Precise homing?

[hub]

Hi, I didn't find much info on the topic doing search.
So, would be nice to share some ideas / pick some brains.

Main point, it should be a "DIY-budget" solution..
And, to be able to home the machine to a very accurate position (+-0.001mm)

- Mechanical micro switches: (+)cheap, precise enough?, (-)will wear with use..
- Inductive prox. sensors: (-)not so cheap, (-)precision?, (+) long life time
- Capacitive prox. sensors: (-)not so cheap, (-)precision?, (+) long life time
- Photo sensors: (-)some are cheap, (+-)precision depends.., (+)long life time
- Laser sensors: (-)expensive, (+)good precision?, (+)long life time(?), (-)bad to eyes?
- Other?

Ideas? Experience?
I would go for a photo-sensor (sender-receiver) somehow making the path of light so narrow that the accuracy would be met.

[Bubba]

hub,
From most of the reading I have done, is to use a switch (such as a mechanical) in combination with an index pulse from an encoder.

[hub]

Originally Posted by Bubba hub, From most of the reading I have done, is to use a switch (such as a mechanical) in combination with an index pulse from an encoder.

That sounds a bit complicated to implement. It would need some sort of microprocessor/PLC to keep count of the pulses?
Sure, that would be good no doubt, but could there be a more simple (and cheaper) solution?

[Bubba]

No, basically, you trip the switch and then back off and wait for the index pulse. I have seen where it can be implemented with a flip flop circuit.

[mfgbydesign]

To expand on what Bubba has written:
The microswitch is not precisely set to the "home" position. It indicates the machine is close.
When triggered, the machine backs off to the next "index pulse" from the encoder. These are periodic pulses that the control uses to verify counts are not lost. In essence, as the machine moves, it counts pulses, and checks the count on the index pulse. If the two don't add up an emergency stop is engaged and the machine must be "homed."
So back to the homing process; after the switch is engaged and the machine is reversed to the previous index pulse, a parameter lookup is performed for the "grid shift" amount which designates the true home position relative to the index pulse. The machine is then moved to that location.
This method allows for a less precise switch to be used. I've seen one case where the home switch was set too close to the encoder index position and the machine would frequently home 0.100 off location. Moving the switch slightly solved the problem.

[hub]

Ok, I understand.
But how would one do that with say Mach3? It has home inputs which are binary on/off. It still would need a DSP in between to figure out the encoder position? And then send the "on" signal?

[Geof]

What are you wanting to home? Something driven by a PC driving steppers where the homing signal to the PC is just a switch closing or opening.

Just a simple microswitch opened by the linear motion of the axis is not going to give .001mm precision but I think you could make a system that would create the equivalent of an index pulse and home very precisely using two switches in series.

This would require an arm (or disc) attached to the feedscrew and one microswitch would be operated by this arm to detect the rotary position of the screw. The other microswitch would be mounted so it was operated by the linear motion.

The homing sequence would start with the axis away from the home position. As the PC drove it slowly toward home the linear motion switch would operate first and when the rotary motion switch operated that would give the homing signal. The precision would depend on the radius of the arm/disc operating the rotary position switch.

Naturally the rotary position switch is going to operate every rotation of the screw but only when the linear motion switch is operated will a homing impulse be generated. The constant operation of the rotary position switch could cause early failure so a system based on optical detection would probably be more durable.

To improve precision it may be possible to program the PC to reverse the axis direction as soon as the homing signal is detected, drive the axis very slowly until the homing signal turns off and call that point home.

[hub]

I want to home linear axis's.
For now, I'm using steppers & 1 servo with encoder, but all will be servos in the future. They are driven by step/dir.

That's a very good idea Geof! Sound like it could work. Simple and effective.

I have some ideas that might work (with a few problems):
- Digital dial indicator (positioned near limits) that outputs reading to PC
- Machine vision inspection with cameras (very expensive)
- Umm..

[samco]

here is the options with emc..

Homing Configuration

I use index homing on the K&T and it works awesome. I have 2540 count encoders with index - so my scale is around 60960 counts per inch. So I am homing within 1 count or .000016.

If you look at the bottom of the page it shows some homing examples. (and it can do the 'fast to switch - slow backoff home')

sam

[hub]

Thanks
But I don't have that option really. If possible, I would like a hardware solution. If not, I can possibly program a DSP to do it, but the ideal solution would be a simple accurate method of switching..

[Bubba]

Ok, did some digging on my hard drive and found a post from Tom Hubin back on ~2005 on the CadCamDro list.

"Here is one way to do it with half of a LS74 dual D flip flop, two 1
kohm resistors and a normally open limit switch.

The normally open limit switch is wired with one contact to ground and
the other in series with a 1000ohm resistor to +5v. The junction of the
switch and resistor is the switch output and goes to the flip flop Reset
line and D input.

The encoder index output goes to the flip flop clock input.

The second resistor is used to pull up the unused Set input. One end of
the resistor to +5v and the other end to flip flop Set input. Some folks
don't use a resistor and just wire +5v to the Set input. My early days
of EE was designing to meet mil spec and we tied NOTHING directly to
power except power pins. Unusued TTL inputs were pulled up via
resistors.

Be sure to wire the flip flop power and ground pins to the power supply
also. There should be a 0.1ufd, a 0.001ufd and a 10ufd electrolytic in
parallel with the flip flop power and ground pins. These capacitors
should be located very close to the flip flop.

How it works: Travel toward the limit switch until the limit switch
closes. The flip flop Q output will immediately go low when the limit
switch closes. Then travel away from the limit switch. On the first
rising edge of the encoder index after the limit switch opens, the flip
flop Q output will go high as the encoder index rising edge clocks the
flip flop.

When you first power up you should home twice since the flip flop may
power up either high or low. This will guarantee a valid home. You
should NOT be within one index pulse of the end of the travel range that
is opposite the limit switch.

Now this is not the ideal solution since you would probably prefer a
normally closed switch. That way the default for a disconnected limit
switch is safer. I did this for somebody several months ago. I'll have
to dig up my notes to see how I did that with a N/C limit switch. I may
just have used the other half of the dual flip flop as an inverter."

[Robin Hewitt]

Originally Posted by hub But I don't have that option really. If possible, I would like a hardware solution.

If you can keep it clean and dark, how about a slotted opto switch with built in Schmitt trigger. Standard part, real cheap.