Author Topic: mg scale - interesting effect  (Read 5215 times)

0 Members and 1 Guest are viewing this topic.

Offline coppercone2

  • Super Contributor
  • ***
  • Posts: 11333
  • Country: us
  • $
Re: mg scale - interesting effect
« Reply #25 on: March 26, 2019, 03:27:57 am »
what material is best to handle those polished weights? like the expensive E or whatever grade ones. (well the small ones anyway, that you might not want to handle with gloves

Ceramic? Polished Steel? Ivory? Some kind of felt lining?
 

Offline JBealeTopic starter

  • Frequent Contributor
  • **
  • Posts: 314
Re: mg scale - interesting effect
« Reply #26 on: March 26, 2019, 07:32:54 am »
I now use metal tweezers to handle my ultracheap weight set. I don't know how to really minimize mass transfer, but I think that's micrograms or below, not few-milligrams like this scale resolves.

I did another set of measurements, this time I used two different "20g" weights A and B. The A unit was about 8 mg heavier than the B unit.
I weighed A,  back to zero, then B, then A&B together, then removed B to measure A again (shown as "A2"), then returned to zero. I did all this 10 times with results shown below.

You can see there is some load cell nonlinearity as A&B together always measure slightly higher than the sum of their separate weights.
On average it was 6 mg more, which as a fraction of the total weight is 0.015% so it's small but measurable. The scale does not offer multi-point calibration AFAIK, but if you have a reference mass of similar value, you can mostly calibrate this away.

Code: [Select]
200 g x 0.001g no-name ebay scale		25-March-2019		

trial mass mass both mass A2-A1 nonlinearity
# A1 B1 A&B     A2      drift  (A&B)-(A1+B1)
1 20.029 20.021 40.052 20.026 -0.003 0.002
2 20.027 20.021 40.055 20.03 0.003 0.007
3 20.029 20.019 40.052 20.028 -0.001 0.004
4 20.029 20.02 40.052 20.028 -0.001 0.003
5 20.027 20.022 40.056 20.03 0.003 0.007
6 20.029 20.02 40.057 20.03 0.001 0.008
7 20.028 20.018 40.051 20.025 -0.003 0.005
8 20.027 20.022 40.057 20.03 0.003 0.008
9 20.028 20.021 40.055 20.03 0.002 0.006
10 20.026 20.02 40.054 20.03 0.004 0.008

AVG 20.0279 20.0204 40.0541 20.0287 0.0008 0.0058
MIN 20.026 20.018 40.051 20.025 -0.003 0.002
MAX 20.029 20.022 40.057 20.03 0.004 0.008
Range 0.003 0.004 0.006 0.005 0.007 0.006
« Last Edit: March 26, 2019, 03:26:03 pm by JBeale »
 

Offline Conrad Hoffman

  • Super Contributor
  • ***
  • Posts: 2077
  • Country: us
    • The Messy Basement
Re: mg scale - interesting effect
« Reply #27 on: March 26, 2019, 12:37:07 pm »
I use a single 50 gram weight at work that was about $800. The tweezers that came with that are useless and would cause you to drop the weight. I handle it with Kimwipes and have considered making a plastic fork that would slip under the head. The tweezer problem seems to be traditional as my little Ohaus weight set at home came with similarly useless metal tweezers about 50 years ago.
 

Offline ebastler

  • Super Contributor
  • ***
  • Posts: 7375
  • Country: de
Re: mg scale - interesting effect
« Reply #28 on: March 26, 2019, 12:50:30 pm »
For 50G a decent American scale would have about 3-4 times more material in the Part Being Weighed and the Load Scale. It looks like junk unless the industry changed (doubtful). It also uses the chassis as stiffener. The chassis of a real scale is wobbly compared to the 'anvil'. Your talking a heavy glass plastic part not some little sheet of whatever and some dull zinc assembly.

How would that help the precision? I should think that a lightweight and stiff internal mechanism (i.e. high internal mechanical resonance frequencies), combined with a heavy weighing table with dampened, soft suspension (to suppress any floor-coupled vibrations above a very low resonance frequency) would be desirable?
 

Offline JBealeTopic starter

  • Frequent Contributor
  • **
  • Posts: 314
Re: mg scale - interesting effect
« Reply #29 on: March 26, 2019, 03:46:02 pm »
Typical vibration can be mostly low-pass filtered out but I think the concern is static deflection. I think the load cell wants a perfectly rigid base, and any flexure in the mounting adds an additional shear, torque and/or levelling change to the loadcell leading to a non-ideal response.  The thinner the mounting surface is, the more it will flex under load. I don't know about plates or webbed structures but IIRC the deflection of a rectangular beam is proportional to the inverse cube of the thickness, so twice as thick = 8x less deflection.
« Last Edit: March 26, 2019, 03:50:03 pm by JBeale »
 
The following users thanked this post: ebastler

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: mg scale - interesting effect
« Reply #30 on: March 26, 2019, 06:04:43 pm »
Load cells are constructed with a geometry to make them insensitive to side loads and such.  This is especially a problem for those intended for platform scales and they may be mechanically or electronically trimmed to provide the proper response under all conditions.

