posted February 12, 2007 12:43 PM            

In my experience, as long as the basics are correct (the mounting hole is machined correctly, and in good condition), the probe is mounted in the hole at proper torque, and the transducer is soaked at processing temperature, then calibration of the transducer to instrument is simply by setting zero (at zero imposed pressure) to read zero on the instrument, and span the instrument to 80% of full scale using either an internal (preferred, for temperature compensation) or external 30.1K shunt resistor in the R-CAL line.

If the probe cannot be calibrated to the instrument in this manner (either because it has actually been broken, or overstressed beyond 150% rated pressure, and 'stretched') then the probe must be replaced, but if the instrument can be zeroed and spanned the probe is good enough.

We're contemplating getting a NIST certified pressure probe tester and doing 10 point calibration curve checks, and I'm wondering if anybody else does this (or the equivalent; pulling probes on a regular interval, and sending them out for testing) as a matter of course?

posted February 12, 2007 12:53 PM            

I've never bothered becasue there is usually a huge amount of temperature drift in the pressure transducers which effects zero and span. As far as I can tell, there is no perfect design.

------------------
Best Regards,

Tom Cunningham

posted February 14, 2007 02:20 PM            

quote:

---

Originally posted by rawelk:
I'm wondering how far do most people go with melt pressure probe calibration?

In my experience, as long as the basics are correct (the mounting hole is machined correctly, and in good condition), the probe is mounted in the hole at proper torque, and the transducer is soaked at processing temperature, then calibration of the transducer to instrument is simply by setting zero (at zero imposed pressure) to read zero on the instrument, and span the instrument to 80% of full scale using either an internal (preferred, for temperature compensation) or external 30.1K shunt resistor in the R-CAL line.

If the probe cannot be calibrated to the instrument in this manner (either because it has actually been broken, or overstressed beyond 150% rated pressure, and 'stretched') then the probe must be replaced, but if the instrument can be zeroed and spanned the probe is good enough.

We're contemplating getting a NIST certified pressure probe tester and doing 10 point calibration curve checks, and I'm wondering if anybody else does this (or the equivalent; pulling probes on a regular interval, and sending them out for testing) as a matter of course?

---

posted February 14, 2007 07:08 PM            

------------------
Steve Hodgson

posted February 15, 2007 03:39 PM            

As far as instrument-to-probe calibration is concerned, I would go a bit farther than every six months (that would be my minimum), and recommend doing calibration during every setup, or, at least during any setup where the operating temperature has changed, for instance, when changing from PS to APET.

I can see where it would be useful to track the calibration date, probe's zone temperature, and "before and after'" values for zero offset and span gain on instruments so equipped. For instance, a Eurotherm 942S instrument has a zero offset term in PSI (or bar, if configured in this manner), and a gain term where 1.000 is unity gain.

During instrument calibration (when the zone has been soaked at operating temperature for at least 20 minutes, and with the extruder empty of material) the zero offset term is adjusted until the pressure display shows '0' PSI, then the 80% span is activated, and the gain term is adjusted until the display shows 80% of probe full span (4000 PSI for a 5000 PSI probe, or 8000 PSI for a 10K probe). Depending on the instrument (how much interaction there is between zero and span adjustment), and how far out of calibration the system was, this sequence may need to be repeated.

In my experience with this model controller most probes end up with a zero offset in the 50 to 200 pound range, and a gain multiplier of 0.920 to 0.960. The gain multiplier typically doesn't change by much, 0.002 or so, from one calibration to the next.

If the resulting zero offset pressure, and gain multiplier terms were tracked along with the zone temperature it would be possible to determine when a large shift occurs in the strain gauge element, and, in those cases, it may be desirable to replace the probe, and send the original out for testing and remanufacture. These details will be different for other instruments, but the same general concept applies.

A typical 0.5% accuracy probe (combined accuracy ... including hysteresis, linearity, and repeatability; I'll use the specs for a Dynisco PT460) has a diaphragm zero shift of 25 PSI/100°F, and a zero balance of +/-10% of FSO (full scale output). The electronics package also has a seperate set of ratings, in this case, a zero shift of +/- 0.05% full scale/°F, and sensitivity shift of 0.02% FS/°F.

In practice, as Tom pointed out, temperature effects tend to swamp out other forms of calibration shift unless the probe has been subject to overpressure, and suffered a permanent shift due to stretching the gage elements.

posted February 15, 2007 05:23 PM            

Over the past year, I've began switching out all of our Dynisco products and replacing them with Gefran ISI tranducers and pressure instrumentation. They are just as good as Dynisco and cost only $450.00. They also have a trade-in program that takes anyones used transducers and gives a percentage off on the purchase of a new model from them. They also have a tester available for testing transducers according to their local rep. Here is their contact information.

