SpeedStep stops only AO of DAQCard-6036E?

[sugimura]

CWBufferedAO.zip program runs for 1 to 80 hours on DELL
Latitude V740 and Toshiba DynabookSS 4000 with Intel
SpeedStep CPU when the SpeedStep mode is off.
When the power save mode or SpeedStep is active, it
stops within 5 to 30 minutes.
When AI routine is added to this sample program, AI
routine runs always although AO (Progress) routine
does not work. This trouble occurs after it runs for
1 to 80 hours.
Does Nidaq32.sys work properly on SpeedStep CPU?

[Russell]

Actually, this is a problem we have been trying to track. At this point we really do not have that much information about it. In other cases, the problem seemed to go away once the power management was disabled. (Perhaps no one bothered to run it for 80+ hours.) So this is not a workaround in your case?

I'm wondering if it has anything to do with the DAQCard possibly sharing an interrupt with the power management hardware. At any rate, we are still in the process of investigating this. Is there anything else you have discovered? Do you have any means of a workaround in the meantime?

Russell
Applications Engineer
National Instruments
http://www.ni.com/support

[sugimura]

AO has atopped after 62 hour run on Toshiba note PC
while AI continues. The longest record was about 76
hours on DELL Latitude V740.

An external fan was used for DELL because its fan does
not work always. The fan of Toshiba works always.
The temperature at outside of 6036E was eye-checked
and AO stopped when its temperature became lower than
about 28 on DELL or about 27 degree C on Toshiba.

The temperature may not be the cause to stop AO, but
the coincidence occurred on both PC's.

In the case of Toshiba, its BIOS has menu to stop the
SpeedStep function. But some background tasks such as
power monitor and sound controller may cause to stop
AO. When these tasks are active, only AO stops within
about 2 hours. AI works always.

[Russell]

Thank you for the information.

By the way, what is the rate of your updates? The maximum for the DAQCard 6036E is around 1kHz, but will vary from system to system. This could be the problem if your rate is close to 1kHz. Do you ever get an NI-DAQ error message after it stops?

By the way, someone with a similar problem was able to solve it by disabling the AGP Video driver in their Dell Latitude. (The AO would just stop after some time without error.) Also, you could check the Windows Device Manager for anything else using the same IRQ as the DAQCard. See if disabling these devices has any effect.

[sugimura]

The sampling rate is 1024Hz for 2 channels of AO
and 1 channel of AI. After the network is disabled
on Toshiba 4000 (750MHz), it runs over 100 hours
and is in the second 100 hour test.
When the performance monitor of Task Manager is
running, AO stops within 2 to 70 hours. CPU usage
is about 30%. When AO stops, error routine is not
called.

Thank you for your new information related Latitude.
I will check this.

[sugimura]

In the case of Toshiba with Win2000, one USB mouse is
connected, When the mouse is heavily moved, AO stops.
If the performance monitor of Task Manager is active,
the move of mouse causes AO to stop quickly.
By the way, the program stopped after it run for 50
hours in the second run.

According to DELL support, there is no way to disable
AGP driver on Latitude V740.

[Russell]

Thank you for keeping me in the loop as far as your investigations. The problem is most likely because your update rate (1024 Hz) is over the specified recommended rate of 1 kHz. You can see this on page A-14 of the NI 6034E/6035E/6036E User Manual. You were probably able to improve performance since this rate is system dependant and since it's only slightly over the maximum rate. Disabling other interrupt devices provided more bandwidth to the DAQCard.

However, nonetheless you should be getting an NI-DAQ error when the AO stops. I realize this is still a problem and I have notified the
right people. So thank you for your help. As for your current situation, if nothing else I suppose the only thing you can do is just reduce the update rate below 1 kHz.

Let us know if there's anything else we can do.

--Russell

[sugimura]

There is no difference between 1000Hz and 1024Hz of AO
sampling rate. The same trouble occurs on Toshiba and
DELL Latitude. This program was developed and tested
on Sharp Note without fan.
200 sec data is displayed in CWGraph and a cursor in
the graph appears to point the current output block.

Maker SpeedStep CPU Win2000
Sharp 300-500MHz Max Microsoft
Toshiba 500-750MHz Off Toshiba
DELL 1.2-1.8GHz Max DELL

The CPU usage is about 30% on Sharp and Toshiba, 10%
on DELL. The performance monitor of Task Manager and
heavy mouse move do not cause AO to stop on Sharp
although these cause AO to stop quickly on Toshiba.

The memory size of this program is about 28MB and its
minimu
m WorkingSet size is set to 30MB.

After CWGraph is replaced with user routine, the
program size becomes about 12MB and the CPU usage goes
down to less than 10%. The performance monitor of Task
Manager and heavy mouse move do not cause AO to stop
on Toshiba. Until now, it has run on Toshiba for 42
hours. We hope it can run at least for 100 hours.

We give up to use DELL because their support does not
provide us any proper information.

The miximum AO sampling rate is dependent on PC. But
it seems to recommend us not to use this PC card for
production run.

[Russell]

Well, the maximum update rate still could be less than 1000 Hz (depending on the system). It would be interesting to see if you have the same behavior with an update rate of, say 500 Hz. See if you get the same behavior with a 500 Hz update rate using the sine wave generator from the "Test Panel" in MAX. As for production use, you would be much better off with a PCI or PXI form factor of the 6036E. These boards utilize DMA transfers and are are much more reliable than their DAQCard/PCMCIA counterparts.

By the way, if you don't mind, please send a reply to me, Russell, via "email" (from option 2:"request support" on http://www.ni.com/support). I have another possible fix I would like for you to try
.

Thanks for your patience,
-Russell