LabVIEW
Shared Variables from PC to RT
Solved!
Hello,
I am working with a myRIO project where data is communicated from a PC VI through Shared Variables to an RT Host VI which writes them for the FPGA Target. The PC Front Panel has a cluster for Settings and a cluster for Data, and each of these clusters are type defined clusters that include shared variables (which are also clusters) inside of them. These shared variables and these type defined clusters are referenced in the RT VI.
If I want the values on the PC VI to propagate through the shared variables to the RT Host VI, should I set the cluster on the PC front panel to have a Data Binding of "Shared Variable Engine: Write Only", or should I have the shared variable clusters inside of the type definitions as "Shared Variable Engine: Write Only"? Both the clusters and the shared variables initially had data binding settings of "Unbounded", but I am unsure if this is correct. Would "unbounded" be correct for this type of application?
On trying to test the initial settings of these VI's, I was able to have type defined boolean controls (not clusters) with data binding set to "Shared Variable Engine: Read/Write", and these communicated data to the RT. However, the type defined clusters that I previously mentioned did not communicate the information written to the front panel.
Picture of PC Front:
Picture of TypeDefined Settings cluster SharedVariable (from PC). This is the shared variable that is inside the type defined cluster:
Picture of RT Block Diagram cluster that I'd like to read the value in the Shared Variable:
Picture of RT Front that displays error since the RT read a 0 due to no values written (this produced a condition to divide by 0, hence an error):
I am working on getting a zipfile to share here so that the code may be used directly, but I am still working on getting it set up.
- Best regards, 2001J
Personally, I think Network Published Shared Variables are evil. Instead, use TCP/IP or Network Streams to pass data between your PC and RT. I personally prefer using the TCP/IP route using the STM library.
I just stated that I like to use the STM library since it creates the message framework for you. You can set up your own protocol.
Hi J,
@2001J wrote:
But may I ask why you believe that network published shared variables are evil?
Because they always provide problems with firewalls. You need to configure access to them on each client as well as on the host.
Because they need quite a lot of resources, which is a huge problem when hosted on cRIO targets. (You need to install one more software component onto the target, and they require a lot of memory. On an old cRIO907x they took ~20MB of the 64MB available RAM…)
TCP or UDP comes "mostly" for free as they are supported even with the base software installation.
You only need to implement some message handling - or install the STM library to get that for free.
@2001J wrote:
I would prefer to refrain from downloadable content for this project.
Why do you?
When you are one of a team of programmers you should use a SCC system anyway, and it's quite easy to include additional downloads into the SCC repository so all participants get access to those functions…
@2001J wrote:
But may I ask why you believe that network published shared variables are evil?
To add to the list:
1. They are SLOW. Just try writing to one and immediately read (you will get the old value). The system has to wait for a buffer to fill up or a timeout (I think in the 100ms range) before the updated values are sent to the server. Then the clients will have to get the update from the server. So EXTREMELY slow.
2. No security at all. Anybody on the network can update the value. This can cause disastrous situations.
3. A common of both of the above: very weird race conditions.
Thank you for the information GerdW, Bob Schor, and, of course, crossruiz.
As for using this communication system, I am a bit intimidated by it at the moment, so I was wondering if I could receive some assistance with implementing it.
If that is alright, I am programming a single 1900-myRIO to communicate 1 cluster of settings from the PC to the RT host, 1 cluster of data from the RT host to the PC, and 3 boolean variables from the PC to the RT host. Both clusters include numeric data. When I say host, I am referring to the host of an FPGA Target, but I am not reading/writing with it through Shared Variables nor STM, it communicates through "Read/Write Control" at the moment.
From reading some documentation (link: http://www.ni.com/tutorial/53683/en/), it seems that I must first create a connection between the client and the target, then write an array of Strings describing the data to the "STM Write Meta Data"s "Meta Data" connection. Afterwards, I must write the respective data's description to the "STM Write Message"s "Connection Info", then write the actual data (converted to String) to the respective "STM Write Message"s "Data" connection. Considering that I am writing numeric data, then these data must all be strings. But when it is eventually read, I must have a Case Structure to convert that specific data instance (as specified earlier by its respective string index that was written to the "STM Write Meta Data"s "Meta Data" connection) to its correct data type. Does this logic seem alright?
Assuming that this is in the PC VI, then I would have part of it doing the writing and part of it doing the reading, as I need it to both write settings and read data. Does this setup seem accurate? I am referring to the following picture for the writing:
Later today, I can post a zipfile of a project that I would like to implement this communication. It currently has shared variables. If you would like to view this code, use the "RT Main.vi" Host, the "LIA_myRIO-1 Control Panel.vi" PC, and the "FPGA_LIA_2ChPhase__Question.vi".
- Thank you, 2001J
