Hi Guys:
I heard of the activity here and thought Id poke in and read up a bit.
I think we may have several things going on. So lets discuss CV a bit so we're all on
the same page. Its best if we all understand what happens in the various CV modes and a bit about why a cv works at all..
First, why have CV. Well, you know that one. doing 100 small moves in a line woudl be pretty nasty if they didnt join together
into one path. The problem of course is corners. So lets discuss that first, since most problmeatic jobs involve thousands of them.
We're fighting physics here. Plain and simple. If you drive down the street and have to make a corner at the end, there are a few ways to
do it. First, you could floor it all the way, but your likely to have a pretty round corner. Probably into the house on the corner.
Second, we could jam on the brakes at the end of the street, and then turn thw wheel, but we'd skid quite a ways (skidding is losing pulses here..
)
Or we can do what we all do, slow down, turn the wheel and quite often, anticpate the turn by turning as we slow. Mach3 kinda does this, but allows for any of the
last two really. But to show that , I need to show it to you visually.. No machine required..
First, makeup a Gcode file with
G1X10
G1Y10
g0x0y0
m30
Load it up and you can see the rigth angle lon the right of the screen. So lets test some things..
First run the file with F100 in he Feedrate box. Note the amount of rounding. Now run it at F600
, more rounding? Lots more? Some of this will depend on your personal accel settings, so you may have
to use different feedrate. Turn off CV tolerance and angular check for this.
So if you experimented with a few runs of this at various speeds, you'll see exactly how Mach3 blends,
as the X starts to slow, the Y starts to accelerate. Mathmatically, this makes the best join at the feedrate
you requested. If you didnt notice an asymetry, then your accel is the same on each axis. Use motor tuning to
slow the Y acceleration down a ways. Try it again, see how the Y is now not symetrical ? Think about the way
as one slows the other speeds up and youll see the math behind the blend. Notice the much less rounding at
various speeds. Pick two, say 100 and 600, and take special note of the difference between them.
Now use general config to set a CV tolerance of 1.
Run the two test again, do they look the same? (They kay or may not depending on your feedrates). But youll notice
only the last 1" is rounded ont he X side, the Y will depend on its accel in the X decel time of its last inch. The CV distance then,
is telling the system not to start blending until we are a set distance from the end of the move. This also translates to a lower
feedrate point at which it will start blneding, hence a lower amount of deceleration time in blend, thus less rounding..
OK, now try config/general to set an angular check of 89 degrees, run the test, no rounding at all , right? Because the
right angle is more than what was set , set the angle to 91 degrees and full rounding will appear because the allowed cv angle is
greater than the 90 degree change at the end of the line. Play wth the accel settings, without changing velocities to see some of the effect.
The thing you should notice is the faster you get in the segment with your F move, the more rounding will result unless corrected by distance tolerance.
Since distance is a function of velcoity and tiime, another way of thinking about distance tolerance, is the speed at which blending will begin. Thats basically
what it affects..
Running a few games with that simple file will allow you to fully visualise whats being done.
I saw above some suggestion of analysing vectors to determine when to slow to get to full speed the fastest, well, thats exaclty whats being done, if you analyse
that to the end, youd find that basically means blending the two moves together in one way or another to smooth the transition of one vectoral velocity into another.
Mathmatically, Mach3 does this very well, as it actually shows the vectorral change as a circle on the corner which is what such a functions waveform SHOULD look like.
All that being said, corner rounding is a problem, very frustrating and annoying. One afflicted with it wonders why more people arent upset with it. Its a function of power.
The more power you have, the less you see it. My Mill would need micrometers to see any rounding, as the acceleration is always maxed out. This problem afflicts those with
heavy gantries, with not enough power to push them. Generally, if it takes a second to get up to speed, thats a second too long from a physics point of view, I mean if acceleration
is infinite, then no blending is necessary, the Y would be up to speed instantly as the X stops instantly, thus no blending is required. Its a sliding scale downwards from ifinite acceleration
,the lower the acceleration, the more blending of velocities is necessary. Its the difference of driving a sports care around the corner , or a mach truck.. one slows more than the other ,
and rounds the corner more..
