some questions about contoured move

[bgrain]

Hi Everone
 I am using a pci-7344 board and i am not very clear about why the time interval in contoured moves is at least 10ms.
 How to make it less than 10ms.
 I am doing a project that requires the time interval to be 1ms how to make it!!
 Indeed I want to use the card to control a system which samples a new target position and
current position every 0.5ms.  contoured move is what i want but its time interval is too long.
 I wonder if I can do some changes to the card to make the time interval shorter or if ni can provide some toolkits to make it shorter.Except contoured move is there other method to implement this type of movement? What about onboard program?
 Can it work with a external hard timer?
 Last question:Is the 7344  partially open to the users for further development?
THANK YOU!!!!!

[Jochen]

Bgrain,

there is no way to tweak a PCI-7344. The 10 ms timing for contoured move is the fastest that you can get out of the board. On the other hand please take into account that the PID loop is running much faster and in order to align the trajectory generation rate ( = 10 ms in contoured mode) and the PID rate (= up to 62,5 µs) the DSP is calculating interjacent setpoints at the rate of the PID loop using a spline function. This results in very smooth moves even with a relatively low contouring update rate of 10 ms. As long as you don't need to calculate new setpoints dynamically (= during the move) at the rate of 0.5 ms the standard contouring mode should do it.

If you can't calculate your contouring data prior to the start of the move the NI SoftMotion development module for LabVIEW could be a good solution. This module provides a hardware independent motion API, trajectory generation and a spline engine. For your task the best hardware platform would be a PXI-system running under LabVIEW Real-Time with a FPGA based RIO-board plugged in.
This might sound very complicated but in fact it offers you complete freedom to design your own motioin controller with all the features you need. I have worked with NI SoftMotion and my system was up and running in less than one working day. Precondition for starting development with NI SoftMotion are solid LabVIEW skills.

Best regards,

Jochen Klier
National Instruments Germany
 

Message Edited by Jochen on 07-19-2005 08:59 AM

[bgrain]

Hi Jochen Klier
 Thank you for your reply
 0.5ms is the system requirement because if the time interval is shorter, the trajectory is more
close to the anticipant signal wave,for example a sine wave and this is what we need.
our another requirement is the new target positions should not be given by the host computer evey
o.5ms but a lot of a trajectory points should be downloaded to the card once then the control is independent of the host computer.the dsp get a new point from the downloaded ones every 0.5ms and compare it with the feedback ,then give the following error to the PID loop.
 Can you give me some advice about how to implement this function with the pci-7344 or with 7344 and other hardwares,because i have a 7344 card only presently.
 If a PXI system is used ,is it not necessary to use a host computer because it has a embedded system pentium for example.
 In order to meet the system requirement can you recommand me some schemes in detail including
which kind of card of ni can satisfy the reqirment at the least cost?
thank you!!!!!

[Jochen]

Bgrain,

again: The PID loop runs with up to 62,5 µs. This is much faster than the 0.5 ms you need. The position comparison runs of course with the same rate and the PID loop gets a new setpoint at the same rate. It's correct that contouring data can be passed to the DSP with a maximum speed of 10 ms from the buffer and there is nothing you could do to speed this up but on the other hand the DSP interpolates between the contouring data points using a cubic spline algorithm so at every PID cycle there is a new setpoint calculated. This is much better than a linear interpolation and will result in a move that's very close to the ideal sine trajectory.
Sorry that I can't explain cubic splining in some simple words but there are people who have written books about this topic...

As long as the period of your ideal trajectory (e. g. sine) is not shorter than about 200 to 100 ms you should get very good results with this method. If you find out that the contouring accuracy is not good enough for your application we still can talk about other software and hardware options. Please note that inaccuracies might also be caused by not optimal tuned servo motors rather than by the interpolation.

Jochen