[rokag3]

Hello.
there is DRO which are equipped with 5 micrometer glass encoder
those encoder are quadrature and most of them differential (a+a-b+b-)with index often
it will be very easy to amplify and recopy the signal
i have made a little prototype with a motor for raising up or down a thread mill equipped with an trapezoidale screw and an impressive gear box
i fixed it on a piece of wood with tape
i use a linear mylar encoder from a printer and i put the signal on my hp uhu working at 24V for this experience
guess that the rigidity of this was (censured)
i put my comparator
precision 1/20 of milimeters !!!
with the motor moving a pure mechanical nightmare !!!
there was some hunting on the begining but i manage to parameter uhu very well . this piece of .... was positioning very well.
So i believe that with a very good encoder it can be done.
i try to use a rotary encoder with poolie but vibrations make it very difficult

[kreutz]

Hello.
there is DRO which are equipped with 5 micrometer glass encoder
those encoder are quadrature and most of them differential (a+a-b+b-)with index often
it will be very easy to amplify and recopy the signal
i have made a little prototype with a motor for raising up or down a thread mill equipped with an trapezoidale screw and an impressive gear box
i fixed it on a piece of wood with tape
i use a linear mylar encoder from a printer and i put the signal on my hp uhu working at 24V for this experience
guess that the rigidity of this was (censured)
i put my comparator
precision 1/20 of milimeters !!!
with the motor moving a pure mechanical nightmare !!!
there was some hunting on the begining but i manage to parameter uhu very well . this piece of .... was positioning very well.
So i believe that with a very good encoder it can be done.
i try to use a rotary encoder with poolie but vibrations make it very difficult

Hello;

Is there any way to get more info on those glass scales?

Thanks,

Kreutz.

[cnc2]

Hello;

Is there any way to get more info on those glass scales?

Thanks,

Kreutz.

Hi Kreutz !

check this out: http://www.cnczone.com/forums/showthread.php?t=55231 it shows a pic of an encoder and its wheel (i think it's mylar, just a kind of rigid plastic) look at the transparent wheel on the toothed gear it has 1200 CT (counts !) and 200LPI (line per inch) this encodes the main shaft of the printer.

Well, on the same printer there's a linear encoder made of the same mateial using the same sensor that the one on the pic, this encodes the printhead movement in HP printers (inkjet ones), the linear encoder is aprox 30 cm long and seems to be 200 LPI (do the math !), i have an HP PSC1510 & it uses the same encoders, the sensor is an Agilent.

Regards,
cnc2.

[rokag3]

Very detailed technical info
http://www.heidenhain.com/wcmsmimefi...3_20_15180.pdf
Unfortunately very expensive scale
I am looking for much cheaper (same precision)
there is some kits very interisting
I have made a roller screw for my lathe my backslash is 4/100 mm this backslash mean 3.6 degres on my screw not a big deal to stabilize and avoid hunting effect.
it is a chinese lathe i am sure that i can compensate with linear encoder
backslash is the main

[ironDigit]

heidenhains will xost you a knee and a rib new.

Unfortunatly i think you will need to consider what quality DRO you want to use for feedback since a DRO that displays 0.00001" 'does not necessarilly say it is always precise ad repettitive to 0.00001".Heidenhains ARE, most of the time when installed exactly correctly but the cheaper ones sure ain't.

I do believe it's the best feedback a machine can have since an encoder on your motor will only tell how many turns the motor/screw made and calculate theoreticalposition from there while a DRO will "take up all backlash" and only count actually moved distance though i can only guess to much backlash will make it a pita to tune the motors and maybe impossible if it''s not very repeatable.

me no expert by the way so please prove me wrong and i will be the first to steel your design.

[bunalmis]

I use linear encoder at my milling machine. My machine doing 0.1 mm error by 5um encoder. (While aluminum machining)

http://cnczone.com/forums/showthread.php?t=56761

[kreutz]

Thanks for the info. I think I could get a couple of Heidenhein incremental scales about 1000 mm each (for free) for testing. Will let you know.

[ironDigit]

i can almost guarantee that of you het the heidenhain's to feedback it will be a kickass system you got running.And there might be some sadness when the owner comes to collect.\


good luck.

[sigma relief]

From a conceptual standpoint, linear scales are the way to go when it comes to determining where an axis is in its travel, but I have to think that they will lead to resonance in the system. If there is a small amount of slop in the assembly, the motor will have to compensate for this in each direction change, but the tuning parameters may have trouble dealing with this dead zone. Even advanced higher derivative PID control loops like the one used in the UHU can have difficulty with this type of system.

Speaking from a strictly academic point of view, when the power to the motor is reversed, the UHU will take Velocity and Error into account, but will not understand the large change in required force to move the motor. With a traditional encoder, the UHU will register an edge much quicker than anticipated and will adjust power output quickley to avoid an out of control motor. With a glass scale, the UHU will not respond untill the axis moves, but the motor will have to rotate farther than anticipated with a less than anticipated force.

