What's new
What's new

HXX optical comparator DRO

JC Price

Aluminum
Joined
Oct 21, 2007
Location
Boulder, CO
X setup.jpg Y setup.jpg

I decided to replace the readout on my new-to-me Starrett/Craigslist HB350 comparator because the original Starrett scales were not reliable and they did not have as much resolution as can be gotten from this machine, even at 20x. Also I wanted to get some of the cool measurement features of more modern comparator readouts. Being the miserable cheapskate that I am, I bought new 1 micron scales and a DC-3000 comparator readout directly from HXX (Shenzhen Hengxingxing Precision Instrument Company) on Alibaba, for a total cost of $324 including shipping. They were easy to communicate with and the package arrived 7 days after I paid them.

So how well does it work? The photos above show the test set up. I used a 250 mm glass calibration scale that is marked "David W. Mann Precision Instruments, Lincoln Mass." I bought it on Ebay and don't know how accurate it is, but D.W. Mann was a high-class enterprise in its day and it looks like the sort of thing to use to test a comparator. The Mann scale is ruled in millimeters, so I put the DC-3000 into metric mode and wrote down the readout value with each ruling on the Mann scale aligned with crosshairs for every millimeter within the range of each stage. The plots show the results. The x-axis on each graph is the position on the Mann scale and the y-axis is the HXX readout value minus the position on the Mann scale, so it is the deviation from the value expected if the Mann scale were perfect.

Y deviation.jpg X deviation.jpg

The X scale deviation fits a line pretty well with an rms deviation of 2.5 microns around the line. That seems to be about how well I can reliably gauge when the image of a ruling on the Mann scale is aligned with the crosshairs. (I used a 20x lens, the only one I have.) But the slope is larger than I expected, about 40 microns over 190 mm, equivalent to 2.5 mills per foot. I don't think this slope is due to the accuracy of the Mann scale, since when I use it to test the Y scale I don't see the same linear deviation. Maybe I am doing something wrong.

Are these scales about as good as can be expected, or does it look like I'm doing something wrong, or could I get better scales (for reasonable cost)? The DC-3000 can use any scale with 5V TTL output. I have not played with all of the features of the DC-3000 yet, but as far as I can see it does what it claims to do.
 
Does this readout have linear correction?
If it does you are all set.
At 20x 2.5 microns is great so simple linear full length compensation will remove your sloping line.
Also the scale and axis must run true but assuming this has been indicated in.
Many compartors allow you to twist the table for measuring threads so this axis has to be straight when checking.
Bob
 
Yes, the DC-3000 has linear compensation, and I agree that should do the trick for the X scale. It also has something they call segmented linear compensation, might work for the Y scale. I have to decide if I trust the Mann scale, or maybe have it checked somehow. I indicated in the scales so they are parallel to the stage motions to a few mills. This comparator does not have a way to rotate the stage axis. Instead it lets you move the light source a little from side to side to better illuminate a thread.
 
Your X axis stage travel may have pitch error that is contributing to the linear error (research "Abbe Error" if you are interested). Older machines, especially with crossed-roller linear bearings, are susceptible to this. You can test for this by lowering your Mann scale as close to the stage as possible, aligning it parallel to the X axis, and retaking a few of the same measurements you took in your original set. 50mm, 100mm, 150mm, and 200mm should suffice. Now raise the scale as high as you dare off the stage and do the same. Compare the results from all three heights. If the measurement errors get worse the further the Mann scale is raised vertically off the stage, your X axis has enough pitch error to cause some concern. You can still enter a linear error correction into the readout to reduce the measurement error, but you will still have differences in the X axis accuracy depending on the height of your measurement line off the stage.
 
Your X axis stage travel may have pitch error that is contributing to the linear error (research "Abbe Error" if you are interested). Older machines, especially with crossed-roller linear bearings, are susceptible to this. You can test for this by lowering your Mann scale as close to the stage as possible, aligning it parallel to the X axis, and retaking a few of the same measurements you took in your original set. 50mm, 100mm, 150mm, and 200mm should suffice. Now raise the scale as high as you dare off the stage and do the same. Compare the results from all three heights. If the measurement errors get worse the further the Mann scale is raised vertically off the stage, your X axis has enough pitch error to cause some concern. You can still enter a linear error correction into the readout to reduce the measurement error, but you will still have differences in the X axis accuracy depending on the height of your measurement line off the stage.

Interesting idea. This one has ball slides. I know because I had to fix the focusing axis. It went bump-bump-bump as it moved because of dings in the rods. I was able to fix it by turning the rods 180-degrees along their axes, so the dings no longer contact the balls. But I understand what you are saying. If the slide motion also rotates the stage a bit that could cause an apparent scale change. I'll try to test for it. I also wonder if maybe the weight of the X stage could cause a little rotation of the stage about the optical axis at the extremes of motion. That could also be revealed by putting the Mann scale at different heights above the table. There are gibs for each slide, so it's possible they are not as tight as they should be.
 
