Tuners: accuracy vs. precision...

[ecguitar44]

Originally Posted by HHP

If the variation is too great and is unpredictable, calibration is not possible.

Not necessarily. Depending on the scenario...and the cause of the inaccuracy...you may be able to adjust (calibrate) the system to increase precision (decrease variation).

I'm not much of a shooter (shootist?) but perhaps altering the rifling or changing the muzzle velocity could decrease the variation. But you're correct that you can't increase precision necessarily by tweaking your scope or sights.

That's why the "target" example isn't the best representation of what's really going on...and you should instead reference my probability distribution curve.

I guess I'm coming at this from an engineering point of view...which doesn't necessarily jive with colloquial uses of the terms. However, I will insist that my usage of the terms is correct...even if I'm not doing the best job in explaining.

[ecguitar44]

Originally Posted by HHP

In the case given, by your definition, the "accurate" thermometer results inevitably in a 10 block radius being lost. Each of its measurements is closer to the known value, but its erratic readings cannot be calibrated or adjusted so that you have any certainty of the reading except in a range of about 10 degrees.

I don't mean to beat a dead horse...but...

Your example of thermometer #2, as was said, can be "averaged" to get close to the actual temperature. The average of your temperatures was 209. Theoretically, if you take 100 measurements...it will be even closer to 207.

This is exactly how measurement instruments work (digital thermometers, speedometers, digital volt meters, etc.). They take a bunch of reading (very quickly) then average them to get as close to the real reading as possible.

Same reason your digital scale at home takes a few seconds to pop up a reading.

[HHP]

Originally Posted by ecguitar44

I don't mean to beat a dead horse...but...

Your example of thermometer #2, as was said, can be "averaged" to get close to the actual temperature. The average of your temperatures was 209. Theoretically, if you take 100 measurements...it will be even closer to 207.

This is exactly how measurement instruments work (digital thermometers, speedometers, digital volt meters, etc.). They take a bunch of reading (very quickly) then average them to get as close to the real reading as possible.

Same reason your digital scale at home takes a few seconds to pop up a reading.

Does averaging work when you know you have an absolute to deal with such as the temperature at which the chemical explodes. You are OK on one side of the variation and vaporized on the other.

In the case of the guitar tuner, you would seem to be working to match an absolute like A440.

[ecguitar44]

Originally Posted by HHP

Does averaging work when you know you have an absolute to deal with such as the temperature at which the chemical explodes. You are OK on one side of the variation and vaporized on the other.

In the case of the guitar tuner, you would seem to be working to match an absolute like A440.

Yes, averaging works in that scenario. It's actually even MORE necessary. However, the overall monitoring system is much more sophisticated. In those cases, the thermometer would be HIGHLY accurate...and HIGHLY precise.

The averaging would be done to average out noise in the wires, temperature differentials from thermocouple effects from dissimilar metals touching on the circuit board, and random electrical noise (browning motion, for example) in the system. All in an effort to be as accurate as possible when measuring. We're talking in 100ths or 1000ths of a degree.

So, in reality...NEITHER of your thermometers would be chosen.

At the end of the day, the choice of your measurement device depends on your use case. Both accuracy and precision (and many other factors) are important. The choice is seldom "one or the other".

Full disclosure here, I'm an electrical engineer in the "test and measurement" industry. I work for a company that designs measurement instruments.

[Bucc5207]

Originally Posted by HHP

In the case given, by your definition, the "accurate" thermometer results inevitably in a 10 block radius being lost. Each of its measurements is closer to the known value, but its erratic readings cannot be calibrated or adjusted so that you have any certainty of the reading except in a range of about 10 degrees.

Tuners typically have accuracy to "XX cents" AND a calibration adjustment to zero it against a known value. If the variation is too great and is unpredictable, calibration is not possible.

I agree in part with what you say. Precision and accuracy always involve tradeoffs with decisions made to fit the intended application. A very accurate but very imprecise thermometer would not be useful for the explosives application. That is because it would take too much time to make the repeated measurements necessary to achieve the required precision. "Just 100 more readings, and we'll have ..." BOOM! The same thermometer might be an outstanding tool for calibrating other thermometers in a standards lab, where accuracy is paramount and time is not an issue.

Calibration is always possible, but it can't produce a better result than the precision of the instrument allows. If a tuner can read frequency to within 10 cents, then you will not be able to 'zero' it against a standard (e.g., a tuning fork) better than that. No matter how carefully you calibrate it, next time you check it against the tuning fork, it will read somewhere between 437.5 and 442.5 Hz (10 cents being about 2.5 Hz at 440 Hz, and ignoring the finer statistical points about confidence levels and such). That's a built-in limitation of the particular device. If it's not good enough for your purposes, you need to get a better (meaning more precise) device.

ETA: I missed ecguitar44's posts while I was writing, but I agree with his comments. In terms of qualifications, I am both a shooter and a former professional scientist.

[Long Jon]

I've got a Snark. It's OK.

[johnny196775]

I tune to a piano most the time.

[drplayer]

Originally Posted by ecguitar44

Not necessarily. Depending on the scenario...and the cause of the inaccuracy...you may be able to adjust (calibrate) the system to increase precision (decrease variation).

