[Elecraft] DSP / filtering thoughts.

[Bill Coleman]

On May 9, 2006, at 10:07 AM, Darwin, Keith wrote:

> Something bugs me about the implementation of DSP filters.  It seems
> we're using DSP to duplicate crystal filters by having them do  
> bandpass
> functions.

Crystal filters are just one way of rejecting unwanted signals. Other  
types of receiver designs have used tuned circuits, ceramic  
resonantors, and mechanical filters. Why would DSP be any different  
from these?

There are certainly some issues in using an audio DSP to implement  
receiver filtering. Fortunately, the K2 has a pretty good crystal  
filter design, and an excellent mixer.

My only beef with the K2's crystal filters is the poor bandpass shape  
and slope factor, especially for the OP1 (SSB) filter. The DSP  
filtering really helps to clean up this filter.

> The other thing we do with DSP is noise reduction which I
> hear works best at wide bandwidths so the algorithm has some noise to
> work with.

This works pretty well with the OP1 (SSB) filter.

> But what about other DSP things?

Don't forget about the automatic notch filter! It makes 40m SSB  
tolerable.

> How about a specific CW filter that takes into account CW speed and CW
> elements length along with pitch?  Imagine a filter that selects a 12
> wpm signal in the presence of 15 - 20 wpm signals.

That would be pretty tough to do. Properly detecting an OOK (on-off  
keyed) is really difficult, because the off state can be easily  
confused by noise, especially on weak signals.

And synchronizing to a particular speeds works better if the speed is  
known and fixed. But CW speeds are highly variable, especially when  
hand-keyed. K6STI had a RTTY decoder that did something like this --  
it synchronized to the sending and could fill in bits that were  
mostly or partially missed.

>   Or how about a
> filter than can de-flutter a signal traveling over the north pole.

This is possible. CocoaModem implements something like this for RTTY.

> Or maybe one that blocks strong signals!

This is tough. Rejecting strong signals is what we attempt to do with  
frequency-based filters.

> How about a filter that corrects bad CW spacing?

Certainly possible, but you have the same problem with building an  
OOK detector, and you have the speed-variance problem.

> Or a filter that uses the CW signal as a trigger to drive a  
> synthesized
> CW signal for virtually infinite S/N ratio! It would be like listening
> to a code practice oscillator controlled by the other op.

This has been done! They call it "regenerated CW". Basically, it  
requires an OOK detector driving a tone oscillator.

> I guess I'm thinking that there are other filtering techniques and
> approaches that could be far more novel than simple band pass or low
> pass and would really set DSP apart from the tried and true crystal
> filters.

Of course, another thing we could do is to use modulation techniques  
other than CW (OOK) that are easier to digitally detect. (eg FSK,  
multi-FSK, PSK, etc)

Bill Coleman, AA4LR, PP-ASEL        Mail: aa4lr at arrl.net
Quote: "Not within a thousand years will man ever fly!"
             -- Wilbur Wright, 1901