Page 7 of 20

Re: 2.6.9

Posted: Mon Nov 18, 2019 12:49 pm
by W4WMT
ramdor wrote:
No, this issue I have found is to do with the seq log data request from the client end. I have fixed it, but I dont really 'like' the fix.

From memory tx and rx transitions sleep the main ui thread briefly, for the rf delay setting, ptt_out_delay and mox_delay. Could be scheduler 'doing its thing' ?

Try changing the GDI to .Invalidate instead of Refresh() in RunDisplay() and test to see if it then happens in gdi with panadapter on. I expect it will.

Richie.


Like this?

GDI Change MW0LGE.jpg
GDI Change MW0LGE.jpg (71.38 KiB) Viewed 15792 times


If so, things are still the same. I'm seeing the DDC/PortAudio threads being blocked when in DX mode OR when in GDI with the Panadapter turned off. No blocking when in GDI mode with the Panadapter running. Maybe I'm just up against the limit of my 4 core i7 :(

I should mention that my computer doesn't have a graphics adapter, I'm just using the built in graphics in the i7.

Still though, I would like to get to the bottom of this thread blocking business before the GDI+ code is deprecated :D

73

Re: 2.6.9

Posted: Mon Nov 18, 2019 12:58 pm
by w-u-2-o
VE7DQC wrote:Hi Scott
Regarding 2.6.9, I am wondering if an issue with channel scanning has been reported....Mike VE7DQC
Mike,

This is the first I've heard of it. I'm assuming this is with PCALE? CAT control works flawlessly for me hear with a wide variety of ancillary applications.

And, as a one-time PCALE user myself, we should be more precise in our language: "channel scanning" just means PCALE sending repeated VFO (and maybe mode) requests via CAT. It's not a unique function.

Have you tried this with 2.6.9? If so, and if the problem remains, please describe it in more detail.

Thanks,

Scott

Re: 2.6.9

Posted: Mon Nov 18, 2019 12:59 pm
by w-u-2-o
W3MMR wrote:A7 working good here so far. I am having a PS issue though.
Perry--what was the last "good" version for you in this respect?

Re: 2.6.9

Posted: Mon Nov 18, 2019 1:09 pm
by W4WMT
ramdor wrote:
No, this issue I have found is to do with the seq log data request from the client end. I have fixed it, but I dont really 'like' the fix.

Richie.


I saw no memory leak during an overnight run with Process Explorer. But, I had already commented out all of the seq log data request stuff before doing the run.

73

Re: 2.6.9

Posted: Mon Nov 18, 2019 4:30 pm
by ramdor
Bryan W4WMT wrote:I saw no memory leak during an overnight run with Process Explorer. But, I had already commented out all of the seq log data request stuff before doing the run.

73


Yes would only happen very rarely with a seq error, it is now resolved. Ran for 12hrs without any problems.

Ref the blocked threads. How are you testing/seeing this? What settings are you using for the RB/PA latency?

Richie.

Re: 2.6.9

Posted: Mon Nov 18, 2019 6:17 pm
by W4WMT
ramdor wrote:
Ref the blocked threads. How are you testing/seeing this? What settings are you using for the RB/PA latency?

Richie.


I'm just looking at the VAC1 ringbuffer events, when it happens the rings are getting blocked from both sides at the same time, just ever so briefly. Using 0 for the PA suggestedLatency and 3 for the ringbuffer latency (which gives me room for 4 sample blocks in the rings at the same time).

QRX one. I'm going to bring an Nvidia graphics card home with me from work today, jack it in to the SDR computer, and see if I get the same results as I'm getting with the built in i7 graphics in DX mode.

Will report back later...

73

Re: 2.6.9

Posted: Mon Nov 18, 2019 6:26 pm
by ramdor
If i run my i7 3770k using the hd4000 integrated gpu, with 0 set for latency on both rb and pa, and force turned on for to/from with 1.00000 set, then I get no underflows/overflows when doing tx/rx cycles in both DX and GDI (also with displaymode off). As soon as I re-enable the re-sampler (tick off force) then underflows/overflows just runs away. The var ratios seeming to rise/fall in an endless cycle and never reach a settle point.

I have never used the resampler because I am yet to see it settle, with auto latency or manual. Traces in ft8 for example just slope one way, then start sloping the other, an endless cycle.

All that said, I probably have everything setup wrong and the above info is about as much use as a chocolate teapot :P

Also, I did notice when using HD4000 that opening Chrome from fresh would cause lots of packet loss seq errors to occur, I don't see that when running through the 680gtx.

Richie.

Re: 2.6.9

Posted: Mon Nov 18, 2019 6:39 pm
by W3MMR
w-u-2-o wrote:
W3MMR wrote:A7 working good here so far. I am having a PS issue though.
Perry--what was the last "good" version for you in this respect?


Honestly Scott, I dont know. I havent been operating much SSB the last 3 or 4 months. I never use PS while operating AM, which is what I mainly operate, or FT8. I go thru spurts of phone DXing. If I was to say a version, I would just be shooting in the dark.

