What is a safe scan time ?

[Jeff23spl]

Hello

I know it would depend of the applicaton but to the older peoples here (That was in the programming when plc was slower than now) what is your usual safe scantime limit to respect and why ?

I personnaly try to keep it average under 100ms and use the right plc size for the application but sometimes we are asked to do bad things or explaining why we refuse to do it....

I have to control protections on a constant running machine and i have a special application that would rise it near to 4-500ms and i'm a bit afraid to do so...

i never experienced a very slow scan time to know well what it cause and how it would react... at which level it will begin to give issues etc...

 

[Binaural]

I've had problems with very high-speed machines with as low as 10ms, but that would probably be rare. Once, I was designing a small processing machine with a basic processor, and could not get the damn thing to run to rate. I upgraded the CPU, and the blasted thing lifted its production rate by 3 pieces per hour, reaching the target just like that. Don't underestimate a quick scan time as a potential source of machine speed increases, basically.

The slowest I've ever seen was a DCS-controlled system which had an effective update time of around 5s. This was a disaster for the process they were using it for.

 

[ndzied1]

If you are worried about safety I have to assume you are worried about hurting people. In this case, a standard PLC is the wrong tool for the job.

First wave of attack: Eliminate unsafe situations by removing the hazard or using hard guarding (removable only with use of a tool)

Then look to locked guards with interlocks like power to unlock or hardwired zero speed circuits to unlock. You can also use timers to unlock.

Then you get to the devices which allow access such as safety mats, pull ropes, light curtains, area scanners. These all must be applied properly or they are worse than not having any safety.

As far as the PLC scan time is concerned, it should only be required to make the parts and not have a role in the safety decision making.

 

[KalleOlsen]

In process systems we use PCS7 (DCS?) where the controllers scans (default setting) at 100ms. It is usually OK, but f.x. rotary monitors must be taken specially care for.

In a machine control this scan frequency would be faaaar too slow.

Kalle

 

[tomalbright]

My first PLC was an Allen Bradley PLC-3. The scan time ran around one second or more, just because. I think the shutdown was at 1.5 seconds? After we found out it wouldn't run timers at that speed, we duplicated the timers for sampling purposes, and everything ran peachy.

It all depends on what you need. Opening and closing valves, running motors, and monitoring alarms is all we needed. ESD worked like a champ! There's some processes where a fast scan time is needed, but with newer processors like CLX, where the IO is always available, it really might not matter much. You have to watch PID timing and things like that, but after the PLC-3, PLC-2, and I *think* we had a PLC-? that was even older, everything is lightning fast!

Work with reason, not with speed.

 

[BobB]

1/2 the length of the shortest input pulse to make sure you get it.
Apart from that i do not like anything over 20ms - if longer use a more powerful processor.
I grew with PLCs from the very early nineties and often stuggled to optimise code to get scan times down.
One old trick was to use a counter for a compare - it used 1 less mnemonic instruction than the compare function. Plenty of others of course but no need these days - just buy a faster processor.

 

[Jeff23spl]

i said ,protection, but isn't related to human stuff but to protect a machine from self destruction if used beyond it's limits...

It control valves, speed and pression but no motion or high speed positionner involved...

I want your

 

[Clay B.]

Scan time is one of those things I only worry about on really fast projecs. That said I worked for a Can manufactor and we had to have scans so fast on some PL5/15 (yes I am that old) that we acually would calculate the time for certain functions and create code to get the job done in 20 ms or less. It was about as much fun as poking yourself in the eye with a blunt stick.

On alot of equipment I have worked on 100 ms scan time would be a bit much. I see 50 ms+ I start to think about what I may need to do to speed things up.

 

[Tom Jenkins]

Good question, so the right answer is "It depends".

I've had some process control apps with a lot of math so the scan time was pushing 500 ms. However, the process was slow and most of my logic was done intermittently at 5 second to 5 minute time intervals, so the 500 ms was just fine. As others have pointed out, if you are doing motion control or have high speed pulses then 50 ms might be too long.

 

[dskohio]

I don't like it if my go over 15 to 20 ms on high speed machines.

 

[daba]

To me, there is no concept of a scan time being "too long", providing that the machine or process is "under control".

It is definitely a "horses-for-courses" thing, and I have seen lame systems with 2 to 3 second scan times that performed their system tasks admirably. Conversely, I have seen the fastest systems around struggle to cope with simple tasks such as counting flow-meter pulses !

It's all about using features of the system you are dealing with, handling fast inputs with Interrupts, or even Data-Change event code. Many systems offer these facilities specifically to overcome "normal" processor scanning.

20 years ago I worked on PLC3 systems that scanned their code about 1.3 seconds. More recently, they have been replaced with ControlLogix that scanned even more code in around 200mS. The PLC3's were not replaced because they were "too slow".

 

[MichaelG]

depends -

I get closely monitor the scan time with motion and try to make sure that it does not exceed 2 to 3 x the motion group coarse update time.
without motion I'll go up to 50ms

I try to time slice code execution if the scan time gets large
eg say 200 PID instructions (controlling Temperature so long time constants) that have a 400 ms cycle time (each PID must execute every 400 ms)
Put them into a Event task that runs every 100 ms but only executes 50 PID's each call, this reduces the "Blocking" effect that an event task can have on the other tasks

I will also do similar stuff for sorting and large calculations