For the Inspection of Rotary Tables …

[Michael GilliganParticipant @michaelgilligan61133]

This offers a humbling level of precision:

http://vintagemachinery.org/pubs/9432/29078.pdf

… and, if I have read correctly, that was in 1965

MichaelG.

[PeteParticipant @pete41194]

Thanks for the link Michael, interesting reading.

[Nigel Graham 2Participant @nigelgraham2]

Very interesting – thankyou for that!

 

Their method for calibrating two rotary-tables against each other seems a version in principle of the industrially well-established “Reciprocity” technique for calibrating sonar transducers that are reciprocal, i.e. ones that can act as both hydrophones (microphones) and projectors (loudspeakers).

Three examples of the same type, nominated A, B and C, are arranged to “ping” at each other in turn a set distance apart, in the order A to B, B to C and A to C.

The transmit voltage to each in its projector mode is known, the receive output volts from each in its hydrophone mode is measured, then with other electrical values added to the mix, in frequency steps over their appropriate band. The numbers are fed into a lot of mathematics to arrive at the transmit and receive sensitivities for all three devices.

.

I was trying to think if similar cross-calibration or reciprocity methods are used elsewhere in mechanical engineering, then remembered it was used to make the first, or master surface-plates:

Three plates, A, B, C, machined as accurately as possible. Then test A against B and C, and B against C; both ways, in scraping them flat.  (Clearly, if you use only two plates you could end up with a perfect concave / convex fit.)

[Michael GilliganParticipant @michaelgilligan61133]

As there has been some interest:

https://emtoolbox.nist.gov/Qms/14030%20-%20Indexing%20Tables%20by%20Subdivision.pdf

.

https://www.nist.gov/programs-projects/dimensional-measurement-services

 

MichaelG.

.

Edit: __ and to address the inevitable question:

https://www.lightglassoptics.com/Moore-1440-Precision-Index_p_1248.html

[Robert Atkinson 2Participant @robertatkinson2]

I’m very surprised that they would still make these. Or that anyone would use one. There are better techniques. Laser interferometry is the current standard for in-situ work and lab work. It is a order of magnitude higher resolution than the Moore index and can be applied directly to the item being checked e.g a CNC rotary axis. It provides continuous readout not the 1440 steps of the indexer. An example of a current system is the Renishaw XL80 https://www.renishaw.com/en/xl-80-laser-system–8268

The measurement head can be placed outside the machine with just small optical components.
It also measures distance straightness flatness etc.
I’ve used the older ML10 system (bigger laser head, same optics) and it is soo easy compared to a colimator. Yes you need a colimator to use the indexer so more obsolete equipment to buy and maintain when using one.

[Robert Atkinson 2]

On Robert Atkinson 2 Said:

An example of a current system is the Renishaw XL80 https://www.renishaw.com/en/xl-80-laser-system–8268

Somehow the forum has destroyed the  link you supply. There should be a double minus sign between ‘m’ and ‘8268’.

I looked at that page and while it was clear on its linear resolution, it was a little wooly on angular measurement.

Further digging led me to this page:

https://www.renishaw.com/en/xr20-rotary-axis-calibrator–15763 (again a double dash between ‘r’ and ‘15763’)

That claims +/- 1 arc second accuracy. I may be misunderstanding the Moore literature but that appears to claim +/- 1/10 arc second accuracy, which to me is an order of magnitude improvement.

I would note that Renishaw feel the need to provide a five page document on ‘error budget’ (bottom of page) explaining how they arrive at that number. Things were so much simpler in 1965.

 

[Robert Atkinson 2Participant @robertatkinson2]

The interferometer accuracy depends on the baseline of the optics used. The rotary axis calibrator has a short baseline and is servo controlled to keep alignment with the laser head as the axis moves. This adds the errors of the encoder and servo system into the measurement.
The Moore claim of 0.1 sec of arc is a bit disengenious as it does not include the auto collimator accuracy. This is typically 0.1 sec resolution and 0.5 sec accuracy. In use if the item is spot on at the angular step the collimator accuracy is minimised but increasing offset causes increasing errors.

[Robert Atkinson 2]

On Robert Atkinson 2 Said:

…

The Moore claim of 0.1 sec of arc is a bit disengenious …

But good enough for NIST [?]

Incidentally: The reason I posted this because it represents, to me, the pinnacle of mechanical accuracy … using a concept which we could use in our own work.

I am almost sure that none of us could aspire to making one, but we could all learn something from studying its  design and execution.

It is essentially a simple device, exquisitely executed by people who understand what they are doing.

MichaelG.

[Robert Atkinson 2Participant @robertatkinson2]

I appreciate the perfomance for the time and the simplicity of the basic idea (implementation is something else). My comments were more about the linked webpage trying to sell one….

The Hirth Coupling, with fewer positions, could be useful for applications like cutter grinders.

Robert.

[PeteParticipant @pete41194]

Likely very few of us require accuracy to + – one minute never mind down to 10ths of a second, but it’s still an interesting problem for how that degree of accuracy can be verified with some level of certainty.

While it’s about checking a very good dividing head with a high end surveyors theodolite and not a rotary table. Both are in reality no different as far as checking there rotational accuracy. https://www.youtube.com/watch?v=3NcIu0EF6lQ

His idea is in my opinion quite clever and valid, but I’ve used high magnification 36x rifle scopes over long enough distances to have some idea that there’s a built in flaw. The unavoidable and varying temperature induced mirage over the distances shown in the video would be well in excess of the 1 second accuracy of the theodolite itself. In other words the target can easily shift optically up, down, left, right or even combinations. What happens is the targets zero point will shift optically to a position where it physically isn’t located despite what your eyes are telling you. That optical mirage can also be present even at night just due to the various temperature gradients as it’s not consistent over distances like that shown in the video due to ground temperature change. Even more so with any roads or water between the theodolite and the target. Wind can also affect those optical effects since it helps to also shift that mirages direction. One thing that does help his measurement checks is the elevation drop between where he was checking from and the target location as that should reduce those effects. Obviously it would be much more time consuming, but doing those exact same readings over a period of at least 10 days and then averaging your results should be much more accurate.

