apache-labs.com

Sure, post an update while I'm away on business, why don't you!

Very nicely done Richie! It's almost a hypnotic experience watching 5 or more peak detectors at once. The weaker stations will fade and the lowest peak will instantly jump to the station with the next strongest signal. Very effective implementation. I found setting more than 5 peak detectors was way too much excitement for me! You are absolutely right that this will make measuring transmitted IMD a breeze.

Thanks for the this fun release! 73, Rob W1AEX

ramdor wrote:Hi folks,
....
However, just a small teaser...
...

Very nice!!! Would be nice to get an alarm if a signal reaches a selectable level within a selectable bandwidth. The alarm should be a sound, jingle or beep from a custom made wave file

Thanks Richie , running at 144 now with acc timing enabled with no red indicators . CPU up by a few per cent but nothing to worry about.

You can get your "alarm" right now, as long as the "selectable bandwidth" is 20KHz or less. Just turn on the squelch and set the level as desired. When a signal breaks squelch you will hear it. I use this all the time to monitor a frequency or two in the afternoon when I am working.

w-u-2-o wrote: I use this all the time to monitor a frequency or two in the afternoon when I am working.

That's funny Scott, I do the same thing in the eve waiting for the group to show up on 1855 KHz. Sometimes it's almost like FM on 160M.

that's how I describe it, too! Attenuator and AGC level set appropriately, NR2 on, 4KHz bandwidth, squelch set properly, and...BAM...near broadcast FM quality over HF when you've got 3 or 4 S-units of SNR.

These radios really are great even if they take a little more skull sweat than the average radio appliance.

73!

Scott

Tony EI7BMB wrote:Thanks Richie , running at 144 now with acc timing enabled with no red indicators . CPU up by a few per cent but nothing to worry about.

What cpu do you use? Could you please tell me the difference per logical processor? I have a twelve core processor and one core goes up from 20% to 90% Of course combining all of them is a few percent in total but very close to saturate one core.

Regards

sv1jso wrote:but very close to saturate one core.

It will, which is why I dont really recommend using it. It should jump around perhaps from core to core as it is not fixed to a specific one, at least that is what I see on this 3800X processor. It is purely experimental and will probably be removed at some point.

Unfortunately using Thread.Sleep for a short time delay, say 16.odd ms for 60fps, whilst removes the load on the cpu does not guarantee that the thread will resume when we want. This is one of the main reasons for the jitters that can be seen sometimes in the waterfall. Games and what not don't really care and just run a tight loop to give timed physics and fps stability.

Also, in cases such as 60fps, the delay required is 16.66666r ms, less the time it took to draw the frame, and Thread.Sleep takes an integer. The next version (b3) will see this fraction accumulated and applied, but still does not fix the fundamental problem that Thread.Sleep is not a guaranteed accurate time delay.

I will keep researching it, but at the mo there doesn't really seem to be an ideal solution. It is only noticeable in the waterfall due to the nature in which it fix rate 'falls'. It is quite impossible to see anything odd happening to the panadapter for example.

Edit: b3 will also see the thread.sleep timed, and any late return from it will be removed from the next frame delay. Blimey... should have thought of this a while ago .. sheesh, lol.

Richie.

ramdor wrote:UPDATE

HI all,

Small update with the peak indicators, and a couple of other bits and bobs. Most of it is only implemented in DirectX at the moment. The peak detect is a simple algorithm, and does not use crossing point/smoothing/slope.

Quite handy for this sort of thing, with PS on :

psON.png

Also, an experimental change to frame rate calculations that can be enabled with 'accurate frame timings' option. However, It sits in a very tight loop until the expected delay has passed and consequently is quite cpu hungry. It is perhaps only worth using if you have a work horse of a PC and want to test if it improves the micro-stutter in the waterfall.

Link : https://www.dropbox.com/s/psqkr8p22m8g1 ... 2.zip?dl=0

Cheers, Richie.

(11/19/19) b2
-add(*): peak markers on the panadapter. Configure through Setup>Display>General. Note: this is a very simple algorithm and does not consider slope
-add(*): experimental accurate frame timing. This will probably have a substantial impact on cpu usage
-change(*): if a timed waterfall update is ‘late’ the next update will be made ‘early’
-change: now using hiperf timer in RunDisplay
(*) only implemented in DirectX atm

Very cool Richie!!

Great Job Richie .
Now for uV readout !

Tnx

gary / / wa1oxt

Using a Ryzen 1700 processor. Sorry I don't have the other information you asked about.

sv1jso wrote:

Tony EI7BMB wrote:Thanks Richie , running at 144 now with acc timing enabled with no red indicators . CPU up by a few per cent but nothing to worry about.

What cpu do you use? Could you please tell me the difference per logical processor? I have a twelve core processor and one core goes up from 20% to 90% Of course combining all of them is a few percent in total but very close to saturate one core.

