How are digital and analog timers calibrated?

Subject: Calibrating Stop Watches
Date: Sat, 18 Sep 1999 16:45:47 -0400
From: "Morrill S. Reynolds III"
Organization: Genzyme Corp.

Hello Marc,

I just stumbled across your website and one of your topics was calibrating stopwatches. We are looking at buying a device that measures the frequency of the crystal oscillator inside. There are no eye hand coordination errors etc. You simply place the timer/stopwatch on the device and it determines the number of seconds per wek or month it is off. It will work with quartz and mechanical type watches. The device model number is a Q Test 6000 and it is made by Witschi Electronics. They may be reached at 1-800-882-7977. We are currently testing the watches and timers we have by comparing them to a frequency counter. This device will save us a considerable amount of time. I just thought you may be intersted in it

Morrill Reynolds III

I just read your post on calibrating timers and stopwatches. There are a few different methods to do this. The simpler method I have used involves first running an experiment with numerous iterations between the operator and the standard wherein you start the stopwatch and standard and the same time and stop the stopwatch and standard when the stopwatch reads exactly 10 seconds. Repeat this process at least 25 times (I did it 100 times for about 10 seconds each). Record the data and calculate a maximum deviation between readings. Add this to your measurement uncertainty as operator error. For a standard, use either a time mark generator (such as used for an oscilloscope calibrator) or other suitably accurate signal source. Use the start stop function on a frequency counter (I used a H-P 5335A counter). Set the repetition rate of the source to either 1 KHz, 10 KHz, 100 KHz or 1 MHz. Reset the counter readout to zero. Press start on the counter at the same time as pressing the start on the stopwatch. Let them run for an adequate amount of time (do uncretainty caluclations for the combined uncertainty of the counter, the signal source and the operator uncertainty), and allow enough time to minimize the impact of operator uncertainty. Be sure you get 4 to 1 or greater Test Uncertainty Ratio. When you stop both, be sure to look at the stop watch display to try to stop at a nominal (or close). Record readings, calculate deviations.

I do NOT recommend measuring the timebase in a stopwatch. You will get very mixed results, and MAY damage the stopwatch.

There is anohter method that I can look up if needed. Some nuclear industry contractors make up test fixtures that start the stopwatch and the standard at the same time. I have not been able to locate one of these.

You can email me at rn0414@email.sps.mot.com for further questions.

Having completed a search for "Timers" and thinking that some of the responses may be overkill, I would like to present the following from our Coating procedure for Pressure Vessels (Tanks), big enough to work in.

Gel Bake-275 Deg. F. 40 to 50 minutes....Depending on vessel configuration. Longer for vessels with false bottoms and/or portsights

Final Bake-360 Deg. F. 110 to 140 minutes...Depending on vessel configuration. Longer for vessels with false bottoms and/or portsights

We have (2) Honewell Digital Temperature Indicators to monitor and set temperature and a Process Timer that turns off the heat to the oven at the time you set it for. It has been proposed that we install a Bi-Metallic Thermometer at the point where the thermocouples are located to compare the meters. To satisfy a potential calibration inquiry. The thermometer would be in 1 or 2 deg increments, calibrated by an outside lab. We cannot afford to lose the time to get the equipment calibrated by an outside lab. They also do not want to purchase extra equipment to install during the time it would take for calibration outside. The timer would be checked against the electric clock nearby or somebody's timepiece. Don't laugh. The reasons given is that the time for baking and the varying temperatures are estimates arrived at through years of performing this process.
Can an Auditor question the methods proposed with such varying baking times and temperatures? The time would be +/- 2 minutes and the temperature would be +/- 5 deg.

Take it easy on me. :ko: :smokin:

energy said:

The timer would be checked against the electric clock nearby or somebody's timepiece. Don't laugh. The reasons given is that the time for baking and the varying temperatures are estimates arrived at through years of performing this process.
Can an Auditor question the methods proposed with such varying baking times and temperatures? The time would be +/- 2 minutes and the temperature would be +/- 5 deg.

Take it easy on me. :ko: :smokin:

Here is a TRACEABLE, FREE alternative, and you can do it yourself! First, go to
http://ts.nist.gov/ts/htdocs/230/235/105-5.pdf and download the NIST standard for stopwatches. Then, ftp to time-b.nist.gov/pub/daytime/ and get nistime-32bit.exe and download this handy little app that will set your computer clock to the NIST UTC timeserver (make sure you do it a few times to reduce the effects of 'net delay). Start your timer, run it 3 hours, stop it, repeat. Make sure you run the NISTime app before the stop and start for full traceability. There you go, no counters, only a computer connected to the net. Your uncertainty will be about 0.2 sec, with a tolerance (at least) ten times greater, so you'll be golden.

