hi again, Ray here. (Not william, btw: that's my legal first name, but no one who knows me uses it, except maybe my parents when I've been bad 8^), Hygrometry is an inexact art at best. Among all the measurements that we take, only hygrometry and altitude remain so poorly understood by their users. Well, maybe I have to change that, since a) I haven't been a calibration technician for 30 years, and b) the advent of GPS systems adds two more dimensions to the altitude problem. Anyway. It is always good to remember the 10% rule, especially for those of us who tune pianos, and therefore are calibration and repair technicians without even knowing it. (Also fundamental to the world of metrology is the fact that you _measure_ something by _changing_ it. Temperature of the air is measured by changing the temperature of something else, the glass and liquid of a thermometer, the sensor in an electronic thermometer, etc.) The 10% rule expresses the difficulty of transferring a certifiably calibrated quantity from one device to another. Generally stated, it says that if you want to calibrate one device to magnitude 10^N, you'd better be doing it from a standard that is 10^(N-1) more accurate. An example would be tuning A=440 on the piano. We all feel that we can match a piano string's fundamental frequency to an A=440 source with flabberghasting accuracy... but that A=440 source may be far from 440Hz. Here's a short description of the trip that frequency standards make from NIST to your tuning fork: NIST determines what A=440 is from a set of rubidium clocks. To decide what their real standard is, they take 10 of them, throw out the extreme values, and average the rest. Then they take the closest clock and use that as The Standard, usually for The Day. (That means that it is possible for The Standard of The Day's frequency to vary from day to day, but fear not, these guys are good to a gross number of decimal places, and if the TV broadcasters don't sweat it, we sure don't have to at 440Hz!) This frequency is transferred to a frequency standard through a calibration process combined with an adjustment process, using a phase comparator which works exactly the same way the rotating displays on Accutuners do. The measurement is taken over many days, which increases the accuracy of the measurement, and after adjustment, another couple of days of measurement are taken. This frequency transfer standard is adjusted according to its measured aging rate: if it 'goes flat' with time, it is adjusted 'sharp', and vs. versa. Over the time of the certified cablibration cycle, the transfer standard ages, and its output frequency changes from one side of the actual claimed frequency to the other. Again, this is to many more decimal places of accuracy than we need. But the frequency transfer standard is certified to be only 1/10th as accurate as the rubidium standard of the day. Any further frequency standards which are calibrated against this standard will be 1/10th as accurate. And any standards calibrated against those third-generation standards will be 1/10th less accurate than their 'parent' standards. Few tuning fork manufacturers have a short direct line between NIST and their forks. (It'd be silly to bother, since that direct line is very expensive and waaaaaaa(insert another 23 'a's here)y too over accurate for their purpose. Instead, they purchase a standard that is (probably) a thousand times more accurate than they certify their forks for and which does have a certification which shows its pedigree all the way back to NIST. This means that it is probably good to six decimal places (I'm guessing here, because I have no direct experience with tuning fork manufacturers, just with time/frequency calibration), which may, in fact, be much more accurate than that (by accident or design) but is _certified_ to 10^-6Hz. They then tune their forks to 10^-3Hz (throwing away 1000x of accuracy, as it were) under specific conditions, which they sell to us branded as accurate to two decimals (10^-2Hz). We then tune a piano and claim (unscientifically) A=440, but probably think in our heads that it's going to be A=440.00. It isn't, because the fork isn't that accurate under changing conditions of temperature and pressure. So even though we probably have a good enough phase comparator system (beats, ears and brain) to match up the pitches to four or five decimals of Hz (although physiologists will swear it's a tenth at best), that's only a match to the temporally-changing state of the fork. So along comes electronics, and we think we're doing better, because after all, it's electronic. But unless a tuning machine has a quartz, oven-controlled clock to produce it's beginning standard frequency, from which everything else is divided out (and no, even computers don't have that) they're only 'better' because they start with a very high frequency and divide down. And that's for digital systems: analog systems (like my faithful old (and I do mean old) Hale sight-o-tuner, from long before SAT (-1) days) have even more ways they react to temperature and pressure and the phase of the moon. However, for frequency, most of this is well-defined, and Old Hat, I'm sure, for all of you. Hygrometry is a different story for most of us, though. We know that we can get relative humidity with a drybulb-wetbulb swing