APF in V1.1.2 does not seem as good

  • 6
  • Problem
  • Updated 6 years ago
  • Acknowledged
The APF function in V1.1.2 does not seem to work nearly as well as it did previously.  The effect of the slider is minimal and the peaking seems significantly diminished over the previous version.

I know there were some complaints that it boosted the signal too much in V 1.1 but it worked much better than V1.1.2 for me. 

I made a video and would like to see if others on V1.1.2 see the same thing.  Also if you are still on V1.1,  can you compare your version with the video?  

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com


http://youtu.be/i3r0W8_4Ym0

The video shows the APF being turned on and off as well as adjustments being made to the slider.  The audio does not change much and almost sounds better off with some settings.



Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes

Posted 6 years ago

  • 6
Photo of Ken - NM9P

Ken - NM9P

  • 4239 Posts
  • 1351 Reply Likes


Have not tried the new update yet.  I guess I am luck that I didn't see the release notice because I was chasing W1AW/8&4...(Got 'em each on 4 or 5 different bands the first night, thanks to the excellent CW performance, splits and narrow filters).

But from your screen pic, I think it makes sense to put the APF in the drip down menu in CW mode in place of ANF, because no one would use the ANF in CW Mode and no one would use APF in SSB mode.  Good move logistically, assuming that both controls are deactivated upon mode change.... 

Looking forward to the re-release of the new update....


Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes
Ken,
Yes, I agree that location of the APF in the drop down makes sense for the reasons you mentioned.  And the slider is a good way to control the boost adjustment.    However, it seems like at the 100 setting the amount of boost should be at least as much as it was in V1.1.  In V1.1.2 it seems like it is significantly lower.

A filter width adjustment would also be nice.  (i.e  a slider to adjust the width of the APF filter.)   I asked about the width in a previous post but don't think it was ever answered.  

Question -  The APF is centered on the side tone frequency in the audio chain - how wide is the filter? 5hz, 50hz, etc

Here is a link to the original posting: 

https://community.flexradio.com/flexradio/topics/apf_ability_to_control_width_gain_and_attenuation_a...
 
Many of us find that also adjusting the width of the APF is a good way to fine tune it especially when trying to pull out a weak DX signal.   I hope this is also considered in the future. 

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com


 
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 323 Posts
  • 138 Reply Likes

For the reasons cited by Jeff, AC0C in his excellent technical analysis of APF, it would be useful to include both parametric tools of Q and gain in a CW APF filter design:

http://ac0c.com/main/page_ft5k_apf_for_the_cw_op.html

In particular, see his sections titled:

"A Historical Note  -  Two Flavors of APF"; and

"Wish List - APF Characteristics."

Elecraft struggled with the same issue -- and inclusion of APF generated a flurry of comments when it was implemented into the K3 transceiver.  It wasn't until someone ran a Spice model of the FT-1000 APF circuit that quieted down the discussion.  CW ops have held the old Yaesu circuit in legendary status but nobody really knew what was so different about the circuit until the analysis was conducted.   

I also like the idea of a drop-down menu reconfiguration to show APF in CW and ANF in other modes. 

Paul, W9AC

  

  

(Edited)
Photo of Ken - NM9P

Ken - NM9P

  • 4239 Posts
  • 1351 Reply Likes
Thanks, Paul.
I read the article and it summarizes and clarifies several points that are noteworthy and are right in line with what I have been thinking.

I like the adjustable Q, but from the article, and my own experience, I would think the slider should adjust the Q from 2 on the low end to about 20-25 on the high end, no more.  And perhaps log-weighted so that there is more adjustment on the low end?
A separate gain slider could adjust filter gain from a flat 0 dB to perhaps 9 or 10 dB.  Probably all that is needed.
I agree that the center frequency should remain tied to the spot/offset frequency. 
Now that we have 1 Hz tuning steps (Thank you!) it is very easy to get the signal centered in the passband.

I don't want to sound like a spoiled, whining, nit-picker....  But FRS almost has it perfectly implemented, and I am looking forward to seeing this amazing rig fully embrace it's awesome potential.

(And this is coming from someone who has never used an APF, other than an Autek "Q-box" AF-1 as a novice in 1974... but now that I have gotten a taste of how much it can help in weak signal work, I have gotten on the bandwagon.)

My new 6500 is making CQ ops fun again, and chasing DX and weak signal 160 Meters is becoming a new part of the hobby for me.  Now I have to spend some time on my day off tomorrow building a coaxial receiving loop for the 160 SSB Contest this weekend and putting up an inverted 'L' antenna....)

Ken - NM9P
Photo of Tim - W4TME

Tim - W4TME, Customer Experience Manager

  • 9198 Posts
  • 3558 Reply Likes
In SmartSDR v1.1.2 and the upcoming v1.1.3, changing the APF control actually changes the "Q" (width) of the filter.  We reduced the overall gain of the filter based on user feedback.  We are looking at enhancing the algorithm to factor in both gain and Q in the APF control for possible inclusion in the next release of SmartSDR.  In the mean time, slightly increasing the slice AF gain upwards will compensate for the reduced gain.
(Edited)
Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes
Tim,
Thanks for the information, that makes sense based on what I've observed.  

Actually the level of GAIN or BOOST in the V1.1 was working well for me.  I noticed I can jump back and forth between the AF gain tab and DSP tab to make those adjustments but it is not very convenient.  

It would be better to also have an APF BOOST slider on the DSP tab or MOVE the APF slider to the SPEAKER tab.  (see snapshots)

BTW, do you know the range of the  filter width that is represented by the slider values 0 thru 100?  

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com






OR






Photo of Ken - NM9P

Ken - NM9P

  • 4239 Posts
  • 1351 Reply Likes


I have only played with it a bit, but have mixed reviews.

Nothing definitive to report until I have more time under the hood.

But initially....

I love the fact that it turns off upon mode change, and the new location of the controls.  Thanks.  It is much easier to use.

My first impression of performance is that it "rings" a lot more than the 1.1 version did.  Especially with any adjustment more than about 1/4 scale.  But after tweaking it, in combination with the AGC-T I could get it to work pretty well.  I am not sure that the enhancement is as smooth as the one in version 1.1, despite the overall gain of the version in 1.1being too high.

But until I have played with it a bit more, I won't make any recommendations at this point.  I haven't been able to determine whether it is a performance issue or an adjustment difference.


BTW, on other enhancements...

I LOVE the adjustable tuning steps.  1 Hz is really nice with the 50 Hz filter and APF in a CW pileup!

Thanks for the ability to select the display averaging method.  I find that I use different settings for different modes and band conditions.


Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes
Ken,
I spent a few more hours working with the new version of APF and agree with your "ringing" comment.  That's why I initially said that on some settings with the new version it sounds better off than on. 

Adjusting the combination of AF gain (per Tim's note) and also the AGC-T helps but it does not seem to work as well as it did in V1.1.   

Trying to make all of the adjustments is not very convenient (or as you put as smooth) and overall APF does not seem to be as effective as it was.  

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com




Photo of Rick Hadley - W0FG

Rick Hadley - W0FG

  • 606 Posts
  • 130 Reply Likes
I'd agree with the comments about observing more ringing with the V1.1.3 enhancement.  The slider is great, but it seems to me that the adjustment rate is logarithmic rather than linear as the change seems to occur almost totally between 0 and 25.  Recognizing a mode switch is a major enhancement though.  It was frustrating to move from CW to SSB and realize the APF was still on with 1.1  I'd like to see 1.1's filter with 1.3's control.
Photo of Stan - VA7NF

Stan - VA7NF

  • 472 Posts
  • 119 Reply Likes

I also agree with the ringing statements.  1.1.3 is definitely more "ringey" than 1.1.

But why does APF work on DAX sourced audio?  Thought DAX was just unfiltered data.

Photo of Steve - N5AC

Steve - N5AC, VP Engineering / CTO

  • 1057 Posts
  • 1097 Reply Likes
There are two groups of IF/audio filters in the FLEX-6000.  There are filters in the signal chain for each mode.  These include the bandpass filters, noise blanker, noise reduction, ANF, APF (new), demodulator, AGC, etc.  There is also a set of audio filters that operate on the combined audio that is going to headphones and speaker (codec filters).  These include the equalizer and the old APF.  DAX audio takes the signal chain output before it gets to the codec filters.  So there is no equalization but your passband settings, etc. are in place.

We would not want to send unfiltered audio to DAX because of aliasing -- in the 24kHz of audio in the signal chain, roughly 40% of it has aliasing* (folding from adjacent frequencies).  If you're not familiar with why this is and you are interested in reading on it Richard (Rick) Lyons has a number of papers and books that explain the basics of DSP.  So we must go through some filter and so we decided to go through the bandpass filter of the slice.  You can widen this up to roughly 10kHz if you like, but we do need to go through a filter.