Gefran ISI Inc.

1-888-888-4474--Laura Noonan is the contact.

I highly recommend anyone looking at hteir products as I introduced them to the melt pump market when I was a Field Engineer with LCI Corp in the 1990s.

Give me a call sometime if you want to discuss.

quote:

---

Originally posted by rawelk:
Jeff, good to hear from you ... Ed mentioned you were back in the area.

As far as instrument-to-probe calibration is concerned, I would go a bit farther than every six months (that would be my minimum), and recommend doing calibration during every setup, or, at least during any setup where the operating temperature has changed, for instance, when changing from PS to APET.

I can see where it would be useful to track the calibration date, probe's zone temperature, and "before and after'" values for zero offset and span gain on instruments so equipped. For instance, a Eurotherm 942S instrument has a zero offset term in PSI (or bar, if configured in this manner), and a gain term where 1.000 is unity gain.

During instrument calibration (when the zone has been soaked at operating temperature for at least 20 minutes, and with the extruder empty of material) the zero offset term is adjusted until the pressure display shows '0' PSI, then the 80% span is activated, and the gain term is adjusted until the display shows 80% of probe full span (4000 PSI for a 5000 PSI probe, or 8000 PSI for a 10K probe). Depending on the instrument (how much interaction there is between zero and span adjustment), and how far out of calibration the system was, this sequence may need to be repeated.

In my experience with this model controller most probes end up with a zero offset in the 50 to 200 pound range, and a gain multiplier of 0.920 to 0.960. The gain multiplier typically doesn't change by much, 0.002 or so, from one calibration to the next.

If the resulting zero offset pressure, and gain multiplier terms were tracked along with the zone temperature it would be possible to determine when a large shift occurs in the strain gauge element, and, in those cases, it may be desirable to replace the probe, and send the original out for testing and remanufacture. These details will be different for other instruments, but the same general concept applies.

A typical 0.5% accuracy probe (combined accuracy ... including hysteresis, linearity, and repeatability; I'll use the specs for a Dynisco PT460) has a diaphragm zero shift of 25 PSI/100°F, and a zero balance of +/-10% of FSO (full scale output). The electronics package also has a seperate set of ratings, in this case, a zero shift of +/- 0.05% full scale/°F, and sensitivity shift of 0.02% FS/°F.

In practice, as Tom pointed out, temperature effects tend to swamp out other forms of calibration shift unless the probe has been subject to overpressure, and suffered a permanent shift due to stretching the gage elements.

If that is the case then installing a fresh probe, and sending the suspect unit back for complete testing, and (almost certainly) remanufacture would be indicated.

---

posted February 17, 2007 08:11 AM            

Thanks for the info. We've been using Gefran/ISI probes for replacement since before ISI became part of Gefran.

It is unclear to me, however, whether you are advocating pulling the probes, and performing testing in depth, or rather just doing a two point instrument calibration every six months.

posted February 27, 2007 12:31 PM            

quote:

---

Originally posted by rawelk:
Jeff, good to hear from you ... Ed mentioned you were back in the area.

As far as instrument-to-probe calibration is concerned, I would go a bit farther than every six months (that would be my minimum), and recommend doing calibration during every setup, or, at least during any setup where the operating temperature has changed, for instance, when changing from PS to APET.

I can see where it would be useful to track the calibration date, probe's zone temperature, and "before and after'" values for zero offset and span gain on instruments so equipped. For instance, a Eurotherm 942S instrument has a zero offset term in PSI (or bar, if configured in this manner), and a gain term where 1.000 is unity gain.

During instrument calibration (when the zone has been soaked at operating temperature for at least 20 minutes, and with the extruder empty of material) the zero offset term is adjusted until the pressure display shows '0' PSI, then the 80% span is activated, and the gain term is adjusted until the display shows 80% of probe full span (4000 PSI for a 5000 PSI probe, or 8000 PSI for a 10K probe). Depending on the instrument (how much interaction there is between zero and span adjustment), and how far out of calibration the system was, this sequence may need to be repeated.

In my experience with this model controller most probes end up with a zero offset in the 50 to 200 pound range, and a gain multiplier of 0.920 to 0.960. The gain multiplier typically doesn't change by much, 0.002 or so, from one calibration to the next.

