Latency 1.4.....Improvement?

Mark Griffin

What is a shape factor filter?

Lee - N2LEE

Google or buy an Amateur Radio Handbook.

Charles - K5UA

Hi Mark, The shape factor of a filter is defined as the width of the filter at its -60 dB points divided by the width of the filter at its -6 dB points. A "perfect" filter would have vertical sides because its -60 dB points would have the same width of its -6 dB points, i.e., a "brick wall" filter. Such a filter is impossible in the real world because it would have infinite latency, but the engineers at FRS have managed to design filters with incredible 1.04:1 shape factors with reasonable latency. Hence my tongue in cheek post "give me 1:1 shape factor filters" earlier to state my objection to degrading the shape factor of these beautiful FRS filters to shave off a few milliseconds of latency. I would not object to some future version of SSDR giving the user an option to degrade the filter shape factor to decrease latency, BUT I strongly object to simply degrading the filter shape factors to decrease latency as a default condition of SSDR. By comparasion, the best analog filters have shape factors of about 1.4:1. Plot out a 1.4:1 filter versus a 1.04:1 filter and you'll see why I don't want to give up those filter skirts for the sake of a few milliseconds of latency. In the CW world filter shape factor is much more critical than for the SSB world. That's why you will hear mny more CW operators talking about filter shape factor than SSB operators.

Walt - KZ1F

I didn't even ask the question, but thanks for the explanation Charles! Way more informative than the previous 'look it up' response.

k3Tim

agreed..  thanks Charles....

Ken - NM9P

Good explanation, Charles. If I remember correctly what I have heard before, the latency is a function of shape factor and width, which is why latency is more of a problem with the contesters, because they often use more narrow filters in the battleground. On the other hand, I don't think any of those proposing a change are suggesting that the solution become standard, only a selectable option. I would welcome a "low latency" option for that purpose as long as it didn't degrade the standard filter slopes, which are the best in the business!

Lee - N2LEE

Walt, the reason for my "look it up" is because frankly I am frustrated with so called hams who are too lazy to learn. It took longer to enter the question than it did to do a search.

If we keep handing people the answer instead of encouraging them to at least TRY to learn then our hobby will die. Sorry for being so terse but it is discouraging to see forum after forum of "licensed" hams who will not lift (literally) a finger educate themselves.

There is no one here to has all the answers, we are all learning no matter our age. But, if our hobby is to survive we have to push back some times and help others by teaching them how to find the answers on their own.

Charles - K5UA

Yes Ken, just wanted to make sure the Community understands why low latency filters should be an OPTION, and not the default condition of the 6000 series filters. Degrading the shape factor of a 2.9 kHz wide SSB filter for the sake of latency is not going to affect its performance in a crowded band AS MUCH as degrading the shape factor of a narrow 100 Hz CW filter in a crowded CW band. In my opinion, I would rather have a 1.04:1 shape factor 200 Hz CW filter than a 2:1 shape factor 100 Hz CW filter. It sounds counter-intuitive, but if these two filters are plotted out it will become apparent why the wider filter with the better shape factor gives superior selectivity in a crowded CW pileup. "Where the rubber meets the road".... "In the final analysis"....... "After all the smoke has cleared" ..... it is always selectivity, selectivity, selectivity that decides if that rare one gets in your log.

Stan VA7NF

The shape factor is just a number that varies with the bandwidth. 

For example, Charles you referenced 1.04:1 at 2.9Khz - 2900*1.04 = 3016 or 58 Hz difference from -6 to -60 on each side.  Take that same filter slope to a 200Hz filter then the shape factor will be (200 + 58 + 58) / 200 =  1.58 slope factor.  Both are the same filter requiring the same processing overhead and the same latency, just the slope factor number changed.

Charles comparison of a 1.04:1 SSB and 2:1 at 100Hz is correctly comparing the same filter.  Latency will drop as more processing power is brought to the algorithm, or an improved algorithm.  Since this software is quite refined and the hardware is fixed for this generation, it appears what is left is slope vs. latency.  Assumption: Processing is already done on the gate array.

Pushing this option to the user would be a slider  "Slope <-- . . . . --> Latency"  Squeeze the marshmallow on the Slope side and the Latency side grows.

Charles - K5UA

Hi Stan,

I may be going out on a limb here, so I don't want to sound too dogmatic about the this filter shape factor discussion. I hope that Paul Christensen, W9AC, will weigh in on this discussion.

It appears that you are implying that the slope of the filter skirt of a 2900 Hz, 1.04:1 shape factor SSB filter is the same as the slope of a filter skirt of a 200 Hz, 1.04:1 shape factor CW filter.  I think you are confusing slope of the skirt with the definition of shape factor of a filter.  Let's look at the math....

As you correctly stated, a 2900 Hz, 1.04:1 shape factor SSB filter would be 2900 Hz wide at its 6 dB points and 3016 hz wide at its -60 dB points.  The slope of the skirt from the -6 dB point to the -60 dB point would be (change in db) divided by (change in frequency), or 54dB/58Hz = 0.93 dB per Hz.

