This may actually be useful for other people.

When I went to test different extrusion speeds for my 0.75mm nozzle on the Ubis13 I was concerned about how fast I could go and still have the heater keep up with the target temperature. (After my experience of being down 8-20C from target depending on speed and the details of the print in question.)

I fired up Pronterface (Octopi is still mediocre for manual control) , positioned the hotend somewhere in midair and started wasting T-glase.

And I found something else: it seems that no matter how high I crank the target temperature (within the material’s limits), above a certain extrusion speed the heat simply doesn’t get into the filament in the same way. Or something. The filament comes out of the nozzle, starts cooling, shrinks lengthwise and immediately swells to roughly double the nozzle diameter.  It also gains a rough surface. Below that speed I get a nice glassy surface and negligible swell (a diameter of about 0.83mm).

Faster extrusion on the left, slower on the right

I’m pretty sure this is about heat (although it could also be about viscosity and some sort of weird adhesion to the rim of the nozzle, because the extruded plastic does a weird side-to-side dance as it comes down). For the first few centimeters after a pause, the filament comes out glassy again, which tells me that a little bit of preheating is enough to get the plastic up to a reasonable temperature.

245C, 500mm/min and up, with slight pauses

The extrusion speed for swelling to start varies a little depending on the hot-end temperature, but not as much as I might have thought. At 220C, anything above about 300mm/min of incoming filament (which my mediocre math skills translate to a free-air output of about 27mm/sec) starts getting weird. At 245C (which is as high as Taulman says is OK) I can crank the rate up to 400mm/min and still get a mostly glassy surface.

What does the swell mean for real prints? I have no idea. When I pulled on the filament as it was coming out, it stretched back down to a normal diameter (but still with the rough surface texture, and I wouldn’t want to bet against any interesting internal stress patterns). At 245C I could crank the extrusion rate up to 700mm/min (which would be somewhere about 65mm/sec for a full-thickness layer) without skipping; at 220C, skipping started around 500mm/min.

This is what a stepper skip looks like feeding T-glase way too fast

I’d be interested to see the same kinds of test from a hot-end with a longer hot zone. Can you go faster and still stay glassy?
Also, I was a little amazed at just how tenaciously PET strands stick together when they’re hot — even just lying on top of each other after falling from the extruder makes a fairly solid bond: