Thetis v2.8.11 - UPDATES and current issues

[ramdor]

*** another test 21k9pre4d

Please give this a try.

Download x64 [md5:] - test complete

- fix: QSKEnabled setter for the disabled state where it did not turn off cwsidetone when disabling qsk if it needed to
- change: DX2 error handler change

Richie.

[DL2XY]

Have another one, maybe related?

Thetis 21k9pre4c:

Switching modes takes a very long time (up to 30s) if mode specific panel have to be changed, e.g. USB/LSB/AM to DIGU/DIGL.
During this time the mode specific panel is only partional updated and GUI is unresponsive.

Starting Thetis in DIGU/DIGL leads to 6s black window (only bachground) after splash screen.

Database reset was futile, no exceptions or error log.

Walter

that is very interesting, perhaps the window issue was a bunch of dead msg boxes related to this occlusion issue spamming up. I added catch all error handling into the render function just to check things were behaving, and resulted in catching..... well.... everything

Yes, the slowness of mode swap, it is a concern. I will build with older wdsp and time the difference. It can take upward of 2-3 seconds to rebuild the buffers/filters/types for a given mode depending on how different they are from the mode you just left. If there is some odd Mode/State swapping at startup then it could be doing this multiple times and the time adding up. It is not too high on my list at the moment, but it is there.

Richie.

It's not the 2-3s i am be used to, this is way more and it's definitly introduced by 21k9.

I measured a few older versions to compare:
21j2 4.13s
21k6 4.05s
21k7 4.18s
21k8 5.36s
21k9c 45.96s

It happens in phases one can clearly monitor on CPU load:
1. DIGU click- sound stops, panel garbled, panadapter freezes, GUI freezes.
2. Panadapter works again
3. Sound starts
4. Mode specific panel is updated
5. GUI is responsive again

mode.jpg (78.44 KiB) Viewed 9103 times

All in all that's nearly a minute of unresponsive GUI, most users would have killed the process before... (I did at first )

hope it helps a little,
73 Walter


Edit: most CPU load during hang is in main thread.

threads.jpg (85.13 KiB) Viewed 9098 times

[Yevgeni]

*** another test 21k9pre4d

Please give this a try.

Download x64 [md5:7B2B6986C10505764EDCDB9B07A2FE7F] - https://bit.ly/3LdC3kF

- fix: QSKEnabled setter for the disabled state where it did not turn off cwsidetone when disabling qsk if it needed to
- change: DX2 error handler change

Richie.

I don't have any mistakes DirectX.
Thanks!

[ramdor]

@Walter

21k9pre4c has been and gone, gone to the code grave. It could have been thrashing out exceptions every frame for all I know at this stage. Move on to next next version

WDSP has changed, that is out of my control, and the changing of buffers/filters/types for the dsp is slower than it was before. I will spend some of tomorrow working through the codepaths when changing modes and filter changes, because I think they are causing some problems.

For the moment align the modes that you are switching between for your DSP filter widths and types. There is some additional delay wdsp causes by the large amount of work it does when everything need swapping around and rebuilding. That is fine if you are calling a 1second delay function once, but not if it is accidentally being called 10 times. I've not had a chance to start looking.

EDIT: the new wdsp was introduced in 21k9, I only had it on the 30th Jan.

R.

[W1AEX]

Thanks for the pre4d update Richie. It came up fine with my Win 7 machine and is running smoothly (so far)!

73, Rob W1AEX

[ramdor]

nice, all a learning curve with these variants of directX levels, different interfaces, etc. Getting there slowly. I did start looking at variable refresh support, but that can wait for a very rainy day

I did a good test last night with my surface pro 8 (i5) which when at 100% scaling is like 3000x2000 odd pixels. It could not maintain 60fps at that res with two receivers, with all the trimmings, but turn down the decimation thing, and boom, 60 or even 120 hz !

R.

[I2NOY]

21k9pre4d is working without DirectX error on my PC.
21k9pre4b had DirectX error after 10-20 minutes.

73,

Maurizio I2NOY

[DL8LAQ]

*** another test 21k9pre4d

Please give this a try.

Download x64 [md5:7B2B6986C10505764EDCDB9B07A2FE7F] - https://bit.ly/3LdC3kF

Got a blue screen when starting this version the for 1st time. PC was running several hours.
No problem after this unintentional reboot.

