2.6.9

w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: 2.6.9

Postby w9mdb » Sun Nov 17, 2019 4:55 am

I ran my test over about 24 hours
Mike W9MDB
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: 2.6.9

Postby ramdor » Sun Nov 17, 2019 5:00 am

nothing found, all flatlined in vs2019. What was the cat connection doing/connected to?

Richie.
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: 2.6.9

Postby ramdor » Sun Nov 17, 2019 5:09 am

w9mdb wrote:I ran my test over about 24 hours


Well you are seeing roughly 1MB per 10mins, if you take your 620 to 765 reading over 2hrs, which just isn't occurring in the memory profiler here. Perhaps the cat connection is doing something all the while? what was on the end of it during that 24hrs?

Richie.
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: 2.6.9

Postby w9mdb » Sun Nov 17, 2019 5:13 am

Not sure what you mean by "on the end of it"
Seems like DirectX is not having the memory problem at least in the short while I've been running it.

One other thing you might consider...don't do any drawing when the window is minimized. I do that in some of my software to prevent CPU usage.
Mike W9MDB
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: 2.6.9

Postby ramdor » Sun Nov 17, 2019 5:17 am

w9mdb wrote:Not sure what you mean by "on the end of it"
Seems like DirectX is not having the memory problem at least in the short while I've been running it.

One other thing you might consider...don't do any drawing when the window is minimized. I do that in some of my software to prevent CPU usage.


was there anything on the end of the cat connection? like ham radio deluxe, wsjt-x etc etc etc.

I have considered it.

Richie.
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: 2.6.9

Postby w9mdb » Sun Nov 17, 2019 5:26 am

FLRig. WSJT-X connected to FLRig, Log4OM connected to FLRig, and my own tuner utility connected to FLRig too.

My utility does power changes, VFO, PTT, frequency, and mode via FLRig.
Mike W9MDB
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: 2.6.9

Postby ramdor » Sun Nov 17, 2019 5:38 am

I guess wait to see if any other reports of it happening.

Richie.
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: 2.6.9

Postby ramdor » Sun Nov 17, 2019 6:57 am

UPDATE

Hi all,

Quick update with a couple of changes and some hidden ;) fixes : link removed, see post #1

73 Richie.

(11/3/10) a7
-change: caching of directX string sizes
-change: DXGI_ERROR_WAS_STILL_DRAWING exception ignored in swapchain present
Last edited by ramdor on Mon Nov 18, 2019 10:54 pm, edited 1 time in total.
User avatar
ULTIMAX
Posts: 50
Joined: Tue Apr 11, 2017 11:16 am

Re: 2.6.9

Postby ULTIMAX » Sun Nov 17, 2019 1:38 pm

Hello Richie,

First able awesome job on the updates and so far all working really good here, just one question which I know it sounds odd
I notice in many of the updates thru out the latest versions & updates that I have to go to calibration to re-adjust because the previous calibration gets shift by few clicks and I have re calibrated and do a bit of factor correction to get it back into point, using 15.000 MHz.

Just wondering if that's normal , I don't mind checking /re-calibrating just wanted to ask you

Also maybe down the line besides other things it would be nice to get better meter maybe like the one on FiFi SDR or something like what KE9NS did , also how about the option of been able to relocated the control boxes on either collapse /expand mode to the users choice, like the Flex Smart SDR do.

I don't want to be a pusher or picky just suggestions mate.

Thank you again for all your efforts.

Al
73
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: 2.6.9

Postby w-u-2-o » Sun Nov 17, 2019 3:05 pm

Running a7. All smooth and serene here. :) Not noticing any memory leaks. Not noticing any reduction in CPU since initially going to DirectX. Seems like additional drawing efficiency may have reached the diminishing returns category unless there is some plan to go thin client/server. On the other hand, anything that reduces the chances of a blocked DSP thread is definitely worth it!

Question: is it time to remove GDI yet? It might be worth putting out an a8 pinned to DirectX just to force the issue and see who screams. If it's nobody then there's the answer. That said, if someone tries to run this on 10+ year old hardware is it worth supporting PC technology that is over a decade old?

Question: is it worth applying DirectX to the drawing of the main console? Resizing the console is certainly clunky, but not a show stopper by any means.

Not sure what's on your hit list, Richie, but here are some ideas:

- VAC1 option to carry the Master AF feed
- Improved resampler stability with small VAC buffer sizes (this may require some help from Bryan)
- Ability to spot from more than one spotting source simultaneously
- Improved RF Drive control linearity (this may require some help from Warren)

Cheers & 73!

