I'm doing some benchmark performance of instruction in LOGO. There I compare a timer set to 30s.

To use the memory flag and latching relay do not produce any delay. So my conclusions is that the delay is from On-Delay or the Up/Down counter. 

In this picture I have to plc function that activate Q1 and Q2. The idea is to make Q1 and Q2 to activate at same time. If On-Delay is set to 6 seconds On and 0 seconds Off it will be slower (Q1) compare to Q2. To set On-Delay to 5s 95 in other words (50 milliseconds) less. Q1 and Q2 activate almost at same time.

This have only been done in software simulation

In software developers perspective 50ms is kind of slow for 2 instructions. And maybe that is not the reality if the PLC program is running on the hardware.

So how accurate is the software simulation in LSC compare to hardware ??

Ps. I did other test with Online simulation.

Q1 and Q2 was activated at same time when On-Delay was set to 6 seconds. It means that the software simulation is not accurate compare to hardware simulation. It make sense to do Online simulation against the hardware but it was a surprise that 50ms slower for 2 instructions. :)

The counter is On at 5 so its will be 50ms x 5 times = 250ms slower.

The test was performed in Windows 11 Pro, but I did the same in macOS no difference.

Hello FredrikS7,

The software simulation is accurate with a few deviations when it comes to the execution of a program. The different processing times are due to the fact that a logo is emulated in the simulation.

The simulation is therefore basically slower.Added to this is the processing speed of the operating system and how many programs are running in the background. Even the time measurement of a block in a program is affected.Even with online queries there is a certain latency.

The program on a real logo will therefore always be faster than a simulation on the PC.

Regards Ella_68

(translated with google)

Hi Ella_68

You are right that emulation of PLC could be slower. On other had a faster computer or CPU could emulate instruction and running the software faster. Example of this would be emulation of old game console running on modern hardware.

In my example I had 2 different function that activate Q1 and Q2 after 30 seconds. There one was using On-Delay set to 6 second and counter to 5. You would think the reset of On-Delay and increase a counter index is fast process.

I do not think its hardware issue (PC vs PLC hardware). I think its the implementation of the software simulator that is not accurate to the PLC hardware. Same instruction in the hardware could be simulated in software on PC with better performance.

Something tell me this is Java performance issue.

Doing the same thing in Python is less and 1ms to execute. The software simulator performance is not accurate to the hardware. I do think Siemens could do it better. And that's why I share this. Doing a simulator in Java maybe have some limitation I do not know.

You are of course right. If you could get a program written for PC/M- or DOS to run on a modern computer, it would be extremely fast. But with today's resource-hungry operating systems like Win11 & Co., that is largely put into perspective.

And you may also be a little right about Java. All the problems with Logo 8.4 and the current LSC only seemed to start with the switch to Java from AWS.

However, I don't really understand the point of a discussion about the different processing speeds between LSC and Logo. Ultimately, the processing speed of the Logo is the decisive factor.

And if the interaction of functions is time-critical, you program in a cycle-oriented manner anyway.

Regards Ella_68

I do understand that a simulator is not same as emulator. Maybe the purpose of LSC simulator is to simulate a execution and functions and not to do it with accurate performance compare to the hardware. Maybe that is enough for most cases and I do agree with that.

That said it doesn't explain why reset of timer and increase counter take 50ms. Performance is always important in software development and if we could increase performance we should. And sometimes when we do we find other things that matter for performance.

I was only surprised that reset timer and increased counter could take 50ms to execute in Java.

FredrikS7

Maybe the purpose of LSC simulator is to simulate a execution and functions and not to do it with accurate performance compare to the hardware. Maybe that is enough for most cases and I do agree with that.

...

That's exactly what it's designed for.

Greetings Ella_68