[W7GES]

Richie,

Thanks for all the hard work you do!

I haven't seen any issues with 21k9(pre4).

The Spectrum display, peak blobs, and active peak lines work well here. Not sure if you only wanted problem feedback but I thought you might like to hear some are not having problems with direct x, etc.

Windows 10 Pro 21H1, Intel i5-10500, 80gb ram... 7000DLE MKII

73 de George / W7GES

[DL2XY]

@Walter

21k9pre4c has been and gone, gone to the code grave. It could have been thrashing out exceptions every frame for all I know at this stage. Move on to next next version

WDSP has changed, that is out of my control, and the changing of buffers/filters/types for the dsp is slower than it was before. I will spend some of tomorrow working through the codepaths when changing modes and filter changes, because I think they are causing some problems.

For the moment align the modes that you are switching between for your DSP filter widths and types. There is some additional delay wdsp causes by the large amount of work it does when everything need swapping around and rebuilding. That is fine if you are calling a 1second delay function once, but not if it is accidentally being called 10 times. I've not had a chance to start looking.

EDIT: the new wdsp was introduced in 21k9, I only had it on the 30th Jan.

R.

You are right, i have tracked down the delays by adjusting everthing manually - it's mostly changes in DSP Buffersize and Filter Size.

thanks for your time,
Walter

[Yevgeni]

Hi

For many this question is not important. The carrier level on the panorama is invalid. When measuring, you can make a mistake if you do not know this.

[W9BHI]

Is there a IK3VIG special button pack update for the new ZTB feature?
Thanks.

[oe3ide]

Just save this 2 images to IK3VIG Special\console

btnDisplayZTB-1.png (1000 Bytes) Viewed 8905 times

btnDisplayZTB-0.png (640 Bytes) Viewed 8905 times

73 Ernst

[wa2cop]

I admit, I could be wrong - but did enabling the attenuator RAISE the noise floor measurements (ie pan adapter and S meter) happen in previous versions?
I have auto AGC off.

73
Mike
WA2COP

[w-u-2-o]

I admit, I could be wrong - but did enabling the attenuator RAISE the noise floor measurements (ie pan adapter and S meter) happen in previous versions?
I have auto AGC off.

Yes, this is a natural consequence of receiver noise figure (NF) degradation (increase in noise figure) due to input attenuation prior to the first gain stage.

The first, and only, gain stage in our receivers is the ADC preamp, which in the 7000 and 8000 provides 14dB of gain and a 6.2dB NF.

The NF of an attenuator is equal to its loss factor (see the above link again). So if you set the attenuator to 5dB the noise figure of the attenuator is 5dB.

The resulting system noise figure is not just a summation of these values. The math for cascaded noise figures, the Friis formula, can be found in the above link. A handy calculator can be found here. However, because an attenuator has a negative gain equal to its NF, its effect on noise figure is simply to degrade it by an amount equal to the attenuator.

You can easily demonstrate this by applying a dummy load or 50 ohm termination to the antenna input to eliminate the environmental noise floor from the antenna. The resulting noise floor represents thermal noise in the measurement bandwidth (e.g. 2.93Hz on the panadapter or 2900Hz on the S-meter) plus the system noise figure. Then simply adjust the attenuator and you can see the noise floor change in an amount equal to the attenuator setting. It won't be exact because our radios are not lab equipment, but it'll be darn close.

With the antenna reconnected, environmental noise is generally (but not always) so much higher than the system noise floor that the effects of NF degradation from increased attenuation are much less, but they are still noticeable, as you have observed.

Best practice is to increase the attenuator in 1dB steps until a slight diminishment in the SNR of the largest signal on the band is seen, then back off a couple of dB. This optimizes the input dynamic range into the ADC for the current band conditions.

[W9BHI]

Thanks!

[W9BHI]

I got the 21k9 pre4 installed as well as the ZTB button.
I see it to the right of the zoom slider, but it doesn't do anything when clicked on.
Maybe I don't have something else enabled?
Don

[W9BHI]

I figured it out, got to be in CTUNE.

[W9BHI]

21K9pre4 is working FB.
I did notice it has a blank gray screen for a few seconds before starting.
Don

[ramdor]

I am glad things seems to have settled today, after the changes. Thanks for the testing and all the feedback.

