If I am standing on a digital scale or measuring the weight of my luggage, the display doesn't show the weight immediately - it takes a few fractions of a second.
Why is that and can you overcome it to always, immediately show the current weight (e.g. weighing the exact weight of a continuously filling cup)?
Answer
Well, let me take the digital body weight to explain why it takes some time, usually less than two seconds to get the results.
The balance usually has four load cells at four corners, each of which connects to a HX711 weight sensor with a sort of Wheatstone bridge measuring voltage.
When you step on the balance (or placing your luggage on the balance), you don't place yourself in a good balance, spreading evenly over the four corners. (For luggage balance, there might be many more load cells.)
The host controller would wait for the slow human's four readings to settle down, and use some simple algorithms (spatial/time moving average etc) to try to "balance" the unbalanced/unevenly spread readings.
In short, it is the unsteady human body, or the "unsteadyily placed" luggage that take time to settle down. The HX711 sensor itself is 24 bit accurate, and only takes 0.1 second to do the analog to digital conversion.
References
(1) Hacking a body weight scale - rpi.org.forum 2018dec16
(2) HX711 weight sensor - Rpi.org.forum 2018nov17
(3) HX711 24-Bit Analog-to-Digital Converter (ADC) for Weigh Scales Datasheet - Avia Semiconductor
(5) Mini Pocket Digital Scale for Gold Silver Jewelry 100g-0.01/200g-0.01/500g-0.01 - AliExpress US$5
Appendices
Appendix A - HX711 weight sensor and load cell
Appendix B - Hacking a body weight scale
Appendix C - HX711 Evaluation Notes
Appendix D - Accuracy and Precision Measurement Results
Wiki says the following:
Accuracy in a set of measurements, is closeness of the measurements to a specific value
Using 50g standard weight as the bench mark, I found the closeness is max 50.00g and min 49.96g. So the closeness is
(50.00g - 49.96g) / 50.00g
= 0.06 / 50.00 * 100%
= 0.12%
For 10g standard, accuracy
= (10.01 - 10.00) / 10.00
= (0.01 / 10.00) * 100%
= 0.1%
Wiki also says the following:
Precision is the closeness of the measurements to each other.
With a very limited sample size of only two 100g/0.01g weight scales, I also found the closeness is about 0.1g, so the precision is also 0.1%.
Of course the cheapy US$5 weight scale's 0.1% accuracy and precision is far from HX711's 24bit accuracy/precision. However for everyday casual use with 6 decimal digits LCD display, 0.1% is good enough.
For higher accuracy and precision, we can use HX711 to connect to the digital scale's torn down load cell, and get higher accuracy and precision.
Reliability, repeatability, and availability casually means the same criteria. In this quick and dirty 100g weight scale tests, all three are about 0.1%
Appendix E - 100g/0.1g / 500g/0.1g Load cell interface with HX711
Appendix F - HX711 and 100g/0.1g Load Cell Interface
Appendix G - Weight scale's response time
For body weight scales, the measurement is is of the order of one or more seconds, because the human body standing on the scale is not steady, so the controller might need to wait for the weight to become steady or the two feet to settle down, more evenly placed, taking time moving averages, spatial redistribution of the 4 load cells etc.
For the mini 100g scale, I try to use my hand to disturb the sample weight and found the response time is only less than half a second. This explains why body weight scale needs more waiting time.
Not an electronic thing, probably, but simply the fact that anything being moved and stood up on a surface has a mass and an elastic volume, so that the force it excerts downwards isn't constant until oscillations have stopped. Mechanical scales suffer the same.
Other than that, if you want a low-variance measurement, you need to filter your observed electrical quantity (a resistance, in this case) with a low bandwidth. But for the weights and accuracies we're talking about here, this is irrelevant.
I would say that the delay mostly comes from the math algorithm used by the microcontroller rather than from the force sensor or force oscillation. Judging by how smoothly the indicated weight changes, I would assume the scale displays some sort of running average spanning a few actual instant measurements, with measurements taken 2-10 times per sec.
May I suggest a simple experiment to rule out the "unsteady body" hypothesis? Try pressing the panel of your scale with your hand (or both), keep pressing it for a while and then remove the hand instantly. How long does your scale take to realize the weight is zero? It took 4-5 indicator update cycles (a bit more than a second) for my household scale to drop back from 50 kg to 0. I'm pretty sure such long delay can't be due to neither oscillation nor elasticity of the glass panel.
Weight is measured by the extension/compression of a spring (in this circumstance). A mass-spring system oscillates with a period dependent on the spring constant and the mass; the initial amplitude depends on the initial deflection, e.g., stepping onto the scales. Then there is an exponential decay of the amplitude of the oscillation with respect to time.
Once the oscillations have reduced in amplitude sufficiently, which takes some time, a fairly accurate value can be given.