Perry

Re: 2.6.9

Posted: Mon Nov 18, 2019 6:41 pm
by W3MMR
BTW, I cant wait for A8 so I can see how far I can push my GPU lol

Re: 2.6.9

Posted: Mon Nov 18, 2019 6:51 pm
by ramdor
W3MMR wrote:BTW, I cant wait for A8 so I can see how far I can push my GPU lol


a8 version
i7 3770k @ 3,9GHz
nvidia 680gtx
1920x1080 thetis client window size (borders included) on a 2k display (2560x1600)
single rx in panafall 50% split
filled panadapter
anti-alias
back ground skin

Completely pointless, mostly as this screen is 60hz, but still :lol: Capped at the 144 limit I set in there.

CaptureDX.PNG
CaptureDX.PNG (14.04 KiB) Viewed 15704 times


Compared to the same settings in GDI+ mode. The difference is somewhat extreme.

CaptureGDI.PNG
CaptureGDI.PNG (9.27 KiB) Viewed 15704 times

Re: 2.6.9

Posted: Mon Nov 18, 2019 6:57 pm
by w9mdb
Where did the antenna selection page go?

Re: 2.6.9

Posted: Mon Nov 18, 2019 7:02 pm
by ramdor
w9mdb wrote:Where did the antenna selection page go?


hasn't moved, but is in setup-general-ant/filters-antenna, or setup-general-alex-antenna

Richie.

Re: 2.6.9

Posted: Mon Nov 18, 2019 7:06 pm
by W3MMR
ramdor wrote:
W3MMR wrote:BTW, I cant wait for A8 so I can see how far I can push my GPU lol


a8 version
i7 3770k @ 3,9GHz
nvidia 680gtx
1920x1080 thetis client window size (borders included) on a 2k display (2560x1600)
single rx in panafall 50% split
filled panadapter
anti-alias
back ground skin

Completely pointless, mostly as this screen is 60hz, but still :lol: Capped at the 144 limit I set in there.

CaptureDX.PNG

Compared to the same settings in GDI+ mode. The difference is somewhat extreme.

CaptureGDI.PNG


Well my i7 4770 16gb ram, 250GB SSD, w/ Nvidia GTX1050 TI GPU, 1920x1027 resolution, should have no issue... Now to upgrade to better monitor lol.

Re: 2.6.9

Posted: Mon Nov 18, 2019 9:27 pm
by w-u-2-o
W3MMR wrote:
w-u-2-o wrote:
W3MMR wrote:A7 working good here so far. I am having a PS issue though.
Perry--what was the last "good" version for you in this respect?


Honestly Scott, I dont know. I havent been operating much SSB the last 3 or 4 months. I never use PS while operating AM, which is what I mainly operate, or FT8. I go thru spurts of phone DXing. If I was to say a version, I would just be shooting in the dark.

Perry
Then there's no way to pin the problem on Thetis specifically. I'd suggest that if you want to track this down, start a separate thread in the newly created Thetis > PureSignal Operations sub-forum. There are just too many things that can affect PS performance. It will be best to sort it out there and, if the culprit does turn out to be Thetis, then we can come back to Richie's main 2.6.9 (or maybe it'll be 2.7.0 by then) topic and post.

BTW, using PureSignal for AM is definitely worth it! It'll make you sound like a million bucks and reduce that nasty AM IMD that always seems to take up 20KHz of the band.

73,

Scott

Re: 2.6.9

Posted: Mon Nov 18, 2019 10:53 pm
by ramdor
UPDATE

Hi all,

Please try : link removed, see post #1

Use everything from the zip.

Source has gone off to Doug.

Richie.

(11/3/10) a8
-fix: rework of the seq log UI retrieval of data
-change: can select up to 144fps for those with systems and screens that are able to do that
-add: you can slide the edge of RX1 meter to resize it. The ratio between max and min is maintained when thetis window size is changed. Flicker will be resolved at some point

Re: 2.6.9

Posted: Mon Nov 18, 2019 11:30 pm
by wa1oxt
Another great job !

Tnx Richie

wa1oxt / / gary

Re: 2.6.9

Posted: Mon Nov 18, 2019 11:33 pm
by W3MMR
Now I just need to buy a 144hz monitor! Great job Richie!

@ 1920x1027
Capture.PNG
Capture.PNG (29.25 KiB) Viewed 15651 times


Perry

Re: 2.6.9

Posted: Mon Nov 18, 2019 11:47 pm
by ramdor
W3MMR wrote:Now I just need to buy a 144hz monitor! Great job Richie!

:lol:

Re: 2.6.9

Posted: Tue Nov 19, 2019 1:28 am
by Tony EI7BMB
Nice job on 144 Richie

Re: 2.6.9

Posted: Tue Nov 19, 2019 4:28 pm
by W1AEX
Thank you for the a8 patch Richie. My monitor maxes out at 60 fps but it's perfectly smooth and holds at the 60fps mark with RX1 + RX2 both running in Panafall mode. I never imagined that smooth operation with no stutters would be possible with that load! The re-sizable s-meter works fine and Thetis has been flawless with the a8 release during a couple hours of operation last night and during 3 hours of operation on 20 meter sideband this morning. It keeps getting better and better!

