calculating backlash on a z axis.

[bmillmine]

Im using mach3 and there is a 20% backlash option. how does one calculate backlash on the z axis.
Because in 16 passes of machining, my z drops 3 inches. It is crazy. Its like the z is possessed.

Similar Threads:

• calculating pulse equivlent factor for a rotary axis in place of the x axis
• X axis backlash
• Z Axis Backlash
• 'Z' Axis Backlash ?
• Calculating "steps per unit" for rotary axis?

[RCaffin]

Your Z axis is not 'possessed', but it is getting one hell of a lot of noise from the motor drive cables into the encoder signals.
You should get ZERO creep.
It has NOTHING to do with backlash.

Cheers
Roger

[bmillmine]

I am having trouble figuring out what dip setting to use on my dm542 microdirver using a 255 motor and 5 threads per inch leadscrew.

[Eldon_Joh]

so mach3 has a few bugs with regard to blacklash.. that 20% default number is the velocity at which the backlash is added to the direction of motion.. and that motion is added after the drive has reversed direction, not before, and not during either.

That 20% of maximum velocity also has infinite acceleration. however if it is missing steps you should see this cancel out on average. you can lose steps because the backlash is added too slow or too fast, but again they should cancel out on average. but the Z axis usually has static weight hanging on it and more work is required to lift it.

Anyhow, I have noticed a number of people have noticed lost steps in only one direction. I noticed this in the Z axis and the Y axis (but not the X axis) of a romax router I was using to cut circuit boards in half for a production job (high feed rate but very low cutting force.)

The fix was to enable half pulse sherline mode in mach3, which makes the pulse lengths about 20 micro seconds long for the slowest kernel speed... apparently the parallel port drivers on many computers these days just don't have a low enough output impedance or their rise time is too long. --but why the step is only added or lost at the direction change, i have no idea.
-this is the only explanation i can come up with but i think a full investigation will result in other problems.

i run mach3 on a relatively old single core desktop, it works fine even on 2 us pulse lengths, so i just don't get it.. parallel port is on the motherboard, not an add on card.
all of my problems with missing steps in one direction only have been with auxiliary parallel port cards.

[RCaffin]

Hi Eldon

Missing pulses in one direction only often means you have your Step pulses set up with the wrong polarity. Try inverting the polarity for that line. It won't alter the number of Step pulses sent.
Yes, this is consistent with the use of Sherline mode - which you should be able to cancel.
As far as timing differences between an LPT and a something else ... nanoseconds, but they can matter.

Cheers
Roger

[bmillmine]

I reversed the direction of the y and x axis as it was cutting backwards but the z is still not zeroing. not sure what sherline mode is or where it is enabled but I am definately missing or adding steps on the y and z axis still. Been smashing my head against this every night for the past month now.

[RCaffin]

NO, NO, NO!
Do NOT reverse the DIR signal. That controls the direction of motion.

I am assuming you are using Mach3 - from your 1st posting.
Go to Config / Ports&Pins / Motor Outputs
Column 4 is for DIR: leave it set for the 'right' direction.
Column 5 is for 'Step Low Active'. Click on that to change it.
The Config / Save Config ! !

What this does is alter the 'rest' state of the Step pin. It won't alter how many pulses you output, but it WILL alter the state of the Step line while the DIR pin changes from Down to Up & v/v. You want the Step line in the not-active state when Dir changes. If Step is active when Dir changes you may get an extra pulse - or lose one.

Cheers

[bmillmine]

Yes you are correct in how I "reversed" the axis.

[Eldon_Joh]

Hi Eldon

Missing pulses in one direction only often means you have your Step pulses set up with the wrong polarity. Try inverting the polarity for that line. It won't alter the number of Step pulses sent.
Yes, this is consistent with the use of Sherline mode - which you should be able to cancel.
As far as timing differences between an LPT and a something else ... nanoseconds, but they can matter.

Cheers
Roger

no, that is a separate problem.
changing to sherline mode fixes the problem because the step pulses are now 20 microseconds long instead of 2 to 5.
the missing steps are not a result of the wrong polarity but rather something else wrong electrically with a slow parallel port or with software.

its a very predictable number of missing steps. like.. all of 1 of them. and only after changing direction.. in one direction only. if one step was lost on direction change then they would cancel out over time.. and i think most people would never notice it.. i know i wouldn't.. one step is 1/40,000 of an inch.

with the add-on parallel port card and my boss's romax router, anything other than 4 microseconds would not work at all. --evidence something is electrically wrong with the parallel port or with mach3's "virus" that is used to run the parallel port.
so, the z axis would fall at the rate of 1 step per up and down motion.
so for example, every hole drilled would be 1 step deeper for every drill cycle, and if pecking then it would be 1 step too deep per peck.