Onto ZTB, to be honest I am not a fan of ZTB as it is all dependant on the sample rate and bandwidth of the band. Also, it becomes pointless when CTUN is off as I have mentioned in a couple of recent videos. I can put extra effort into expanding up to an edge (right/left) if it can be viewed at the same time as the vfo freq, but... better things to do *grin*

Perhaps we can throw some ideas around regarding it. The other thing to keep in mind is it that it has to reverse calculate the zoom slider value and reset the pan position, so it is all related to the 'sleepiness' of that zoom slider.

In other news, far from complete, but getting there this evening :

9lives.png (45.45 KiB) Viewed 8798 times

Richie.

[W9BHI]

Richie,
So is the delay in the startup an issue or the new normal?
Don

[ramdor]

Richie,
So is the delay in the startup an issue or the new normal?
Don

Time will tell really. There were important changes/improvements Warren made that need to stay, and have added slight additional delay to the building of the DSP filters in WDSP. When you change mode, depending on the mode you are leaving, if the buffers/filters/types are different then they have to be rebuilt/adjusted, this takes time.

Yes, work needs to be done to the client to improve start up speed. At the moment for example, before you even turn the radio on, SetModeRX1 has been called five times during startup. This is not good, this is actually quite bad, and it will need time to resolve. It has been like it since day one, but as operations were speedy this issue was ignored. It is largely a by-product of way way too much logic being integrated into the UI.

Richie.

[w-u-2-o]

In other news, far from complete, but getting there this evening :

Cool!

[W7GES]

In other news, far from complete, but getting there this evening :

Cool!

I will second that!

Also, I haven't really noticed slower performance with 21K9pre4 but the computer running Thetis has lots of RAM and the rest of my Ham software is running on another PC.

73, George / W7GES

[Yevgeni]

I am glad things seems to have settled today, after the changes. Thanks for the testing and all the feedback.

Onto ZTB, to be honest I am not a fan of ZTB as it is all dependant on the sample rate and bandwidth of the band. Also, it becomes pointless when CTUN is off as I have mentioned in a couple of recent videos.

Richie.

Hi

According to the ZTB. When I applied the markup on the zoom slide, I stopped pressing the zoom buttons.
I set the slider to the desired optimal zoom value.

[ramdor]

Hmm going to make a variation to the ZTB button, with an option in setup to switch it to this :

right click : stores zoom / pan for that band
left click : restores the stored zoom / pan for that band

Will work in CTUN off/on modes. It would be more use to me than how it is at the moment. If you like looking at the CW section, you could set that as the ZTB button for that band as an example. You can use bandstack overlay to do similar, but I think it would be nice to have a dedicated feature.

Richie.

[oe3ide]

Hmm going to make a variation to the ZTB button, with an option in setup to switch it to this :

right click : stores zoom / pan for that band
left click : restores the stored zoom / pan for that band

Will work in CTUN off/on modes. It would be more use to me than how it is at the moment. If you like looking at the CW section, you could set that as the ZTB button for that band as an example. You can use bandstack overlay to do similar, but I think it would be nice to have a dedicated feature.

Richie.

Richie, that would be a big improvement. So we can adjust the span/zoom for the desired portion of the band and store this. Then zoom in to a specific signal and with the push of a button, zoom back to the desired portion of the band ... I like that idea

73 Ernst

[ramdor]

yep, you could set up the dx portion one day, cw the next. Seems like nice thing... but need sleep Will look at it tomorrow.

Fixed a bunch of other stuff today though, including the grey window/box at start up. Also all the duplication of events related to filters/modes at startup. Takes around 11 seconds on my machine. There are 4 main areas that are slow, each around 3 seconds or so and not much I can do to improve it. I have set them as the stages on the splash screen.

Richie.

[ramdor]

*** 21k9pre5 another test release

Download x64 [md5:] - test complete

- fix: clean up of start up process and the grey window is finally banished !
- add: TCPIP CAT server (under CAT+). Any number of connections can be made, each client has their own cat 'answer' strings. Currently the connection is terminated after 30 seconds of inactivity [todo: option to disable this/increase time]

Richie.

[Tony EI7BMB]

Nice to see the grey window gone:-) Thanks Richie