vac buffer overruns during 2 tone testing

VK4BXI
Posts: 62
Joined: Sat May 22, 2021 11:44 am
Location: Brisbane

vac buffer overruns during 2 tone testing

Postby VK4BXI » Thu Jul 14, 2022 7:31 am

Hi to the group,
I've been working at slowly reducing the latency of my audio in Tx and have reached a point where I have what it consider to be a "nice" stable setup. So I only get buffer over/under flows every few 10's of minutes (and usually when the system over all cpu useage "blips" ) and the graph traces monitoring the VAC percentage settle within two or three overshoots. However when I do a 2 tone test (or change the output drive when I'm tuning up) I get the 2 tone display "flair out" every couple of seconds to a much wider display. When this happens I get the vac buffer counting up overflows/underflows (40 or 50 counts at a time). I think that this is the Pure Signal software doing something. I can get much the same thing if I am using the tune function and make a less than slow change in drive level. I don't seem to be able to find a setting of VAC buffers etc. that will stop this happening.

Is there a setting or something that I am missing ? or is this an artifact of thetis?

My setup is a 7000 mkII black edition running the latest versions of Thetis and P2. The computer is an I 5 running windows 10 professional with the throtling fix installed. It has 4 cores running and runs at an overall cpu % of ~21 ~23. I use voice meter banana and everything is at 48 kHz. My microphone a logitech usb headset. Buffer size is 64, buffer latency is 10 millisec and the driver is ASIO.

The rig drives an Ldmos 1kW amp at the 400 watt level. I have pure signal always always on with ~ 50db neg of feedback from the amp giving a stable S-ATT attenuator of setting of ~20db. When the 2 tone display is stable I'm getting ~ 50db before the first distortion products. I generally get good or very good audio reports from contacts. So I don't think there is much wrong with "normal operation"

The drive levels are scaled in the PA settings for 400 watts at 100% and "linearised" at each 10% setting giving good linearaty on the PS gain display.
In pure signal map is set but not pin or stbl. Auto attenuate is set as is quick attenuate responce but not relax tolerance.

So the setup works, am I worrying over nothing, or is there still something to fix.

Regards Bob
dynamicfusion
Posts: 57
Joined: Thu Oct 14, 2021 5:23 am

Re: vac buffer overruns during 2 tone testing

Postby dynamicfusion » Thu Jul 14, 2022 8:24 pm

Do you suppose it's RF related coupling to some of your equipment? Running ferrite on all cables?
VK4BXI
Posts: 62
Joined: Sat May 22, 2021 11:44 am
Location: Brisbane

Re: vac buffer overruns during 2 tone testing

Postby VK4BXI » Thu Jul 14, 2022 10:05 pm

Don't think so I'm a bit OCD about that. I've installed ferrite cores on all of the supply leads (AC and DC). A ferrite core between the rig output and the amp. And a ferrite core on the amp to rig feedback line. Nothing on the amp output...........but thats a 1 metre double screened coax into the dummy load. (all of my coax's are double screened). Everything seems very stable in normal operation.
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: vac buffer overruns during 2 tone testing

Postby w9mdb » Thu Jul 14, 2022 10:17 pm

Another thing you can do is twist your power cables.
I've started doing that on all mine. A built-in choke.
Mike W9MDB
VK4BXI
Posts: 62
Joined: Sat May 22, 2021 11:44 am
Location: Brisbane

Re: vac buffer overruns during 2 tone testing

Postby VK4BXI » Tue Oct 11, 2022 5:14 am

A few months onward and I've gotten around to a few more tests.
Using a separate receiver I've been doing some more low power tests (10 ~ 15 Watts) in transmit variously in voice, tune, and 2tone. The hiccup has been happening in all of them with PS on or off not making much difference. From the receiver side its clear that the Hiccup on the transmit waveform is actually a break in the transmission. The Hiccup is typically occurring every few seconds (up to 10 seconds apart). When it happens the VAC monitor both in and out kick up by 50 to 100 counts. Not sure if that's the cause or effect.
HOWEVER
I use a WiFi connection to the internet with the Ethernet connection used for the rig. In desperation (because no combination of buffer size or buffer latency would fix it) I tried disabling the WiFi connection to the internet...............that worked !. To check the fix I turned it back on .......still fixed !. I'm not sure what happened but I beginning to suspect that I have something like ET phone home running in my machine (probably a windows program) but cycling the WiFi stopped it. At least for now ...probably until I turn the machine off.
Suggestions please of the probable culprits ?
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: vac buffer overruns during 2 tone testing

Postby w-u-2-o » Tue Oct 11, 2022 12:00 pm

VK4BXI wrote:I use a WiFi connection to the internet with the Ethernet connection used for the rig. In desperation (because no combination of buffer size or buffer latency would fix it) I tried disabling the WiFi connection to the internet...............that worked !. To check the fix I turned it back on .......still fixed !. I'm not sure what happened but I beginning to suspect that I have something like ET phone home running in my machine (probably a windows program) but cycling the WiFi stopped it. At least for now ...probably until I turn the machine off.
Suggestions please of the probable culprits ?