[RCaffin]

Hi Eldon

the missing steps are not a result of the wrong polarity but rather something else wrong electrically with a slow parallel port or with software.

Well, that is possible, but I have my doubts. The parallel port may be slow or fast, but as far as I know they ALL used the same MOSTEK chip, and the output register was loaded from an internal register in parallel. It was the only chip of that sort made, and it is now obsolete.

its a very predictable number of missing steps. like.. all of 1 of them. and only after changing direction.. in one direction only.

Yes, that seems to be the normal fault. It's not new.

if one step was lost on direction change then they would cancel out over time.

Ah, but that is not what happens in practice. It usually happens on a direction change in just one direction. Yes, that probably has to do with the slew rate of an antique CMOS output pin being slightly different in the two directions, but starting from the wrong logic state is the underlying problem afaik.

Could it be due to a SW problem? Well, unlikely, because thousands of users of the direct parallel port connection do not have that problem. However, if using the Sherline Mode solves the problem for someone, that's fine too.

What you do when the PC or the add-on card dies is a good question. The parallel port chip is now obsolete, dead and unavailable. The current crop of 'parallel port adapters' do NOT work with Mach because they do not implement the internal registers of the LPT chip the way Mach used them. So, we upgrade to an external motion engine. Incidentally, Mach is not the only SW package to have this problem: others have been caught too.

Mach4 does not drive a paralle port any more in its default configuration. A very clever 3rd party developed the Darwin driver so it could use a parallel port adapter, but you have to licence that separately.

Cheers
Roger

[Eldon_Joh]

If my count is right, two people on this form and myself have solved this problem by switching to sherline mode and all were probably pci pp cards

so yes it could be the risetime, however i have not spent a great deal of time investigating the matter and i suspect there will be more interesting problems discovered. perhaps some cmos non inverting buffers could fix the problem.

[RCaffin]

If my count is right, two people on this form and myself have solved this problem by switching to sherline mode and all were probably pci pp cards

Yes, it's a known problem all right. And Sherline does work for some, while for others inverting the logic levels on the Step pin works. Just how a novice with minimal electronics knowledge is expected to manage ...

perhaps some cmos non inverting buffers could fix the problem.

Buffers by themselves may not be sufficient, but I do not know. Ideally one would try to latch the outputs in a register, but the way Mach uses the pins is seriously at odds with the definition of the PP. Pin 1 is the Strobe for an LPT port, but Mach uses it for M3.
I suspect you would need a day or two with a multi-channel storage oscillscope or a logic scope. Or persuade the mfr of the PCI/PP card to give you some really detailed engineering drawings and timing results. Good luck with that one!

Cheers
Roger

[Eldon_Joh]

i have access to some very expensive 4 channel scopes... but unless mach3 becomes free.. it will die in a few years and so will the rest of the machines it can run on.. and everyone will move to external dedicated motion control boards.. as they should.

[RCaffin]

but unless mach3 becomes free.. it will die in a few years

Well, I am sure Artsoft would like to see it replaced by Mach4, and that within a few years. You may be right.

so will the rest of the machines it can run on..

Are you trying to predict the death of the PC or equivalents? That could take a bit more than 'a few years'.

everyone will move to external dedicated motion control boards.. as they should.

I certainly agree here. Given the low cost of the Raspberry Pi, and the even lower cost of the next gen design (CHIP, $9), it seems obvious that external engines will be the way of the future. Ha - buy a KB and a screen and get the 'PC' thrown in for free. Only free Linux and OS-SW will match the price.

BUT - will there be a standard interface design? I suspect we may see the DB25 (and 26 pin ribbon connector) with the Mach pinout remaining in place for a very long time. Thing is, it works, it is widespread, it is easy to implement, and it is hard to see any benefit to a new mfr of going down a different route. No-one buys Teletypes or VT100s anymore, but the serial RS232C interface is still with us as well.

Cheers
Roger