Ultra High Speed PLC with Analog input

[mamo]

Update...

I talked with NI and they suggested the NI cRIO-9073.
http://sine.ni.com/nips/cds/view/p/lang/en/nid/205621

The high speed analog input module, digital output module, and chassis is about $1500. The software is $7,000. Which puts this option just out of reach. I am going to pursue the OPAMP idea which Peter suggested.

Thanks everyone again for your help.

 

[kamenges]

You might also want to look at products from Tern, Z-World, or Mosaic. I've personally never used these but they look like they might do the trick for you. At least they are much less restrictive than a plc will be.

In looking at your drawing I'm not sure why the dS/dt of the shallow hole would necessarily be less magnitude than that of the deep hole. I would expect both hole flanks to have the same slope so they should produce the same dS/dt, especially with an op amp. You may be able to trick a digital system into acting this way since you have a discrete sample time. All you can really do is a per-scan difference. But it may be tough to get an analog system to act the same way unless you do some pre-filtering.

Keith

 

[milldrone]

Mamo,

You may have allowed for this in your 450Hz spec, but maybe not. If the "pulse" of the hole is significantly shorter than the non hole reading, you may have to up your Hz spec. I'm presuming that you counted the holes per second and not double the hole width as a cycle.

 

[mamo]

In looking at your drawing I'm not sure why the dS/dt of the shallow hole would necessarily be less magnitude than that of the deep hole. I would expect both hole flanks to have the same slope so they should produce the same dS/dt, especially with an op amp.

Keith

The change in S from 1 sample to the next on a shallow hole should be far less than the change in S from 1 sample to the next on a deep hole. I may not have drawn it perfectly but the both good and bad holes have a straight down drop off as a side wall after an initial chamfer.

Mamo,

You may have allowed for this in your 450Hz spec, but maybe not. If the "pulse" of the hole is significantly shorter than the non hole reading, you may have to up your Hz spec. I'm presuming that you counted the holes per second and not double the hole width as a cycle.

In fact the hole should be visible to the sensor for about 3ms which should ensure that at least one reading will catch the hole.

 

[ndzied1]

I'm not sure if they are fast enough but you may want to look into trying a pneumatic proximity switch with a pressure wave switch.

Something like this:
http://www.clippard.com/catalog/Page 155.pdf

I have been trying to find a mtg I used to know about that had a more precision version of this type of sensor.

The idea would be you would get a bigger swing in pressure change over the defect hole than the good hole.

Again not sure it would work but might be worth looking into...

 

[kamenges]

Originally posted bu mamo:

The change in S from 1 sample to the next on a shallow hole should be far less than the change in S from 1 sample to the next on a deep hole.

That is correct. But that isn't the same as the derivative of the measurement, especially in the analog world. Remember, there is a change in t that corresponds to the change in S. And in a straight hole, that change in t will be extremely small. Small enough that you will probably bump into the slew rate of the amplifiers both in the output of the sensor and in the device you will construct. At that point both holes will look the same.

But as I said, you should be able to look at it simply as a digital delta. This is really what I think you were shooting for. All you know digitally is that the value changed from one point to another and it change somehow across the duration of the update. It may have changed instantaneously or it may have changed in a ramped fashion. You really don't know. But you do know how much it changed. And that is really all you care about.

So this is less or a derivative and more of an incremental measurement. And that is a much easier proposition since there is almost no processing to it. You just need to be able to read quick enough.

As I said, I would look to one of the little CPU boards for this. It should be easy to program and it should be cheap.

Keith

 

[mamo]

Thanks Kamenges,

I think you are right that incremental will be much simpler and will accomplish the same.

 

[Paul T]

The round hole is 2-3mm in diameter and is drilled into 60-90m/min multilayer pipe and must only puncture the top layer. It must be 2-3mm deep. if it is 10mm deep then i want to sound an alarm. But for robustness when using different pipe diameters rather than use the actual distances I want to only look at slope. That way I don't need any adjustment if I am looking at 32mm diameter pipe or 20mm diameter pipe.

