "context" bandwidth (along
> with a very high level of ultimate attenuation), while APF provides
> the "focus." The combination of the two sometimes works better than
> brick-wall filters at an equivalent narrow bandwidth, at least in some
> QRN/QRM scenarios. For example, I was listening to 40 meters last
> night with S5 background noise. Copy on many signals was better with
> WIDTH=400 and APF on, than with WIDTH set to 100 or 50 Hz and APF OFF.
>
> Many of you sent suggestions for how APF should be turned on and
> controlled (thanks). We're weighing various alternatives.
>
> Wayne
> N6KR
>
>
>
> ______________________________________________________________
> 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

[wayne burdick]

Working on that....

Wayne


On Oct 27, 2010, at 12:41 PM, Guy Olinger K2AV wrote:

> Goes without saying that a lot of us would love to have this for SS CW
> and especially CQWWDX CW, even if it's only a field test.

______________________________________________________________
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

[P.B. Christensen]

> One explanation for why APF improves copy is that it is *not* a brick-
> wall filter: it has a narrow peak, but the skirts are very broad,
> leading to a more "natural" sound.

And the reason why some of us have been stating that roofing filters and
brick-wall DSP filters, as  good as they are for their intended purpose,
cannot equate to the results obtained by the APF.

Wayne, any idea as to the "Q" of the equivalent DSP filter?  One typical
aspect of the simple APF circuit as used in the FT-1000/D is that Q changes
as F changes.  More sophisticated parametric circuits keep Q relatively
constant regardless of F.  Perhaps Lyle's implementation in DSP preserves Q
as F is changed?

I believe the FT-1000/D's APF had a minimum Q of 10 at the lowest F setting,
then progressively increased to about 30 at the highest setting.

Paul, W9AC

______________________________________________________________
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

[Steve Ellington]

I'm wondering if the sidetone could be routed through the APF? Using a
simple audio filter makes the sidetone sound sharper and cleaner and somehow
improves my sending ability.
N4LQ
Steve

----- Original Message -----
From: "Wayne Burdick" <[hidden email]>
To: "Guy Olinger K2AV" <[hidden email]>
Cc: "Elecraft Reflector" <[hidden email]>
Sent: Wednesday, October 27, 2010 3:47 PM
Subject: Re: [Elecraft] K3 APF (audio peaking filter) progress report #2

> Working on that....
>
> Wayne
>
>
> On Oct 27, 2010, at 12:41 PM, Guy Olinger K2AV wrote:
>
>> Goes without saying that a lot of us would love to have this for SS CW
>> and especially CQWWDX CW, even if it's only a field test.
>
> ______________________________________________________________
> 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]

Paul,

The Yaesu hardware-based APF design resulted in essentially constant
bandwidth of about 30 Hz (6 dB points) and a peaking gain from 7.5 dB to
10 dB  depending on its peaking frequency (200 Hz to 1 kHz).

The present K3 APF we're testing has a constant bandwidth of 30 Hz at
the 6 dB points and constant 9 dB of peaking gain. It is an IIR filter,
closely mimicking a hardware approach.

73,

Lyle KK7P

> I believe the FT-1000/D's APF had a minimum Q of 10 at the lowest F setting,
> then progressively increased to about 30 at the highest setting.

______________________________________________________________
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

[Steve Ellington]

The Smiths:
 Routing the sidetone through a filter resonant at the sidetone's pitch always cleans up the waveform and improves articulation. The filter must not be so narrow as to cause ringing however. The K3's sidetone is quiet clean but external audio filtering indeed improves the quality of the note which helps, at least this cw op, send faster with fewer errors. I suppose it could be considered psychoacoustic but it works.
N4LQ
Steve

  ----- Original Message -----
  From: The Smiths
  To: [hidden email]
  Sent: Wednesday, October 27, 2010 4:22 PM
  Subject: RE: [Elecraft] K3 APF (audio peaking filter) progress report #2


  Steve,
   
  Maybe I'm missing something here, or you are...but the APF has NOTHING to do with sending... It has NOTHING to do with SPOTTING someone, it has NOTHING to do with TONE even, it has to do with Peaking a signal within your current IF passband area regardless of where the IF SHIFT is set, regardless of where your PITCH or TONE is set.. None of these have anything to do with an Audio Peaking FILTER... The on/off button is going to be the Dual PB button.. since that has to do with Filters.
   
  > From: [hidden email]
  > To: [hidden email]; [hidden email]
  > Date: Wed, 27 Oct 2010 16:17:10 -0400
  > CC: [hidden email]
  > Subject: Re: [Elecraft] K3 APF (audio peaking filter) progress report #2
  >
  > I'm wondering if the sidetone could be routed through the APF? Using a
  > simple audio filter makes the sidetone sound sharper and cleaner and somehow
  > improves my sending ability.
  > N4LQ
  > Steve
  >
  > ----- Original Message -----
  > From: "Wayne Burdick" <[hidden email]>
  > To: "Guy Olinger K2AV" <[hidden email]>
  > Cc: "Elecraft Reflector" <[hidden email]>
  > Sent: Wednesday, October 27, 2010 3:47 PM
  > Subject: Re: [Elecraft] K3 APF (audio peaking filter) progress report #2
  >
  >
  > > Working on that....
  > >
  > > Wayne
  > >
  > >
  > > On Oct 27, 2010, at 12:41 PM, Guy Olinger K2AV wrote:
  > >
  > >> Goes without saying that a lot of us would love to have this for SS CW
  > >> and especially CQWWDX CW, even if it's only a field test.
  > >
  > > ______________________________________________________________
  > > 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

[Kok Chen]

On Oct 27, 2010, at 1:06 PM, Paul Christensen wrote:

> And the reason why some of us have been stating that roofing filters and brick-wall DSP filters, as  good as they are for their intended purpose, cannot equate to the results obtained by the APF.


Indeed.

If you squint a little at the Yaesu APF filter shapes, the "skirts" fall off with a more or less 1/f shape -- perhaps coincidentally, the spectral envelope of a periodic On-Off Keyed signal follows a 1/f profile.

In a sense, that is what a matched filter does in the signal processing world -- and we know that matched filters, in the absence of QRM, optimizes SNR in an AWGN channel.  Although I won't go as far as extend linear matched filter theory to the highly non-linear ear-brain system, it still raised my curiosity when I first saw the Yaesu filter plots which someone kindly posted earlier.  

When there is strong adjacent QRM though, you may still have to "roof" the APF with a filter that has more abrupt skirt -- similar problem exists with matched filters for RTTY: they are great for pulling signals out of the noise.  But you need to engage a more abrupt filter when QRM creeps closer in.

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