However, a 200 Hz, 1.04:1 shape factor CW filter would be 200 Hz wide at its -6 dB points and 208 Hz wide at its -60 dB points.  The slope of the skirt from the -6 dB point to the -60 dB point would be (change in db) divided by (change in frequency), or 54dB/4Hz = 13.5 dB per Hz.

As you can see, the difference in the SLOPE of the filter skirt of the 200 Hz, 1.04 shape factor CW filter is much steeper than the SLOPE of the filter skirt of the 2900 Hz, 1.04:1 shape factor SSB filter. I think you are correct when you say "it appears what is left is slope vs. latency", but from the above calculations it should be clear that the SLOPE of the two filters are dramatically different.

Only Steve from FRS engineering can confirm or deny your statement that ..... " Both are the same filter requiring the same processing overhead and the same latency".  My guess is that they are NOT, based upon the above calculations. However, I do not claim any knowledge about Gate Arrays or DSP programming, soooooooo...........

Steve, please help us with this thread, we always learn a lot when you clarify things for us.

Walt - KZ1F

@Lee, you never taught or mentored, did you? That's not how one encourages another to research. It is likely, very likely, not everyone is as knowledgeable about online resources as others are. This doesn't require a conversation between you and I, nor does it require you to defend your remark. I,merely, drew a comparison between your answer and Charles's.

Bill -VA3WTB

Lee, I do understand you, but is this not what we do here, or is this place only for the astute?
Some times I like someone explain things to me, If I look it up I may find it confusing if I don't understand what they are saying. We are all on different levels here, sorry to say I'm on the lowest level I think....

But I agree, most things can be researched on line to increase understanding.

caravanken

Here is some perspective: .111 is faster than Ashley Force Hoods average reaction time .115 in a Funny Car. It is .026 slower than her dad's at .085. 

Walt - KZ1F

But to reiterate what you said Bill, sometimes the Wikipedia answer was drafted by an engineer, not what does it mean to the average person. Not every ham is a EE. It's akin to "ok, could somebody say that again, in English this time?".  I understood what Lee meant too and sometimes I catch myself before I say, um, less than gracious things because I am not in the mood. Rarely, I don't catch myself before I commit it to type. I merely wanted to give a shoutout to Charles for, what I thought, was a really good English answer.

Lee - N2LEE

I actually mentor and speak to class rooms all the time in another industry, but not ham radio.

I will do everything I can to help someone who wants to learn. But, when they are too lazy to open a book, do a simple search or at least try then NO, I will not spoon feeding those that just want the answer.

A teachers job is to teach students how to learn, how to find solutions to questions and to retain that information. I wonder how many people who hand out answers are actually do this for their own self gratification and ego vs those that are actually trying to teach ?

I am honestly not picking on Mark because I see this on every forum. But if he would have posted a question along the lines of "I read a WIKI page and now have a better understand of shape factor but still not see how DSP filters achieve this, you can help me understand ?"  that would have totally different.

There are a percentage of hams who are formally trained engineers. I am not one of those. I believe many hams get into the hobby because they are inquisitive people who love to learn how things work. And a huge percentage are self taught through trial and error, reading, mentors or asking questions.

• How would handle a new ham that came to you and said I want to get on the air, can you give me radio ?
• I don't know anything about antennas, will you build me one and come put it up ?
• Don't know a lot about computers, can you install logging software and then show me how to use it ?
• Then called you day after day asking question after question, would you consider this mentoring ?

I will restate me initial point. If we don't teach hams how to learn, find the answers and make mistakes, then our hobby will die.

Walt - KZ1F

And, implicit in what you just said, (I am really surprised you dbl'd down on this), it is YOUR job to 'teach' them. Actually, Lee, maybe I am not surprised. But it is good for all to hear it from you that you are the new sheriff in town. Lee, my wife and I both have MEd degree, I suspect we have the 'what does teaching involve' covered. Do you?

Lee - N2LEE

Darn Walt, I give up, I thought I had some good points but you win

I admit defeat and I am certainly no match for an over educated know it all that insist on contributing to the dumbing down of the the hobby.

The hobby is all yours. Go for it.

James Whiteway

Lee, having had Walt offer me guidance on several things I have asked about concerning the Flexlib api, I can truthfully say, he is NOT "contributing to the dumbing down of the hobby". Instead, he has encouraged me to take things one step at a time and given hints instead of outright answers.
  If there is any "dumbing down" in that, I fail to see it. 
james
WD5GWY

KY6LA_Howard

What relevance has any of the above nonsens to ADP.

Walt - KZ1F

Howard, I believe it has to do with the question about a shape filter. You are, of course, welcome to read the thread in order to answer your question. I should point out, as I know you would, your post just now added zero value.

KY6LA_Howard

I stand corrected

Jay Nation

I tried reading this thread "in order."  I failed! Too much time traveling required. 
73, Jay - NO5J

Walt - KZ1F

I totally agree Jay. Posts 9 mins old nestled between two posts 6 months old. It would be nice if the software prevented that, it is, after all, just software.