New Myford 7 DRO article

[JasonBModerator @jasonb]

Chris has just let me know that he will have a stand at the Midlands MEX so anyone else with questions will be able to ask in person.

He also tells me that he is happy to sell sensors and tapes to those who don’t want to fit the product to a Myford and that he is also working on a simple to use but fully featured display console that won’t require a PhD in Chinese -English to understand, hopefully a prototype will be on display at the show.

[Michael GilliganParticipant @michaelgilligan61133]

A very welcome posting, Chris

Thank You

MichaelG.

[JasonBModerator @jasonb]

So Bob and Robert, now that your queries about accuracy have been answered what are your thoughts?

[Robert Atkinson 2Participant @robertatkinson2]

Chris’s response on accuracy seems entirely reasonable. It is hard to quantify accuracy at these levels without access to sophisticated equipment. I’ve designed equipment using 1um optical encoders that we checked with s Renishaw laser interferometer. A DRO can improve the “accuracy” of a machine particularly if care is taken to locate the scale in relation to wear and lack of rigidity.
I think it would be good if he put a similar explanation on his webpage.

The other question I had was about CE/UKCA marking which remains unanswered.

Robert.

[Bob WorsleyParticipant @bobworsley31976]

Ah, not quoting an accuracy or repeatability, no one can complain then. Why not? Well, you tell me. These figures are the basic information about the DRO, it is irrelevant how they are mounted, it is the best possible figure for the DRO its self.

My 2 thou was an error band of +/- 2 thou. Because I used a drum with wire wrapped around then the predominant error was cyclic, so I tested every sensor over 400mm to ensure this band was not outside +/-2 thou, so error could be over 10thou, o 10 inches. Repeatability is another measure, and has to be guaranteed by the design and manufacture, this is why I mentioned the Price book. I note that Chris hasn’t done any tests and provided the results.

So the weasel wods about accuracy depends on the fitting of the DRO and the state of the machine are just that, weasel words, and mean nothing. What I wannt to know is what the best possible accuracy and repeatabilty of the DRO, not the machine, produce. This requires something beyond the Myford 7 lathe.

Do you understand the tests that gave a figure of 0.5um? No, I don’t either, has the method been described to allow you to repeat the tests? No, of course not. Do I? Well, I pointed you at the Price book and the two step test piece, what I forgot to say was that each step was machined by power feed.

Using slip gauges to measure a dynamic measurement is to fail to understand what slip gauges are used for, static mesurements, not dynamic measurements. A pile of slips are wrung together to produce a fixed distaance to very close tolerances, less than 1 micron? Might be optimistic there. A DRO measures dynamic distances, not static distances.

Another problem is that the sensors seem to use magnetic sensing. Mention is made of the Newall ball bearing sensors at 5um. There is a problemm here, assume a ball bearing diameter of 5mm, so to get to a resolution of 5um than each 5mm needs to be interpolated, interpolated, not measured, by a factor os 1000. My sensors use a 22mm diameter slotted disc with 120 slots, I then use a 96 times interpolation, so each 90 degrees of the sine wave is interpolated by 24 giving about 6 microns resolution. I know, having done it, that this is an interesting problem because the two sine waves are not at 90 degrees +/-6 microns. I also know, having done it, that i then have to stitch these two sinewaves together to produce a smooth output that doesn’t jump when moving from one sinewave to the 90 degree sinewave as the sensor moves. This is not easy at X24, so how is this done at X250, answers on the back of a postage stamp please.

To expand on this. You have two sinewaves at 90 degrees, but you only use one, then the other, for position measuring. Why? Because as the sinewave goes over 60 degrees then the slope of the wave reduces rapidly, and moving from, say, 75 degrees to 76 degrees then the wave only changes by 0,004 so interpolating that to a level of X250 is, tricky. So the sinewave is only used from 0 degrees to 45 degrees but then the position sensing has to shift to the other sine wave, now at 45 degrees and moving down to 0 degrees. This is where the stitching together of the two sinewaves has to be done, This also has to be done by every interpolated sensor, whoever makes it, including me. It is a bugger! This is why you need static slip gauge measurements to move the sensor acrosss this interpolation distance and just see what happens, and weep. Now you can see why having very large interpolation values are a two edged sword, you cab gain ny amount of reolution, as here, but the accuracy is left in the dust, possibly here as well but no information is provided, unlike me.