If the resulting zero offset pressure, and gain multiplier terms were tracked along with the zone temperature it would be possible to determine when a large shift occurs in the strain gauge element, and, in those cases, it may be desirable to replace the probe, and send the original out for testing and remanufacture. These details will be different for other instruments, but the same general concept applies.

A typical 0.5% accuracy probe (combined accuracy ... including hysteresis, linearity, and repeatability; I'll use the specs for a Dynisco PT460) has a diaphragm zero shift of 25 PSI/100°F, and a zero balance of +/-10% of FSO (full scale output). The electronics package also has a seperate set of ratings, in this case, a zero shift of +/- 0.05% full scale/°F, and sensitivity shift of 0.02% FS/°F.

In practice, as Tom pointed out, temperature effects tend to swamp out other forms of calibration shift unless the probe has been subject to overpressure, and suffered a permanent shift due to stretching the gage elements.

If that is the case then installing a fresh probe, and sending the suspect unit back for complete testing, and (almost certainly) remanufacture would be indicated.

---

I have worked quite a bit with the 942s controller and have found the same results. The only reason I can think of the calibrate the transducer specifically is to test for linearity of output.

The Eurotherm 2704 is different, in that it has an auto calibration funtion......

posted February 28, 2007 07:53 PM            

quote:

---

I have worked quite a bit with the 942s controller and have found the same results. The only reason I can think of the calibrate the transducer specifically is to test for linearity of output.

The Eurotherm 2704 is different, in that it has an auto calibration funtion......

---

The 942S user interface was nice in that they put the zero offset and span gain on the same menu screen. I haven't played around enough with 2704s yet to be sure, but I'm guessing Eurotherm still allows users to configure instrument UI menu access levels, and (if this is true) it'll still be possible to fairly easily track zero and span.

Don't ask me how I do it (my mind is set up to operate in a more procedural rather than a rote memory fashion; I keep my eyes open, and try to tune in on what the developer was driving at in regard to how the UI should operate, if you follow my highly parenthetical drift), but at access level 3 there is a calibration submenu that walks through several parameters. These include at least the zero offset, and, if memory serves, a term representing the span (although it's not a gain multiplier term as in the 942S) that have been generated by the auto-calibration routine.

posted March 27, 2007 03:06 PM            

quote:

---

Originally posted by rawelk:
The 942S user interface was nice in that they put the zero offset and span gain on the same menu screen. I haven't played around enough with 2704s yet to be sure, but I'm guessing Eurotherm still allows users to configure instrument UI menu access levels, and (if this is true) it'll still be possible to fairly easily track zero and span.

Don't ask me how I do it (my mind is set up to operate in a more procedural rather than a rote memory fashion; I keep my eyes open, and try to tune in on what the developer was driving at in regard to how the UI should operate, if you follow my highly parenthetical drift), but at access level 3 there is a calibration submenu that walks through several parameters. These include at least the zero offset, and, if memory serves, a term representing the span (although it's not a gain multiplier term as in the 942S) that have been generated by the auto-calibration routine.

Maybe next time I get one hooked up to a computer it'll be a good idea to move these parameters down to security level 1 (if possible) for easier access.

---

2704 has multiple inputs, so you can, for instance, display melt temp, barrel pressure, die pressure, and pick your control loop. Also has auto-calibration routine for transducer inputs.

Nifty instrument package.

posted April 14, 2007 10:24 PM            

quote:

---

2704 has multiple inputs, so you can, for instance, display melt temp, barrel pressure, die pressure, and pick your control loop. Also has auto-calibration routine for transducer inputs.

---

I was impressed, and especially liked that they tamed the 'setpoint up/down' software routine ... on the 942S one needed to be very careful when setpoint values were approaching different 'decade' points. More than once I shut down a machine, or caused a process upset once a setpoint started blazing through '100's or '1000's.

posted April 26, 2007 10:38 PM            

quote:

---

... Why you would want to calibrate the pressure transducer?

---

I don't want to calibrate the transducer itself. To my knowledge, this is a process suited only to the manufacturer, or similarly qualified shop, and involves laser trimming the bridge resistance elements themselves. However, a pressure probe must be calibrated to the instrumentation so the raw mV output from the bridge is scaled and displayed in terms of engineering units such as PSI.

These days this is a relatively simple process discussed in greater detail earlier in this thread, although, before manufacturers had the capability to easily trim the bridge elements they would often mark the probe label with full span pressure output. For instance, a 10,000 PSI probe, (which is spanned to 8,000 PSI) would have a value like "7940", "8120", etc. showing the probe's actual full span scaled output.

quote:

---

... Are you in need of an exact pressure measurement, or just a variation?