Mamo,

You may have allowed for this in your 450Hz spec, but maybe not. If the "pulse" of the hole is significantly shorter than the non hole reading, you may have to up your Hz spec. I'm presuming that you counted the holes per second and not double the hole width as a cycle.

OK... I'm confused (g!).

90 m/min = 90,000 mm/min = 1500 mm/sec

Wow. Flying, that. So, a 2mm diameter hole is going to pass the laser in 1.3 ms. This seems to be more like 770Hz, not 450Hz. Or do smaller holes run at slower speeds? Even so, a 3mm hole is going to go by in 2 ms, so not a lot of help there.

Either way, the laser itself has to process quickly enough to "notice" the hole. The laser, and it's internal electronics probably can. The sensor's interface to the outside world probably can't - check that part with the vendor. Sure, the "laser" can run at a pretty high bandwidth. But how about the the 4-20mA output? What are the specs on that?

Even given that the laser has an analog output, it probably can't slew quickly enough to keep up, and why bother anyway? If your criteria is that 3mm is OK, but 10mm is too deep, then there have got to be much cheaper laser sensors that can measure a distance and set a discrete output if the return signal is out of range. Just setting the discrete output is going to take around 0.5 ms to 1 ms, even if the laser sensor itself can take and process multiple measurements every millisecond.

Also, how often does a hole come along to be measured, and what do you do about it if a bad hole is found? How do you keep track of the flaws, and mark or cut or whatever downstream? That stuff is likely to be harder than finding the flaw, especially given that your good/bad criteria is so well delineated (3mm vs. 10mm).

 

[ndzied1]

Here is a sensor with a 1kHz response on its discrete output...

Balluff BOD 6K

 

[dash]

With the op-amp idea and if this is fairly dedicated and you may produce more than one you may want to look at a micro-controller like NetBurner. They have a nice realtime OS (based on micro-C OS II), IDE and cost effective options that are industrial rated.

www.netburner.com

Darren

 

[Homer_BL]

No need to use analog input at PLC. Use OMRON E3X series optical amplifier with adjustable detection treshold. With E3X you can go 80us respones time. You need also to select proper head for amplifier (E32-L series). See catalogues.

 

[whirlair]

Maybe I don't understand what you are doing, but I see no reason to analyze slope.

I think there is a control logix analog card with a 300µsec update rate, but I have never used it so I have no idea how the processor uses the data.

 

[Peter Nachtwey]

How about a $2000 motion controller?

It would be cheaper than any of the suggestions provided if you count the time spent designing the op-amp board. This would required 3 op amps anyway. One for converting the signal to the derivative. One to do a peak hold on the derivative and this may require another resistor for the reset. One more op-amp is required for the comparator to convert the peak rate to a simple 1 or 0. Time will be required to order the parts, bread board and test the design, layout a small board. Stuff it and test it. That sounds like $2000 worth of work to me. Oh, and then I forgot documentation so the next guy knows what the heck is going on. Even then it isn't programmable.

So how much time and money are budgeted for this project?

 

[Longbow]

I am searching for the fastest possible way to analyse an analog signal with an output frequency of 450Hz. I need to evaluate the slope and give a discrete output if the slope of the analog signal is in a specific range.

Is there a PLC which can do the above without missing any data due to scan times?
Is there an alternative analog signal analyzer which can do it and give discrete output that would stay on long enough for a slower PLC to see?

It was mentioned already:

25µs (40kHz) analog sample time, 250µs (4 kHz) PLC Cycle:

VIPA Speed7 CPU with Fast Analog In Speedbus with 8 Channels and FIFO Buffer
(<2100 Euro)

 

[mer]

This may be what your looking for???

http://www.automation.siemens.com/m...0/function-modules/FM352-5/Pages/Default.aspx