OK, dont get upset yet,
, I m not picking on your power. I just want to make sure we're all on the same page when it comes to CV and what it does, as well as how it does it. And
why your type of systems represent so much more of a challenge. We need to understand all this because Im getting undercurretns of perhaps other things going on with this type of
problem. Someone mentioned I think that when their motros were going in a certain range, they sounded good, in a lower range, they sound bad. Thats bad. Shouldnt happen. Its
a suspicionof resonance, and if it is resonance, it can really mess things up when your on the edge of performance already. So if its your system thats doing it, you need to run some tests..
G1X10F10
G!X0F20
G1X10F30
etc..
Where is the sweet spot, better high, worse low? better even lower? any particualr speed bad, if so, you need to dampen some vibration, may change the pinion arrangement,somethign to steady
things up a bit, resonance can be a nasty thing that takes away smoothness no matter what you do..
OK, but lets say it isnt resonance. Its just CV , and it drives you nuts. (It has driven me nuts before..never did recover..). Notice form the test we did above, that without changing acceleration, the rounding is
less and less the lower the feedrate the run is made at? Theres a key there thats important. Lets say for example, that we're cutting the roadrunner test file. I command 1000IPM, it runs at 60 or so. The reasonit runs at 60 or
so instea dof 1000 is that there is never an opportunity to get up to 1000. By the time it starts the accelerate, it has to decelerate. So it never reaches 1000. SO we're safe to set th efeedrate at 500 right? I mean its only
going to reach 60 in any event? Not a good idea. What happens, is that it actually reaches higher velocities in the middle of segments so it needs to decelerate sooner than if trying to go slower, so the rounding effect grows
exponentially with the higher demande dfeedrate, even though yourr average feedrate will be the same as a F60 to run the file. I guess hwat Im trying to say, is that not only do you have a feedrate you want to maintain
for a given chip load on your tool, but you have to understand there is a phsysical based feedrate limitation based on your tables accel and vel settings, and that if you exceed that physical limit, the CV will not act very good for you at all.
Thoise with high acceleration in their system can demand whatever they like from a file, they may get it, they may not, the file will limit the speed in the end, a feedrate is only a maximum speed, in most hard files the tools ends up
goiing much slower much of the time. So demanding a file run only to the limits of your systemns ability will sometimes make differences in orders of magnitude of the performance. Now thats for complex files with hundreds or thousands
of segments were talking about there. If you simply cutting squares in wood, then its different. Nope. Even there slowing down the requested feedrate will make big differences in the rounding. Depending on how your settings are,
it can go al ong ways to fixing it. Many play with accel, angualr, CV dist..etc.. and never slow the commanded feedrate int he file down. But feedrate has the greatest contribution to the problem. So slow down. Youll
usually find a speed where it actually seems smooth in the corners with little rounding. Thats the real, true cv limit speed of your gantry or machine base don your velocity and acceleration. CV distance, and angular rejection
can help you command somewhat higher speeds than that, but Id find that natural spot just so you know what it is. If its 5IPM, then you know that when you go over 5ipm, your gonna have to add helpers like cv distance and such,
or add another motor, or get a bigger one for that axis...
I am continuously experimenting and trying to makie better CV algorithms, and technology slowly catches up. This is a very complex area though, I wont bore you with the technical detaisl of what CV takes and such, this has been long enough,
but I hope the suggestions help, le tme know if you see any anomolies that my explanaion does seem to jibe with. Im always on the lookout for bugs in this area, but I haven tseen any so far. I hope to be able to make it better as I go,
but I cant say I have any brainstorms at the moment to make this better, other than the suggestions above. Now thats Ive posted on this one, I'll be notified of any additions..
Later,
Art
Guest
