Thetis QSK and the new default low-latency filters

User avatar
W1JA
Posts: 28
Joined: Sun Apr 09, 2017 3:19 pm

Thetis QSK and the new default low-latency filters

Postby W1JA » Sun Apr 21, 2019 10:51 pm

Starting with Thetis 2.6.4, the release notes recommend that we reset our database to get new, improved default settings. One of the new settings is a filter type change from Linear Phase to Low Latency.

I have found that QSK is better with a linear phase filter.

To demonstrate the issue
Operate in the new QSK mode using a dummy load. For this test, you want to be able to hear your transmitted signal clearly in the Thetis receiver. To this end, turn off any sidetone from any external keyer, and make sure the Sidetone checkbox in the Thetis console is unchecked. To start, make the problem as obvious as it can be by setting the CW RX filter row in Setup > DSP > Options to Buffer Size = 64, Filter Size = 16384, Filter Type = Low Latency, Filter Windows = BH-7. (Except for Filter Size, these settings are the new Thetis defaults.) Set the receive selectivity to a narrow value, say 250 Hz, and send some CW. You should hear significant audible ringing on your CW signal.

Now put the Filter Size back to its default value of 4096. The ringing is reduced but still present. Even at the shortest Filter Size of 1024, I can still hear a slight artifact, especially at a selectivity of 1.0k.

I find that nothing equals the Linear Phase filter type to eliminate these artifacts at all selectivity settings. With the Linear Phase filter, I tried Filter Size = 1024 for the lowest latency, but filter skirts are noticeably degraded. Changing Size to 2048 produces a negligible increase in latency, but a big improvement in skirt selectivity.

(I should add that these artifacts are present on all received CW signals. They're harder to hear on signals weaker than one's own transmitted signal, but they're there.)

73, John W1JA
User avatar
w-u-2-o
Posts: 1895
Joined: Fri Mar 10, 2017 1:47 pm

Re: Thetis QSK and the new default low-latency filters

Postby w-u-2-o » Mon Apr 22, 2019 1:25 am

I was waiting for someone to complain about ringing on CW, but I didn't think it would be associated with linear phase versus minimum phase (low latency). The filter skirts are very sharp even at the smallest filter sizes either way. BH-7 is already less steep than BH-4, and with better stop bands. Maybe we need a size 512 for the low latency filters?
User avatar
w9ac
Posts: 65
Joined: Sun Apr 09, 2017 4:01 pm

Re: Thetis QSK and the new default low-latency filters

Postby w9ac » Mon Apr 22, 2019 2:46 am

John,

What you're describing shows just how good performance is with the new QSK changes. It's oh-so-simple to blank audio for a period of time in order not to deal with it. Apart from Elecraft, that's what every commercial manufacturer does.

You're right about the slight ringing. My default CW filter size is 64/1024 and that produces the lowest latency and no artifacts as you noted.

It's nice to have the ability to change these settings based on conditions so that we -- the end-user can make these decisions and not someone else making a decision that gets locked down in firmware.

This is by far the best QSK system I've used. Quite honestly, I'm blown away that it's even become possible due to so many inherent SDR/DSP limitations.

Paul, W9AC
User avatar
W2PA
Posts: 78
Joined: Sun Apr 09, 2017 6:34 pm
Location: LaGrangeville, NY
Contact:

Re: Thetis QSK and the new default low-latency filters

Postby W2PA » Sun Apr 28, 2019 1:21 pm

Guys,

I don't believe this is a ringing issue in the usual sense. It's more like a resonance in the Low Latency filter that happens in *all* filter widths and tracks the cutoff edge furthest from the carrier (i.e. the highest audio freq).

I first thought it was a resonance in my headphones that I'd never noticed before. I was hearing it on voice peaks in SSB during a QSO. But on going back to Linear-Phase filters, it disappeared entirely.

