Adjusting The Crystal Filter Settings On The K2 And Achieving Maximum Intelligibility with Narrowband SSB

When I first built my K2 about 8 months ago in November and December of last year, I initially set up the variable bandwidth crystal filter as per the suggesting settings in the K2 manual.  Shortly afterwards, I downloaded Spectrogram v5.17 and followed the filter set-up procedure described by N0SS on this page.  Incidentally, as of the time this post is being written, the page that is linked to for downloading Spectrogram no longer has the download. Version 5.17, the last free fully free version, is more than adequate for setting up the K2 filters and it’s still available if you look around the internet a bit.

I did reasonably well aligning most of the filters but noticed while on the narrowest CW setting – 200Hz, that it attenuated the signal quite a bit. On looking at the response curve in Spectrogram the other day, it became obvious why. Tuning the receiver so that it received a tone of 500Hz (my selected frequency offset) placed the signal halfway down the lower skirt of the pass-band of the filter. Heaven knows that I was thinking. With every other bandwidth I had set up, the filter pass-band was centered nicely on the 500Hz mark, but the 200Hz filter pass-band was way off.

This incorrect filter setting had arisen as a result of  a misunderstanding of the way that the K2 filter adjustment process works. I don”t understand exactly what is happening inside the K2 during this but the best way I can currently describe it is as follows: When adjusting the BFO frequency for a particular filter setting in order to move the passband, the CW note (or pitch of the SSB signal) also changes. When you move to the next filter setting, or hit the menu button to finalize that particular setting, the BFO frequency is re-calculated, and you will hear the signal at the correct pitch again. Call me slow, but I wasn’t noticing that recalculation step. As a result, I was reticent to move the filter pass-band very far during adjustment of the settings for fear that once they were finalized, as I stepped through the different filter settings, the CW note (or SSB signal pitch) would change significantly. What a handicap my misunderstanding presented!

The other difference in the way I was adjusting the filter settings this time around was that instead of using band noise, I decided to build a simple wideband noise generator. The circuit is described by N0SS on this same page on the Elecraft site.

Boy, I like simple circuits. The ubuiquitous Altoids tin was the way to go for this one. I could have built it ugly-style, but the MeSQUARES are so easy to use.  No need for a circuit board – I just glued the MePADS directly to the tin and used the tin as the ground plane:

Here are the Spectrogram plots for the 4 CW filter settings I chose.  For the widest one, 1000Hz, I didn’t center the filter pass-band on the 500Hz mark, as that would have placed part of the filter response curve on the other side of the BFO signal,  meaning that the receiver would no longer be a single-signal receiver. For this bandwidth, I placed the 500Hz mark (indicated by the vertical red line) a little to the left of the center of the pass-band. Although the nominal bandwidth of this setting as reported by the K2 was 1000Hz, I estimated the 3dB b/w as about 500Hz and the -6dB b/w about 690Hz:

The nominal 700Hz b/w was closer to 335Hz at the -3dB points and 440Hz at the -6dB points:

The nominal 500Hz bandwidth looked to be about 250Hz at the -3dB points and 380 at the -6dB marks:

While the 200Hz nominal bandwidth checked in at the only slightly lower figure of 240Hz at the -3dB points and 340Hz at the -db points, it sounded a lot narrower than the 500Hz nominal setting. This was only an estimate.  The shape of the curve in Spectrogram is not static – it does move around a little,  so there is room for variation in the measurements:

I’m really happy with the way the 200Hz setting sounds. It’s a shame that it was incorrectly adjusted for so long.  I had read that the signal is attenuated if you wind the filter down to 100Hz, so I assumed that the attenuation I was experiencing was normal. Turns out it was normal – for a filter in which the signal is centered halfway down one of the skirts! This got me thinking about the small number of complaints I’ve read from K2 owners who say that their K2 sounds terrible, and can’t help wondering if they have not yet learned how to set up their filters properly.

Which brings me to the next part – adjusting the filters for SSB. I never used to think of anything other than bandwidth when thinking about filters. For some reason (and this is evidence of my particularly inflexible way of thinking) all I thought about was the width of the filter. “How narrow is it?” was my only question.  I never gave much thought to the importance of exactly where in the pass-band the BFO is placed. Once again, I don’t know why. I seem to never pay attention to things until I’ve been beaten over the head with them many, many times.

Things aren’t so critical if your SSB filter is 2.5KHz or wider, but as you dial the pass-band of your SSB filter down it can get pretty hard to retain good intelligibility.  I Googled the general subject and found this really interesting article by G8JNJ titled “Improving the Intelligibility of SSB Transmissions”.  Originally published in Radcom in Feb 2009, it gives some good tips for achieving the maximum clarity of SSB transmission and reception in the limited bandwidths we amateurs use.  To simplify, the article says that the range  up to 8KHz is all that is needed for intelligible speech.  Trouble is, we amateurs use bandwidths of much less than that.  In western languages, most of the energy in the vowels takes place under 500Hz. The vowels are what help another person determine that it is you who is speaking – they do a lot to give your voice it’s unique identity that makes it sound like your voice.

That’s all very nice (and it is) but the vowels don’t contribute anywhere near as much to intelligibility as the consonants do. Consonants occur at higher frequencies than the vowels, and the range 800 – 5000 Hz is particularly important. That still represents a bandwidth of 4200Hz though.  Going further, Martin says that the area around 1600-2000 contributes the most in terms of consonants.

Now we’re getting somewhere.  So the range from 800 – 2000 Hz is particularly critical. That’s a bandwidth of just 1600Hz!

I didn’t follow these figures faithfully when adjusting my settings for the SSB filters in the K2, but I did follow some general rules based on what I had just learned. Incidentally, I don’t yet have the KSB2 SSB option for the K2, so am using the CW filter to receive SSB:

The wider bandwidths were easier to adjust. I set them to allow more bottom end and as a result, they are more pleasant to listen to. By the time I got down to the 1600Hz (nominal) setting, I had to set the lower cut-off at about 800Hz to achieve maximum intelligibility.  Without having first read the Radcom article by G8JNJ, it wouldn’t have occurred to me to set the lower cut-off so high, but I was surprised at how clear the audio is at that setting, even if it’s not really that pleasing to listen to.

In the following screen captures, the markers are set at 300Hz and 2500Hz. That doesn’t mean anything – they’re just markers to show you where the 300Hz and 2500Hz points are, in case that helps you to interpret the filter response curves. The 2490Hz (nominal) setting was easy. The -3dB point at the lower end looks to be around 300-400Hz. It sounds fine:

As we go down in b/w to the (nominal) 2200Hz setting, the lower -3dB cut-off looks like it’s around 450-500Hz (my rough estimate only). It definitely sounds more restricted at the lower end but still has plenty of clarity:

The 2100Hz response curve looks very similar. I’m not sure why I picked 2 bandwidth settings that are so close to each other. I may well change this:

Look at the 1600Hz (nominal) setting. I’ve set the lower -3dB cut-off point to what looks like around 800-1000Hz.  There’s a lot less fidelity than the wider filter settings, but not much less in the way of intelligibility:

I’m pretty sure that at some point in the next few months, I’ll be building the KSB2 SSB option; not so much because I want to become active on phone, but more because I’m interested to see how the filter compares to the CW filter in the basic K2. The CW filter does have quite a bit of ripple in it at the wider settings, and it’ll be good to have an SSB filter with a flatter top to the response. I’m also keen to spend a little time setting up the transmit audio to see how it sounds. It wouldn’t hurt to have the SSB option fitted.