If it makes sense, we could bypass some filters in the signal chain, APF for example.  DAX IQ has only an anti-aliasing filter and no other filters.
Photo of Stan - VA7NF

Stan - VA7NF

  • 472 Posts
  • 119 Reply Likes


Thanks Steve,  I prefer the 1.1.3 version of having the APF(New) optionally active on the DAX.  It just surprised me hearing the AFP while listening to the DAX output. 

Your planning sessions must be "interesting" with the built in "Scope Creep" that your users request. 

Photo of Steve - N5AC

Steve - N5AC, VP Engineering / CTO

  • 1057 Posts
  • 1097 Reply Likes
We set aside a certain amount of time to do things we have promised.  We work to get those done as early as we can in a release to "bake" them and tweak them until we are happy with them.  But we reserve time to do things that our customers are asking for or to implement new ideas.  Engineers often say "wouldn't it be neat if ..." and then we can say "yes, that would be neat!  How much work is it and when can we do it?"  

Having a completely software defined radio and an easy way to update it means that we can dream up new ideas along with our customers and implement them in a short period of time.  Early in the development of this radio there was a lot of very hard work to get the foundation laid.  But now, we can implement lots of neat features and capabilities.  For us, this is fun.  I refuse to call it scope creep ;-)
Photo of K1UO - Larry

K1UO - Larry

  • 886 Posts
  • 135 Reply Likes

Not sure what happened but just got around to loading Ver 1.1.3 this morning and I agree with the statements regarding APF now.  As opposed to Ver 1.1 I notice ringing immediately upon activating APF and no setting of the slider makes much difference to me..Ive tried wider filtering and various settings of af/rf gain etc. but nothing seems to pop the signal like the initial APF algorithm did... even with its slightly high gain...  I got used to that.   I do like the  control location of the APF and also the auto disable when switching out of CW mode.  Even the fact we now have an APF under construction is good news to me.    Now for diversity capabilities......
PS:  I like Al / NN4ZZ's APF and Gain idea posted above.
(Edited)
Photo of Sergey, R5AU

Sergey, R5AU

  • 862 Posts
  • 117 Reply Likes
Well, we can see different types of the APF realization on F6000 in 1.1 and 1.1.3 , Paul (W9AC) thank you for nice link, I never seen it before(? Strange), in any way I guess every body already thinking and looking what APF should be different depends on the band conditions and other things around, I mean Q, Width, Gain should be accessable for users for set up in future release of SSDR, BTW , FRS thank you for be in touch with real Implementation/Testing/Modofication process
Photo of Carl/K5HK

Carl/K5HK

  • 47 Posts
  • 9 Reply Likes
1.1 APF implementations brought out the weak signals better in my view.  I will continue to play with APF width and AF/T-AGC settings but not be able to achieve that striking presence that APF gave to weak signals in v1.1 yet.   

I am sure Flex will have more tweaks and perhaps adjustments in future renditions.
Photo of Steve - N5AC

Steve - N5AC, VP Engineering / CTO

  • 1057 Posts
  • 1097 Reply Likes
Official Response
I read ACØC's notes on APF.  Good stuff -- thanks for the link Paul.  

I've not had a chance to do a lot of operating in the last few months, but my impression of the v1.1 APF was that it worked well -- perhaps too well -- but that for me getting the signal into the filter was tedious.  Once I turned APF on I would listen and then if the signal wasn't peaking I would move a little up and down the band trying to get the signal in the filter.  Perhaps this is not a problem.  

The APF filter is a different kind of filter to what we use for bandpass.  The bandpass filters have a rectangular response designed to have a gain of one inside the filter and a gain of zero outside the filter.  Of course this is never achieved and there are skirts, but this is the goal.  The APF is different in that it has more of a triangular response as seen in Jeff's notes.  It is designed to boost a narrow range of frequencies while reducing others (but not totally rejecting them).  The higher the Q (controlled by the slider now), the sharper the filter and the narrower the peak.  

Al asked about widths and we could measure the 3dB width of the filter, but this is not a flat-topped filter.  I agree with Jeff that there should be some gain in the APF to reduce the "the volume went DOWN?" effect.  I recall giving the engineer direction that the gain of the filter should be one and this was a mistake.  Of course, everything is software and we'll tweak things until we get consensus that it's the best stuff out there.