This is easy enough to test for yourself; what is the reading with the same weight place at different and extreme points on the platform?
 
The following users thanked this post: ebastler

Offline JBealeTopic starter

  • Frequent Contributor
  • **
  • Posts: 314
Re: mg scale - interesting effect
« Reply #31 on: March 26, 2019, 08:13:32 pm »
Load cells are constructed with a geometry to make them insensitive to side loads and such.  This is especially a problem for those intended for platform scales and they may be mechanically or electronically trimmed to provide the proper response under all conditions.
This is easy enough to test for yourself; what is the reading with the same weight place at different and extreme points on the platform?
My "100g" test mass reads on average 100.098 at the front edge of the scale, 100.104 in the middle and 100.110 at the back edge running through the front-middle-back sequence four times, with a repeatability of 2 mg at each location.  So the effect is small but it is above the noise level.  When I did my previous experiments I tried to get the masses as close to center as possible.
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 16385
  • Country: za
Re: mg scale - interesting effect
« Reply #32 on: March 26, 2019, 11:34:17 pm »
Those eBay mass sets are reasonable, the one I got went through a verification traceable to the SI kilogram, and actually was well within the required specifications for accuracy. The Ohaus set is quite stable, and has been that way for decades as well, because they only get handled with plastic tweezers for the small masspieces (and those have a measured mass down into the nanogram range for the foil pieces, and the 1g is known to 4 decimals as 1.0000g, with an error of 80ug) and the larger masspieces I have a set of cotton gloves and cloths used to wipe them before use, and the gloves for handling them. They are stable as well, because they are handled with care, and not scraped or dropped. They were used for the certification as well, later date (previous week instead of 6 months for the ones on the van) and had been certified by the same national laboratory.

For massmeters the cheap ones are really only 12 bit converters, as the 16 bit ADC's inside are too noisy at DC, and even massive oversampling is not going to make them any good, the reference drifts too much, and the internal IC temperature changes too much as well, they are just relying on digital clamping to show a stable zero, while internally it is drifting slowly.

Only way to get better accuracy is to power it on for a few hours in a constant temperature environment, and then run your tests. Hard as most will power off after 10 minutes or so. You need a commercial or laboratory massmeter, as those in general will have selectable power behaviour, and in general most will stay on 24/7, so they are the most stable. As well they will have a better ADC inside, probably 18 bits, with much better references, and much better drift performance as well, as in service mode ( however you enable it, varies from manufacturer to manufacturer, but in most modern ones either a set of jumpers on the PCB or a key sequence, thankfully the days of bespoke dongles are seemingly over) you can often get the raw ADC value without zero compensation, so you can see the slow drift.

But yes, for your once off application probably better to weigh out 10 times the mass, weigh out the mass of distilled water (more accurate than volumetric measures in most cases, especially without calibrated glassware) 10 times as much and dissolve, then discard the rest, or use it as fertiliser for the garden. 
 

Offline texaspyro

  • Super Contributor
  • ***
  • Posts: 1407
Re: mg scale - interesting effect
« Reply #33 on: March 27, 2019, 02:57:06 am »
what material is best to handle those polished weights? like the expensive E or whatever grade ones. (well the small ones anyway, that you might not want to handle with gloves

Ceramic? Polished Steel? Ivory? Some kind of felt lining?

Mine are made of polished Unicorn horn.   
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: mg scale - interesting effect
« Reply #34 on: March 27, 2019, 04:08:58 am »
For massmeters the cheap ones are really only 12 bit converters, as the 16 bit ADC's inside are too noisy at DC, and even massive oversampling is not going to make them any good, the reference drifts too much, and the internal IC temperature changes too much as well, they are just relying on digital clamping to show a stable zero, while internally it is drifting slowly.

If that kind of design is common, then current engineers need to study how it was done 30 years ago.

Reference drift is irrelevant when you use do the conversion ratiometrically.   Then a 7815 or 7805 will be fine because the drift and flicker noise cancel out during the measurement.

Integrate the input over a whole number of 0.1 seconds to cancel out 50/60 Hz line frequency noise; forget about using a sampling converter unless it is part of an integrating converter and even that is a bad idea.

The cheap ones use 12 bit converters because they are the most common high resolution converter available in microcontrollers.  They are terrible for this application though with perhaps 10 bits of linearity at best.
 
The following users thanked this post: helius

Offline nixxon

  • Frequent Contributor
  • **
  • Posts: 342
  • Country: no
Re: mg scale - interesting effect
« Reply #35 on: May 08, 2019, 02:48:29 pm »
I recall I watched a similar video by Applied Science on YouTube. I found it. It is a 2014 spin off of the 2011 video mentioned earlier in this thread.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf