Key clicks on CW

[JJ4SDR]

For the first time ever, someone told me that I generated awful key-clicks on CW. Also, I was told that it seemed like there were more than 1 CW signal at the same time. The station I worked told me afterwards that I was "clean" on his end. I have embedded some images below.

Even barefoot, with amp, the CW looked about the same into a dummy load. It is flakey so I don't want to have the amp in line when putting RF from amp into it.

Any ideas?

CW at 1200W on Web SDR.jpg (79.72 KiB) Viewed 766 times

CW at 40W on Web SDR.jpg (24.17 KiB) Viewed 766 times

[JJ4SDR]

Other images

DSP_CW settings.jpg (76.33 KiB) Viewed 764 times

DSP Options.jpg (99.84 KiB) Viewed 764 times

CW at 1200W.jpg (241.24 KiB) Viewed 764 times

CW Options1.jpg (136.2 KiB) Viewed 764 times

[JJ4SDR]

Nobody has any comments or advice? Perhaps, most bought this product for the amazing Audio quality ? To me, I need quality on CW too

Juha
NI2M

[W2PA]

Juha,

I'm not seeing anything wrong with the settings you posted, except maybe your MOX delay is high. See my settings below - my CW sig has always been click free. Also, it's difficult to completely judge your panadapter photo without its vertical scale.

Are you really running Thetis v2.10.3.4? If so, why not upgrade - the current test version is v2.10.3.12.

[w2gs]

Juha,
Have tried a database reset and a computer restart?
73, George
W2GS

[JJ4SDR]

Hi Chris!

Vertical scale:
Max -20
Min -100

Will try to play with MOX.

The only reason for me using a very old version is that on SSB, any version that features cmASIO (I don't need to use Voicemeeterpotato as I don't need/want to) is that when I have Digital Audio Workstation (DAW) in the audio chain, latency increases.......Truth be told, I haven't actually tried it as the current ("antique") version has worked well
for years. I have also taken part in CW contests and have never been made aware of any "clicks."

I assume the latest "test version" can be had on github?

George,
I appreciate your suggestion, too.

Juha
NI2M

[w9ac]

Under General>F/W Set, try reducing Rx1 Sample to 192K to first see if that has any effect on the problem.

Paul, W9AC

[JJ4SDR]

Hi Chris,

RX1 sample size is already 192k. Reduced it by a notch. No appreciable difference. Will also see if changing MOX setting makes a difference. Tried and no joy.

Juha
NI2M

[ramdor]

try keying with cwx form, and check if it is same/different

[JJ4SDR]

Hi Richie,

Thanks for the suggestion. I have never used that form, but I guess I can figure it out. Will report back.

Juha
NI2M

[laurencebarker]

The firmware "shapes" the CW amplitude so that it ramps up slowly to avoid clicking. That assumes that the PA is already turned on, so the RF signal builds up through an "already on" RF chain.

You mentioned an amplifier - how is that keyed on please? Any delay in turning it on may have an impact. Similarly there are timing parameters in Thetis for when the TX ramp starts, to make sure that it is after the RF chain is ready.

[w4bf]

Some solid state amps (Mercury III) keep bias off until it detects rf is in band. This puts hard edge on rise time of first dit - actually chops a bit off.

[JJ4SDR]

Hi Richie and Laurence!

Tried CWX form and don't see much difference. See embedded images below. About 40 W out and 1300W.

NOTE: The scale used to display the X axis is 4x!

Amplifier is a German RF2K-S full-legal limit and it is keyed with a PTT line between the ANAN and the amp. The RX/TX switching in the amp is via a PIN diode and according to the specs, the time from RX to TX is <1ms.

I have provided images from some of the CW settings that I am using using currently above, in this thread! Would be wonderful is someone who also uses an RF2K-S amp would take a look at my settings. The reported CW clicks are likely due to poor CW settings in Thetis!

I will REALLY appreciate pointers!!

CWX Form at 40W.jpg (288.65 KiB) Viewed 418 times

CWX Form at 1200W.jpg (295.88 KiB) Viewed 418 times

General Options Form.jpg (142.47 KiB) Viewed 418 times

[w9ac]

Juha,

What's being missed here is that you're hyper-zoomed-in on the CW waveform spectra -- and it looks worse than it is. You've expanded each horizontal division to 200 Hz. At 200 Hz from Fc, your keying sidebands are approximately -40 dBc.

I conducted the same test, 200 Hz/div. using an Rx1 Sample of 384K, and 40W into a Bird Termaline load. I also measure -40 dBc relative to 200 Hz from Fc (see attached image below). The waterfall image shows bandwidth mostly held to +/- 200 Hz (i.e., red coloring). A break can be seen between two bandwidth samples.