hygrometer, but may not have a clue (or want to) about the accuracy of the thermometers involved. But I'd bet no one uses a swing hygrometer anymore anyway. Instead, we use these neat electronic, digital hygrometers, and think that, because the specs say "3%" or "+/-3%" that we're getting accuracy or even a clue what we're reading. Unfortunately, all digital readouts have built-in problems in the last digit. You have no clue if a reading of 50.0 is actually 50.0000000 or 50.099999 (ie, effectively 50.1). This is a fundamental problem, but it's not a big one, because we intrinsically should distrust readings of 1/10th of a percent in hygrometry anyway! Further, the +/-3% spec doesn't tell you whether the readings will be the same amount off or a varying amount off from one end to the other. This is a common misconception in meters: that if they're one percent low at the left end of the scale, they're one percent off at the right side of the scale. Analog meters are always more accurate on the high (right) end of the scale than on the left side... but digital display meters can be as off as their twisted calibration curves show... or they might be very linearly off, or that line might be a nice smooth curve. All of this hides behind those reassuring numbers. But with Hygrometry, it's far worse than all that. First of all, Hygrometry relies on both temperature and humidity measurements. These two measurements might be built into the same structure, but each will have a built-in 'calibration curve' describing the inaccuracies at different values. They may coincide or cancel or do either and or both at various stages... And their manufacturers will claim that they are accurate despite changes of temperature and pressure. But they don't say a thing about how these pores might get filled by cigarette smoke (or wood smoke from homefires, oil from cooking, or perfumes and colognes). The swing hygrometer was the first indirect measure of relative humidity. Other physical forms involved human hairs or other moisture-sensitive physical bodies which contracted or expanded with moisture. These moved needles on bearings... and like tuning pins, those physical rotating joints suffered from friction. This is why people reading dials always are tapping them: to 'knock loose' the stiction in the bearings. But what are the 'modern' forms of hygrometry? They are usually thin-film silicon circuits, where an insulating layer is used to form a capacitor, and the moisture in the pores of the thin film structure determines the Rh reading. Unfortunately, reading out capacitors is also a complicated and inexact art. Oh, the science is all there. And using high frequencies and carefully built dividers and counters make accurate readings... but they're still capacitors. And the capacitance curves still have flat areas at one end, where differentiating one percentage reading from the next one up requires differentiating an extremely small change in capacitance. But the most fun part of Rh readings is that they are affected by so many things. The good old swing hygrometer required a specific period of swinging to get its measurement, and this averaged out all sorts of things...which these loverly new thin film circuits respond to quite readily. Here is an illustration of this: I was tasked to take Rh, temperature and air-movement measurements on a compartment. The actual application isn't important. In fact, the Rh measurements were to be made to +/-20%, temperature to +/-1degreeF, and air movement was just relative (fan on high, fan off, etc). I chose my sensors carefully, thermocouples with a proper reader for temperature, Omega duct-mount Rh sensors (the thermocouple reader had settings for reading out this sensor as well), and a Ford Escort motor airflow meter! I set it all up in my office...and discovered that the thermometric and Rh readings would vary as much as 10 degrees and 20%Rh depending on whether we were being visited by the secretary, or if my office-mate left and brought back a hot cuppa. (I left it running, and yes, I affected the readings by my presence and absence too. I'm not a vampire or anything!) While I took measurements on the compartment that discovered the problem we were after, I didn't mention that I could tell the person in charge whether the people who said they'd been in the compartment had been there when they said, and where they stood, when! Anyway, that means that it is likely that you are affecting your Rh readings by how you take them. And even if you leave your Rh meter at a distance from people, going to read it can well change the reading. All of which is to say that some of our measurement tools should be taken with a grain of salt, and some of them should be taken with Morton's Salt Mines. By the way, for information, 'calibration' means 'measuring accuracy of'. If you take your torque wrench to be calibrated, and they measure it's accuracy and hand it to you and say, "throw it out, it's inaccurate", they've calibrated it. If they take your Rh meter, measure it, and adjust it to match its readings with reality to some level of accuracy, that is 'calibration and adjustment'. I hope that sheds some light on the Rh discussion, and may I someday be able to speak as usefully about Piano repair! raybro
This PTG archive page provided courtesy of Moy Piano Service, LLC