watching both marlin and klipper progress percents, i do not believe they are using anywhere near an accurate algorithm;
what i believe is that they are using a “layer-based” algorithm;
that is to say , for this discussion purpose, if a model uses 100 layers and the first 5 are complete then the progress indicates 5%;
But, if the model, like so many, is conical/pyramidal in shape then those last few layers should be much faster than the first few which are at the base;
i do appreciate that getting a true percent would be quite algorithmically involved, i do not believe such is being employed at present
HI @vernd
I do agree with you, I have spent many hours looking at prints completing. I have noticed the same thing. It does only happen I find the % displayed in the printer.
If you populate the Jerk and Accl settings in your slicer the time estimate becomes very scary accurate. I hypothesized that these settings in the slicer do not actually affect the printer itself, In other words, do not hand the settings onto the print via gcode, they are used by the slicer to calculate what the actual movement times are based on the parameters you input.
I have not tested this theory however by actually looking in the Gcode for any modification settings but the print physical time does not seem to change at all, Just the time estimate from the Slicer itself.
Just my $0.02 - the only printer I have that produces accurate progress percentages along with expected print times is my Zortrax M200.
The issue is that the slicer information and the printer firmware are fully integrated in the Zortrax products whereas with virtually all other printers they are separate and can have settings that affect the times that aren’t properly recognized together.
I find that the slicer prediction is generally low by about 10% for traditional firmware and high around 15% for Klipper and just run with that.
that makes total sense, If everything is integrated it probably means they have the accl and Jerk settings the same as the printer.
Is that something you can see in the slicer or did they hide it from the user?
It’s not visible to the user.
1 Like
Hi there!
I too have noticed this on almost all of my printers. Luckily for me the estimation is usually much longer than the actual printing time is! Usually when I am trying to think about how much longer a print will be I take a look at the amount of layer done vs. the amount of layers left to go. I also take into account the overall size of the model and how large each section of it is.
This usually ends up giving me a decent estimate of how much longer I can expect the print to take. But I would be curious to see how it could be improved (Maybe a summer programming project 
)
If you have any ideas of possible things that could be done in order to improve this process I would love to hear it.
Thanks,
Matthew
For casual users, I don’t think it’s a big issue. As I run Klipper on all my printers now, with the exception of the Zortrax, I generally go by what Cura says and know that the print will be done a bit faster than that (as I said, about 15%) which is anywhere from a half hour to an hour or so for the prints I normally do.
I’d be interested in hearing what companies that do contract 3D printing say about the issue as any improvement in their print time means more money or a cost avoidance.
Thanks for your response,
That is certainty an interesting idea which I have not considered before. But I can definitely see them having some problems with it, especially scaled up to such large print farms.
I wonder if they have any ways of dealing with it themselves, after enough tuning you might be able to get the estimate much closer than it would normally be.
Definitely something interesting to think about.
Matthew