AFSK or FSK in a K3: What's the tradeoff?

or "Why I tried to blow up my K3"

Andrew Flowers, K0SM
1/20/2013

This is not the usual "how to transmit clean AFSK" article. This is much more fun than that. This is about doing horrible things to an AFSK signal to see just how bad we can make it when it comes out of a K3. It's about bad AFSK practices and examining the potential wreckage strewn across the band.

Why on earth would I do this? It seems that some people are unhappy with the key clicks coming from their internal FSK generators and are wondering whether they should try to do any better with waveshaped FSK. I think this ultimately distills down to two things:

1. Can I ensure that I transmit a signal that is intelligible to the person I am talking to?
2. Can I ensure that I do #1 without causing problems for people around me.

A nice thing about using the internal FSK generator is that you know exactly what you are going to get. You really can't mess it up, but on the flip side, there is not much of anything you can do to make it better. The main issue with the FSK Generator in the K3 is that it does no filtering or shaping of the keyed waveform. As a result, it generates "key clicks" that may cause problems to other operators on the band. This is not anything unique to the K3--all transmitters to my knowledge have this issue. (Please see the article on waveshaping RTTY and why it matters for a review.)

So, if you use the K3's internal FSK generator, you can rest assured that your spectrum will look almost identical to Figure 1 at any power output. It is more-or-less the standard, and you should do your best to understand the benefits of (1) doing better than this and (2) the risks of accidentally doing worse. The waveshaping article discusses (1); this article is about (2).

Figure 1: K3 internal FSK generator at 100W.

As it stands now, there is no easy (or Elecraft-sanctioned) way to improve the clicks generated by the K3 in FSK-D mode. The decision to transmit a waveshaped FSK signal using audio modulation (AFSK) requires more work in the sense that it may be possible to make things worse, not better. That is the great irony of AFSK transmission--it can produce a much cleaner signal than the internal FSK generator, but it also has the potential to do much worse if you mess it up. There is an economic decision here that the operator must make, and the risks are different for different people; a contest CQ-machine with 4x4x4 and a kilowatt can wreak much more havoc with a wide signal than a casual operator with 50W and a dipole. The purpose here is to help people understand, by example, these risks as they pertain to the K3 so that they can make an informed decision on what to do. The waveshaping article demonstrated how good things can potentially be...now, lets find out how bad they can be within the tools that the K3 gives us to keep our signal clean.

Disclaimer--I am about to do bad things with my radio. I do not suggest that you try any of the horrible things I am about to do. This is for information only and should you choose to reproduce them you do so at your own risk.

As we go forward, keep in mind that this is only with a K3 and the behavior is likely very different with other radios.

In all of the following tests I used a normal (non-attenuating) audio cable from the sound generating device directly to the line-input of the K3. No intermediate interfaces were used.

Audio Overdrive

The instructions in the K3 manual simply state to use the "voice" settings for ALC: adjust for 5-7 "bars" on the "ALC" meter on voice peaks. This does beg the question of what that means for a mode that doesn't have amplitude peaks like SSB, and apparently some people advocate setting AFSK for 3-5 bars on the meter. My understanding is that the "ALC" meter is not a hardware ALC but rather indicates that the DSP audio processor ahead of the RF chain is able to adjust the drive optimally for the RF stages later on to avoid problems. In practice I found that with all levels of line-in gain I could not clip any of the audio stages in the radio so long as I kept the ALC meter off full scale.

But we're not here to transmit nice signals....

You are probably dying to know how much audio it takes to clip the line input of the K3 with the gain level set to its minimal level. I found that I could just barely do it with my equipment. It took about 1160mV RMS / 1700mV peak from MMTTY's sound output (measured on a DSP-599zx). At this point we are far beyond 7 bars on the ALC meter. This was near the maximum output on the headphones port of my Presonus FSM, and many if not most consumer sound cards won't do this on a headphones or line-level output. If you manage to do this, and do not have the AFSK TX filter in the K3 enabled, you will end up transmitting something like this without the AFSK TX filter engaged (and it sounds awful in the monitor audio). Yes, you can hear those spurs on the air, and if you were to transmit this in a busy contest you probably would upset a few folks.

Figure 2: K3 AFSK with approximately 1700mV peak, clipping the line input. Power is 100W in this and all other figures. Reference is about 3dB lower than in Figure 1, but that is a result of me having slightly more attenuation in line when this picture was taken. Add about 3dB to this spectrum to compare with the FSK-D signal in Figure 1.