For your temp, I wouldn't use a bimetal, but there are some very reasonably priced digital thermometers (<$80) with a k-type thermocouple included. It is cheaper to have a digital thermometer calibrated, and they are much more accurate, so you're better off all around.

Ryan

Ryan,

Great response. I'll let you know how it goes.:bigwave: :smokin:

Energy

Can you verify the coating after processing anyway ?

i.e. do you measure thickness, adhesion, etc ?

We measure thickness with an Instrument designed to measure coating thickness on ferrous materials. In addition, we apply apprx. 20,000 volts through a wire "Fan" in the interior lining to eliminate the possibility of having pin holes (holidays). This can occur if there is some porosity in the weld. As for adhesion, no. We know it adheres because we never had rejection due to peeling. There is a five year warranty on the lining and we have equipment that was coated in a previous life-same process-that has been in the field over 20 years. We are anticipating an Auditor witnessing the usage of instruments without any signs of Calibration or status. You know how you guys are!:p :ko: :smokin:

Energy

It could be argued that if you can verify the output of the process with calibrated equipment then you need not verify the parameters of the process itself with calibrated equipment.

For example we do not 'calibrate' in the strictest sense of the word the lead screws on our machine tools, or the dials on the slides, or the display of motor rpm, etc - what we do is measure the actual piece produced.

With such a large band on your baking times I would imagine that its unnecessary to calibrate a timer, any wristwatch should do the trick (verified with the bloke stood next to you if you wish).

P.S. Despite what Jim thinks good auditors would always consider the context in which the equipment is used in consideration of the effectiveness of the system in an internal audit - not just 'oh its measuring equipment it must be calibrated'.

Reminds be of that thread back along about calibration of a PC !!!

M Greenaway said:

Energy

It could be argued that if you can verify the output of the process with calibrated equipment then you need not verify the parameters of the process itself with calibrated equipment.

With such a large band on your baking times I would imagine that its unnecessary to calibrate a timer, any wristwatch should do the trick (verified with the bloke stood next to you if you wish).



That approach is what I would hope would satisfy an Auditor looking at 7.6 Control of Monitoring and measuring devices. Maybe it's my experience in anticipating questions that have caused problems in the past. Anytime you list a value in your procedure, it had to be measurable. If it doesn't matter, leave it out. Unless, like you said, there was a Final Acceptance Test that covered any sub-processes. We only verify that the lining is "Spark-Free" and the required thickness is there. Even the thickness tolerances are broad. For example: 20 to 40 or 40 to 60 mils thickness is the requirement. So how do we know that it is not under-baked/over-baked? (Fully cured). Only by adherence to the process which specifies values-no matter how broad, can we guarantee a consistant product. I just don't think there is any way around it. I wouldn't equate this with the calibration of computers. I suspect that you, being an Auditor Emeritus, would look at these three instruments and question their usage with an eye on what is their importance. Not seeing calibration stickers or instrument status on them is an open invitation for futher digging.

I once had a Gov't Inspector doing Final Acceptance Inspection on a sophisticated piece of electrical equipment. He tapped the face of each instrument used as test equipment. He asked, "Do you know why I do this?" Not overly impressed with him at first blush, I said "Not a clue." "Well, a few years ago I found a meter with the glass missing." You can only imagine what this highly skilled Quality Representative would say about a meter with no calibration status on it. Even instruments incidental to the process had to be labelled "No Calibration Required". The equipment could have been built upside down and he would not have noticed. But, he had the power/Final say. One has to be prepared to meet this this type of individual and ready to explain why things are not the way he/she imagines them to be. I'm not comfortable with allowing that opportunity. Heck, I may open the door to design validation or some other silly thing like that!:vfunny: :ko: :smokin:

Regarding verifying the accuracy of a process oven timer:

ISO 9001:2000 section 7.6 states: "Where necessary to ensure valid results, measuring equipment shall a) be calibrated or verified at specified intervals, or prior to use, against measurement standards traceable to international or national standards..."

A wall clock or personal wristwatch would not meet this requirement.

Verifying the accuracy of process equipment, if in fact it is used to provide evidence of conformity to quality specifications, should be a formal, controlled process. Measuring and Test Equipment used for such activities should be in the plant's calibration control system. The fact that the process tolerance is broad does not justify using an uncontrolled, untraceable measurement standard to calibrate/verify the process equipment.

Stopwatches are inexpensive, and can easily be calibrated against the NIST time signal available by calling (303) 499-7111.