My question is about the width of the filter and if any of you are bothered by it.  There are alternatives to the shape  to make it look more flat-topped so that you get equal response in a range of frequencies that is easy for you to hit.  Another alternative would be to have a control or method to lock to a specific CW signal.  If we did this (somewhat harder) we could make the filter very peak-y since we would know that you would always have the signal where you need it.  My inclination is to:

1. Flat-top the filter for a 10Hz or so range and that this combined with (add a couple of more poles)
2. Set the Q range in the slider from 5-20
3. Include 3-6dB of gain in the passband of the filter

We'll take another pass at this and I'd like to have some of you test it before we release this next time.  What do you think about this set of changes?  Full disclosure -- I've not talked to the engineer yet so I could be off base in some of this.
(Edited)
Photo of Ken - NM9P

Ken - NM9P

  • 4239 Posts
  • 1351 Reply Likes

Steve,

Your suggestions sound good.  I really appreciate FRS's responsiveness to customer feedback... the best in the business in my book.

The neat thing is that you CAN "tweak" things until we get it right!  If we don't like it, you can always "tweak" it again.  Better still, as you did in the last release, add some controls so that we can tweak it ourselves according to band conditions and personal preference.

I think the Q adjustment range sounds like a good place to start. 

Gain of 3-6 dB would probably be OK, even better if brought out to a control (eventually) for individualization with a range of flat to 9 or 10 dB.

Bandwidth sounds good, but only experience will tell.

BTW... I never thought I would say this, but using the 6500 in the past two 160 contests (CW & SSB) and the ARRL International DX CW contest and many others, I have actually enjoyed the CW contests MORE than the phone contests! 

This is due to a large degree to the fine filtering, APF, zoom on panadapter, smooth QSK, and quiet receiver.  When I operated 160 SSB contest last night I said, "Where did all the acreage go?"  The filters and zoomed panadapter on CW made the 160 band seem HUGE compared to SSB!

Thanks for a great rig.

Ken - NM9P

Photo of Carl/K5HK

Carl/K5HK

  • 47 Posts
  • 9 Reply Likes
Steve,

I like you idea going forward with 10hz width and 3-6db gain (maybe more possible if variable gain is implemented?) on the peak.  Flat response each side of center.  I don't like the idea of locking on a signal as sounds like a good way for it to want to lock on the wrong signal that the one we individually desire.  

Really I think the biggest loss between 1.1 and 1.1.3 is the reduction in peak gain which you seem to acknowledge was a mistake.  
 One thing I noticed in playing with audio gain and t-agc gain was that when gains are up where I had them for v1.1 I seem to get some peak/ringing.   I had to reduce t-agc in particular considerable to get a normal sounding tone whether APF was on in 1.1.3 or not.   Could something about that APF still be active even at a reduced level even when APF is turned off?

73, Carl / K5HK


Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes
Carl, 
I also had some questions about the locking feature. I don't know of any radio that does that but locking could be very nice if it locks on the right signal.  The concern is how well would it work in a pileup or when there are close by signals.  And some folks might just prefer manual tuning.  

As long as it is an option that can be turned on/off we can have it both ways.   Use it when we want, when conditions are right, or not at all. 

What to do think, does this make sense?

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com

Photo of Carl/K5HK

Carl/K5HK

  • 47 Posts
  • 9 Reply Likes
Al,  It would be nice to try especially if the lock on option can be turned on/off.  Yes that makes sense.   
Photo of Carl/K5HK

Carl/K5HK

  • 47 Posts
  • 9 Reply Likes
Steve,

Like to be part of the test group for the APF tweaks as well.
73,  Carl / K5HK
Photo of Ken - NM9P

Ken - NM9P

  • 4239 Posts
  • 1351 Reply Likes

BTW.  If I use offset of 500 Hz, and set up the RX EQ with all sliders at bottom except the 500 Hz all the way to the top, the APF works noticeably better on noisy bands. 

I also forgot to mention.. I would also be willing to help test the new version.

Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes
Steve,
The APF in V1.1 was very good once you got used to the extra gain.  Your feedback is appreciated and your ideas (1, 2, & 3) sound good.  

Does #3 mean we would get a gain slider to make the adjustment?

An option to "lock on" to a CW signal also sounds neat but I wouldn't want to delay the other changes to get it. .  

I would be glad t help test the next version of APF (or any CW related changes for that matter).