Scott
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: 2.6.9

Postby w-u-2-o » Sun Nov 17, 2019 3:12 pm

One other idea (or perhaps it's a misconception of mine?): would it be possible to have the database fully updated when the power (really run/stop--can we rename that?) button is in the off (stop) state? Right now I often find myself quitting and restarting the app after making a number of band stack related changes so that a crash won't accidentally lose them.

Thanks!

Scott
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: 2.6.9

Postby w9mdb » Sun Nov 17, 2019 3:33 pm

a7 doesn't work here very well on my ANAN 100

Hitting Tune gives a multi tone output as seen in the attachment that my MFJ-998 can't tune with (even though showing 20W it says "increase power" because the power is so spread out.
And doing a Tune from WSJT-X behaves much the same way.

Mike
Attachments
Capture.PNG
Capture.PNG (94.39 KiB) Viewed 16778 times
Mike W9MDB
User avatar
w2ner
Posts: 98
Joined: Wed Sep 11, 2019 12:14 am
Location: SC

Re: 2.6.9

Postby w2ner » Sun Nov 17, 2019 4:12 pm

Working very well here on my 8000DLE, tune is clean single tone.
User avatar
NJ2US
Posts: 106
Joined: Sun Oct 29, 2017 9:39 pm

Re: 2.6.9

Postby NJ2US » Sun Nov 17, 2019 4:25 pm

A few of us have observed a "TUNE" anomaly where a wide band spurious signal is transmitted in TUNE WHILE PURE SIGNAL IS ENGAGED. This behavior has been observed in "Pre Richie" versions of Thetis. It is infrequent, but when it happens....it is a wide nasty signal. Disabling PURE SIGNAL before a TUNE sequence fixes it every time.

Jeff
NJ2US
Last edited by NJ2US on Sun Nov 17, 2019 6:27 pm, edited 1 time in total.
W4WMT
Posts: 325
Joined: Sun Apr 09, 2017 10:12 pm

Re: 2.6.9

Postby W4WMT » Sun Nov 17, 2019 4:37 pm

w9mdb wrote:a7 doesn't work here very well on my ANAN 100

Hitting Tune gives a multi tone output as seen in the attachment that my MFJ-998 can't tune with (even though showing 20W it says "increase power" because the power is so spread out.
And doing a Tune from WSJT-X behaves much the same way.

Mike


Some operators report seeing that same multi-tone effect happening if they don't turn off "PS-A" before hitting "TUN".

73
User avatar
W1AEX
Posts: 425
Joined: Sun Apr 09, 2017 6:17 pm
Location: Connecticut, USA
Contact:

Re: 2.6.9

Postby W1AEX » Sun Nov 17, 2019 5:09 pm

Thetis has been running smoothly with the 269_a6_2 patch since it came out and all morning today with the 269_a7 patch. I have seen the "Tune" anomaly now and then as described in the thread but it corrects for me as Jeff indicated by toggling PS off.

Thanks Richie!

Rob W1AEX

(ANAN-200D with F/W v1.8 and Thetis 2.6.9_a7 - direct connection to second computer NIC on separate static assigned subnet)
"One thing I am certain of is that there is too much certainty in the world."
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: 2.6.9

Postby w-u-2-o » Sun Nov 17, 2019 5:16 pm

You can create a similar tune anomaly by setting your RF drive level very low, at least you can on the 8000. It has been this way for me well prior to Richie's contributions.
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: 2.6.9

Postby ramdor » Sun Nov 17, 2019 8:58 pm

One reply for all :)

--AL--
Al, not sure about that re-calibration thing. I dont see anything like that and I am always changing / running new builds. However I use one of these http://www.qrp-labs.com/progrock.html with 30m lpf (not really needed) and their gps kit. Works quite well. Yes, we need a better meter, agreed. Also, yes when the drag/dock stuff/layout stuff is in, everything will be configurable, but I am not planning on starting that work until most/all issues ironed out and a line can be drawn under current implementation.

--SCOTT--
Scott, there is a very small memory leak in there somewhere (thanks w9mdb) and will be running an extended memory profile run when I go to bed later. I have a feeling that it might be something on the vac side of things, not 100% sure yet though. I have been running for about 6hrs with display only, no cat, no vac/audio, and things haven't moved. If confirmed perhaps Bryan can look at it as I am not up to speed on any of that side of things really. Previous test with Vac1 enabled Thetis has chewed through an additional 50MB in roughly 8hrs.