While this seems desirable, the problem is servos are an error feedback system and only respond when an error is registered from the encoder. When an axis is not receiving movement commands, the controller is still hunting to ensure that it stays between the two edges of the encoder. With linear scales, this means it must rotate the motor thru the entire slop of the system and move the actual mechanical components.

As far as what option is more controllable and or accurate, I have no idea. Either system could be accurate, but I highly doubt that linear scales will compensate for any significant mechanical slop without adding controll problems. If you really want an ideal system, use rotary encoders to drive the UHU and use linear scales to interface with the control software and use feedback from both the linear scales and the error parameters on the UHU to compensate for any error. The linear scales can easily be fed into a PC, but I have no idea what software can do the number crunching and motion control fast enough to be of any use.

[rokag3]

hello,
still looking for the cheap one.
but i can say that in order to do it you must be used to trim servo i have seen that the response must be less nervous if you have mechanical backslash
i can imagine that a lot of tester have been trapped by that because you like to have a very fast feeling this is a wrong approach , make it slower to respond so your servo will not hunt and then make it a bit more nervous but never forget that this is real motion and not an elasticity or a compression it's a full load real move !!! that's why it should be cool

uhu is perfect for this job Mr. bunamis you should try it and i am sure that you will have something much better 1/10 th mm

[ironDigit]

i've seen some old anilam controlled bridgeports that used lineair scales as feedback and they rocked *ss and these were setup in the 80's.They did use both tacho as lineair feedback though and i can barely imagine a backlash-free bridgeport.

I can imagine a servo rocking back and forth to compensate for backlash whilst attempting to hold a position under load. If this feature can't be turned off and "assume" the cuttingloads don't exceed the drag of your ways(almost nil with rails)or some breaking sytem for when a specific axe is not moving. When going 0,0005 a breaksystem might be a usefull feature anyway.

[bishb25]

Hi all,

I run a machine called a prototrak DPM. It's a bed mill that makes use of both linear scales and rotary encoders. The ball screw in the X axis currently has about .003" backlash. The machine compensates very well. The motor tuning is automatic. I just type a service code into the control and it jogs the axis and measures rotary feedback against the linear feedback and the machine configures itself. Not too shabby for a DOS based system.

Motor hunting is an issue in this type of system. While taking heavy cuts across the Y axis I can see the X hand-wheel jogging back and forth to comp for the slack in the system.

I use the machine for 3D surfacing of permanent mold cavities and it performs very well. Even with the backlash it will still interpolate a circle within .0005 out of round. It's all in the tuning.

I just wanted to let you all know that it can be and has been done. How to apply it in a home built system I don't know. It would appear that you will need real time feedback monitoring by the control not just a closed loop driver with a "blind" control.

hope this gets the wheels turning

[rokag3]

hello,
for sure the combinaison of 2 encoders is the best solution now about doing it in diy
well i have got an idea
let say i send an order one step on the left to my uhu
i can sense if the order has been executed(through sense resistance ? or someone know more simple)
if yes(the motor move) then i read the linear encoder
if the value of the linear encoder is none(the tool has not move) then my little interface will generate a copy of the last order send here one step on the left to my uhu input signal and i go back to the loop
when the move is seen on the linear i stop the process
of course the electronique of the interface must be very fast
Is it too simplistic ?

[kreutz]

Dual encoder use is not new, the encoder attached to the motor is used as speed feedback, while the encoder attached to the load is used as position feedback.
The firmware (taking care of PID/PI loops calculations) should take care of using each of encoders' information on the right loop (outer position loop and inner speed loop). Aditionally, a third, faster loop (torque loop, inner to the speed loop) could also be present.

[rokag3]

there is some cheap electronic to do that ?

[kreutz]

there is some cheap electronic to do that ?

In principle, it should not be difficult to design such controller (it could be a little more expensive, because of the processing power required to calculate two or three PID /PI loops at different sample rates, e.g: as fast as 50 uSec or less for the inner torque loop, 0.5 mSec or less for the speed loop, 1 ms for position...) The problem comes when trying to tune them....

[samco]

http://jmkasunich.dyndns.org/cgi-bin...-02-20-08.html

sam

[rokag3]

i am making research on "dual encoder" "rotary" "linear"
i find this
http://www.pmdcorp.com/downloads/Mic...ge_7-30-08.pdf

[LMSC]

Hello rokag3,

I can offer very inexpensive scales to you that either have A and B signals or A/ A not & B/ B not signals. The scales are 5 volt TTL square wave out of the encoder. The resolutions are 5um/.0002 inch or 1um/.00005 inch. My e-mail address is [email protected] if you want more information.

Regards,
Tim

[rokag3]

hello,
i have seen cncbrain very interisting

http://www.cnczone.com/forums/showthread.php?t=59763