1. Have you properly set your network metric settings? See steps 6 and 7 at this link: http://www.w1aex.com/anan/anan.html#network

2. Strip out all the unnecessary protocols from your Ethernet interface to the ANAN. See an example of all I stripped out at this post here: viewtopic.php?f=22&t=4342#p21860
JJ4SDR
Posts: 436
Joined: Fri Jul 30, 2021 10:09 pm
Location: TEXAS, USA

Re: vac buffer overruns during 2 tone testing

Postby JJ4SDR » Tue Oct 11, 2022 10:46 pm

Scott,

Not that I have any real serious issue with "superfluous" Ethernet connections, but there are some unnecessary connections that I may want to strip.

On the Ethernet LAN, there are no connections (all cables terminate in an unmanaged T switch) other than:
1. CAT 8 cable from the modem
2. CAT 8 cable from PC
3. CAT 8 cable from ANAN 7000 DLE MKII
4. CAT 8 cable from RS2K-S SSPA

My PC has two WiFi radios, one that is built-in and the other is extra, external one that I have plugged into a HUB hub.

I don't really need any WiFi functionality (I have another PC on another desk that I can use) so I would imagine that keeping those switched off is fine.

Don't fix it if it ain't broke is a good principle to go by, but perhaps I should make some changes to the current configuration. At least, if they help with getting the ANAN to work even better.

Below, see a few images.

Ethernet.jpg
Ethernet.jpg (212.47 KiB) Viewed 1955 times


Best,
Juha
NI2M
PC: 8 Core i7-10700 CPU @ 2.90GHz, NVMe SK Hynix 512 GB SSD, 32GB RAM
Windows 10 Home, Version 22H2
Thetis v2.10.4.3 x64
Protocol 2 v2.1.18
VK4BXI
Posts: 62
Joined: Sat May 22, 2021 11:44 am
Location: Brisbane

Re: vac buffer overruns during 2 tone testing

Postby VK4BXI » Wed Oct 12, 2022 3:21 am

Well I've been bouncing around looking for anything that might be a culprit or at least reduce the hiccups in transmission. I'm beginning to think that Pure signal maybe involved is some way.
The reason I say this is that under the conditions that I am testing (two tone on pure signal on with ~ 12 watts into a dummy load). I am monitoring the CPU utilisation under task manager. With my PC that runs at about 30% when I'm transmitting with pure signal. What appears to be happening is that when I change some of the pure signal settings "quick attenuate response" seems to be one of the worst. The cpu utilisation drops about 10% for a couple of seconds, then as the utilisation returns to its previous level I get a hiccup in transmission......every time. If I deselect quick attenuate response this happens once and then does not seem to happen again i.e. hiccup free. If I go the other way and enable quick attenuate response then I get the above changes in utilisation and the resultant hiccup in transmission .....but this carries on repeating every 10 seconds or so. changing auto attenuate or relax tolerance will also cause the hiccup "cycle" to operate once and then stay stable.
This is by no means conclusive but I suspect the latest changes to pure signal may not be quite as robust a first thought. It would seem that changes to pure signal settings should not be done while transmitting. And that there may be an issue with the quick attenuate response option with something accumulating up and then overflowing or something similar.

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

Re: vac buffer overruns during 2 tone testing

Postby w-u-2-o » Wed Oct 12, 2022 12:08 pm

VK4BXI wrote:Well I've been bouncing around looking for anything that might be a culprit or at least reduce the hiccups in transmission. I'm beginning to think that Pure signal maybe involved is some way.
PureSignal is not to blame. Either your CPU or your NIC is not keeping up with the demands of the radio.

Have you tried any of the suggestions I made above?