You will only notice improvements if you have been pushing DirectX version to the limit, say at 60fps with two rx's in panafall on a 4k display that set to 100% scaling. I have to work hard to get the warning to show now even on this 'old tech'.

GDI removal. I think you are right for the main display area. It would potentially still be used for the meter etc not sure yet. We are limited to 8 render target surfaces with the level of directX that we use at the moment, so care will need to made when moving over to the DDC per window approach.

Resizing console, 50% of that clunky feeling is from moving all the controls around in code, the other 50% updating hpsdr spectrum pixel width every resize event, which in turn calls SetAnalyzer. It would probably be best to shut down draw of the spectrum, move controls etc and then when the resize has finished update hpsdr spectrum pixel width and resume rendering.

Spot system needs a rewrite, simple as that really. Some of the code can be used and tidied up, but it really does need sorting out. As a stop gap, to achieve multi nodes I just run ve7cc client and get Thetis to connect to that. Then add more spotting source nodes into ve7cc.

Setup->Apply will do an update to the DB and should save all bandstack, not 100% on this, would have to test. If you click ok, the update to the DB will only happen when you shutdown as part of DB.Exit() . Odd I know, but that is how it is implemented atm.

--w9mdb--
Cheers for identifying the leak after 24hr usage. I am still investigating this.

and lastly...

The tune issue........ *searches for shrug emoji*

Richie.
User avatar
ULTIMAX
Posts: 50
Joined: Tue Apr 11, 2017 11:16 am

Re: 2.6.9

Postby ULTIMAX » Sun Nov 17, 2019 9:22 pm

ramdor wrote:One reply for all :)

--AL--
Al, not sure about that re-calibration thing. I dont see anything like that and I am always changing / running new builds. However I use one of these http://www.qrp-labs.com/progrock.html with 30m lpf (not really needed) and their gps kit. Works quite well. Yes, we need a better meter, agreed. Also, yes when the drag/dock stuff/layout stuff is in, everything will be configurable, but I am not planning on starting that work until most/all issues ironed out and a line can be drawn under current implementation.



Richie.




Thank you Richie, no worries

Al
73's
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: 2.6.9

Postby w-u-2-o » Sun Nov 17, 2019 9:28 pm

Thanks for the comprehensive reply, Richie. I am limited to 30Hz monitors here, but there are three of them and they are all 4K. I can span the window across all three of them no problem. But I can't push the limit in terms of refresh rate, which I think you've limited to 100Hz at the moment?

I will look more carefully for the memory leak here. I routinely run VAC and CAT.

Interesting idea about stopping the render for resize. I wonder what that will "feel" like from a user experience perspective.

Excellent work around for the cluster business. I'm not a big DX'er but when I wanted more than one simultaneous spotting source I would run to DXLabs SpotCollector, but that didn't buy me anything on the panadapter, although it does integrate nicely with DXKeeper.

73!

Scott
W4WMT
Posts: 325
Joined: Sun Apr 09, 2017 10:12 pm

Re: 2.6.9

Postby W4WMT » Sun Nov 17, 2019 10:18 pm

ramdor wrote:One reply for all :)

--SCOTT--
Scott, there is a very small memory leak in there somewhere (thanks w9mdb) and will be running an extended memory profile run when I go to bed later. I have a feeling that it might be something on the vac side of things, not 100% sure yet though. I have been running for about 6hrs with display only, no cat, no vac/audio, and things haven't moved. If confirmed perhaps Bryan can look at it as I am not up to speed on any of that side of things really. Previous test with Vac1 enabled Thetis has chewed through an additional 50MB in roughly 8hrs.

--w9mdb--
Cheers for identifying the leak after 24hr usage. I am still investigating this.

Richie.


What tool do you guys use to keep tabs on memory usage, leaks, etc. ?

73
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: 2.6.9

Postby ramdor » Sun Nov 17, 2019 10:52 pm

Bryan W4WMT wrote:What tool do you guys use to keep tabs on memory usage, leaks, etc. ?
73


I am using the vs2019 profiller, and process explorer from sysinternals/microsoft. Vac looks fine, but there is something odd going on every so often (rarely). There is a brief pause in the front end and memory spikes (unresponsive UI).

