Volcano/Bulldog step rate limitation

Excellent information,  thank you for spelling it out so clearly.  When I had the problem I understood the basics of the issue, now I have the science to support my theory.  At the time I did in fact reduce to 1/8 stepping to help solve the problem.  I am fairly certain the Volcano wasn't the issue as I could hear the motor fall flat on its face and see very erratic motion of the drive gear.  As a test I simply removed the filament to eliminate that variable and the BulldogXL did the exact same thing.  Now I realize that I was pushing the speed limits of the extrusion process as we know it and will continue to do so.  It is my belief that new technology and equipment will bring faster printing while maintaining quality is just around the proverbial corner.
It is conversations like this that will spark those new innovations.

Marlin is not very good firmware, to be honest. It has had features hacked in on what was meant to be a experimental basis, often times using poor coding practices. Those issues generally are not the cause of this particular limitation, though.

The interrupt handler that creates the pulses cannot be run much faster than 10khz. The interrupt handler can generate 1, 2, 3, or 4 steps per axis each time it is run, but if you start generating more than 1 step per interrupt, then the period between steps is unequal since you are getting 2, 3, or 4 really close together, than a pause, then 2/3/4 again.

It may be possible to further optimize the stepper handler, but I doubt there's much to be gained or it would have been taken care of by now. I haven't looked into it much myself.

Some helpful reading:

http://reprap.org/wiki/Step_rates
https://github.com/MarlinFirmware/Marli … tepper.cpp

Link fixed in OP.

Follow-up on this. Are you sure it is 10kHZ? This isn't really important for any reason, but it's driving me absolutely bonkers.

So let's say the step rate is 10,000 steps per second, and the steps per mm is 88. This would be the conditions for a stock Solidoodle running MXL belts on a 1.8 degree stepper at 1/16 microstepping. Given this, we can find the maximum millimeters per second via 10,000 steps/sec * 1 mm/88 steps =~ 113 millimeters per second. This isn't the correct answer; at a minimum I am capable of 150mm/s travel, but can actually get to 200mm/s and above. I haven't been able to drive it fast enough to cause any anomalies, so finding the breaking point isn't a way for me to find the actual value.

To achieve these speeds I'm seeing is it using quad stepping mode?

EDIT: Nevermind, I see the solution in Marlin/stepper.cpp, lines 538 to 548. It essentially skips every other step if above 10kHz, and skips steps 0-3 but keeps 4 if step rate is above 20kHz.

Essentially the reason I wanted to know is because this is telling me my retractions are too fast. I should probably stick to within the 10kHz step limit for these. Currently  using a hobb goblin at 285 steps per mm and 100mm/s retraction, which is 28.5kHz. Should probably back down to around 35mm/s retraction, which would be approximately 10kHz (9975 Hz).

Your assumptions seem right to me. thanks for making this thread, I actually leaned quite a bit trying to understand this. If you break down the math at the print speeds I typically use (> 100mm/s), Marlin is actually already overloaded (> 10kHz) and skipping every other clock tick. I can only imagine this has negative impacts on precision. Really need to upgrade to a Smoothieboard, or another board that can genuinely produce the steps.

The maximum speed I can use before Marlin starts skipping on the clock is 10000/88 =~ 114 mm/s.

Decided to check it out on the RepRap calculator just out of curiosity. http://prusaprinters.org/calculator/ Now unfortunately this calculator doesn't include jerk, not sure how much that will influence the graphing.

My travel speed of 300 mm/s yields ~25mm before maximum speed is reached.

My lowest printing speed of 100mm/s yields a fairly short distance before speed is reached, about 2.5mm