[rokag3]

Since i make a long post and....I loose it, I am going to write it step by step


Everyone who start in CNC, one day or another have the obvious idea to use the DRO linear encoder to drive the table !

Then, the poor guy enrich his vocabulary with the word "hunting" as soon as he is explring the multiple forum.

As soon as he experiment it by himself he the understand the meaning of "hunting"

In fact this is a Graal, everybody admit it, BUT it is (i can say wrongly since 3 days) considered as impossible!

[rokag3]

What are the problems to do it ?

Mainly backlash and a bad understanding of the way servo work.

The rule:
The backlash must be smaller than the resolution of the encoder !

Our machine is a Lagun FTV2 manual that we have transformed 100% ourself.

The DRO read an encoder of 20µ lines (so 5µ resolution)

after having try to make a roller screw we (for the time been) give up and buy 4 chinese C7 ball screw with 8 nuts !
25 mm lead 5 mm

the screw was off line by a generous 2 mm and we straight it to 1/100 mm

We the machined it ourself ( my friend is very suspicious about making precision work for little money) the surface is hardened so carbide is a must.

we then use it for a while with only one nuts. the backlash was 2 (X)and 5/100 mm(Y) all coming from the nuts (we have angular bearing and the belt is very tight)

with such backlash forget about linear encoder (chair)

[rokag3]

We then decide to cure the backlash :

George have to design a new nuts holder for the Lagun (having done it for our roller screw it has not been so difficult) it's a big work I will post the CAD later.

George use 8 bellevilles 4 against 4 of 15 kg each so 60kg preload max due to lack of choice of belleville in Athens trimmable. with a nuts this piece has been made VERY precisely < 1/100mm

George also change the balls of the nuts, we discover that 4 to 5 balls was missing for each nuts ! and the variation of diameter for the balls was important (George get the info)

George find balls of 3.175mm and we measure each ball one by one with the micrometer and put it (use grease to do that job and stay cool (much more easy to do than say BUT for a confirmed machinist )



after installing the screw we measure a backlash of ....we could not measure it despite that we have a micrometer 10 turns for 1 mm !!!
It was possible to make hand variation of the micrometer without disturbing the DRO (5µ)

video soon

[rokag3]

[ame=http://www.youtube.com/watch?v=bDk-4gnhuIw&feature=related]l'atelier de mon copain à Athènes - YouTube[/ame]

This is the workshop of Georges , Hi is this type of great machinist that work in Piraeus port and can save huge amount of money to the shipyard by making the impossible ! since there is no more work he now make some bike fixing, welding aluminium or rebuild motors !

You can see the nuts holder the round piece with holes is the adjustment for the bellevilles in black on the middle
the x table is off we are making the measurement with a simple comparator on this video (the other one is fragile and we did the measurement after)

[rokag3]

[ame=http://youtube.com/watch?v=jTYb79g98jo&feature=related]essai - YouTube[/ame]

you must watch y axes only

here we are working with the rotary encoder

[rokag3]

After this success we start to having the willing to measure the screw accuracy and then we discover that our C7 was a C8
ok not a big deal we must enter the table of compensation in Linuxcnc

But before I have the idea to test the Graal

Since I made the db9 connector of my cnc compatible with my DRO the test was easy to do
I just have to put 200 steps per mm in the ini

the test with x axis work fine immediately the servo (keiling) was stable we start with jogging very slow speed ready to push the emergency button !

everything was fine (I didn't have the dro at that time since I cut some wire in my home made signal duplicator)

we increase the speed and everything was perfect.
No hunting!
We make the test for the linearity of the screw and despite the lack of linearity of the screw the measurement was EXACTLY what we ask, we have the accuracy of the DRO.

We then try to test the dynamic compensation. George Lagun is not specially in chewing gum but, if we put the table at the extremity we can get some lateral jog 2/100mm and ......the servo compensate!

[rokag3]

Encourage by this success we make the try with the y axes !

It did not came smooth as the X axis at all ! the servo went in a tough oscillations and we have to stop it.

Then I wanted to adjust the gecko 320, I choose to give a large dumper and ......it work!

since I fix my wire i could see the result on the DRO and I discover that we have in fact at a very low frequency a "hunting" effect 2 oscillations per minutes of 5 µ .

This is normal in fact a servo is often "testing" the bounce if you have a linear encoder with 1000 steps/mm you cannot see it but with 200 steps/mm you can see it.


So we decide to make a circle on a piece of aluminium .

Complete success we could not make better with the lathe !

The video is in my camera !

I will wait someone to ask me to show it !

[Mike Everman]

Great work! Let's see the video!

[rokag3]

ok , I just finish the mail to get a sponsor to give us a linear encoder 1µ since the price of the linear encoder is ridiculous comparing to the price of a high precision ball screw.

