Does my bathroom scale use a trick to appear more accurate, by recognizing that a weight is close to the last one measured and displaying it as exactly the same, so that if you keep getting onto it again and again you will not see any variation?
Here are my recent daily weights:
188.6
188.6
185.2
(missing)
185.4
185.4
187.2
188.6
188.6
184.8
184.8
184.8
184.8
184.8
It seems obvious to me that there are too many daily repeats given the large changes that sometimes occur. The daily differences have what I think is an implausible distribution.
Why on earth would it do that?
If I weigh myself, and then weigh my daughter, the last thing I’d want is for the machine to tell me that she weighs the same as me!
Similarly, if I was on a diet, I wouldn’t want the thing to tell me that I wasn’t losing weight, even if I in fact was.
Your machine might be broken, or your weight really may be as it was measured, but I can’t think of any reason why the machine would fake it.
It seems more likely that repeated ‘exercise’ of the load mechanism might have rendered it more nearly linear in response (by dislodging bits of dirt and grease in the mechanism or some such)
Try to fool it by weighing yourself, then weighing yourself again with a book in your hand. By your theory, the second weighing should register the same. Otherwise, you have easily disproven your theory.
The balance is not engaging in any “funny business.” But Mangetout has a point: certain values may be unstable due to mechanical or electrical irregularities. Do a sensitivity measurement (as detailed by The Controvert) to validate proper response.
Looks like the scale’s only measures to ± 0.2 pounds.
I work for a company that sells bath scales (and other products), and I do believe that some scales will repeat the same weight if you step on multiple times. These scales simply couldn’t be accurate enough to record the same value every single time (we’re talking about inexpensive bathroom scales, here); I suspect that some of them simply store the value for a brief period of time, and if a similar weight appears, they show the original value. This isn’t necessarily to deceive the customer, though. I’d suspect the issue is that if people see that their scale is recording slightly different values each time, they’ll get overly worked up about it. Most people who buy a bathroom scale don’t realize that accurate scales are expensive, precision instruments, and an inexpensive bathroom scale isn’t going to be the most accurate measurement device. Therefore, I suspect its done (if it is – I have no cites, unfortunately) because it simply makes sense to do so, to satisfy the consumer.
It doesn’t make sense, because it would cost the scale manufacturer more money. Extra wires, extra juice from the battery, a small amount of non-volatile memory, all for a questionable benefit. The “mechanical or electrical irregularities” explanation seems more plausible.
Yes, use something that you know the weight of–a handful of coins or a cup of water. In a small plastic cup, 4 fluid ounces should give you about .25 pounds. I’m betting a dollar to your donut it works.
Wait, I’m going to go try mine first.
Actually it probably wouldn’t cost them any extra money. I’ve never taken apart a bathroom scale, but I can’t imagine that it’s anything more than a load cell connected to a microcontroller that drives the display. I doubt that the microcontroller is fully utilized, so adding an extra “featuire” would only involve a bit of programming. However, I can’t imagine that they would do anything like that. Their goal is to make a machine that is as accurate as possible while being as cheap as possible. Fudging the numbers would make it less accurate, which goes against what they are trying to do.
There might however be a “mechanical irregularity”, as was previously mentioned, that makes the machine stick when it is close to the last weight in the table. Your actual weight might be varying by a couple of pounds, but the machine always reads it the same because the load cell tends to get stuck on that value and won’t read a higher one until a bit more weight is added. I wouldn’t trust a bathroom scale to be accurate to the nearest 0.1 percent, even though it display indicates that it is. As long as it is within a couple of percent it’s probably good enough for them. Remember that part of their goal is to make it cheap. If you want something calibrated to NIST standards it’s going to cost you a lot more.
OK, tried it. My scale only goes down to the .5 pound. I got on and off a few times, with a half cup of water. Most of the time, it worked (I lost a half pound when I didn’t have the cup, gained it when I did), but once in a while it would jump up a half pound. I suppose I could have been right at x.25 pound, and quantum jiggling could have done that. Sigh. More experiments…
- Yes, the scale has always appeared to read out in 0.2 pound increments, though the instructions don’t say what its resolution is supposed to be.
- Of course I tried holding various weights. If they are very much different, 5 lbs or so, it registers different results. So you and your daughter aren’t going to weigh the same unless you’re so close you wouldn’t know anyway. However when I try smaller weights I observe confusing behavior that I have not figured out - it will repeat the value a few times, then jump to a new value for a few times, or jump randomly around. I haven’t figured out how to guess what I am going to observe next.
- It won’t cost any more wires or juice or memory - there’s a microcontroller in there, and it obviously remembers things because it uses a rezeroing from last time when you step onto it (there’s no poking it first like my last one), and rezeroes itself when you get off (for the next time). These microcontrollers are pretty cheap places to implement the sort of treacherous scheme I am describing.