BREAK

JJ4SDR wrote:Not that I have any real serious issue with "superfluous" Ethernet connections, but there are some unnecessary connections that I may want to strip.
Juha, I'm not sure why you are posting in this topic. Are you having problems with VAC buffer overruns during 2 tone testing?
JJ4SDR
Posts: 436
Joined: Fri Jul 30, 2021 10:09 pm
Location: TEXAS, USA

Re: vac buffer overruns during 2 tone testing

Postby JJ4SDR » Wed Oct 12, 2022 9:11 pm

Scott,

I posted here as I thought that what VK4BXI is writing may help make experience with Thetis even better. I have internet connection active on the PC my Thetis runs on. Not sure where to post ;) .

Juha
PC: 8 Core i7-10700 CPU @ 2.90GHz, NVMe SK Hynix 512 GB SSD, 32GB RAM
Windows 10 Home, Version 22H2
Thetis v2.10.4.3 x64
Protocol 2 v2.1.18
VK4BXI
Posts: 62
Joined: Sat May 22, 2021 11:44 am
Location: Brisbane

Re: vac buffer overruns during 2 tone testing

Postby VK4BXI » Thu Oct 13, 2022 7:47 am

Drat....puzzlement and relief. The machine started up clean and hasn't put a foot wrong since. I did do a fairly heavy clean up of any programs not actually required to run Thetis before I shutdown yesterday , including any windows programs that the the machine would let me delete, that may have helped !

I did return to AEX's documentation and the Apache user manual and worked for several hours trying to get a sub set of 192.168.XXX.XXX going to no avail. AEX's comment that if one had a 169.XXX.XXX.XXX IP address then there was probably something wrong. A look on the internet noted that anything 169.XXX.XXX.XXX was reserved for direct connections only i.e. rig to computer direct, which was what I was running. Note that I already had a 192.168.1.118 IP address for the WIFI. So I copied down the auto assigned IP addresses of the rig and the computer (169.254.16.55 (computer) and 169.254.230.241 (rig) ) I then put these values into static IP addresses for the rig and computer and rebooted. Everything works now, through 4 or 5 reboots. I really am not sure if this was the right thing to do, but the setup works and has achieved the goal of going on to static II addresses. I did add in the other bits and pieces that he suggested, so all good there.

I don't know if this will help anyone else. It's been a voyage of exploration....Chivied along by Scott's comments (thanks you do a great job Scott).
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: vac buffer overruns during 2 tone testing

Postby w-u-2-o » Thu Oct 13, 2022 12:04 pm

VK4BXI wrote:Drat....puzzlement and relief. The machine started up clean and hasn't put a foot wrong since. I did do a fairly heavy clean up of any programs not actually required to run Thetis before I shutdown yesterday , including any windows programs that the the machine would let me delete, that may have helped !

You act like it's some mystery on how things "fixed themselves". And then you go on below about how you actually changed some things. It is most likely those changes you made that fixed your problems, not magic or luck. Or the Windows programs you removed.

So I copied down the auto assigned IP addresses of the rig and the computer (169.254.16.55 (computer) and 169.254.230.241 (rig) ) I then put these values into static IP addresses for the rig and computer and rebooted. Everything works now, through 4 or 5 reboots. I really am not sure if this was the right thing to do, but the setup works and has achieved the goal of going on to static II addresses.

All of the network savvy folks are cringing right now. Given your failure to put in proper static IP addresses, e.g. 192.168.3.x, you probably failed at this again and both the PC and the ANAN are still using automatically arrived at APIPA addresses.

I did add in the other bits and pieces that he suggested, so all good there.

If you mean you changed the network metrics as he instructed, then this is the one change that most likely made the difference for you, as I alluded to in my post above.

I don't know if this will help anyone else. It's been a voyage of exploration....Chivied along by Scott's comments (thanks you do a great job Scott).

Thanks! I recommend that you do more studying on Ethernet networking and, once you have educated yourself a bit more on the subject, revisit this problem again in order to achieve correct and proper static IP addressing.
VK4BXI
Posts: 62
Joined: Sat May 22, 2021 11:44 am
Location: Brisbane

Re: vac buffer overruns during 2 tone testing

Postby VK4BXI » Fri Oct 21, 2022 8:25 am

Close out report.

static IP's set up at 192.168.2.200/201 all working nicely (a friend who is in the business came by and assisted). No change in operation.

Final conclusion was, that it was not an IP issue, but rather too many program "fleas" running in the background. Not sure if it was a windows program or one I've collected along the way or a combination. It required several reboots afterwards to finally clear (not totally sure about this but one restart didn't work, but was cured when switched on from "cold" the next day)

Resolution to self, do a regular clean out of programs running on the PC that are not necessary for operation of thetis.

Bob

Return to “Thetis”