I'm not much of a shooter (shootist?) but perhaps altering the rifling or changing the muzzle velocity could decrease the variation. But you're correct that you can't increase precision necessarily by tweaking your scope or sights.

That's why the "target" example isn't the best representation of what's really going on...and you should instead reference my probability distribution curve.

I guess I'm coming at this from an engineering point of view...which doesn't necessarily jive with colloquial uses of the terms. However, I will insist that my usage of the terms is correct...even if I'm not doing the best job in explaining.

I'm with you ecg, except that at some point improving precision--although possible--becomes increasingly more difficult, to the point of diminishing returns. The amount of variation is irrelevant to accuracy in the traditional sense, as statistically, accuracy (or lack thereof) is generally a measure of the sample mean (or population mean in rare cases) as compared to the target value (i.e. bullseye). Variation on the other hand, has a direct affect on precision...so that even if the sample mean is identical to the target value--in essence making it perfectly "accurate"--the sample can still be imprecise. "Precision" simply defines the amount of variation (usually in standard deviation units); whereby the terms "precise" or "imprecise" are relative terms based on what is being measured and the allowable amount of variation that can be tolerated. Building contractors use tape measures, which are precise enough for their use; however, is you're making watch parts...not so much. Nothing, absolutely nothing, is "perfectly" precise, so we just establish an amount of variation that we can live with...

[ecguitar44]

Originally Posted by drplayer

I'm with you ecg, except that at some point improving precision--although possible--becomes increasingly more difficult, to the point of diminishing returns. The amount of variation is irrelevant to accuracy in the traditional sense, as statistically, accuracy (or lack thereof) is generally a measure of the sample mean (or population mean in rare cases) as compared to the target value (i.e. bullseye). Variation on the other hand, has a direct affect on precision...so that even if the sample mean is identical to the target value--in essence making it perfectly "accurate"--the sample can still be imprecise. "Precision" simply defines the amount of variation (usually in standard deviation units); whereby the terms "precise" or "imprecise" are relative terms based on what is being measured and the allowable amount of variation that can be tolerated. Building contractors use tape measures, which are precise enough for their use; however, is you're making watch parts...not so much. Nothing, absolutely nothing, is "perfectly" precise, so we just establish an amount of variation that we can live with...

Indeed!

I'm trying to strike a balance between helping others without overburdening the discussion with too much statistics, theory, or general nerdery.

P.S. "Nerdery" was initially autocorrected to "berserk" on my iPad.

P.P.S. I haven't a clue what the original point was of this thread.

[Long Jon]

Sorry, Should have read : I've got a Snark. It's OK. . . wabi-sabi

[Long813]

Originally Posted by ecguitar44

Yes, averaging works in that scenario. It's actually even MORE necessary. However, the overall monitoring system is much more sophisticated. In those cases, the thermometer would be HIGHLY accurate...and HIGHLY precise.

The averaging would be done to average out noise in the wires, temperature differentials from thermocouple effects from dissimilar metals touching on the circuit board, and random electrical noise (browning motion, for example) in the system. All in an effort to be as accurate as possible when measuring. We're talking in 100ths or 1000ths of a degree.

So, in reality...NEITHER of your thermometers would be chosen.

At the end of the day, the choice of your measurement device depends on your use case. Both accuracy and precision (and many other factors) are important. The choice is seldom "one or the other".

Full disclosure here, I'm an electrical engineer in the "test and measurement" industry. I work for a company that designs measurement instruments.

Don't forget that we also include a safety factor, which would be greater than the deviation in the example. It would be quite a rare case in which you would need to be operating just below a failure point - or in this case a exploding point.

[drplayer]

Originally Posted by ecguitar44

P.S. "Nerdery" was initially autocorrected to "berserk" on my iPad.

As it should have been

Originally Posted by ecguitar44

P.P.S. I haven't a clue what the original point was of this thread.

Something about getting a tuner that is both ultimately accurate and ultimately precise, so that the one person in 10,000 that is capable of noticing it won't leave the room when you play...

[ecguitar44]

Originally Posted by Long813

Don't forget that we also include a safety factor, which would be greater than the deviation in the example. It would be quite a rare case in which you would need to be operating just below a failure point - or in this case a exploding point.

Sure...uncertainty analyses, guard banding, system redundancy, etc.

I didn't want to point out that the thought experiment was absurd in any real scenario. Fundamentally the points/discussion are applicable.

I was actually chuckling to myself imagining a scenario where a system is within 1 degree of exploding and it's being monitored by a dime-store mercury thermometer by some schmo in a lab coat!

"Looks like about one-nineteen point... BOOM!!!"

[philjs]

Originally Posted by drplayer

Something about getting a tuner that is both ultimately accurate and ultimately precise, so that the one person in 10,000 that is capable of noticing it won't leave the room when you play...

Actually, it was to make people think twice about tossing around the word accuracy when they mean precision (or vice versa). Tuners should be both accurate AND precise...a tuner can be accurate but readings can vary considerably around a target frequency unless it's also precise!

Phil

[Landru]

Yeah - but one tuner has only one bullet, so why bring precision into the convo? Did you mean six tuners = six bullets? The illustration is precise but not an accurate representation.