- I don’t see the bit of dirt in the mechanism idea working. It’s not like those old fashioned scales with gears and whatnot inside - there’s a bit of metal that flexes, probably flexing too little to see by eye, and it’s got resistors glued to its sides so that their values shift a little when strained, and there’s a battery and a microcontroller and a display (and I bet the microcontroller and display are integrated together to save another fifty cents).
- They are not trying to make a machine as accurate as possible and as cheap as possible. What they are trying to make is money. If those two goals coincide, hooray for accuracy, but if not, we’re on our own.
- My previous scale, which read out to 0.5 pounds, was also suspiciously good at repeating itself, though it wasn’t as obvious because the increments were bigger anyway.
The way to make the scale most accurate for the cost includes, among other things, displaying more digits than are accurate, because if the display limits the accuracy then you’re throwing away something more expensive than display digits! In other words it wastes money to have the display be the limiting factor because they’re cheaper to upgrade than the other parts. But you can bet people would complain if the last digit kept varying moment to moment.
I think occ got it right, but it’d be grand to hear some more specific evidence.
Wait, here’s more:
Remember that microprocessor-based scheme a few years ago with gas pumps?
The pump makers knew that 5-gallon and 10-gallon standard containers are used to randomly check gas pumps for accuracy, but they wanted to cheat consumers, so they had the pumps start at 0 gallons, then gradually overrepresent the amount pumped by more and more till they got to 2.5 gallons, then they got closer and closer again till they were correct at 5 gallons, and continued this pattern so that integer multiples of 5 gallons would always be correctly metered but odd integer multiples of 5/2 gallons would be optimally overrepresented.
At least, this was the gist of the scheme - I might misremember what the standard containers were, but they were definitely choosing when to look accurate and when to cheat so as to avoid detection.
So a scale maker getting tricky with microprocessors - which after all would be an easy thing to keep secret and a hard one to prove - isn’t all that unbelievable, is it?
My weight can drop 2-3 pounds overnight. Some through urine, some through perspiration and exhaled water vapor, and some through burning of calories just in sleeping.
The difference before and after using the toilet is often a pound.
So I don’t think your numbers are all that believeable to begin with. You’d have to be a lot more regular than most.
My comment about 'dirt, grease or some such) was something of an oversimplification and it is rather difficult to elaborate, but I believe it is the case that materials that have been recently subjected to repeated flexing (albeit perhaps only microscopically) are more likely to perform in a way that is (closer to) the ideal than those that have not.
There’s also things like thermal/capacitive/inductive equilibrium being reached after the electronic components have been powered up for a while.
I’m still not explaining myself properly.
Oh, there’s also the small chance that radiated heat from the soles of your feet is adding something to the system.
if you want to empirically test it, it would be better to exclude yourself from the loop(as it is impossible for you to truly stand still); get a sack of potatoes, weigh it repeatedly, then take a few potatoes out, or add a few and weigh again. etc…
“thermal/capacitive/inductive equilibrium” ???
“radiated heat from the soles of your feet” ???
“is impossible for you to truly stand still” - well, sure, it is now…
You can’t stand still unless you have been the first-hand recipient of a taxidermist’s services. It really isn’t possible.
OK, I think I have them now. After 5 days in a row of 184.8, this morning I measured 182.0 (I am dieting so it should be dropping). Then I tried it holding a 0.5 pound weight and got 182.0. Then without weight, 182.0, and with, 182.0, and without, 182.0, and with, 182.0.
This can’t possibly demonstrate anything but that they have a fake accuracy cheating scheme! Can it? What alternative is still plausible???
Now try a two pound weight
A little information on scale theory:
The inaccuracies you are seeing probably come from the fact that the measurement is analog and the electronics are digital.
One of the determinates of the accuracy of a digital scale is the number of weight segments that the scale range is divided into. 5000 is a typical value for a mid-range accuracy industrial scale. This means that if a scale using a division of 5000 is to weigh loads from 0 to 10,000 pounds, you won’t get a precision greater than 2 pounds. Depending on the electronics, it may or may not try to represent a higher precision. If such a scale shows a weight of 1042.7, the 0.7 is not reliable. 1042 is.
The other way such a scale could display numbers is to only increment them when a change has reached the threshold of accuracy, when another of the 5000 divisions clicks, so if adding pounds to a scale reading 1042, you won’t see any change until the scale displays 1044.
I think there are two problems in your bathroom scale. The first is that there aren’t many weight divisions – more divisions cost more money, plus who besides you and scale geeks are going to notice? These divisions plus the range it was programmed for may have an accuracy of 0.2 lb, or more likely the programming was designed to register in 0.2 lb increments, even though this is not the accuracy of the scale.
The other problem is the conversion of the analog signal from the load cell to a digital increment. The load cell should give a nice repeatable signal given different weights. However, the D/A converter could be taking an overly wide range of miliamps and lumping them into the same scale increment, thus giving you the same weight, even though your weight is changing.
So, I’d say that the decimal point is giving a false sense of the precision of the scale, and the D/A converter due to cheap electronics may also be taking a larger range of signals and funneling them into a single scale increment, giving an inaccurate and repeating reading.