Thanks again,

Rob W1AEX

Re: 2.6.9

Posted: Tue Nov 19, 2019 5:20 pm
by ramdor
Thanks for feedback Rob. I only quickly put in the resize for rx1 because I knew it wasn't everyone's cup of tea to have it so wide.

Richie.

Re: 2.6.9

Posted: Tue Nov 19, 2019 5:58 pm
by w9mdb
Running a8 now but seems the last few versions the VAC1 Enable is not being remembered across restarts.
Where is the source code for this branch? I can do some debugging myself if need be.

Re: 2.6.9

Posted: Tue Nov 19, 2019 6:22 pm
by ramdor
w9mdb wrote:Running a8 now but seems the last few versions the VAC1 Enable is not being remembered across restarts.
Where is the source code for this branch? I can do some debugging myself if need be.


vac on/off states are stored in the txprofile. Consequently you will need to set vac on, then save the profile.

TX profile is now applied after the setup form state is restored to all the checkboxes/controls. Yes this is not ideal as there are technically two sets of data pertaining to one checkbox, the window state, and the tx profile state. State saving/loading could do with a re-work, but it is how it is atm.

From database.xml :

This block block below is the form control state (ie the checkbox), saved by iterating through all controls on the setup form and dumping them into the database.

Code: Select all

  <Options>
    <Key>chkAudioEnableVAC</Key>
    <Value>True</Value>
  </Options>


and then in each TX Profile.....

Code: Select all

<VAC1_On>false</VAC1_On>


So yes you can see it is not ideal, duplication left and right. It has been like this since powersdr as far as I can tell. As you can see, the checkbox chkAudioEnableVAC would get checked on when setup form state restore, and then checked off if the above tx profile was used.

Source available in due course I guess through Doug's personal github, until it moves over to TAPR. Good luck picking through it all, where initialisation flags are turned on/off so that changes get made/dont get made, etc etc etc.

Edit: and the reason why an iteration through all the controls is done is because there are over 1800 of them in the setup form, and hand crafting a save system for all that becomes a bit boring :D Reading them in, and then overwriting with selected profile seems the least mind numbing solution even if people forget to save the profile ;)

Edit2: you can also use 'auto save tx profile on thetis close' but that opens up the possibility of saving changes you dont want

Richie.

Re: 2.6.9

Posted: Wed Nov 20, 2019 2:14 am
by w-u-2-o
it is with great sadness I bring forward a bug report that crept into 2.6.9 a8.

Environment: 2.6.9 a8. 8000DLE running pre4. Win10Pro.

Problem: CTUN state is not saved when program is quit and restarted.

Steps to reproduce:

Set mode to LSB or USB, Enable CTUN. Tune passband off center in spectral display. Quit and restart Thetis. CTUN is off and passband is centered in spectral display.

73,

Scott

Re: 2.6.9

Posted: Wed Nov 20, 2019 2:22 am
by ramdor
will look at it in a mo Scott. Odd though, yes reproducible here.

Richie.

Re: 2.6.9

Posted: Wed Nov 20, 2019 3:16 am
by ramdor
UPDATE

To fix the issue Scott reported.

Please try : link removed, see post #1

Note: use all files from zip.

Edit: this issue was uncovered due to changes in the way the main form is initially positioned/presented in a8, but has always been there, lurking.

Edit2: this fix is not in the source now available over on Doug's personal github space. It will be included in next source release in the future.

Richie.

(11/3/10) a8_1
-fix: btnBandHF/VHF/GEN functions were setting CTUN button when they shouldn’t have been. Not only that they were basing that on the incorrect boolean. Code removed.

Re: 2.6.9

Posted: Wed Nov 20, 2019 2:50 pm
by w-u-2-o
a8_1 working great, Richie. Thanks for the super-fast bug-fix! :)

73!

Scott

Re: 2.6.9

Posted: Thu Nov 21, 2019 11:19 am
by W3MMR
w-u-2-o wrote:a8_1 working great, Richie. Thanks for the super-fast bug-fix! :)

73!

Scott

Hey Scott, i figured out the Pure Signal issue. I reinstalled the 1.8 firmware and all is A-Ok at this point. Pure Signal held in for an hours long QSO yesterday with no real issues. Every once in a while it would drop out but then it would come right back. After talking to some others running these later Alpha versions, they say thats normal.

And Richie, A8_1 is working great!

Perry

Re: 2.6.9

Posted: Thu Nov 21, 2019 1:36 pm
by Tony EI7BMB
I maxed out the main display FPS to 140 and get the odd red flash , the green indicator shows 124 /125 , everything working great but I'm going to use this as a lame excuse to upgrade my GTX950 video card :D

Re: 2.6.9

Posted: Thu Nov 21, 2019 2:03 pm
by ramdor
Anything above refresh of screen is somewhat pointless as what you see on screen is essentially locked to vsync being that it is being rendered into something on the desktop.

Also keeping it lower rather than turning it up to 11 reduces the chance of the rest of Thetis from missing a beat.

All good fun though in the name of 'science' :)