RE: [mill_drill] dro reccomendations
Certainly, go right ahead. I have already posted it on a completely public web site.
I need to check out the rest of your site.
At 07:35 AM 3/31/2011, you wrote:
Now I recall some of this past discussion. If it is OK with you, I would like to copy your entire email and put it into a pdf that will be next to the original article on my web site.
From: firstname.lastname@example.org [ mailto:email@example.com] On Behalf Of Paul Alciatore
Sent: Wednesday, March 30, 2011 10:30 PM
Subject: Re: [mill_drill] dro reccomendations
I have seen your discussion on this before. In fact, I am one of the ones that you gave credit to at the end for contributions. I don't remember my specific contribution, but thanks for that.
Your discussion of accuracy is a bit simplistic. I suspect the HF instructions were written to be simple and really do not attempt to analyze the various sources of error individually. Nor do they do a good job of breaking those errors down by distance. You go to some trouble examining the first few hundredths of a mm from the zero point. In this, you try to apply the +/- 0.02mm spec to that short range. I don't believe that it applies to that low range. You state that the spec sheet reads "After moving a distance, mm" and next to 0 to 100mm in that column, they give the +/-0.02mm spec. I believe what they are saying is that if you move a FULL distance of 100mm, from 1 to 100mm on the scale, then you can have that as the maximum error. Without going into an analysis of the actual sources of error in such an instrument, it would be a lot more reasonable to assume that the actual error is approximately linear from 0 to 100mm with the value at or very near 0 at the 0mm point and gradually increasing to the max at the 100mm point.
You could then extend in a linear fashion from +/-0.02mm at the 100mm point to +/-0.03mm at the 200mm point. And again, continue the linear estimate from 200mm to 300mm. Actually, with the numbers shown, the line would be straight from 100mm to 300mm.
But, as I said, this is a very simplistic analysis and does not consider the factors individually. First of all, it is a way of lumping all the factors into one simple figure and virtually guaranteeing that the error will never exceed the amount given. That does not mean that it is going to ever approach those values at any particular point on the scale, least of all in the area that is near the zero set point. Also, in all digital devices there is always a +/- one count error added to the other error estimates. This is to account for the step nature (which you illustrate nicely) of any digital device. But this is not the least count shown on the display, but rather the internal least count which is much smaller. It is omitted here or just included in the stated figures.
I imagined a linear graph of POSSIBLE error values above. In reality, if real devices were tested, the actual error values would probably follow a smoother curve. It may raise more rapidly near zero and then level off to reach the 100mm value given. Again, from 100 to 300mm it would also be slightly curved, rising faster near 100mm and slower at 300mm.
The real error factors would include, but not be limited to:
Temperature: This would be fairly linear due to expansion.
As a subset here, there would changes in the sensitivity of the pickup devices with temperature. This would not be linear.
Accuracy of the internal steps in the manufacturing process. This would have two components which would add together:
The overall or long term (distance) accuracy. This would also be fairly linear.
The individual variation of the individual steps. This would be more or less constant at all scale values.
The =/- one count error
Long term ageing effects
Errors due to parallax of the jaws. This would depend on technique and the object being measured and would be fairly constant over the full range for similar objects and users.
As you can see, a really good graph showing all of these and any other factors that I have failed to think of, would be very complex. This is why they just provide the three simplistic values you puzzled over. For parts of the stated ranges, these values are way too high for any real caliper ever produced but true to their word, the real calipers do meet these specs. Part of this is the CYOA philosophy.
To apply the above to your analysis of the first 0.01mm step, I would estimate that temperature expansion would be all but undetectable. The error due to the change in sensitivity or the pickups is probably within +/-5% for a wide temperature range. The long distance error of the scale at the 0.01mm point would also be all but undetectable (I would say 0.0001mm at the very most). The step variation would probably be well within +/-10%. The one count error would depend on the size of the internal counts, perhaps another +/-10% if they are 0.001mm steps. Long term ageing effects would be very small for this short distance and again undetectable by anything either you or I are likely to have. Assuming we are measuring shim stock with good technique, jaw parallax would also be very low. So, the total possible error would be the sum of the above factors: 5% + 10% + 10% and perhaps a generous allowance of 1% for the undetectables. That gives a total possible error of 26% or about +/- 0.0026mm. And this would be the maximum error I would ever expect. At the maximum error in the negative direction, this would give a reading of 0.01mm for real values of 0.0024mm to 0.00124mm instead of the theoretical range of 0.005mm to 0.015mm. If all errors were in the positive direction, this range would be 0.0076mm to 0.0176mm.