So, still not impressed because Chris has answered none of the questions, and has not even done somme basic tests to give the results, I have.

As I said, a DRO is trivial, it is the sensor that is the problem. I started out with a Heidenhain 2um spar as my calibration standard, and using slips as a fixed distance standard confirmed that the calibration rig as a whole could be believed to +/-5um. My catalogue didn’t say what grade 2, 1 or reference slips accuracy was, but I recall is being less than 1um, so where Chris’s 0.5um came from who knows. It is woth pointing out that 0.5um is the wavelength of orange light, net really very much at all.

 

[Chris IrlamParticipant @chrisirlam72008]

Robert, thanks, and I agree. This thread has changed my mind about how I communicate accuracy. When I get some time I’ll add a page to the website explaining what’s set out below, including the raw test results for anyone who wants to interrogate the numbers, along with the caveats around machine wear.

Apologies for missing your question on CE/UKCA marking. Yes, the kits are CE compliant. UKCA marking isn’t currently required, as the UK has extended recognition of the CE mark indefinitely.

Bob, I’ll take your points in turn.

Method. Each reading was taken by zeroing the axis against a hard stop, winding it back, placing a slip gauge stack between the axis and the stop, and winding the axis up until the gauge was held, so every reading is approached from the same direction and backlash and hysteresis drop out. 71 positions from 0.5 mm to 100 mm, including 0.05 mm steps within a 2 mm magnetic pole pair to look at the interpolation error directly.

It isn’t a gold-standard laser interferometer measurement, but it is a simple, repeatable test against a good length standard. I’m happy to share the raw results with anyone who wants them.

I have a demonstration rig set up the same way and I’m bringing it to the Midlands Model Engineering Exhibition, so anyone can put a gauge in and see the reading for themselves.

What the numbers mean. As in my previous post, the +0.5 µm is the mean (or average) of the 71 errors. That is the measurement bias, not the accuracy, and it is small because positive and negative errors cancel. The figures that describe the accuracy of the read head are the standard deviation of 3.6 µm (0.00014″) and the extremes of −6/+10 µm (−0.00024″/+0.0004″). These were all in my previous post. I’ve given every number rather than a single headline so nothing is hidden, but that does mean each one needs reading for what it represents.

The sensor itself. The encoder IC manufacturer specifies a best case of ±5.6 µm under ideal conditions, increasing to ±10 µm in real-life conditions. My test results agree with this.

Regarding interpolation, this is done inside the encoder IC, so it is already included in their published accuracy figures. My understanding is that it reads sine and cosine together and takes the arctangent, so there’s no stitch point.

I don’t offer this product as any kind of criticism of what you built – your product is objectively impressive and it’s interesting to hear how you achieved it. But please give me the benefit of the doubt and allow me to provide the explanation / data before jumping to conclusions about my product.

Chris

[JasonBModerator @jasonb]

The sensor itself. The encoder IC manufacturer specifies a best case of ±5.6 µm under ideal conditions, increasing to ±10 µm in real-life conditions. My test results agree with this.

I think that says it all in response to “What I wannt to know is what the best possible accuracy and repeatabilty of the DRO”

As you are really only reselling a sensor and tape together with a bit of hardware and drawings for working out how best to fit them the DRO accuracy is going to be as good as the maker of the components has tested as they are the two bits that make up the DRO.

As I said before must people don’t seem worried about knowing the accuracy if sales from the likes of Machine-DRO are anything to go by who don’t state accuracy so quoting 10um could be considered just as misleading  as not stating it at all.

And putting it into prospective that bit of steel I machined this morning when the workshop was a bit chilly will have grown this lovely sunny afternoon by at least 16microns😉

[Bob WorsleyParticipant @bobworsley31976]

Measuring slip gauges by winding the cross slide up to them? No mention of any method of getting consistent loading, this is, after all, the cross slide and a screw thread will easily put significant load on the gauge. Not a problem, but the load will stress the saddle and headstock, so how much movement is there in them?

So the magnetic sensor uses 2mm spacing on the poles. So to get to a 2um resilution we have a x1000 interpolation, 250 for each quadrant, and each step needs severaal bits on the ADC so looking at, what, an 11 or 12 bit ADC. This generated from a signal possibly only in hundreds of mV, in an industrial environment.