But what happens if the AFSK TX filter is engaged in this scenario? This places a ~400 Hz filter around the audio tones we have defined in the "PITCH" setting. The distortion products outside this window--the ones that are likely to cause QRM to others are filtered out in the DSP. Remember, we haven't changed the audio level--we are still clipping the ADC before the filter. You might wish to compare this awful scenario to the FSK-D signal in Figure 1 to see just how much more adjacent-channel QRM this could create than if you were to transmit with the FSK-D signal. You might be surprised.

Figure 3: Same signal as Figure 2, but with the AFSK TX filter engaged on the radio. Again, add about 3dB to this spectrum to compare with the FSK-D signal in Figure 1.

We are already at "full scale" on the ALC meter. But what happens if we open the flood gates, and set the line in to 60 when we are already clipping the ADC? Remember, the AFSK TX filter is still enabled. This is as much power as I can give the radio (about 1700mV Peak), and as much gain I can ask of it. You might want to compare this to the FSK-D signal above for reference to see how much worse we've gotten.

Figure 4: Same signal as Figure 2 and 3, but with line in gain set to maximum on the radio. Add 3dB to this Figure to compare with Figure 1.

Now, let's say we clip something in the audio chain even before it gets to the radio. Let's clip really badly:

Figure 5: Badly clipped AFSK. Here the damage is done even before we get to the radio.

Let's do something even worse...we'll use low tones so we get of lots of harmonics in our passband. We will inject this into our radio and adjust it "properly" for 3-5 bars on the meter. Without the AFSK TX filter engaged, we get exactly what we might expect (see Figure 6). The sound of this in the monitor vaguely resembles RTTY, and will probably earn you no friends if you transmit this over the air. Garbage in, garbage out.

Figure 6: Transmitted spectrum of badly clipped AFSK shown in Figure 5. This is without the AFSK TX filter enabled on the radio. This figure uses the same reference as Figure 1 (no adjustment necessary to compare).

Now lets set our AFSK TX filter to the low tone pair and engage it. You may compare this to the FSK-D waveform above to see just how bad things have gotten.

Figure 7: Transmitted spectrum of badly clipped AFSK shown in Figure 5 with the AFSK TX filter enabled. Same reference as Figure 1.

Just for good measure, we will now go "sliders up" with this signal so that we presumably clip both the ADC in the K3 with this already clipped audio input. This is going for maximum damage: full power from the audio source, maximum gain on the K3 line input, AFSK TX filter turned on in the radio. Again, you may wish to compare this to the FSK-D spectrum at the beginning of this discussion to see how far the signal has degraded.

Figure 8: Same as Figure 7, except we cranked the line-in gain up to "60" on the K3. This is a clipped waveform coming into the radio, clipping again in the ADC, and then filtered in DSP using the AFSK TX filter. Reference is the same as Figure 1 for comparison.

Buzzes, ground loops, and noise! Oh my!

Let's move on to another bugaboo: noise on the audio input. This can come from, 60 or 120Hz buzz, noises from ground loops etc. Even some really poor sound cards have a tendency to pick up electrical signals at a low level or possibly generate their own spurs. All of this has the potential to go out over the air if it falls in your TX passband.

Let's move out of the land of the ideal and go to the realm of the completely broken. Let's say the ground popped off your audio cable somewhere introducing a horrendous buzz across your passband at 120Hz intervals. Without the AFSK filter, all of these 120Hz harmonics will show up as charming carriers around your signal, which in turn will earn you much endearment from your neighbors. You can listen to this signal here. The highlighted spectrum indicates the passband of the receever in the audio recording. It is obvious that something is horribly wrong on the K3 audio monitor as well.

Figure 9: 120Hz sawtooth wave generating harmonics across the passband about 15dB below the signal. The AFSK TX filter is not engaged. Reference is the same as Figure 1.

Using the K3's AFSK TX filter in this scenario, you would transmit the following spectrum, still with the broken wire. You can see one or two of the harmonics that make it through the passband of the DSP filter. You may listen this filtered version here.Again, you may wish to compare this to the FSK-D spectrum at the beginning of the article.

Figure 10: Same as Figure 9, except that the AFSK TX filter is engaged. Reference is the same as Figure 1. The tones are of equal amplitude--this picture happened to catch a character that was "space-heavy".