PLEASE DO NOT REPLY TO THIS POST!!
The purpose of this post is just to document some facts about a particular (and not very cheap!) scale where the manufacturer told the truth, and had no idea what the implications of his statement were. It's likely that most of the other cheap digital scales out there with an accuracy of .01 gram/0.1 grain are as bad or worse, and to get an accurate scale you need a jewler's scale with an accuracy on the order of AT LEAST .005 gram, and preferably .002 or .001 gram. All of this has been re-hashed over and over, and there's no need to respond to this post. It's only here so a convenient link can be posted to it in topics where the subject of electronic scales or data from them comes up.
Here's the initial post by Ahrens/Rugersol:
Not to open a big can of worms ... I think it's already been opened ... and shut, again. I'm only posting this as I've actually received a response from the manufacturer ... and figured it was worth sharing.
I think most folks 'round here are aware of the whole "0.1gr/0.01g" deallie. Fer thems that ain't, the conversion from grams to grains is 15.4323584 ... and most such scales (if not all) actually weigh in grams, and then do the simple math, to show grains. At which, a scale that measures 0.02g might show (quite accurately) 0.3gr ... but when it measures 0.05g, it would most likely show 0.8gr (0.77, rounded up) ... and when it measures 0.04g, it would most likely show 0.6gr (without dropping places, after conversion, that's 0.617294336). Even if it didn't round, and rather truncated, there would still be an issue with the scale "skipping" tenth grains.
I had called Dillon, who referred me to CED (the manufacturer of the scale). I talked to a couple women, initially ... they weren't making much sense ... and told me to send an email, which I did. That was over 2wk ago ... I finally got a response, yesterday ... here it is:
“The current D-Terminator electronic scale is measuring in gram and then convert the reading to grain. The error due to unit conversation will be within +-0.05 grain (equal to 0.003 gram)
No matter how, the error due to unit conversion will always be there. It is just a matter of which unit measure is more important for the application. “
Respectfully, Charles Hardy - CED
This may seem like an awkward response ... that's because the question I asked him was ... "is the scale measuring to 0.01g, and then converting to grains, or is it measuring 0.1gr, and converting to grams?" Oddly enough, if the scale accurately measured in 0.1gr, and then converted to grams, and rounded to hunderdths, ya'd see relatively few gaps in the hunderdths for grams, but it would otherwise be accurate to 0.1gr (it would then actually be accurate to 0.00648g, so conversion aside, anyone else happy with 0.01g, would never know the difference). So, by his response (above), even without admitting the "error due to unit conversation", by stating that it measures to 0.01g, and then converts to grains, he's confirmed the "error" ... and in all, has confirmed it, twice.
However, his statement, while he admits there's an "error due to unit conversation" of "+-0.05 grain", is not the whole story. The 0.05gr "error" is simply due to the conversion. As the scale only displays to tenths ... not hunderdths ... you don't get the benefit of seeing when/where this "error" occurs. Instead, you see jumps of 0.2gr, instead of only 0.1gr.
At which, I see this as confirmation from the manufacturer that Dillon's $140 electronic scale (that I was dumb 'nuff to buy, to start with, 'cause it said RIGHT ON IT, was good to "0.1gr" ), is in fact, not one bit better than any other $30 electronic scale which has the same "error". Point of fact ... you would have to get a scale that's good to 0.005g (technically, roughly 0.00648), in order to get accurate readings to 0.1gr.
I'm in no way suggesting that this "error" isn't otherwise acceptable. If you're loading that hot, you need to be working up slowly, long before that point, and watching for pressure signs. And you could otherwise double or triple the number of drops, and divide back out for an "each" weight.
I'm only suggesting one needn't pay an extra $110.
Anyone who'd like to call "BS" on me, is more 'n welcome to pay me $80 for my "as new" Dillon D-Terminator electronic scale ... an absolute BARGAIN
First response from me on Oct. 2, 2010:
Okay, I created a spreadsheet of values in .01 gram increments, and then converted those to 0.1 grain weights, meaning that I calculated the exact grain weight and then rounded off the value to 1 decimal place like the scale does. The results are WAAAYYY worse than I thought they would look like. And remember, ALL the possible gram weights that the scale can measure in this range are represented, so there aren't any holes in the gram data.
So, this is the error JUST from going from .01 gram accuracy to 0.1 grain accuaracy!! Even if the scale is reading the weights absolutely perfectly, this is the error that can NOT be taken out of the system.
Add to that the fact that the scale is going to have problems with weights that are nearly halfway between .01 and .02 (such as.014, .015, .016) and for those weights the scale will report a value in grams that is between .004 and .006 grams off the real value.
Then add to that additional error from static electricity and noise from flourescent lights, and possibly drafts. (Note, if there is a draft with a beam balance, you can SEE it. With a digital, you probably won't!)
Add that all together, and you have a real mess on your hands.
NOTE: In 52 gram data points, I count no less than 29 skips in the calculated grain weight!!! That's over a 50% rate of the grain data being mis-reported just from the calculation roundoff from the gram weight!! To put it another way, in a series of 52 linear gram weights that are .01 grams apart, there are 29 grain weights that the scale will NEVER show on its readout!!
After some more jawing back and forth, the other shoe dropped inside my head, and I posted this:
Thank you mitch, and we were both missing something, and it was bugging the hell out of me until I figured out the right graphical presentation of the data. It took me four tries to get the graph to show what I wanted.
The statement that the actual grain weight varies from the displayed weight by a maximum of .05 grains is in fact true. That statement, however, carries with it some assumptions that the variability is centered about the actual weight (that is actually true) and the variability is a normal gaussian distribution, which in this case is hideously false, and the BIG lie as far as this data is concerned. So: I did a chart of the variance between actual grain weight (four decimal places, based on the four figure conversion factor of 15.43 grains per gram)) calculated from a gram weight (with a precision of .01 grams) and the displayed grain weight with a precision of 0.1 grains. (1 decimal place)
The graph of the variances is shown below:
OHHHH!!! Bet you weren't expecting that!! (I wasn't either... )
The variance is NOT gaussian, but linear, and much worse that that, it flip-flops from a erroneous high weight to an erroneous low weight with every increase of .01 gram, and then repeats the cycle with every increase of .1 gram. So, at the worst, an increase of .02 gram will take your from a grain weight that is .05 grains high STRAIGHT to one that is .05 grains low!! And let's be clear about the real problem here: You have NO IDEA where you are on this chart with any given powder charge!! Maybe you're going up and down by .01 grains, and maybe you're going up and down by .05 grains, You will NEVER have a clue without a reference balance to check your data, which means this digital balance is NOT reliable.
And consider the most likely scenario for loading up a string of loads over a range of 0.1 grain increments: Chrono data!! You load up ten strings of 5 loads each in .1 grain increments, and you are GUARANTEED that you will cross over one of those up .05 - down .05 transitions. That means that in actuality, you will have TWO 5 shot strings that are the same weight (when they are supposed to be 0.2 grains apart), and they will either be on the high side or the low side, but in any case your chrono data will be thoroughly screwed up.
The WHOLE point of load development and chronoing your loads is to establish a predictable relationship between powder weight and velocity, and with this kind of ramped flip-flop behavior in variance you might as well never even bother to take you gun out of the safe.
So, that extra powder isn't dangerous per se, but it COMPLETELY blocks you from doing any RELIABLE load development work. For THIS you want to pay good money??