(solved) weird behaviour from modbus block in RTU to TCP multiple slave

Hi all,

I am using third party software to monitor the communication as shown in the attachment

for your information i have a gateway which convert TCP to RTU, i am using TCP RED library and my slave is RTU slave.

the 1st case is a single slave case, as you can see

1. TCP request is when my CPU request the gateway for data

2. RTU request is when the gateway request data from the slave

3. RTU respond is when the RTU reply the gateway data 

4. TCP respond is (i think) CPU has obtained the data from the gateway

so if i managa to get the data from a RTU slave, this is the monitoring result that i expected.

However in multiple case, especially when all other slave is not connected except slave 11.

As you can see, if the slave is not connected, it only has TCP request and RTU request which is reasonable since there is no slave to respond.

However for the slave 11, since it is the only connected, I expected all the 4 request and respond, but as shown in the image, the TCP respond which is the CPU respond is missing and I was not able to get the data.

Still, this is not always the case, it only fail occasionally, normally it did manage to get the full respond.

And this only happened if, the previous jobs are error. If the previous job is DONE, it will not cause the next job to behave abnormally . 

i forgot to mentioned that each slave have 3 to 5 requests, for slave 11 it has 3 request

however if the 1st request of each slave fail, i will skip all of the slave remaining requests so i can save a lot of time.

I am using job list and i OR the ERROR and DONE output of modus block to the DONE input of the job list

The reason i do this is, if the job list receive input at it's ERROR input, it will skip the rest of the job, which is ok for single slave but totally frustrating for multiple slave, so i use only the DONE input of the job list.

hmm seem like using error input of the job list is also ok

=======

i think the problem is within the modbus block, everytime when there is error, i need to wait around 4 seconds or else the next job will behave abnormaly, for example, skipped.

it work well when every device connected, but i am not sure about the error handling

Kaulquappe

Hello Tylim,
which monitoring time (RECV_TIME) do you use? When a slave is not connected, the monitoring time will elaps before the job is finished with error.

Regards,
Kaulquappe

 hello expert,

at first i also think it is monitoring issue, my monitoring time is default value 1.2 seconds

my understanding of monitoring time is, the system will wait for 1.2 seconds for responds

i check the the time needed for the block to process the request in error before, it is within 1.8 second till 3 something seconds (i do this by start my timer when i receive ENQ signal and stop the timer when i recieve error signal). The time increase when i increase the monitoring time.

i do notice that everytime the modbus block output an error signal, it need to "rest" for at least 4 seconds resulting total time need to process the error without skipping others job is 5.8 second till 7+ seconds for 1.2 seconds monitoring time,

if we try send a new ENQ signal while the modbus block is "resting", the block immediately output an error signal without any processing, and nothing can be viewed (skipped) in the traffic.

everytime there is error signal i will delay the next signal for at least 4 second, which is why you can see in my traffic monitor, between each of my error job there is around 7 seconds delay, 4 of 7 seconds is the time i delay, 3 second is the time modbus block processing (include the monitoring time)

The longer the time i delay, the more stable it is, if i cut my time to 2 seconds delay, it will skip one job after we process one job,

This is not a problem if every slave is OK, the modbus do not need to "rest" after it output an DONE signal, it can immediately carry out the next task.

I try it with job list and job list that i have modified, same thing happend, i also try conventional way (using counter and other combination of basic block), same thing happend, and if i control it manually (insert value in test mode and if do it fast enough), same thing happend.

logically speaking i should solve this after-error effect by delaying the next request but however i still encounter the issue CPU does not repsond (missing TCP resp in the traffic monitor) although i have gave it enough time, and it happens occasionally. 

Kaulquappe

Hello Tylim,
which monitoring time (RECV_TIME) do you use? When a slave is not connected, the monitoring time will elaps before the job is finished with error.

Regards,
Kaulquappe

 Dear expert, 

this image will explain the thing better about the job skipped issue, i also attach my chart where you should able to reproduce the error.

overall i am facing two issue

1. job skipped if error which can be solved if i delayed the enq signal but i want to find better solution.

2. occasionally no CPU resp which i described in 1st thread

image to explain my problem

I solved the problem

After I enable the Modbus TCP exception on the gateway side, everything become fine now, the Modbus block response the error within expected time, around 1.5 seconds for 1.2s monitoring time.