Norm said:
Stopwatches are inexpensive, and can easily be calibrated against the NIST time signal available by calling (303) 499-7111.

My real concern is with the Temperature Meters which also control the temp settings. You set it. The Oven responds. The meter reads the temp through thermocouples. We want to breach the oven wall with a calibrated "comparison" instrument, located in the immediate proximity of the thermocouples. I want to use a thermometer of the same increments to make sure the other instruments are close. Yes, I said close. We are not requiring +/-.05 Deg. More like within +/- 5 degrees. Not critical to the process to have an exact temp. Certain vessels contain thicker or more metal and take longer/less to arrive at the curing temperature range. We know by holding the vessel +/- 20 minutes at the temperature range of +/-5 deg is sufficient to complete final cure. I wouldn't classify them the same as temperature needed to actuate, say a temperature switch. We want those @ +/- 1 deg, in our industry. :ko: :smokin:

Energy,

Installing a calibrated thermometer is fine if that is the one you will use to verify the quality of your product/process. But merely using it to get a warm fuzzy that the temperature controller is "close" is not appropriate. Why not use the thermometer to actually calibrate the controller?

Be aware that if your process tolerance is +/- 5 degrees, the temperature controller should be four times more accurate, or +/- 1.25 degrees (albeit 4:1 is sometimes difficult to achieve in temperature calibration; 2:1 is more common). The instrument you use to calibrate the controller should then be four (or at least two) times more accurate than the controller.

I agree with Ryan that a digital thermometer rather than a bimetal should be used as the "standard" by which you judge the performance of the controller. The rated accuracy of most bimetals is +/- 1% of span. For a thermometer in the range of your process, say a 0 to 400 degree model, that would be +/- 4 degrees. With a process tolerance of +/- 5 degrees, a bimetal would be inadequate for even monitoring the process, much less for calibrating/verifying the controller, even, if I understood you correctly, you would only be looking for the two instruments to agree with +/- 5 degrees. The bimetal could read say, 4 degrees low, and the controller read 5 degrees lower, and everything would appear hunky-dory to you. In reality, your process could be 9 degrees off the setpoint, almost twice your process tolerance.

Norm

Let me start by apologizing for coming in almost a month after the last post - Energy's problem may have been resolved by now - but all I can do is plead the super-busy excuse. (8a to 6p at the office, then 8p to midnite at home, blah blah blah) :frust: But on to the issues at hand ...

Temperature measurements: when you are using thermocouples, always remember that there are two parts (at least!) to the measurement system - the meter and the thermocouple. Both parts have to be calibrated if you want to know how hot it is, even to "only" +/-5 degrees. There are many manufacturers with fantastic digital thermometer specs - readouts with resolution of 0.01 degree, and accuracy of +/-(0.05% of reading + 0.3 deg.C) and so on. BUT, when you connect that super-whizz-bang meter to an uncalibrated plain old type K (for example) thermocouple that you just whipped up, your system uncertainty instantly goes to the thermocouple spec of +/-4 degrees Fahrenheit. (Provided I am remembering the ASTM tables correctly ... they are at the daytime office.) In most cases, the accuracy of the digital meter is trivial when compared to the thermocouple. On the other hand thermocouples are cheap, and pretty reliable when used correctly. For "better" accuracy over a limited temperature range, thermistors might be another choice to consider. For "best" accuracy of course (and kilobuck co$t$) you can use a standard PRT system ... but it's overkill for Energy's needs.

(By the way, calibrating those "super high accuracy" digital thermometers is a real pain. I don't know of a shipping DC microvolt volt thermocouple calibrator that can be automated and has the required uncertainty. [Does anyone else?] That means going back to the old-fasioned way using high-end DC calibrators, and ice point, and reference junctions - manual & slow.)

Time Interval measurements: (as opposed to time of day). I have some alternate opinions about some of the things mentioned. First, Energy is using a controller (as I understand it) that is set for a time interval. The actual time of day is irrelevant - the only consideration is the duration of a period of elapsed time. That is, what is the uncertainty associated with measuring an interval of 120 minutes?