Regards

Thanks Richie,

Great update!
I am only using a little I5 Intel Nuc dedicated to my ANAN 7000 MK I but its very amazing to see the calculation speed even when 10 measurements are open. It gets a bit nervous on the screen then (especially if averaging is low)

I just wanted to share my first experience after a 2 hour SSB test on 80m this morning - Thetis worked super stabil, not one problem regarding Pure Signal and/or Seq. Errors...…

Thanks and 73,

Andy

ramdor wrote:

sv1jso wrote:but very close to saturate one core.

It will, which is why I dont really recommend using it. It should jump around perhaps from core to core as it is not fixed to a specific one, at least that is what I see on this 3800X processor. It is purely experimental and will probably be removed at some point.

Unfortunately using Thread.Sleep for a short time delay, say 16.odd ms for 60fps, whilst removes the load on the cpu does not guarantee that the thread will resume when we want. This is one of the main reasons for the jitters that can be seen sometimes in the waterfall. Games and what not don't really care and just run a tight loop to give timed physics and fps stability.

Also, in cases such as 60fps, the delay required is 16.66666r ms, less the time it took to draw the frame, and Thread.Sleep takes an integer. The next version (b3) will see this fraction accumulated and applied, but still does not fix the fundamental problem that Thread.Sleep is not a guaranteed accurate time delay.

I will keep researching it, but at the mo there doesn't really seem to be an ideal solution. It is only noticeable in the waterfall due to the nature in which it fix rate 'falls'. It is quite impossible to see anything odd happening to the panadapter for example.

Edit: b3 will also see the thread.sleep timed, and any late return from it will be removed from the next frame delay. Blimey... should have thought of this a while ago .. sheesh, lol.

Richie.

Ive noticed the same thing, but even with "AFT" unchecked, one core is still taking the brunt of the load, 65%, while all the other 7 are running anywhere from 10-20%. Can this be adjusted in the bios? Running an i7 4770.

Awesome job however on B2. Working great here. The "Blobs" are fun to play with. And thank you for putting an "enable/disable" check box. I wish there was a way to turn off the CPU percentage and the sequence error icon as well. Not that I get a lot of errors, maybe 1 in an entire evening of operating, but its just something I seem to be stressing myself out about for no reason. Same with the CPU percentage. If I see it go up 5%, i start trying to figure out why. Meanwhile, i'm only running at 12% total lol. You're doing an awesome job Richie. Thanks for the hard work.



Perry

Hi Perry

I had a quick look at the cpu% code and it is currently reporting total cpu% usage of the entire system, which is why we tend to use resmon or something similar to find out how much Thetis is actually taking. I have quickly changed it to reflect the process %cpu that is being used by Thetis. This calculation is an averaged value based on 'cpu count' in the system and will more accurately display how much % thetis is actually using. It now shows the same value as ResMon +- 1%

I have been thinking about the warning symbol for seq errors and may change it to only show it if there are -ve numbers. Positives alone are somewhat meaningless, only telling us that there has been some missing packets. I don't think we have yet seen any major repeatable widespread occurrences of -ve value patterns that could be fixed by an implementation of the pack reordering system.

I wouldn't worry about where the threads are running, we don't set any affinity, so they get scheduled onto which ever processor (core/virtualcpu) the OS thinks is best. As a result you may see a peak on one compared to others(s).

73 Richie.

ramdor wrote:Hi Perry

I had a quick look at the cpu% code and it is currently reporting total cpu% usage of the entire system, which is why we tend to use resmon or something similar to find out how much Thetis is actually taking. I have quickly changed it to reflect the process %cpu that is being used by Thetis. This calculation is an averaged value based on 'cpu count' in the system and will more accurately display how much % thetis is actually using. It now shows the same value as ResMon +- 1%

I have been thinking about the warning symbol for seq errors and may change it to only show it if there are -ve numbers. Positives alone are somewhat meaningless, only telling us that there has been some missing packets. I don't think we have yet seen any major repeatable widespread occurrences of -ve value patterns that could be fixed by an implementation of the pack reordering system.

I wouldn't worry about where the threads are running, we don't set any affinity, so they get scheduled onto which ever processor (core/virtualcpu) the OS thinks is best. As a result you may see a peak on one compared to others(s).

73 Richie.

Thanks RIchie, the CPU loading typically is a function of the OS as you pointed out.

quote="ramdor"]Hi Perry

I had a quick look at the cpu% code and it is currently reporting total cpu% usage of the entire system, which is why we tend to use resmon or something similar to find out how much Thetis is actually taking. I have quickly changed it to reflect the process %cpu that is being used by Thetis. This calculation is an averaged value based on 'cpu count' in the system and will more accurately display how much % thetis is actually using. It now shows the same value as ResMon +- 1%

I have been thinking about the warning symbol for seq errors and may change it to only show it if there are -ve numbers. Positives alone are somewhat meaningless, only telling us that there has been some missing packets. I don't think we have yet seen any major repeatable widespread occurrences of -ve value patterns that could be fixed by an implementation of the pack reordering system.