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com







Photo of Ken - NM9P

Ken - NM9P

  • 4239 Posts
  • 1351 Reply Likes

Same here on "lock on."  Nice feature, but I wouldn't delay the rest for it.  Put it on the list for the future after we get the basic APF perfect.
Photo of Ned K1NJ

Ned K1NJ

  • 319 Posts
  • 83 Reply Likes
       Steve, your ideas seem good.  The peaking of Vs 1.1 was good, but I also like
the filter method of Vs. 1.13.  Putting them together as Al suggests would give us
the best of both worlds.  A great match-up.
Ned,  K1NJ

Photo of Charles - K5UA

Charles - K5UA

  • 319 Posts
  • 89 Reply Likes
I agree with others about not delaying anything related to APF for the "lock-on" feature. With 1 hz tuning steps we can center the signal ourselves easily. Put "lock-on" on the back burner for now. Nice discussion and nice explanation AGAIN Steve.
Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes
Steve,
Any news on the fix and testing plans?  

From above, Steve H wrote:   "We'll take another pass at this and I'd like to have some of you test it before we release this next time.  What do you think about this set of changes?  Full disclosure -- I've not talked to the engineer yet so I could be off base in some of this."

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com

Photo of Ken - NM9P

Ken - NM9P

  • 4239 Posts
  • 1351 Reply Likes
Al. Just for feedback. When you first started discussing APF and EQ shaping for CW I didn't see the point. But now that I have spent a great deal of "key time" with this rig in contests, DXing and hunting W1AW/x on CW, and have used the two different versions of APF, I must say "thank you" for pushing the issue. I have found that both "tricks" have made the difference between no copy and a good contact on MANY occasions! I am at 97 confirmed on LOTW for DXCC!
Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes
Ken,
Congrats on being so close to DXCC it won't take long now.  And thanks for kind words,  I can't  really take any credit through.   There were a lot of us that have used APF before and know how much it can help.   Anyway, it's one more feature that make this a nice CW rig,  so looking forward to the fix and the next version.  

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com   
Photo of Ken - NM9P

Ken - NM9P

  • 4239 Posts
  • 1351 Reply Likes
Once I understood the concept and tried it out, I caught on quickly!  It really helps.  So does the 500 Hz EQ trick.  In fact I have changed my usual offset to 500 now because of it.  Now when I hear code at 600-700 like I used to, It sounds really high!
Photo of Ned K1NJ

Ned K1NJ

  • 319 Posts
  • 83 Reply Likes
Steve,
           I'd also like to help test it.
Ned,  K1NJ
Photo of Steve - N5AC

Steve - N5AC, VP Engineering / CTO

  • 1057 Posts
  • 1097 Reply Likes
Our general approach to releases is that we do surgery at the start of the release and then heal wounds near the end of the release.  In other words, we have carved up the software world to work on waterfall first.  By doing this, anything we do early that has risk has time to settle out before we finish a release.  We spend the last part of a release tweaking our major improvements and working on outstanding issues.  I expect it will still be another week or two before we get a chance to look at this.  
Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes
Steve, Thanks for the update.....

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com
Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes
Steve,
Are you still planning to have a few of us test the APF fix before the next release?


From your reply in the question about the V1.2 release, it sounds like that may not happen now....

https://community.flexradio.com/flexradio/topics/version_1_2_ssdr?topic-reply-list%5Bsettings%5D%5Bf...

Steve wrote:  "we are in Alpha test and testing is going well.  It's going to be mid-April before we have a build we feel comfortable shipping as a Beta and it may not be worth the effort to go through an entire release cycle for a two-week Beta test."

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com


Photo of Steve - N5AC

Steve - N5AC, VP Engineering / CTO

  • 1057 Posts
  • 1097 Reply Likes
We had two folks from this list, Al NN4ZZ and Ken NM9P, test the new APF and I believe we are good on the changes we've made.
Photo of Bill N5TU

Bill N5TU

  • 50 Posts
  • 15 Reply Likes
Excellent!  I'm looking forward to the new version.

73, Bill, N5TU

Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1853 Posts
  • 672 Reply Likes
Steve,
Yes, the new version of APF is working well.   Here are my comments to share with others that may interested.  The gain is back and it sounds much better overall.   Although I couldn't do an A/B comparison to the earlier versions it is as good or maybe a little better than the original version.

I'm using APF together with the RX Equalizer for CW.  I set the 500 and 1K sliders set to +8 and the others set to -10.  

The shape slider settings are good from 0 thru about 70 without any ringing,  Beyond 70 there is some ringing.  I've been using it set to about 50 although lower settings seem to work well also. 

Thanks for the fix, CW ops will appreciate it.

Regards, Al / NN4ZZ  
al (at) nn4zz (dot) com