 

[Michael GilliganParticipant @michaelgilligan61133]

That video is surely worth 15:35 of watching … thanks Pete.

Lovely kit, and very well demonstrated.

MichaelG.

[Robert Atkinson 2Participant @robertatkinson2]

Yes, very interesting video. It is a technique I have considered. I have a “total station” type theodolite which has electronic rather than scales and also has a laser distance measurement built in. Mine is a odd make, Stonex,  model STS5R. This has 1 arc second resolution and 5 arc second absolute accuracy so good enough for checking any practical rotary table or indexer. It uses resolvers (multiphase transformers with relative movement between windings. These have infinite resolution unlike optical encoders. The distance measurement is 1mm resoution and 2 ppm + 2mm accuracy from 1m to about 3km so 2mm at short range and 5mm at 3000m. Lots of interesting measurements you can make.

Going back to the original subject, I happened to be looking for information on another instrument and saw what appears to be a Moore 1440 indexer being used to check a polygon with a modern automatic collimator on page 3 of this document.
https://www.taylor-hobson.com/-/media/ametektaylorhobson/files/product-downloads/autocollimators/autocollimators-en.pdf?la=en&revision=0777311d-fe86-40fc-8a33-c1cf93ff44c2

Robert.

[Robert Atkinson 2]

On Robert Atkinson 2 Said:

…

Going back to the original subject, I happened to be looking for information on another instrument and saw what appears to be a Moore 1440 indexer being used to check a polygon with a modern automatic collimator on page 3 of this document.
…

I think that demonstrates the point nicely, Robert

A Taylor Hobson polygon would be accepted as a ‘reference’ [transfer standard] by most users of good ordinary indexing and dividing equipment.

MichaelG.

.

https://www.taylor-hobson.com/support/ukas-calibration-services

 

[Robert Atkinson 2Participant @robertatkinson2]

The UKAS document says:

“Index and rotary tables

Various techniques are used for the calibration of rotary devices, and uncertainties can be as low as ± 0.3 seconds of arc.”

Interestingly that is 3 times tha accurcy claimed by Moore in the original document, but see comment on NPL below who can acheive an order of magitude better..

The 0.3 for indexer plus the ± 0.2 seconds of arc for the autocollimator give the 0.5 total for the Polygon.

Obviously Taylor Hobson are going to use collimator techniques because that is what they make.

That said NPL stil use comparison with another index table using mirror and auto collimator but their tables are ± 0.02 to 0.05 seconds of arc and ± 0.01 seconds of arc for the autocollimator

https://www.npl.co.uk/products-services/dimensional/angle-standards-instruments

Their primary standards for angle are Sine Bar based. For example the Small Angle Generator used for autocollimator calibration
http://resource.npl.co.uk/docs/science_technology/dimensional/moy-scmi/moy-scmi-95-print.pdf

The critical slip gauge thickness, mirror flatness etc are now all measured by laser interferometry.

 

[Robert Atkinson 2Participant @robertatkinson2]

NPL also use interferometry to measure nanoradian angles including dynamic performance

https://www.diamond.ac.uk/Science/Research/Highlights/2022/getting-measure-tiny-angles.html

 

[Mark RandParticipant @markrand96270]

It should be noted that the 0.5 second accuracy of the autocollimator is almost irrelevant in the measurements, since that is the error over the ±5 minutes range of the equipment. In the calibration of the polygon, the range of measurements will be seconds of arc, not minutes.

I’m satisfied that when lapping my surface table the repeatability of the autocollimator and the baseline of the mirror gave me a resolution of 2µ inch. Controlling the temperature profile of the shed was far more difficult than measuring the flatness of the surface table.

[Mark Rand]

On Mark Rand Said:

It should be noted that the 0.5 second accuracy of the autocollimator is almost irrelevant in the measurements, since that is the error over the ±5 minutes range of the equipment. In the calibration of the polygon, the range of measurements will be seconds of arc, not minutes.

I’m satisfied that when lapping my surface table the repeatability of the autocollimator and the baseline of the mirror gave me a resolution of 2µ inch. Controlling the temperature profile of the shed was far more difficult than measuring the flatness of the surface table.

Hi Mark, your experience with the surface table could make the basis of an interesting article for MEW…

Neil

[Mark Rand]

On Mark Rand Said:

It should be noted that the 0.5 second accuracy of the autocollimator is almost irrelevant in the measurements, since that is the error over the ±5 minutes range of the equipment. In the calibration of the polygon, the range of measurements will be seconds of arc, not minutes.

I’m satisfied that when lapping my surface table the repeatability of the autocollimator and the baseline of the mirror gave me a resolution of 2µ inch. Controlling the temperature profile of the shed was far more difficult than measuring the flatness of the surface table.

Hi Mark,

Not quite 5 minutes, according to Taylor Hobson. On the document I linked to eariler the accuracy line in the specification table has a * note “* Normally over a centre portion of the range for visual this is the central 60 seconds”
And you have to add the mirror errors to that…
Yes, you can get very precise relative measurements but absolute is another matter.

Robert.

 

[Author]

Posts