I can make it happen at will by introducing a spike-like noise in the receiver. I do it by switching my antenna switch repeatedly. For example, in LSB, with Low Latency filter type and the width set at 3 kHz, I hear a "tink" sound that has a pitch in the neighborhood of 3kHz. With a wider filter, it's a higher pitch - narrower, a lower pitch (corresponding to the upper edge, audio-wise). And it disappears completely by switching to Linear Phase. This happens in every filter width.
73,
Chris, W2PA
Bryan W4WMT
Posts: 80
Joined: Sun Apr 09, 2017 10:12 pm

Re: Thetis QSK and the new default low-latency filters

Postby Bryan W4WMT » Sun Apr 28, 2019 3:08 pm

Hi Chris,

I agree with everything you posted above. I noticed the funny sound you describe from the very beginning, but nobody else agreed, so I just forgot about it and ran the linear phase filters. Luckily, with Warren”s new pipelined overlap/add buffers, the linear phase filters are almost as fast as the minimum phase filters.

I can even hear the artifact you describe when listening to QRN on an unoccupied low-band frequency. My guess has been that it’s some sort of a psycho-acoustic effect where your ears, having grown accustomed to the sound of a mathematically perfect receiver, can pick out the slightest defect - like the tiny phase distortion from a minimum phase filter. Pseudoscience BS, perhaps? Maybe.

73, Bryan W4WMT
jeffreydoran
Posts: 61
Joined: Sun Oct 29, 2017 9:39 pm

Re: Thetis QSK and the new default low-latency filters

Postby jeffreydoran » Tue Apr 30, 2019 8:12 pm

I've noticed the "tinking" ring at higher filter lengths beyond 2048, but now they are very evident with 2.6.6
I am also hearing a ringing artifact on Side Band transmit monitor. I operate with headphones almost exclusively with Monitor "ON".
Thought I was hearing things. I'll experiment with the Linear Phase filter later today.
Thetis 2.6.6 P2 v1.9
Jeff
NJ2US
User avatar
w-u-2-o
Posts: 1895
Joined: Fri Mar 10, 2017 1:47 pm

Re: Thetis QSK and the new default low-latency filters

Postby w-u-2-o » Tue Apr 30, 2019 10:41 pm

I guess my hearing isn't so good because I'm not hearing any of this?
Bryan W4WMT
Posts: 80
Joined: Sun Apr 09, 2017 10:12 pm

Re: Thetis QSK and the new default low-latency filters

Postby Bryan W4WMT » Wed May 01, 2019 8:10 pm

Scott, it’s certainly not objectionable.
Pretty subtle actually.
Bryan W4WMT
Posts: 80
Joined: Sun Apr 09, 2017 10:12 pm

Re: Thetis QSK and the new default low-latency filters

Postby Bryan W4WMT » Wed May 01, 2019 8:19 pm

Jeff,

What you are hearing in the MON is probably something else altogether.
The monitor feed isn’t audio at all, it’s the actual baseband IQ stream being sent back to the DUC for transmit. Sounds kind of like funny out of phase audio. Also, if you have PureSignal running, you’re hearing the predistorted version of the stream as well.

73, Bryan W4WMT
User avatar
W2PA
Posts: 78
Joined: Sun Apr 09, 2017 6:34 pm
Location: LaGrangeville, NY
Contact:

Re: Thetis QSK and the new default low-latency filters

Postby W2PA » Wed May 01, 2019 8:33 pm

The artifact isn't immediately noticeable on SSB, but over time, you begin to hear it as a resonance at the high end of the filter. If you can't hear it at 3kHz, go down to 2.5 or so and listen for it at the high end of your QSO partner's audio. As you pointed out, Bryan, it's also possible to hear it on background noise - but you have to listen more carefully.

To me it's objectionable enough to cause me to avoid using those filters entirely on phone, and it's fatal on CW especially on a band crowded with strong signals. I don't think it matters at all on digital modes. I'm also confident it's an entirely solvable problem.
73,
Chris, W2PA

Return to “Thetis”