And i extract my files

[rokag3]

[ame=http://youtube.com/watch?v=x8I00doLNuw&feature=related]essai - YouTube[/ame]

here we are

[rokag3]

Well , so far so good we are conducting different test, the principal problem we face is that we cannot measure long distances with precision.

We have a 500mm micrometer but the resolution is "only" 1/100 mm

Now we are going to test under vibrations, since we cannot scrap tons of metal, we are going to generate vibration with a bad tool

The goal will be to evaluate the behaviour of the servo, of course we are not expecting precision work from such experience, but we want to see if we are going to amplify the vibrations, reduce the vibrations or be neutral with the phenomenon !

Since we work with this DRO from many years we know that the BIG vibrations affect the display so of course it will be transmitted to the gecko too ! we also know that the linear encoder is a very resistant device (if well fixed on the machine)

[rokag3]

The rule:
The backlash must be smaller than the resolution of the encoder !

But and mainly maybe, the force produced by the momentum of the servo must be such that it can stop on the first state change of the encoder(which is our case)

this is also the reason why the servo must be trim specifically for this purpose !

WHAT ARE THE CONDITIONS TO TEST THIS ON YOUR MACHINE ?

IF you can reach a position distant of 1 mm to 2 mm by hand reading your DRO in less than 5 seconds it mean that your backlash is Very small and you can test it on your system

[rokag3]

I have discover a wonderful product, the AS5311 from Austriamicrosystem.
It's a new chips less than 10 euros on the net

It read a magnet stripe which is made of 1mm pole . the standard accuracy of the magnet is is 40µ/meter(about 30 euros/meter) so equivalent to A C5 precision but there is a 10µ/meter (about 70 euros/meter) which is between C0 and C1 precision

the chips give a quadrature signal A B (512 steps/mm or 1.953125µ per step) and an index signal, it also give an absolute value with 488nanometer so there is a real possibility to avoid the DRO and copy the value to linuxcmc(EMC2) DRO.

The level of knowledge in electronic is well in my range (despite it's low)

the speed of reading is an amazing 655mm/seconds !!!

I give more details here!
http://www.cnczone.com/forums/genera..._%B5_very.html

I am planning to use this encoder on a project of non contact profile measuring for large object!

[harryn]

Thank you for posting this. I am still digesting it all, but many people would like the ability to use a linear encoder with their servos.

Please keep the videos and pictures coming.

Harry

[rokag3]

Thank you for posting this. I am still digesting it all, but many people would like the ability to use a linear encoder with their servos.

Please keep the videos and pictures coming.

Harry

Backlash, this is the point if your table move when your motor turn then you can do it!

so:
A belt with a good tension
A very straight alignment (to avoid flexibily )
Angular bearing well loaded

the connection with the table must be VERY rigid (we have used the whiter grey cast iron as you can see on the video) the size of the connection has been maximized .

your preload of the ball screw must be easy to ajust ! only one ball screw will not work and if it does not for long

balls of the ball screw changed they have been measured and get the same diameter at the precision of our tool (1µ)

Of course if the object to move is light it's much more easy to achieve.


there is an alternative to the double ball bearing :

The roller screw! It work too BUT the diameter of the rolls must be at least 10% bigger than the diameter of the screw (if you have enough room bigger is better :devious

http://www.cnczone.com/forums/451351-post257.html

[ame=http://www.youtube.com/watch?v=PuBtaV_L0fk&feature=player_embedded]no backslash !!! - YouTube[/ame]

by not respecting this rule we have spend hours to make something that does not work ! on the video the system is perfectly working it was a prototype made on a small lathe!

[rokag3]

We have mill a prototype of our AS5311 circuit

2*10 pins 0.27mm space and 0.36 mm coper
the tools is a home made 10 mm carbide vith an angle of 10 degres with one tooth

the pictures are taken with the microscope X30

I train my skill in soldering on this i use fr that old memories of pc , unfortunately the pins of the memory chips are larger than the AS5311 but after welding the height circuits i will have a good skill

I remind you that our table is driven in X Y with our linear encoder

the 4th picture is taken in transparency

in fact out tool should be 9 degrees not 10 but i prefer have a bigger space in between the line in fact i weld it pretty easily !

after milling i have put some varnish and i put grease on the part to be welded.

You can see the eccentricity of the tool on the small dents that appear on the edge of the piste the tools rotate at 3500rpm and the advance was 500mm/min it take 4 minutes to make it

[harryn]

I know it is difficult to obtain in Europe, but if you can find some SnPbAg solder paste, it works wonderfully.

I just solder it by heating the pcb in a frying pan.

Very nice work you are doing.

Harry

[rokag3]

I know it is difficult to obtain in Europe, but if you can find some SnPbAg solder paste, it works wonderfully.

I just solder it by heating the pcb in a frying pan.

Very nice work you are doing.

Harry

If it's difficult in Europe say it's almost impossible in greece !