How good are YOUR actuals compared to CV_M's estimates?

General CutViewer questions

How good are YOUR actuals compared to CV_M's estimates?

Postby cnczane » Tue May 22, 2012 3:36 am

My actual runtimes are running about 70% longer than what CV_M is predicting.

And pretty much always have. I run it (and MeshCAM for that matter) under Wine in Ubuntu Linux, but just thought to give it a go under Windows 7 with--I'm proud to say for Wine's sake--identical estimates. But still, both low.

There doesn't seem to be much in the way of configuration options to "influence" CV_M towards a better estimate.

If I'm all alone out here in Left Field, I can live with that, but Mother wanted me to ask.

Thanks.
--
David
If you have not received a reply from me in over a year, I am not ignoing you: more likely I am fallen asleep under a tree. Again. Please poke me if you think it worth your trouble.
cnczane
 
Posts: 277
Joined: Sun Sep 21, 2008 4:15 am

Re: How good are YOUR actuals compared to CV_M's estimates?

Postby Randy » Wed May 23, 2012 6:20 am

David, I think that the big factor is acceleration. Neither MeshCAM in its estimates, or CV_M, in its estimates, know how fast your machine accelerates. I view it like the time estimate from my wife's GPS unit--if the speed is 55 mph you are traveling 55 mph. If you enter a zone of 70 mph, you are instantly traveling 70 mph. In CV_M's case, the more short discontinuous moves you have, the farther off the estimate will be, because your machine might not even accelerate to full cutting speed on the short moves.

I cannot imagine the overhead that MeshCAM or CV_M would incur in actually simulating the dynamics of the cutting, segment by segment. And if you are using Mach3 and have CV (Constant Velocity in this case, not CutViewer...) which cuts corners and doesn't always fully decelerate, all bets would be off anyway.

Personally, I use MeshCAM's and CV_M's estimates as a minimum estimate of how long I'll need to stay up and hover over the mill. :) Come to think of it, I've never thought to compare the two estimates and see how they compare for a given job...

Randy
All opinions in this post are mine alone. I am not a MeshCAM employee, I do not have a financial interest in MeshCAM, nor do I speak for MeshCAM. MeshCAM user since Beta 5 in 2003. viewtopic.php?f=11&t=15333 :ugeek:
Randy
 
Posts: 1812
Joined: Wed May 14, 2008 9:50 am
Location: North Texas, USA

Re: How good are YOUR actuals compared to CV_M's estimates?

Postby cnczane » Wed May 23, 2012 4:40 pm

Randy,

I see this is a different facet of the topic I asked elsewhere here; regarding how "trajectories" are created from MeshCAM's G-code results.

In fact what prompted that topic was my suspicion that my machine wasn't reaching top speed (or that it was being decelerated just as quickly as it reached it) on the sequences of a few hundred thousand segments of length ~0.01 in.

As you pointed out there, MeshCAM has no dog in the fight about how a sequence is rendered.

In my opinion however, it's hard to understand the rationale for a "simulator" not accounting for the dynamics of the thing it's simulating. :)

I wanted CV_M for that capability alone (not for the pretty DOS-era graphics :lol: ).

I'm sure I aggravated things by trying to tweak-up the performance (in TurboCNC v3.1a) with outlandish accelerations (1,000,000 in/s/s), but it would be nice to know what CV_M assumes.

Thanks a heap. BTW, to be fair: TCNC often reports things like "2 hours, 33 minutes, 33 seconds" for a job that's just run over 9 hours... And IT'S supposed to have been in charge!
--
David
If you have not received a reply from me in over a year, I am not ignoing you: more likely I am fallen asleep under a tree. Again. Please poke me if you think it worth your trouble.
cnczane
 
Posts: 277
Joined: Sun Sep 21, 2008 4:15 am

Re: How good are YOUR actuals compared to CV_M's estimates?

Postby cnczane » Sun Dec 16, 2012 1:05 am

One of my favorite pasttimes, in addition to listening to recordings of me talking to myself, is to revisit my old posts all over the web and give myself a well-thought-out and considerate reply.

(Actually, I stumbled face-down into this post while looking for something else, and realized it could benefit from an update.)

"Here's what works:" I am getting reasonably good approximations by:
* editing the output NC files to replace each Feed instruction with one 180% slower (e.g., 18.33 instead of 33ipm)
* multiplying Cutviewer_Mill's time estimate (at the original, unedited feed rate; e.g., 33ipm) by 180%

The 180% evolved over time as I compared my "actual" times with CV-M's "estimated". If your mileage does NOT vary, I will be flabbergasted.

Cheers!
-
c''zane
If you have not received a reply from me in over a year, I am not ignoing you: more likely I am fallen asleep under a tree. Again. Please poke me if you think it worth your trouble.
cnczane
 
Posts: 277
Joined: Sun Sep 21, 2008 4:15 am


Return to General

cron