Welcome Fellow digital Scale Owners.
Some of whom like me are going to be disillusioned or relieved real soon.

There is a quirk (it was mentioned on Google groups) with some perhaps older digital scales and I would like your help to see how widespread it is.

My old scales ( Philips HR2385) and some other brands and models are said to have it. Lets see how your scales perform.

I thought something was up with my scales but until now had not pinned the problem down. This quirk may not detract from a scales performance in normal use and I am delighted with the way my Philips scale works in practice. The display counts quickly when adding weight and never jumps over a reading like some scales do. Now I know about this quirk I will avoid it.

For the test you need:-

Two light containers like a plastic cups.
About 40g of free flowing product like sugar.

Place a cup on the scale.
Reset/Tare the scale.
Very slowly sprinkle in the product. Stop when the display reads 20g   (The slower this is done the more dramatic the demonstration.)
The hard part is over.

Now weigh the product you just measured.
Place an empty cup on the scale and Tare/Reset the scale
With the cup off the scale put in the product and replace on the scale.
(if your scale does not have a tare feature  just check the weight as you would normally do)
Read the weight. Check the weight.

Results from some scales reported by members of the saltmatters mailing list. 
My old scales are first on the list.

Type Capacity Grad. Result____ Used________ Auto_Turn_off Show_neg
Kitchen 5000g 1g 36g Polenta 3 m Yes
Jewelry 100g .01g 19.99g *2
Polenta 30s Yes
Kitchen 2000g 1g 35g Rice ? ?
Kitchen 2000g 1g 23g  *3
Polenta 45s Yes
Kitchen 5000g 1g 20g Sugar 30s Yes
Kitchen 2000g 1g 20g Rice, Polenta
3m40s Yes
Kitchen 5000g 1g 20g Sugar 3m Yes  *1
Kitchen 6000g 1g 20g Sugar 1m Yes
Kitchen 2000g 1g 21g *2 Rice 1m ?
Kitchen 4000g 1g 25g  *3
Polenta 1m Yes
Kitchen 6000g 1g 20g Sugar 1m Yes
Kitchen 7000g 1g 20g Sugar 2m,5m,no off Yes
Notes:       *1     negative values shown but no - sign
                   *2    Within +- 1 unit is a perfect result
          *3    Slight  error may be due to auto zero see below
If the scale has automatic zero and most do then it will keep making internal adjustments to correct any drift while the display shows zero and this can lead to the first amount of weight not being counted if it was not enough to get the scale off the mark. This may have been the reason that tests marked *3 were slightly out.  You can avoid the auto zero causing errors by ensuring that the first weight added is a couple of grams or so.
Thanks to MB
who has brought my attention to this reference http://www.rightonscales.com/web/learnmore.htm#autozero to the auto zero function causing the lower reading.  ....and to this reference  ahttp://www.jscale.com/support.html about halfway down the page. Which explains that " Adding light loads or "dribbling" can cause display problems. " .
The problem that I think my old scales have a correction that is applied by the scale no mater what the display shows.

Digi  Scale-duggery. (how to get around a problem)
Seems to apply to older digital scales.

The heart of a digital scale is the load-cell which converts the load on the platform into a signal that is processed by the internal computer and the result is shown on the display. The load-cell, the platform support mechanism and associated circuitry has some problems that can be overcome by the software in the scale's computer.
One problem is drift. If drift is not "solved" in some manner it could lead to customer dissatisfaction. Uncompensated drift causes the display slowly change when the scale is just sitting there with or without weight on it.

How a manufacturer could handle the "problem" of drift.

By careful design it may be possible to eliminate drift to the extent that it is not apparent on the display in the time usually necessary to complete a weighing operation or the scale turns off. In this case the drift is less than the smallest display unit. If the display increments in larger, say 2g steps drift is less critical. If the range of weight that the scale has to handle is less, drift is less. If the scale turns itself off quickly there is less time for drift to occur.

If the cost and complexity is too great to fix drift it could be just left as a visible problem. In this case occasionally the display may slowly drift in one direction.

