@CCS86 Pressure advance is by no means perfect. It's an approximation to try and deal with most cases that may not cover the edge cases and extremes very well. Any suggestions for improvement are welcomed.
@toddel Not really. Every machine is different. If it's a machine that was originally a marlin based printer you may be able to get the default values from the old firmware. Or if it's a common design you might find suggestions from other users of the same machine. Alternatively you could ask for suggestions in the forum by describing your machine and someone can suggest a good starting point and then you'll have to do some testing and tuning on your own. General rule of thumb is to start low and work your way up.
Some of the figures in that table are misleading, and some are plain wrong:
Thermistor resolution "Up to 0.16°C" doesn't mention that for a wide-range thermistor, the resolution is likely to drop to 10C or even worse towards the extremes of the range. This makes it difficult or impossible for the firmware to know whether the thermistor is present before turning the heater on.
Thermistor accuracy "1°C without calibration" is only likely to be true over a very limited temperature range, e.g. close to 25C (or whatever temperature the nominal resistance is specified at)
RTD resolution "1.2C" is completely wrong, for the Duet3D PT100 daughter board it is 0.03125°C
Thermocouple resolution "0.5C" is likewise wrong, for the Duet3D thermocouple daughter board it is 0.0078125°C
The response time of any type of temperature sensor is completely dependent on the packaging
Thermistors are for budget 3D printers. If you want to know your hot end temperature accurately, use a PT100, or a thermocouple if you need to measure very high temperatures. A PT1000 may also be a reasonable option if you don;t mind the resolution being a little lower than for a PT100. See our advice at https://duet3d.dozuki.com/Wiki/temperature_sensors.
@h975 ... There is a further complication: I don't believe the compression of filament within the Bowden tube behaves like a linear spring. So there are two non-linear effects to compensate for.
I think the solution will be to use a non-linear PA algorithm, which is more or less what you are suggesting. However, before we can say what sort of nonlinear PA curve is needed, we need to do some more measurements on systems with and without long Bowden tubes, and work out a good method of calibration. Your suggestions will be welcome!
I agree that the filament in the Bowden tube will not behave as a linear spring.
The reason I say that is because I am proposing that as a system with certain parallels, the Bowden tube could be viewed as a loaded structural column with fixed-free end conditions and lateral support. The buckling of such a column will produce a buckled shape which is a string of half-waves in the form of a sin wave. The number of half waves produced is associated with defined critical load limits being reached in the column. Each successive critical load limit that is reached will produce another half wave. The relationship between the number of half waves and the critical load limit is non-linear. For that reason I would say that the filament will not load up linearly.
The equation describing the critical load points is : F(cr) = n^2Pi^2EI/4L^2 where F is the load, n is the number of half waves produced, L is the length of the 'column' and EI is the product of the Youngs modulus and the second moment of area of the filament.
As the above formula is derived for statically loaded columns and the filament is in reality moving in a tube thereby adding friction effects to the driven load, this will not allow an exact application of this formula to predict all variables, but the point is that the same principles of column loading still apply and therefore an analysis of it can, I believe provide valid insight into the system behaviour.
As far as the friction effects are concerned it is my view that these will also not behave linearly. Each half wave will produce another contact point with the Bowden tube which will have the effect of increasing the frictional resistance to movement in the same non linear way as for the spring effect. The friction effects will stack up as you go up the length of the tube thereby adding to the load carried by the filament column.
Note that the critical loads are discrete quantities meaning that the filament would load up in a stepwise fashion as each new critical load level is reached.
I have an idea for a test that can be done to characterize the dynamics in the Bowden tube in order to gain greater insight into this part of the system. I will post this later in a response to your proposal on using a force transducer to test the response of the system.
I agree that a non linear PA algorithm would be the correct way to go. Hopefully the way forward will become clearer once the above hypothesis has been tested.
@fcwilt Hey thank you very much for your feedback as well I went ahead and tried what you advised above and it worked very well! Makes alot of sense to use that approach to keep things simple and I am now happy to report all points are probed and it gets quite close to the edge on each side still.
I used @jay_s_uk tip on changing my probe point to X35 and that also fixed my ramming of my X endstop when executing G32.
Thank you for the help, looks like probing is figured out for now!
P1 would allow jerk to be used between all moves and not just print moves. In practice this would mean between print and travel moves. So with your high acceleration it could be a bit much for the transition. Could try increasing motor currents or reducing acceleration a bit to see if it makes a difference. 15mm/s may also be a bit aggressive.
What print speed and travel speed do you have set? Do you have a separate accel value set for print and travel with M204 P T?
@phaedrux Im unfortunately having that issue on straight walls. so I can confirm its not STL deresolution. honestly pulling my hair out trying to find out why I get vertical lines so blatant on straight walls.
@dc42 I wonder if there is some merit in using two accelerometers for input shaping? One attached to the bed, and one attached to the hotend. Then the actual trace you care about to drive input shaping is the difference between them?
I'm not sure if it is as simple as subtracting one trace from the other in the frequency domain or is it more complex...?
Ideally, yes. However, the bed and the print head are likely to ring at different frequencies, so I doubt that it makes much difference.