Separate Prime and Suck Speeds

[626Pilot]

Same model ("vampire skull-to-pus" from Thingiverse)
Same PLA (Prototype Supply translucent gold)

Top: Printed with 20mm/sec prime and retract, 3.0mm distance
Bottom: Printed with 20mm/sec prime and 50mm/sec retract, 3.25mm distance

Results: It printed noticeably better. There is still some blobbing, but it's less pronounced, AND the filament bears fewer stress lines. It's not as easy to see in the photo, but the top print (20mm/sec both) has a bunch of layers that reflect light from wider angles, suggesting that they didn't all flow with the same ease. This suggests to me that retracting at 50 and priming at 20 is better for the "thixotropic" effect than doing both at 20, which is puzzling. It could be that retracting it quicker makes the extruder spend less time with hot filament higher up, decreasing the heat soak effect.

I wasn't able to run 3.25mm retracts before, but I can now. Anything over 3.0 usually causes jamming after about 20 minutes, but the second print ran for several hours before Repetier screwed up. I'd have to say this change is a win, but I think there's still more that can be done about blobbing. There are a couple threads here about slowing down toward the end of a path, and beginning retraction just before the end of a path. I think those put together would probably take the blobbing down to such a low level that it isn't noticeable anymore.

Another thing that occurred to me: What if, when you're drawing a perimeter, instead of retracting right at path-end, you immediately deflect "inward" to the infill space, and THEN retract? At that point, if it oozes, it oozes inside the part, where it doesn't matter.

Retraction.jpg (95.96 KiB) Viewed 2283 times

[Dreide]

Top: Printed with 20mm/sec prime and retract, 3.0mm distance
Bottom: Printed with 20mm/sec prime and 50mm/sec retract, 3.25mm distance

Hmm, interesting. I wonder if the faster retraction gives better results because of the physical processes going on or just because the retraction finishes earlier and thus the nozzle can get away earlier. What happens if you retract 1.3mm at 20mm/s instead of 3.25mm at 50mm/s (both taking about the same time)? Is this with a bowden feed, 1.75mm filament? Otherwise, 3mm sounds like a lot of retraction.
I think we face at least two fundamental problems when it comes to blobbing. One is the pressure in the extruder which needs to be released at the end of a path and increased at the beginning of the next path. Because the necessary pressure depends on the extrusion speed (extruded volume per time), it is problematic to take care of this with a simple prime and suck that does not take into account the extrusion speed before the retraction and after priming. This is likely something for the firmware, because only there all the necessary details will be known (speed limitations, acceleration, deceleration, etc.). Repetier has the optional ADVANCE feature which is supposed to do just that (that is, adding extra extruder control for increasing or releasing the pressure according to the current nominal extruder speed), but the algorithm is flawed (inherently so, meaning no easy fix in sight, IMHO). Also Marlin has it (or had it), but when I once tried it, it was broken as well. Even if it was correctly implemented, I think it would require different settings for different materials (for example, ABS and PLA seem to respond differently to the advance algorithm).
The other problem is oozing of the idle extruder during travel moves. That is probably something that can be tackled at the slicer level with suck and prime (static, as we have it already, but also dynamic).
My bottom line so far; a lot of Voodoo and guesswork is involved here, and one cannot easily take insights found with one machine/material and draw general conclusions that hold true for other machines/materials. For example, Davide suggests, in his valuable KISSlicer manual, to set prime >= suck. In my experience, however, prime < suck works always better.

Another thing that occurred to me: What if, when you're drawing a perimeter, instead of retracting right at path-end, you immediately deflect "inward" to the infill space, and THEN retract? At that point, if it oozes, it oozes inside the part, where it doesn't matter.

Isn't it this what KISSlicer is doing (see "Depth" parameter for seam hiding)? Or do you mean wiping perpendicular to the recent path instead of along which, I agree, could be beneficial?

[mhackney]

626Pilot, nice example of differential suck/prime speeds!

So, thixotropic fluids work like this...

Have you ever played with Silly Putty? It is the classic example of a thixotropic fluid. If allowed to flow without pressure, it will flow nicely. If you grab two ends and then pull apart briskly, it will snap cleanly, like a solid. Monte PLA works very much like this. So the fast suck "snaps" the small strand of filament in the nozzle. If I were to guess, I would say this happens right after the narrow bore where the nozzle widens. This removes all/most of the back pressure from the small remaining filament in the bore - which is held in place by capillary forces. Now, at prime time (no pun intended!) the PLA is pushed slowly (allowed to flow without exhibiting thixotropic behavior) back into place, ready to print.

The issue with priming fast is that for a split second, the molten PLA "locks up" due to it's thixotropic behavior and that creates a fair amount of back pressure. This is the effect that I've called "hydraulic plugging" on other forums that leads to a hot end plug. It is easily observed with parts that have a lot of quick suck/primes. I have a part (the orange thing posted on the previous page) that I use to test this and it will plug most hotends with 50 mm/s suck/prime on the first layer. Slow down, and you will print nicely.

I'd love to have the time and lab to test and validate and study this. I think a lot could be learned about hot end design and slicer code generation.

I'd also like to go on record by saying that I have studied this phenomenon for over 3 years on at least 11 different printers from RepRap Huxleys to other Cartesian printers and deltas like Rostocks to Kossels and with a variety of extruders and hot ends and I've shared this and helped literally dozens of others solve PLA printing problems by understanding how suck/prime affects this. There is no doubt that other factors contribute to the issue (Bowden tube hysteresis for instance). I go out of my way to hold all of these other factors constant so I can focus as much as possible on this issue.

Before Jonathan added this feature at my request (thanks again!) I am not aware of any other slicer that allows you to differentiate suck/prime speeds. I have been using Python post processing scripts to alter suck prime speeds for over 18 months, that's how I've been able to get the results I get. I understand when someone comes into a forum and makes "this claim" or "that claim" that there is little or no context for the validity of the claim. Folks that know my work on other forums (particularly the SeeMeCNC forum) know that I take a very methodical and thoughtful approach to 1) figuring out what is going on 2) designing experiments to test my hypotheses and 3) using that knowledge to effect change. I don't expect and wouldn't advocate anyone taking my (or anyone else's) word as gospel, that's how we all learn and advance - especially at this early stage in the technology.

One last point, fast suck/prime on ABS works perfectly well, it is not as thixotropic as PLA and a lot less sensitive. I think many people fall into the trap of thinking all plastics (i.e. PLA and ABS) are the same and tend to treat them that way. It's also true that the 80-20 rule likely applies - even what I consider to be poor suck/prime speeds will work under many circumstances for many parts (the 80% case) but for those other more persnickety parts and situations (the 20% case) they do not. If I can print those difficult parts cleanly and reliably, I am very confident everything will print much better over all.

Sorry for rambling a bit but I thought it important in this case since the concept of controlling suck and prime speeds is not something a lot of folks have thought about.

cheers,
Michael

[Dreide]

So, thixotropic fluids work like this...
[...]
Sorry for rambling a bit but I thought it important in this case since the concept of controlling suck and prime speeds is not something a lot of folks have thought about.

Nothing to be sorry for, I think. Thanks for the insights. Very helpful.