Do you need an EXACT pressure reading or just the changes as they occur?

Why you need an exact pressure reading - wouldn't a pressure change be of greater interest?

---

I'm interpreting these as different phasings to equivalent questions. 'Exact' is a question-begging term, but I'm shooting for probe/instrumentation system accuracy within their combined error limits.

Of course, short term variation is the most immediately usable criteria, and can be assessed even when the accuracy of the system is poor. For instance, a 600 PSI head pressure rise in 2 hours is useful to know, regardless of whether the system started out showing a pressure of 1400 PSI, or 2000 PSI.

However, it is important to maintain system accuracy for at least two reasons - to make sense of longer term variation, and for proper limit operation.

In the latter case, for instance, say we have 7500 PSI rupture disks, don't normally encouter head pressure of greater than 4000 PSI, and have set the head pressure shutdown limit to 5800 PSI ... a nice, comfortable value in between the two. If the probe/instrument calibration is allowed to go way out of whack we'll end up either shutting down for no good reason, or, if a real problem exists, and calibration has drifted higher, blowing rupture disks instead of shutting off the extruder.

In terms of the first reason (longer term variation) - we maintain setup sheets which record head pressure for 'clean screen', and 'time to change screen'. If system accuracy is allowed to drift too much over time then these values are of minimal utility.

[This message has been edited by rawelk (edited April 26, 2007).]

posted April 27, 2007 08:54 AM            

System accuracy within the combined error limits is a great goal, but what will it ad to your process? For example, some in the industry “calibrate” thermocouples, and then check them in the system that they are used. Typically, those calibration accuracies are very tight, tighter by several factors of the actual temperature requirements from batch to batch, or processing rage within a batch.

You mention that a 600 PSI head pressure rise in 2 hours is useful to know, regardless of the starting pressure, but I would think that any process that showed that kind of pressure rise would also exhibit size or product quality variations downstream. Assuming of course, that a closed loop system to control output is not in place.

I certainly can see that attaching a probe designed for an instrument with a different “range” can indeed “scale” the output such that the readings may not provide “safe” shutdown, but the instruments and probes are easily identified as to which goes with which.

When I was running materials that required screens, we never changed them on time, just product quality, and of course, cost to run. Slowing down a line can result in substantially higher costs, but they may be outweighed by the time to change screens, re-string and re-size; when that was the case, we opted to complete the run.

But back on topic, the question was “what do most people do for melt probe calibration. Our short answer is we do not calibrate. But now I am wondering how many do.

posted April 29, 2007 12:39 AM            

quote:

---

System accuracy within the combined error limits is a great goal, but what will it ad to your process?

---

Lets say the combined measurement error for the system is 0.7% (or 70# for a 10K probe). This isn't too difficult to achieve. Once the error gets to 2% we're talking real numbers (200#), and, for long term purposes such as maintaining accurate setup data is edging upon excessive.

I should mention, however, that I'm not the proponent of using a pressure probe calibrator (such as a Dynisco PSC2000 or the heated block variation, the PPS1200), and asked the question to see if anyone else here does much more than transducer-to-instrument calibration.

My gut feeling is this level of calibration is sufficient for our needs, and, although it would be interesting to use a probe calibrator (I love to dig deep, and see if what I think I know bears any resemblance to reality) it isn't worth the investment.

From my perspective, for such a tool to have any benefit you'd need to check each probe on at least a yearly basis, keep meticulous records, and (after several years of records have been gathered) you might find it useful in a predictive maintenance role.

It certainly would be useful in a forensic sense, but even then, my gut sense tells me it wouldn't tell us anything substantially more than the 'go/no-go' test (a probe can't be calibrated to the instrument, so either throw it away, or send it for remanufacture), and opens us up to potentially more headaches (each time a probe is removed is an opportunity to screw up the pressure port threads).

In short, from a cost/benefit standpoint, I don't see a need to go farther than probe-to-instrument calibration, although I do think this level of calibration is important and necessary.

[This message has been edited by rawelk (edited April 29, 2007).]

posted May 16, 2007 10:17 PM            

I can see where this would be useful for probes built with "PLC compatible" outputs (i.e. - 0-10 VDC, and 0-20 mA) - the integrator wouldn't need to add programming for probe zeroing - but am having a harder time visualizing how it benefits mV/V output probes.