You could redefine the "problem" from one that affects accuracy to a "problem" that causes customer feedback. The new "problem" then is to eliminate the customer feedback. This could be done by cloaking the drift. Drift will occur slowly so the computer is programmed to watch for any slow change in weight and to presume it is drift and to apply a correction. Basically the computer notes that the weight has slowly risen by say 1g and concludes it is drift and subtracts 1g off the weight. The display does not change and the customer is unaware of the correction. You can see now that slowly adding weight to the scale has brought this "feature" of the scale into play and impacted on the accuracy.

Perhaps all scales make drift corrections and the ones that give incorrect values in the test above apply the correction too quickly and/or fail to undo the correction when it turns out to not be needed.
After a lot of searching I have found a reference to  a "Minimum add weight" specification of  >1g but it was not for a kitchen scale.

What does this all mean practically.
Do the test on your scales rejoice if you do not have the problem. If you do then:-
However much you are weighing out it is perhaps only the last 10g will be added slowly and now you know not to sprinkle in small amounts you should only be over by no more than a gram regardless of the amount you are weighing.  The "problem" makes "weighing out" small amounts say up to 20g difficult to do accurately but a 1g division scale is not meant for such work. I "weighed out" 10 gram lots of yeast on my old scales and the results were 11.77, 11.94, 11.07, 10.65. the more careful i.e. the slower I was the more the problem affected the result.

My New Kitchen Scales

My new scaleAfter discovering that other peoples scales did not have the problem and knowing that my scales were inaccurate after a long hard life and not able to be recalibrated (a 1kg weight showed as 1028g)  I thought it was time to update. There is no knowing if the problem above is a feature of a scale until you try it but I have a feeling that modern scales have improved a lot so I took the chance and paid $72.50 delivered for a My Weigh KD-7000 from Multipoint Technologies www.multipointtechnologies.com.au  For a look at the scale go to www.myweigh.com  The scale seems perfect for what I want. The display is fast with 3 speeds programmable,  auto off is adjustable 2m, 5m, or never, it is accurate ( 100g & 1kg weighed correctly) and it can be recalibrated with a standard 5kg weight but weighs to a big 7kg. It is specially designed for use with liquids and powders i.e. cooking as it came with a protection flap and 4 replaceable keyboard covers. Weighing out small amounts can not be expected to be accurate on scales where the smallest unit is 1g but weighing out 20g gave 19.94,20.10,19.93,19.92 a great result and the accuracy was helped by setting the display to fast. These scales definitely do not have the quirk with slow adding of weight. I made myself a 5kg test weight from an old barbell disk weight so that after many years of use I will be able to return the scales to their original accuracy. By the way I just weighed out some butter for Christmas hard sauce from an old recipe and used the oz. mode. Most products you can find something to complain about but not this scale it is TERRIFIC.

Bathroom scales-another Quirk

With bathroom scales drift is not a problem as they are reset each time and the weighing takes place in a few seconds. There are a couple of problems that may cause customer feedback. When the weight just happens to fall right (or almost) on the boundary between digits it would be a toss up which number is shown on the display for example 89.9kg or 90.0Kg. To help the customer the scale picks a number and to reduce your confusion should you weigh yourself again immediately some scales remember the weight for some time and if the next  weight on the scale is "near" the previous one it does not provide its new figure but regurgitates the previous one.  The interesting thing is that "near" need not be as small as one may think.

Try this on your bathroom scales.

Find something that weighs roughly 200g. Weigh yourself  then holding the 200g weigh yourself again. You may be surprised that you get the same result. Weighing your self multiple times on a scale with this "feature" and getting exactly the same result could lead you to think that the scale had repeatable results that were accurate when in fact the weight shown each time was only  the result of the scales first effort. To test such a scales repeatability it is necessary to give it something to weigh that is not near your weight between the tests or to wait some unknown amount of time.  

Bathroom scale with wooden topMy bathroom scale has this quirk, it remembers the last weight and repeats it unless the new weight is different by at least 250g.

My scale when new gave the same answer each time I weighed myself  i.e. the inherent variation was corrected by the quirk described above but as it aged it started to give different answers each time and I had to average the results of about 4 readings. The problem was traced to the top lid flexing and was solved with the addition of some 5 ply. Note there is no display as it talks and that it is not sitting on the carpet as the edges would touch and cause errors.

I hope you found this interesting.