Does Edge IO NTS and CRA Hybrid Architecture Support 50ms re-healing and time stamp

[Matthewwsh]

Does anyone has experience in using edge IO NTS along with x80 rack with CRA in a M580 PLC system. Does mixing edge IO NTS has any trade off in performance of the CRA? 

Such as will this affect the timestamp and re-healing feature of CRA ? And does the edge IO NTS supports the above mentioned features as well? 

Thank you

[Trinxs1 · Accepted solution]

Edge I/O NTS is considered distributed IO and hence will have a lower priority than a CRA drop.  Therefore, excluding any abnormal issues such as bad cables, poor connections, or cyber-attack incident the recovery time of the ERIO drops should not be affected.    
When Edge IO NTS is connected to any ERIO ring it must be counted in the number of hops allowed for the ERIO network main ring so that we do not exceed the number of drops permitted in the main ring this also ensures the recovery time is maintained.

For edge IO NTS there is no guaranteed recovery time.  However, it is likely to be of a similar order of time.

With respect to timestamping X80 timestamping is managed by the CRA and is at 10ms recovery.  For the Edge I/O NTS this is dependent upon the scan time of the CPU and the module that manages this. If it is managed by the CPU the inputs are updated at the start of scan, the outputs at end of scan.  If managed by the NOC module it may be further processing time as the NOC module must receive the message and then update the CPU memory at the end of scan. In both cases the scanning time of the Edge IO module must be taken into consideration as well. 
If the NOC module scans the Edge IO NTS at a 100ms rate, then the new values are only available in the CPU memory every 100ms plus the CPU scan time.  The assumption here is that there is only 1 NOC module.  If there are multiple NOC modules then the settings of the CPU for the number of requests served by the CPU must also be considered. This will depend on the CPU used and the settings of system word %SW90.

[ciupol]

---

@Trinxs1 wrote:

The assumption here is that there is only 1 NOC module.  If there are multiple NOC modules then the settings of the CPU for the number of requests served by the CPU must also be considered. This will depend on the CPU used and the settings of system word %SW90.

---

This is a little out of topic but I was wondering if there are any good reading resources that explain this subject of requests and system words like %SW90. I know that system words are described in manual but the description is little vague. How does the setting of %SW90 goes along with %SW87. I was told that %SW87 shouldn't ever exceed 16 but %SW90  can be set higher than 16 (depending on CPU).

[Trinxs1]

Refer to the manual https://www.se.com/au/en/download/document/EIO0000001578.  Page 32 provides a table which explains how many requests the CPU processor can manage per scan.  %SW90 is by default the number in column 2 (requests per scan).  It can be increased up to the maximum shown or reduced down to 2 to potentially modify the overall scan time depending on the requirements of the control system.
The table also lists the various sources of requests that can be serviced by the CPU.  For example BMEP585040 can handle 40 requests per scan.  3 of these may come the USB connection, 20 directly to the CPU and only 16 can come from one, two or three NOC modules, which is a total of 39 so there is one request spare out of the 40 it can handle.
Refer to the  EcoStruxure™ Control Expert System Bits and Words Reference Manual
%SW87 is a measure of how many requests are processed by the CPU server per master (MAST) task cycle.  This is a read only value.  Effectively you can calculate the server load by dividing %SW87 by %SW90.  This gives an estimation of the load on the CPU.  

The reference to 16 per scan (I think) is the maximum number of requests / scan that can come from all of the ethernet modules (i.e. 1, 2, 3, 4, 5 or 6 depending on how many ethernet modules the CPU allows).  For example, if there is only 1 NOC module it can make up to 16 requests.  2 NOC modules can only make 8 requests per module per scan....

[ciupol]

@Trinxs1  I assume you are admin here. Can you put our discussion into separate thread? I do not want to hijack this thread as really my questions are off topic.

Thank you for the explanation. I am little confused with it though.

@Trinxs1  wrote:

 For example BMEP585040 can handle 40 requests per scan.  3 of these may come the USB connection, 20 directly to the CPU and only 16 can come from one, two or three NOC modules, which is a total of 39 so there is one request spare out of the 40 it can handle.

---

Indeed in the table it says that this CPU can handle 40 requests per scan. But for the USB the table says 4 not 3. Also you mention 20 connection directly to the CPU but I do not see that number in the table.

---

@Trinxs1  wrote:

For example, if there is only 1 NOC module it can make up to 16 requests.  2 NOC modules can only make 8 requests per module per scan....

---

I assume it does not have to be divided equally. Ig I have two NOC modules 4 requests can come from one module and 12 from the other.

Also I looked in the manual that you linked and on page 33 there is this example:

This again confuses me.

I understand that gray dots are X-Bus connectors and red dots are ethernet connectors on the backplane. In point 4 it says that those are request made to IP of controller. Point 5 says that those are request to IP of NOC module adding that in this case ethernet backplane port needs to be enabled. Shouldn't this be opposite and ethernet backplane port needs to be enabled when making request to controller IP?

Also it is my understanding that NOC301/311 are using X-Bus to talk to CPU whereas NOC302 uses ethernet. In this example there is no differentiation and all are treated together. Does that really make no difference in all those deliberations? How NOC321 with IP forwarding fits into all this?

Going back to my initial question it seems that there is nothing wrong with %87 going over 16 (in my case I use BMEH586040) because those are ALL requests that CPU handles. Now is there a way to know how many of those requests come to the IP of comm modules (which are limited to 16)?

Once again sorry for hijacking the thread.

[Chaox]

Yes you can put Edge IO NTS inside the M580 RIO (with CRA) and it does not impact the 50ms performance !