S7-1500 modbus tcp speed?

Hi guys!

I was wondering if there is a way to speed up the modbus tcp server in the S7-1500?

I'm talking about the time it takes from a request is received in the S7-1500 (the modbus server) to the time it takes for it to send back the response to the modbus client.

Right now it takes just over 200ms to read 125 consecutive 16-bit words from the S7-1500. It's all in one request.

Somewhere in the back of my mind I think there was a way to allocate how much cpu cycles/time would be allocated to communication. I can't remember where and I don't know if that is applicable in this case.

Thanks,
Pete

Usually that is done on the Hardware configuration.
Click on the PLC and further down there should be a communication load option.

I was nder the impression that Modbus was done via some function blocks... perhaps you're not calling them often enough?

is your modbus client capable to send several TCP querys same time to 1500?

cardosocea said:

Usually that is done on the Hardware configuration.
Click on the PLC and further down there should be a communication load option.

I was nder the impression that Modbus was done via some function blocks... perhaps you're not calling them often enough?

Click to expand...

S7-1500 has a block called MB_SERVER that you should put in OB1. So it's called every scan.

I'll check out the communication load option.

Lare said:

is your modbus client capable to send several TCP querys same time to 1500?

Click to expand...

Yes, it's possible to set it up like that. Do you think that will be an advantage?

I don't know what the bottleneck is, but I'd be curious what you see if you do a wireshark trace. The actual sending of the data happens asynchronously to the scan in the communications time, but the processing may still have to happen in the PLC scan. It is possible that there is a bottle neck like being able to only send one Modbus TCP packet per scan of the server block.

I'm curious A) what your scan time is and B) if calling it out of a faster cyclic OB speeds things up.

mk42 said:

I don't know what the bottleneck is, but I'd be curious what you see if you do a wireshark trace. The actual sending of the data happens asynchronously to the scan in the communications time, but the processing may still have to happen in the PLC scan. It is possible that there is a bottle neck like being able to only send one Modbus TCP packet per scan of the server block.

I'm curious A) what your scan time is and B) if calling it out of a faster cyclic OB speeds things up.

Click to expand...

From someone that managed to get my hobby electronics playing well with Modbus TCP, Wireshark would definitely provide a bit of insight or at least more detail on the transactions going on here.

I don't remember an instance of a cyclic OB being faster than OB1 in any of the Siemens applications I've seen in the past. Is this common i9n your experience?

cardosocea said:

I don't remember an instance of a cyclic OB being faster than OB1 in any of the Siemens applications I've seen in the past. Is this common i9n your experience?

Click to expand...

It isn't the common use case, but it's certainly possible. Typically you see cyclic interrupts used for safety code or PIDs, and it is slower but higher priority. I have seen a few projects that were ALL cyclic OB and no OB1, in the process world. Sometimes you need to balance both frequency and priority.

The best example of a fast cyclic OB I can think of is MC-Servo, used for the TOs in 1200/1500. It basically acts the same as a cyclic interrupt, usually at a rate much faster than the PLC scan. I just checked in the project I have open; you can create a cyclic interrupt in a 1515 as fast as 500us.

mk42 said:

I have seen a few projects that were ALL cyclic OB and no OB1, in the process world. Sometimes you need to balance both frequency and priority.

Click to expand...

Yes, this is typical in PCS7 and I suppose the process industry as a lot of the logic doesn't have to run at high frequencies. Though, in that case, OB1 is still there but only to run clean up functions if any.

mk42 said:

The best example of a fast cyclic OB I can think of is MC-Servo, used for the TOs in 1200/1500. It basically acts the same as a cyclic interrupt, usually at a rate much faster than the PLC scan. I just checked in the project I have open; you can create a cyclic interrupt in a 1515 as fast as 500us.

Click to expand...

This makes sense, I suppose. I haven't seen the response of S7-1500's, but something running at 500us, probably will make the system stutter?

cardosocea said:

This makes sense, I suppose. I haven't seen the response of S7-1500's, but something running at 500us, probably will make the system stutter?

Click to expand...

Not 100% sure what you mean by stutter. It won't actually be visible as random pauses to the outputs. It shows up us a lengthening of the overall scan time, similar to communications.

Lets say you need some code running every .5ms, and it takes .1ms to execute. Lets say your overall PLC scantime would have been 10ms. Once you add in the cyclic interrupt, now your typical scantime goes to 12.5ms. If you double the amount of code in the cyclic interrupt, and it now takes .2ms, that lengthens the scantime to 16.6ms. (I think I did that math right...)

Moral of the story is that I certainly wouldn't recommend running much code that fast. I've really only seen it make sense for motion control type situations where you have really fast control loops.

Pete.S. said:

Yes, it's possible to set it up like that. Do you think that will be an advantage?

Click to expand...

It depends how many modbus querys you have. If you have only 125 words, then this probably won't make difference

if you have something like 500 words, then you can make 5 different modbus TCP querys to slave at same time (now 1500PLC). So you get also reply for 500 words at same time (5 times faster and only one interval)

Siemens PLC have max connections limit on PLC hardwire settings and also PLC type limits connections. (Connection resources)

https://support.industry.siemens.co...ic-s7-pn-cpu-as-modbus-server-?dti=0&lc=en-WW