Richie.
W4WMT
Posts: 325
Joined: Sun Apr 09, 2017 10:12 pm

Re: 2.6.9

Postby W4WMT » Sun Nov 17, 2019 11:19 pm

ramdor wrote:
I am using the vs2019 profiller, and process explorer from sysinternals/microsoft. Vac looks fine, but there is something odd going on every so often (rarely). There is a brief pause in the front end and memory spikes (unresponsive UI).

Richie.


Okay, I use Process Explorer also. I wonder if what you are seeing (brief pause) is also causing the ChannelMaster and PortAudio threads to be blocked briefly during Rx/Tx transitions? It happens very fast and can only be noticed when the VAC1 ringbuffer is set very, very small. Also, it only happens in DirectX mode not GDI+. But here's the crazy part, if I turn the "Panadapter" off, it happens even in GDI mode. Counterintuitive at best :?:

73
User avatar
NJ2US
Posts: 106
Joined: Sun Oct 29, 2017 9:39 pm

Re: 2.6.9

Postby NJ2US » Sun Nov 17, 2019 11:35 pm

ramdor wrote:One reply for all :)

The tune issue........ *searches for shrug emoji*

Richie.


Is it possible for a simple stupid tune fix that calls a procedure that "turns off" Pure signal when TUNE is applied, then "turns" Pure Signal back on when Tune is ended? Just thinking out loud. I know it ignores whatever the deeper issue is, but it solves the problem. And actually, since Thetis, I don't think I've seen this problem is some time.

Jeff
NJ2US
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: 2.6.9

Postby ramdor » Mon Nov 18, 2019 12:58 am

Bryan W4WMT wrote:Okay, I use Process Explorer also. I wonder if what you are seeing (brief pause) is also causing the ChannelMaster and PortAudio threads to be blocked briefly during Rx/Tx transitions? It happens very fast and can only be noticed when the VAC1 ringbuffer is set very, very small. Also, it only happens in DirectX mode not GDI+. But here's the crazy part, if I turn the "Panadapter" off, it happens even in GDI mode. Counterintuitive at best :?:
73


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.
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: 2.6.9

Postby w-u-2-o » Mon Nov 18, 2019 1:36 am

I tried looking for memory leaks using ResMon. Let it all ride for 4 hours and results were inconclusive (numbers were too close to tell, some larger, some smaller). Perhaps it's not a good tool for the job.
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: 2.6.9

Postby ramdor » Mon Nov 18, 2019 1:50 am

as far as I can tell, there is nothing significant Scott, other than the issue I found with pulling the seqlog over to the 'ui'. I have just reworked a bunch of it and it is behaving as expected now. VS2019 memory profiler reports zero issues.

Richie.
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: 2.6.9

Postby w-u-2-o » Mon Nov 18, 2019 2:22 am

Smooth sailing here from a stability standpoint :)
VE7DQC
Posts: 1
Joined: Thu Apr 20, 2017 3:59 pm

Re: 2.6.9

Postby VE7DQC » Mon Nov 18, 2019 6:18 am

Hi Scott
Regarding 2.6.9, I am wondering if an issue with channel scanning has been reported. I was using 2.6.8 and went back due to form errors and running our of memory while scanning a channel bank. I scan using ALE software.

This worked in previous issues (prior to 2.6.7) and the last time I saw it was with powersdr.

I use a 200D and the software I use to control the radio hasn't changed. the changes that have occurred are MS updates (win10 x64) and .net to accommodate 2.6.8

I can't say 100% its Thetis but highly suspect it.

Any comments are welcome.

Mike VE7DQC
User avatar
W3MMR
Posts: 145
Joined: Thu Jul 11, 2019 8:36 am
Location: Springfield, PA
Contact:

Re: 2.6.9

Postby W3MMR » Mon Nov 18, 2019 11:53 am

A7 working good here so far. I am having a PS issue though. On SSB, which is the only place I use PS-A, it will be working fine, then all of the sudden, it stops working. PureSignal 2 message is still there, but both the correcting and feedback lights dissapear. Its not a matter of the mic level not being high enough or anything like that where the correcting light would be lit, but no feedback being received. Both go away like they do when youre not transmitting. Then all the of the sudden it'll start working the next time I key up and I dont cycle it on or off. Sometimes when I cycle it on or off, it will fix it, sometimes it wont.

Not a big deal because I dont use sideband that much, but It is weird...

**EDIT**- I am using a 200D, firmware 1.8, Thetis 2.6.9 A7

Perry

Return to “Thetis”