This analysis also shows the rather rapid increase in error values at points near zero. We have seen an immediate increase to 1/8 of the 100mm value at the first point where the scale shows a value above zero. There are 9999 more 0.01mm steps to get to 100mm. This case of the first 0.01mm step is the worst case error on a percentage basis and that percentage error will decrease as you go higher on the scale.
I guess my main point is that you can't take the data sheet values on face value. You must add some common sense along with a realistic analysis of the sources of the errors. A further analysis of some additional points near zero (0.02mm. 0.05mm. 0.1mm. 0.2mm. 0.5mm, etc.) and perhaps out to 50 or 75mm would very likely give you an error curve as I described above. You would probably reach the +/-0.01mm point somewhere around 33mm point and not get to the full +/-0.02mm until you get to the 100mm point.
My analysis above is not based on any measurements of actual calipers, but rather on assumptions that are reasonable. So, don't be surprised if the actual calipers vary from it. Perhaps someone can buy 100 or 1000 Chinese calipers (different brands I would hope) and make actual measurements.
At 11:33 AM 3/29/2011, you wrote:
The Chinese scales do have higher resolution internally. I wrote some about this on my web site. You can get a resolution of 0.00005" but the accuracy is still around 0.001".
If interested, see
For details of these calipers, and
For the internal resolution. I encourage you to read this one with a grain of salt.
Rick (via iPod)
On Mar 29, 2011, at 8:53 AM, Paul Alciatore <palciatore@...> wrote:
Thanks for the detailed explanation: it's far more than I expected.
You bring up another question in my mind. I have heard that the cheap Chinese caliper scales actually work to a much finer resolution internally. A quick engineering analysis shows that this is necessary or at least highly desirable if a single scale is to function in both inches and mm. I have also read that this higher resolution (note I said "resolution" not accuracy) readout is available via the connection port that they all seem to have. So, it would appear that they MAY be capable of working to a finer resolution than is normally shown on their built in displays.
I have wondered about the possible use of this higher resolution output. It would seem to me that any DRO that uses these scales could take advantage of this output. I plan to do some experiments in this direction. Perhaps develop translation devices if they are necessary.
At 10:40 PM 3/28/2011, you wrote:
You ask a good question. I find it difficult to give you a straight answer
but will give it a try.
1. The accuracy of any machine depends on the stability of the machine and
the skill of the operator.
2. There are two kinds of commonly available scales: ones accurate to +/-
0.001" plus a given error per inch of travel. These are commonly called
Chinese scales. Then we have scales accurate to +/-0.0002". I'm not very
familiar with them but assume they have an additional error per inch of
travel. Of course you can get even better scales but that would be just
3. I own a RF30 copy and if I try very hard, can achieve an accuracy of +/-
0.0002" but +/- 0.002" is more common for me.
So, assuming that the RF31 is similar to a RF30 and the operator has a skill
level similar to me, having a scale good to +/- 0.001" is plenty good.
Having one +/- 0.0002" would be hard for me to justify.
If I am going for maximum accuracy, I use the DRO to get close and then
switch to my mic.
From: firstname.lastname@example.org [ mailto:email@example.com] On
Behalf Of Paul Alciatore
Sent: Monday, March 28, 2011 9:48 AM
Subject: RE: [mill_drill] dro reccomendations
At 09:47 AM 3/27/2011, you wrote:
>..... The accuracy will be
>consistent with the performance you can expect from a RF31.
Since I am not familiar with the RF31, what is that accuracy?
Yahoo! Groups Links
- --- In firstname.lastname@example.org, "tinapalian" <pdolb@...> wrote:
>hi group ..i'm the Original poster of "dro reccomendations" with some follow up....thanks for everyone's thoughts and opinions..i was able to find an old d-ann manufacturing (aka Dynamics Research Corporation) dro with glass scales locally for a great price .
the system is installed and working...not 100% accurate when compared to a dti ..but the price was right and i had fun mounting the scales also i ended up with 2 nice chunks of aluminum that were used to mount the scales to it's former machine ....my thanks to all who input their time ..paul
> i have an enco rf31 clone with a power cross feed .i'm looking to buy a new 3 axis dro , but i'm not opposed to buying a really good quality used 2 axis setup ,as i can always add a seperate z readout later .
> is oldtime used better or are the newer produts the way to go ?
> i'm confused with pricing and quality..the ebay items are 50% of the price of the dro setups i see in the big book catalogs
> is it all the same stuff rebadged ?
> what should i stay away from and what should i look for as a must have option ?
> any suggestions and feedback of your experiences would help ..thanks paul