I wouldn't worry about where the threads are running, we don't set any affinity, so they get scheduled onto which ever processor (core/virtualcpu) the OS thinks is best. As a result you may see a peak on one compared to others(s).

73 Richie.[/quote]

Richie...
For what it's worth, I have found that seeing total CPU usage alerts me to a possible problem, as when CPU usage increases from the normal 5-8% to 15-20% is when I will see "seq" errors. If CPU usage is at the high end, pages of "seq" will load and a lock-up is probable. Otherwise a few lines of "seq" errors are insignificant. I have identified Carbonite as one of the worst offenders of CPU usage causing me potential lock-ups. The up side of all this is FW 2 pre4 and Thetis 2.6.9 a8 have been running reliably for several weeks now, with no explanation after having continuous problems previously. It's been great enjoying your evolution of Thetis.
Kudos for all your efforts.
Carl
NX5T

Hi Carl,

I have added option to pick which one you want. Right click the cpu in status bar.

Also working on this at the mo, for some extra eye candy. Probably all released later tonight or tomorrow.



73 Richie.

ramdor wrote:Hi Carl,

I have added option to pick which one you want. Right click the cpu in status bar.

Also working on this at the mo, for some extra eye candy. Probably all released later tonight or tomorrow.



73 Richie.

Thanks Richie!
great stuff ur doing!
Cheers

Your the BEST Richie !!!
I was afraid to ask for to much. Although I figured you would catch on .

tnx

gary / / Wa1oxtRadio

Hi all and Ritchie,

at first 2.6.9 b2 is running perfekt here, many thanks !!

Right Click on CPU in Status-Bar ?? No funktion here??

Sometimes I have a problem to save my Setting in the Database, here it is on DSP - Options, the Filter-Type. Sometimes it is saved, some
time not?? (B1 Version).

73 Erwin

w-u-2-o wrote:You can get your "alarm" right now, as long as the "selectable bandwidth" is 20KHz or less. Just turn on the squelch and set the level as desired. When a signal breaks squelch you will hear it. I use this all the time to monitor a frequency or two in the afternoon when I am working.

That does not give an alarm on a frequency not within the 20kHz! I want this for Sporadic-E and meteorscatter observations.

DO2ZA Erwin wrote:Hi all and Ritchie,

at first 2.6.9 b2 is running perfekt here, many thanks !!

Right Click on CPU in Status-Bar ?? No funktion here??

Sometimes I have a problem to save my Setting in the Database, here it is on DSP - Options, the Filter-Type. Sometimes it is saved, some
time not?? (B1 Version).

73 Erwin

cpu status bar change as not yet been released. Saving issues, probably related to TX profile, save the TX profile and should be ok?

Richie.

UPDATE

Hi all,

Please try : link removed, see post #1

This version includes a work in progress historic meter data. It can be enabled in setup>display>general in the multimeter group. Colour/alpha is set in appearance> meter. Readings are currently spread over 2 seconds.

Also, the regular frame timing code has had a work over, and you will notice that fps displayed in top left of spectrum will be closer to main display fps configured via setup>display>general.

Use the drop down/click the cpu% in the status bar to select displayed value.

NOTE: seqlog error warning on status bar will only show for -ve numbers. There is now an option to control this in the seqlog window.

73 for now, Richie.

(11/19/19) b3
-change: frame timing improvements, and removed waterfall late changes made in b2
-add: cpu% can now show thetis/system cpu%, use the drop down on the cpu% in status bar
-add: option in SeqLog to only show status bar warning when -ve seq errors detected (default)
-add: rx signal history over two seconds shown on smeter. Options to turn on/off and set colour/alpha

very cool eye candy , thanks

WOW , so much fun Richie.
Look's great.
What's next ......a 3D Waterfall ?

tnx.

wa1oxt / /garyradio

wa1oxt wrote:WOW , so much fun Richie.
Look's great.
What's next ......a 3D Waterfall ?

tnx.

wa1oxt / /garyradio

Hah yes it has been running around my head for a while. A 3d landscape that moves into the background over time, a new 'row'/'land' added at the front. On my personal list, perhaps never to be crossed off, but it is on there

Richie.

Richie...

Just loaded and running b3. All new options verified operational, and very cool!
Great stuff!

Thx!
Carl
NX5T

Ah.......... Richie , all the Anan motherboards are crying for a combo VNA and SA software. Matter of fact , the SA part is almost done.
Maybe Apache Labs should think abt that T&M market.

Great Job Richie.

wa1oxt / / garyradio

ramdor wrote:

wa1oxt wrote:What's next ......a 3D Waterfall ?
wa1oxt / /garyradio

Hah yes it has been running around my head for a while. A 3d landscape that moves into the background over time, a new 'row'/'land' added at the front. On my personal list, perhaps never to be crossed off, but it is on there

Years ago Cathy - I don't know her name or email - did that with SDRMaxx. I don't know if the sources are available. I still have the executables for use with my QS1R.