An AC electric wall clock that uses a shaded-pole synchronous motor (most do) is electrically locked to the AC power line frequency. In North America, that frequency is 60 Hz +/-0.02 Hz. That makes a pretty good secondary time interval standard, actually. Over 120 minutes the maximum possible error is 2.4 seconds, and that is only in the unlikely event that the power line frequency is all the way to one limit for the entire time. For Energy's purposes, it seems that a couple of seconds here and there is trivial.
There is an easy way to get NIST-traceable time interval, and time of day traceable to the US Naval Obervatory (where NIST gets it from!) for $50 or less. Go to your neighborhood consumer electronics store and get one of those "atomic" clocks. Get one with a signal quality indicator. Provided you can put it near an exterior wall, preferably oriented towards Denver, it will pick up and lock to a digital subcarrier broadcast by WWV radio and you are traceable as long as its signal indicator is OK. Check its data sheet for it's basic unlocked "digital clock mode" drift, and how many times per day it updates to WWV. Simple math will give you the maximum possible error. On the one I have it is much less than the 1-second resolution, so it's irrelevant. AND it is the only clock in the house I do not have to adjust for Daylight Savings Time! It is easier than the computer thing and you don't have to make a long-distance phone call to Boulder, Colorado.
I believe I have a random memory that average human response time for actuating a stopwatch is something less than 0.1 second. When a long time interval is measured that quickly becomes irrelevant. It is 0.1% at 100 seconds, and in the time period Energy is talking about it is about 14 ppm for on + off response. The response time is important only when things are happening real close together, such as at the finish of a horse race -- that's why they use fully electronic timers now. Once again for Energy's purposes, that response time factor may be irrelevant -- assuming his process does not end in the same way as a horse race! :) The main factor of concern is the accuracy of the chronometer (a fancy word for a time interval measuring device) over the period you are measuring. Norm is correct in quoting the ISO 9001 standard, but I think it is possible to demonstrate that any timer - even a personal digital watch with a chronometer function - can meet this requirement provided you document the process of checking it against a suitable standard. For instance, start the timer while noting the indicated time of day on an electric wall clock. Some period later, stop it while noting the time of day again. Determine the difference in seconds, and see if it meets your needs.


Notice that I have been attempting to differentiate between time interval and time of day. They are related only by the definition of the second.

Time interval is the reciprocal of frequency, and is defined by physics - a particular vibration of some highly agitated Cesium atoms. NIST disseminates time interval and frequency through the WWV radio stations. It is also correlated with the GPS system, because every one of those satellites has a couple of Cesium primary frequency standards on board.
Time of day is determined by astronomy - measuring the rotation of the planet on its axis and its orbit around the Sun. That is done by the US Naval Observatory (and equivalent observatories in other countries.) The counting interval for time of day is the second (see above) but the number of seconds in a year varies. (Ever hear of "leap seconds"?) Time of day is disseminated by the USNO, and one way they do that it by the time codes and voice announcements on WWV. It is also available from the GPS system, and by telephone and over the Internet from the observatory.

I assume the job thing has squared itself away. Great post. So now I have to worry about the thermocouples themselves. Oh Boy. Can the meter and the thermocouples be calibrated as a set?
Sorry about missing your post. I was occupied elswhere. :confused: :ko: :smokin:

energy said:
Can the meter and the thermocouples be calibrated as a set?


Yes, they can be calibrated together, and that can be a good thing to have done especially if you are using them over a limited temperature range. If a cal lab can do the thermocouples, they should be able to match them to a specific meter. Your problem afterwards is one of controlling two items as if they are one. But be careful, as there are a number of labs (such as mine) that can do the METERS by DC voltage substitution but cannot do the thermocouples.

It may be possible to match them yourself, if your meter has a control and if you are using the thermocouple over a very limited range. Some meters, such as the older Fluke 51 & 52, have an operator "offset" adjustment on the front that is intended for that purpose. (But is often mis-used!) You need to have a "known" temperature reference that is independently verified, and is close to your process temperature. Put the thermocouple in the temperature reference, let everything stabilize, and then adjust the control to make the meter match your reference thermometer. This will match THAT meter and THAT thermocouple at THAT temperature ONLY. (or fairly close to that temperature.)

If you use a wide range of process temperatures, it may be worthwhile to obtain a good quality dry-well temperature calibrator for the shop or toolroom. I have seen some with two wells so you can have two points set up. (I can give you references offline.) For example, one well can be set to the ice point (an inter-universal reference for practically everything) and the other adjusted to different process temperatures as needed. Your cal lab will need a standard platinum resistance thermometer to calibrate it, but it only takes a few minutes if it is done on-site.


BTW, I checked my ASTM manual on thermocouples. For a new standard grade K thermocouple (a very common type) the tolerance is +/-2.2 degrees C or +/-0.75% of the output, whichever is greater, when used above 0 degrees C. If you are using Fahrenheit, you have to calculate the tolerance in C and then convert to F. Also, thermocouples do not improve with age and use. Just about everything you do to them (other than letting them sit on a shelf) helps them deteriorate. They should definitely be treated as a consumable item, not a fixed asset.


Graeme