Stress testing the network.

[JRBarrett]

Last night I did a flight from LAX-JFK as QFA107, with the specific intention of testing what would happen with running a "full load" of ancillary programs in conjunction with PSX. I run both PSX and either FSX or X-Plane on a single computer, using two monitors.

Previous flights have been done using the VisualPSX and TrafficPSX software to drive FSX visuals and a VATSIM connection. Those flights were done using weather injection into FSX directly from PSX, and have been gone quite smoothly.

Last night, I disabled the PSX weather injection, and ran Active Sky Next (ASN) on the FSX side - mainly to obtain a more graphically-rich depiction of cloud layers and visibility effects.

I think this may have pushed the network (more specifically FSX simconnect) into the realm of instability.

Heading east over the desert toward Arizona, the Los Angeles Center VATSIM controller informed me that my ground speed reading on his radar client was periodically dropping to zero - though my track and altitude readouts appeared stable. This was just about the time I was entering an area of enhanced cloud cover, which were likely the remnants of earlier thunderstorms. I left controlled airspace shortly thereafter.

At about the same time, I started having some issues with PSX itself, which may have been an indication that the network or perhaps CPU were struggling. Specifically, the autothrottle system started acting very strangely. I noted this first when the aircraft started turning off course, even though LNAV was selected and active. I switched to heading select mode to try to turn back on course, but the turn continued.

It was at this point that I realized that my two left engines had advanced to full power, while the two right engines had dropped back almost to idle - which certainly accounted for the uncommanded turn. I disconnected the autothrottle completely (turned the AT switch on the MCP off) but still had a devil of a time getting the engine power levels stabilized. The throttle commanded positions displayed on the N1 tapes, and the actual positions of the hardware throttles on my CH quadrant were way out of synchronization, and changing continuously. I brought up the instructor menu and turned the throttle "humanizer" off, which helped a little, but still did not appear to have good control over the engine power settings even in manual mode.

At this point, I shutdown the ASN weather client on the FSX side, which immediately cleared the clouds and visibility. Once I did this, I was finally able to regain control of the PSX throttles in manual mode. Once I had all four N1 levels set close to the calculated value on the FMS cruise page, I was able to re-engage the autothrottle and get back on course with LNAV/VNAV engaged.

Though I had no further throttle issues for the rest of the flight, my ground speed output to VATSIM continued dropping in and out for the rest of the flight. This was undoubtedly a simconnect problem, for as long-time FSX users know, once simconnect becomes unstable, it tends to remain unstable until FSX is re-booted.

If you do a search on Vataware for QFA107 under my pilot ID 819073, you can see the intermittent ground speed problem in the vertical plot, and my temporary directional "excursion" in the track plot, as I was heading eastbound towards TRM.

None if this is intended as a complaint - just an observation. I did intend to see if I could over stress the combined PSX/FSX system, and I believe I did do just that.

As nice as ASN-generated weather looks in FSX, I am afraid that trying to run it, VisualPSX AND TrafficPSX all at the same time, may be asking a bit much - at least on a single-computer system. Having FSX and PSX running on separate computers would probably help, though simconnect would probably remain the weak link. I believe that ASN simply overwhelms the network.

I am thinking that the temporary loss of PSX throttle control may have been caused by the CPU cores being maxed out - hard to say for sure.

[Hardy Heinlin]

First thing to check:

Any random malfunction slider out of OFF?

Any items activated on the malfunction pages?


Regards,

|-|ardy

[JRBarrett]

First thing to check:

Any random malfunction slider out of OFF?

Any items activated on the malfunction pages?


Regards,

|-|ardy

Yes, it was the first thing I checked when the problem started. No malfunctions were set, but I went ahead and pushed the "clear all" button on the first page, and went to each individual systems tabs and selected "repair all" on each. It didn't help. It was only after I shutdown ASN and cleared all weather that the throttles started to respond normally.

I didn't make note of the PSX frame rate, which is almost always locked at 72, but I'd wager it was probably down.

I really do think that I was simply asking too much of my single-computer system to run FSX (at pretty high graphics settings) + PSX + Visual and Traffic PSX, + ASN (generating multiple cloud layers). It seemed to be the addition of ASN that was the final straw - "der Tropfen, der das Fass zum Überlaufen bringt" so to speak.

With all that was going on, I didn't want to bring up Windows Task Manger to look at core loading, but if I had, I probably would have seen all four cores running full out.

Fortunately, PSX did not actually fail under this condition, it was more like an aviator suffering from hypoxia - still functioning, but at reduced efficiency. Once I gave e PSX more "oxygen" (by killing ASN) it settled down, and the throttles started working properly.

[John Golin]

For future reference, if you can save a situation file (or 15! ) when you have problems it really helps find out what was going on.

The situation files are tiny and you can either save them from the Instructor dialog, or press the EVENT RCD above the upper EICAS on the MIP which will autosave a file for you.

[JRBarrett]

For future reference, if you can save a situation file (or 15! ) when you have problems it really helps find out what was going on.

The situation files are tiny and you can either save them from the Instructor dialog, or press the EVENT RCD above the upper EICAS on the MIP which will autosave a file for you.

I'll definitely remember to press the EVENT RCD if I have any issues on future flights, and will post here. I had forgotten about that feature!

[Hardy Heinlin]

The throttle commanded positions displayed on the N1 tapes, and the actual positions of the hardware throttles on my CH quadrant were way out of synchronization, and changing continuously.

Maybe a mechanical problem in the USB hardware, or a software problem in the USB caused by another software?

[JRBarrett]

The throttle commanded positions displayed on the N1 tapes, and the actual positions of the hardware throttles on my CH quadrant were way out of synchronization, and changing continuously.

Maybe a mechanical problem in the USB hardware, or a software problem in the USB caused by another software?

That's quite possible. It's probably not a mechanical issue, but could well be software-related. Of course, the USB controller inputs are also read by FSX, which was active on the same computer, though not SET by FSX. I do not use FSUIPC in FSX, to process any axes or buttons.

Perhaps if all cores were running at maximum effort, Windows own USB processing might have suffered in some way, creating instability at the time. After landing, and after shutting down FSX, PSX, and all the associated programs - (but before rebooting Windows itself) - I did check the calibration of all USB controller axes in the Control Panel, and at that time, they were all correct.

Even though I wanted to see how ASN weather would look on the FSX side when run in parallel with PSX, going forward I'll probably stick with PSX weather injected via VisualPSX. It's easier to do it that way, and it has the advantage that cloud buildups and storms seen outside the FSX aircraft, should directly correspond to what is displayed on the PSX weather radar.

If I see any instances of unexpected behavior in PSX on future flights, I will be sure to remember to make use of the EVENT RCD button to capture a .situ for later analysis.