The spectra is a string of dots at 20 WPM. Bandwidth doesn't change with keying speed. It's the slope and contour of the rise/fall waveform that determines bandwidth, notwithstanding issues like ALC envelope control, for example. In that case, ALC can have an effect on bandwidth as keying speed changes, but that's not at all evident in the Apache Labs SDR transceiver designs.

The station complaining of clicks may have had a noise blanker engaged. Did you ask?

I rarely look at spectrum when assessing things like RF hot-switching, a common cause of key clicks. I'd rather look at the time domain. Just my opinion, but all of us should own an oscilloscope for various transmission measurements. We think nothing of spending nearly $1K on a fancy wattmeter with needless high-precision accuracy, then neglect something truly useful like a scope.

By the way, there have been discussions between the ARRL and transceiver manufacturers to comply with a proposed RF bandwidth mask. For CW, it's a modified Gaussian curve, not a strict raised-cosine shape. I've no idea how far along commitments have come from manufacturers.

Paul, W9AC

[JJ4SDR]

Paul,

As always, you provide gems of information.

I was always under the impression that as provided by AI:
"Sending Speed (WPM): A faster sending speed increases the rate at which dots and dashes are sent, increasing the bandwidth and the spacing of the signal's sidebands" (as per AI Overview) + RF Rise and Fall Time: The time it takes for the CW signal's envelope to turn on and off (its rise and fall times) is crucial. Faster, sharper transitions create more harmonic content and thus occupy more bandwidth, regardless of the sending speed + Harmonic Content: The on-off keying of the signal inherently creates harmonics, which are essentially sidebands of the carrier frequency. A faster sending speed and shorter rise/fall times increase the harmonic content and the overall bandwidth."

So I think I need to look at my Thetis settings some more. I often send at 30WPM-35WPM, even when chewing the rag.

Reporting station was gone fast so I didn't have a chance to ask about NB. I may play with RX1 sample rate some, too.

You are dead-on regarding the way we spend money. i.e., I have not acquired an oscilloscope.

73!
Juha
NI2M

[w9ac]

Keyed bandwidth and its relationship to information rate is best answered by these authors on the subject:

https://www.w8ji.com/cw_bandwidth_described.htm
https://www.ivarc.org.uk/uploads/1/2/3/ ... sion_1.pdf
https://w9cf.github.io/articles/click.pdf
https://www.w8ji.com/cw%20bandwidth%20analysis.htm

Paul, W9AC

[w-u-2-o]

Warren Pratt has been working closely with the ARRL on the ARRL's "Clean Signal Initiative".

https://www.arrl.org/arrl-clean-signal-initiative

As a result, he has proposed a further optimization of the shaping of the CW leading and trailing edges for openHPSDR radios. Note that shaping is already hard-coded into the software and firmware that substantially meets the ARRL goals. I'm not sure/can't remember what the status of this is, but I will endeavor to find out and report back.

Here is a snippet from the QST May 2024 article (worth a read, see the link at the URL posted above):

Capture.JPG (108.96 KiB) Viewed 191 times

[w9ac]

Now compare the proposed RF mask with my measured spectrum. At 200 Hz from Fc, the current raised-cosine shape from my 7000DLE with Thetis *just* fits under the stepped curve. But at -60 dB, it just barely falls short. It's so far down, that it has a de minimis effect.

From G3OTK's article at https://www.ivarc.org.uk/uploads/1/2/3/ ... sion_1.pdf

Referring to Figure 5:

"The part of the Gaussian response between the amplitude 0.1 and 0.9 is a close approximation to the raised cosine with a maximum slope equivalent to the 5ms ramp but with an additional delay of about 1.2ms. The initial rate of change is less and the so required bandwidth is less. This Gaussian response is particularly attractive because it has almost no overshoot."

This is the League's new target.

Digressing back to signal rate versus waveshape and the effect on occupied bandwidth - the keyed information rate can have an effect on bandwidth. But it's because the waveshape slope must decrease as the information rate increases. That said, at normal CW keying speed, the bandwidth is only determined by the waveform rise/fall time and contour shape. That's because until we reach a rate where the normal 4-5 ms rise/fall becomes a dominant factor, bandwidth is not determined by the CW speed. I think W8XR explains it better:

"Some might still argue that the increase in the modulation rate is causing this. Of course, there’s a kernel of truth there. If you increase the rate of modulation, you increase the steepness of the modulating waveform (as is evident in Figure 6.) But, it’s actually the steepness of the rising and falling envelope that causes the sidebands (and consequent bandwidth), not the signaling rate. If you had only one wave at this frequency, the bandwidth required would be exactly the same as the continuous train of waves in our example."

Paul, W9AC

[w-u-2-o]

Or, to simplify, Thetis provides the best spurious output performance of any radio available for both CW and phone by means of proper CW waveform shaping and PureSignal linearization.