I find it amazing that you average the displayed distances measuring the gauges. So if two gauges, one is +5um and the other -5um gives an average error of 0um.  Of course if the two slips were +/-10um, or +/-20um or even +/-100um you get the same average error of 0um. Umm.

You give the standard deviation as 3.8um, so normal bell curve statistics say that 6xSD gives the statistical error of 1 in 1000 above or below these values, so the achieved tolerace is 6×3.8um or 25um. The other given SD is 16um (6+10) so achieved tolerance is 6x16um or 96um, or 0.1mm.

So where are your results to justify the manifacturer’s +/-10um? Why not do your own tests?

These are simply done by getting a bar of metal, brass is easiest, and holding in the chuck with an end stop in the tailstock drill chuck. Touch the bar to the stop, turn about 1mm off the diameter using power feed, then part off a disc from the bar end, about 3-4mm is enough. Then number the disc so you know what order they were cut off in. Slide the bar out to touch the stop, turn downn to the same diameter as displayed and cut another disc, repeat for 20 times or so. The end stop means the same length of bar is parted off eah time, eliminating one variable. Now you need a book on statistical process control, like the Price book, and a fiducial micrometer. You have 20 discs all turned to the exact same diameter according to the DRO, measuring them and the SPC will determine what you can achieve. It is alos worth repeating this exercise but using the cross slide hand wheel markings to determine what the lathe is capable of.

Using the sinne and cosine assumes that they are exactly at 90 degrees? The sensor seems to be a length of plastic with magnetised spots which are read. The Hall effect sensor measures mv, and how does it average out multiple spots? The advantage of optical sensors is that the mask has mutiple slots so reads mutiple disc slots, averaging the inevitable slot to slot variation. What effect does not having the strip exactly in line with the sensor head so it reads at an angle, likewise variation in height from strip to sensor head?

I wouldn’t need to jump to conclusions if I had full information about what the sensors can do. I note that you still haven’t spent the 15 minutes to do the double disc test I mentioned above, without knowing what the machine is capable of all the rest of the numbers mean nothing. With the exception of the code discs and masks made by Photofabrications I made all of my sensors, so I know exactly how and why they work, you have just bought in something and what do you really, really, know about them. Sales bumf? Do not insult me to cover your own lack of knowledge.

 

 

[Chris IrlamParticipant @chrisirlam72008]

Bob, one correction, then I’ll leave it here.

Six times the standard deviation is the full width of the band, so 6 × 3.6 µm is a spread of 21.6 µm. As a ± figure that is ±10.8 µm, which tends to agree with my worst-case results of −6/+10 µm and the IC manufacturer’s figure.

The 16 µm is the spread between the two worst readings, not a standard deviation, so multiplying it by six doesn’t make sense; there is no 0.1 mm figure. The mean was given as the bias, next to the SD and the extremes, precisely because a mean on its own hides the spread.

The accuracy page and raw results will go on the website when I get some time, and I’ll have the rig at Midlands for anyone who wants to try it.

Chris

[JasonBModerator @jasonb]

Bob one moment you want to know the accuracy of the DRO the next you are asking what the lathe will cut  a disc to when yourself you have said this is no real indication of the accuracy of the actual DRO.

I have not seen Chrisi’s set up but if it make use of a dti up against his gauge blocks then a constant pressure will be applied to get the dial to show zero. But then again how hard do you push a bit of metal up against a tailstock and can it be done consistantly?

As for saying it is weasel speak there must be a lot of people happy to listen to it, taking Allendale/Machine DRO as an example again. They too seem to suggest that you will get exceptional accuracy from fitting their products. Yet a thread in the last couple of weeks had the majority of replies recommending them plus the odd direct from china thrown in.

I assume you have raised similar points with them. As well as how bulky their bolt on kit for a Myford is?

I’ll leave it to others to decide but comparing the two then Chris’s offering is a vast improvement on that and now that we have some figures likely to be at least twice as accurate as although sold as “high Accuracy” tape theirs is actually +/- 20 micron basic tape.

I have also seen some tests where the reading was affected by the wire from DRO to console dragging, again chris seems to have addressed this by either having a rigid connection on his built in option or supplying suitable cable clamps so again likely to be a better option.