Interesting idea. This one has ball slides. I know because I had to fix the focusing axis. It went bump-bump-bump as it moved because of dings in the rods. I was able to fix it by turning the rods 180-degrees along their axes, so the dings no longer contact the balls. But I understand what you are saying. If the slide motion also rotates the stage a bit that could cause an apparent scale change. I'll try to test for it. I also wonder if maybe the weight of the X stage could cause a little rotation of the stage about the optical axis at the extremes of motion. That could also be revealed by putting the Mann scale at different heights above the table. There are gibs for each slide, so it's possible they are not as tight as they should be.

I concur with the other posters about doing the linear error compensation in the DRO unit; that should take care of the slope of the error plot. Note as well that both plots show the same type of point distribution of error relative to best-fit line; i.e., positive error value at each end and negative error values in the central portion of the plot. This says to me that you may be seeing sagging of the stage at each end of travel, which would also then indicate that the gibs need a tweak, as you mention. Ball slides are not as stiff as (for instance) crossed roller slides, but you may be able to take a little of the observed error out with gib-tightening.
 
To take this a little farther I got a high-accuracy glass slide with a 50x50 array of 0.5 mm diameter dots, spaced by 1 mm. It is Edmond part #57983, found on Ebay for $120. Comes with a cal certificate claiming better than 1 um accuracy. I made a plate holder for it and stuck it on the stage:
Edmond slide on HB350.jpg Edmond 50x50 array on screen.jpg
These tests don't cover the whole range of the stage, which has about 8" X and 3" Y travel, but right now I need to measure small parts that are less than 2" by 2". I measured the position of every 5th dot and got these results:
raw data deviation XY plot.jpg
This is the raw data with no scale or skew corrections. So you can see the errors, I have plotted each point at its nominal position but with the deviation expanded by a factor of 100. Here is a scatter plot showing just the deviations:
raw data deviation.jpg
So with no scale corrections or skew (rotation) I see a standard deviation of 5.3 microns in X and Y. See next post for what happens if I apply a scale factor to each axis and a rotation...
 
Last edited:
If I take the raw data shown in the previous post and scale the X coordinate by 1.0003 and the Y coordinate by 1.0005, and then apply a clockwise rotation by 0.01 degree, I get this improved plot of the XY coordinates:
scaled and skewed XY plot.jpg
Here is the scatter plot of the deviations with these corrections applied:
scaled and skewed deviation.jpg
Now it looks like I can get a standard deviation of 2.1 um in X and 2.8 um in Y. I think that is a pretty good result. I'm sure it would not be as good over the whole travel of the stage. I think some of the remaining errors are coming from the HXX scales, and probably some from the stage as well. I would love to see some interferometer data showing how good these scales really are. It seems like someone out there must have tested them since they are so common.
 
CarbideBob, thanks for the kind words.

One more post here just in case any of you are interested in the DC-3000 readout. I had to take mine apart because there was something partly covering the display:
screen.jpg
It looked like a protective film that should be removed after shipping, but it was not on the outside of the display window. So I went inside and sure enough, they had forgotten to remove it from the LCDs before assembly. Here is the scene inside:
inside 1.jpg inside 2.jpg
Yikes. The main board looks okay, but there is an ugly nest of point-to-point wiring and lots of hot glue holding things together. Not exactly good looking electronic assembly, and not promising in terms of reliability. I also had to poke around to figure out how to connect a foot switch. There is a DB9 for it on the rear panel, but the manual they sent does not give the pin-out. I found some other versions of the manual by hunting around on the web, and they did give a pin-out, but not the right one. I found that the foot switch works when you connect a normally-open switch between pins 4 and 5 on the DB9.

The feature set seems to be very similar to the Metronics Quadra-Chek 2000. I would say it is a copy, but some things are implemented a little differently so it is necessary to read the manual, especially if you want to change system settings or write a program. The manual is pretty badly written, but you can generally figure out what to do. The software seems to work, although I found that it often hangs up if you try to examine the program steps. Despite the issues, I have been able to get it to do everything I need to do, and it has been reliable so far in normal operation.

I'd say I got what I paid for. Probably not the thing to buy for a shop where a lot of different people will use it and you need clear documentation, but it does the trick for cheap in my one-man R&D operation.
 
Hello
Maybe I'm on tbhe wrong forum, but maybe someone can point me in the correct direction.
I have a milling machine with an Anilam wizard dro but my x scale is broken, I bought a new scale on aliexpress with the same 9 pin plug thinking it will work but it is not working. I had a look in the plugs and the wiring of the pins is different from my original scale, any ideas on how to figure out if I can get this scale to work with my anilam unit. The new scale is the same voltage (5V).
Thanks for your help.
 








 
Back
Top