A few other thoughts members may wish to ponder when deciding on a DRO.

It has been mentioned in this thread that a lathe carriage can twist under load which I agree with. So is it better to have the dro sensor placed as close as possible to the cutting tool which can be achieved by mounting it under the slide or will having your read head off to one side make it more suspeptable to giving inaccurate readings?

Following on from that does using the carriage to set the thickness of a disc give as accurate a reading as using the topslide set parallel to the lathe axis to set the length. I know if I’m doing things like cutting discs or fins on an IC engine that I lock the carriage and use the topslide to set the cut. Easy with a nice out of the way sensor and summing used in the console. But I would not want a big bulky wire sensor stuck out the end of my topslide as it woul dget in the way of teh handwheel and tailstock

I mentioned temperature yesterday. What is going to be more affected by fluctuations in air temp, a wire that is surrounded by the fluctuating air temp or a tape stuck to a big lump of cast iron which will tend to have a damping effect on any temperature fluctuations?

 

[southernchapParticipant @southernchap]

I do think there’s something peculiarly British about inventors who’s products become superseded by better designs and the availability of better hardware, becoming irrationally defensive and hostile towards the progress that has made their product obsolete.

I expect this happens in the US and Germany too, but I suspect the fewer occurrences of this ugly behaviour that defeats the whole raison d’etre of engineering, are drowned out by the general pleasure at innovation in those nations.

@chrisirlam72008

I have a Warco 918 (one of the previously ubiquitous Taiwanese 9×20 Compact 8 ‘evolutions’*) and obviously the glass scale for the Z axis I recently fitted, was relatively easy (although the motor was in a buggeration of a place, so I had to take off the motor fan and fan shroud, and replace it with a 12V DC supply driven PC case fan and make my own smaller shroud out of a muffin tin!), and with the use of an ultra slim glass scale, fitting the compound scale was a walk in the park.

I completely dismissed the idea of a top slide scale at the time, but having seen your product, I rather think I may drop you an email in the future.

It won’t be soon, as I have to save up for the lathe version of the TouchDRO adapter (I already have the mill version) to replace the frightfully cheap and nasty 2 axis display I have at the moment.😳😄

As for milling my top slide, eh… it’s a component on a mass produced machine tool, not a bleedin’ holy relic. 😄

Given some of the laughably absurd asking prices on eBay for 7 Series lathes, I suspect some people think their lathes (lathes that are, I’ll grant, definitely the prettiest of the common hobbyist lathes, and have a belt change that’s about as convenient as it’s possible to make) have been touched by God Almighty, but again it’s just a machine tool, and arguably not as capable a lathe as a Boxford or Raglan…

(I just heard the sound of a collective cry of outrage from old men in sheds all over the country😄)

In any case, ignore the miserable old sods on here attempting to reduce their cognitive dissonance by tutting at your work.  They’re a minority of a minority. 😉👍

[southernchap]

On southernchap Said:

 

(I just heard the sound of a collective cry of outrage from old men in sheds all over the country😄)

 

In their sheds, can’t be right that would mean getting out of the armchair😋

 

I was thinking about the “weasle speak” of improved accuracy when fitting a DRO. Forget how many microns you can cut a gnats wisker into there are other real world things it will help you with.

• My carriage handwheel is divided into 0.020″ divisions, 1.2″ to a turn. Now I can only best guess what cut I might be putting on by eyeballing between the division. Having a DRO on the carriage would improve that no end.
• I know most of the time we are feeding in one direction on a lathe but there are times when an axis needs to be moved in both directions. We all have some backlash so again a DRO will take that out of the equation, which should improve the chance of hitting the size you want.
• Chris from his photo looks to be the type of chap who was brought up and educated in metric. His choice of metric cutters and fixings for his kit would seem to confirm that. Now most Myfords are going to be imperial so it’s out with teh calculator to convert things to suit the handwheels which can lead to errors in calculation, rounding up to avoid multiple decimal places and trying to split between handwheel readings when you do try and dial in that 0.03937″ cut. Oh how much things would improve if you could just wind the wheel until the screen showed 1.000
• Then there are those functions of the DRO console that will make things like setting an angle or forming a curved profile a lot more likely to turn out correct than working out co-ordinates and feeding in cuts that again need the divisions splitting.

[Author]

Posts