I was not trying to cast aspersions on you guys at all (But had to ask very tongue in cheek lol).
Mine just works as expected and is now even better since I converted to Flying extruder (Only change was to reduce motor current to around 400) oh and reduce my retraction from 9 to 2 mm again I am using a Genuine E3D V6 hotend with 0.4 mm nozzle and pt100 sensor. oh and use ABS or PETG usually. and only 1.75mm filament.
I will keep an eye on this thread and hope you guys find the answer.
It's strange indeed. I'm leaning towards the fact that it's not Titan related - other users have no probs with different geometry machines.
Ref the difference you are seeing between the two - could it be due to nozzle diameter? Has one got a bit worn compared to the other?
No probs. I've got a flying extruder variant on my CoreXY. The 3 Titans are suspended above the centre of the bed with a counter weight/pulley arrangement. It means I can have Bowden tubes around 300mm long with a bed that's 400mm square. Like you, my retraction is low - 3mm (but I've yet to refine it).
I'm also using PT100. Printing PLA at 195 which seems best on my machine, with that particular filament in tests I've done. Also, only using 1.75 mm filament (and yes, I've checked the diameter).
Ian just going back to basics. If you have the extruder set to move 100mm of filament through it when you command G1 E100 then your steps are calibrated correctly for the Titan. At that point if the amount of filament being extruded is less then you would suspect an issue with high pressure causing the extruder to skip steps/strip the filament, or in your diamond setup the hot filament taking an alternative path other than out the nozzle. All that does not apply though because you are experiencing too much extrusion, rather than too little.
With that in mind can you explain why you believe it is over extruding? I suggest going back to printing a single wall cube with no base, using only one extruder with no extrusion multiplier set and see what results you get. Also try a different slicer (say cura) on the same simple object and then compare the gcode to see how much extrusion is being asked for.
For a very simple print (say a single wall, single layer height square) you can measure the amount of filament that goes through the extruder for the.print and compare it to what the gcode extrusion amount was.
Basically trying to locate if the error is with the slicer settings or the firmware having eliminated the mechanics of the extruder.
Duet Wifi Hardware Designerwww.duet3d.comwww.think3dprint3d.com
Thanks for the input. For these tests, I'm only using a single extruder so we can discount any effects of mixing ratios. Although the large, multi coloured vase I posted was printed using an extrusion factor of 0.7 so I'm fairly confident that the issue is common to all 3 extruders.
Yesterday I started using DJDemon's calibration object from Thingiverse - (can't find it again otherwise I'd post a link). Using an extrusion multiplier of 1.00, the surface finish is crap with blobs all over it, there are bulges on the corners, lumps on the sides and lack of detail. The dimensional accuracy is good though. Printing again at with an extrusion factor of 0.9 makes little difference. Dropping to 0.8 gives a much better surface finish and good detail. Dropping it to 0.7 with this particular object shows signs of under extrusion with gaps appearing between the lines.
I think what I need to do next is accurately measure the actual nozzle diameter.
Not sure what you mean by the single wall, single layer height thing - can you elaborate please?
I mean print a cube that is only 1 wall thick with no base and top. If you grab a model of a cube, set base and top layers to 0, set the wall thickness to 1 and the infill to 0. You then get a very simple start point to pinpoint where your issue lies.
Something like this:
Although I would scale it up to 40x40 at least
I'm aware of the single wall cube method for adjusting the extrusion multiplier but not a big fan. There was a recent post on the RepRap forums which detailed why it's not a good idea.
What confused me was this
"For a very simple print (say a single wall, single layer height square) you can measure the amount of filament that goes through the extruder for the.print and compare it to what the gcode extrusion amount was."
end of quote.
Are you saying a cube one wall thickness and only one layer - effectively a square shaped string. Then how to measure the filament that goes through the extruder? The same way as for setting the extruder steps per mm?
Sorry if I'm being thick.
I was not thinking of the single wall cube to measure extrusion width, rather that as simple print rather than complex calibration object it would be easier to isolate the source of the issue.
For the single wall single layer print, or other simple short print, you can make the filament a known distance from the extruder input, run the print, measure the new distance and so see how much was extruded. Comparing this will the gcode will then isolate if the issue is with the printer - I.e it's extruding more than it should be, or with the slicer - I.e. the amount of extrusion requested is greater than what is appropriate.
Be careful with slicer setting that increase extrusion widths etc for first layers so they don't muddy the waters.
So, here is an update.
I made a simple cube 10mm x 10mm x 10mm in OpenScad. Then sliced it in Slic3r with 100% infill, no skirt or brim. Checked the gcode file and at the end Slic3r reported that it would need 417.8 mm of filament and the volume would be 1.0cm^3. All good so far. I checked Slic3r's maths. 1.75mm dia filament so (1.75/2)^2xPi which gave me the cross sectional area of about 2.405something and multiplied this by 417.8 and got 1000mm^3 (=1cm^3). Still good. Then I pulled a load of filament off the reel and marked it at 700mm above the extruder. Then I printed the object (which looks bloody awful) and measured the filament to the top of the extruder and got 263mm. So 700-263 = 437mm. That's what actually went through the extruder.
So on that basis, my steps per mm are about 4% high. Also, if I measure the 10mm cube, it is a bit lumpy and bulging but as near as I can tell, it's about 10.4mm in all three directions which correlates remarkably well.
Actually, I think setting the steps per mm this way under "real world" printing conditions is probably not a bad idea. It'll take into account the effect of retracting and un-retracting which may wear the filament down as it goes back and forth across the hobbed bolt and also the pressure and friction effects of the hot end and Bowden tubes.
So, I'll set the steps per mm down from 424 to 405.
In conclusion (so far) it seems that my extrusion isn't far out - maybe 4% high which I'll rectify. I need to look elsewhere to find why I need to use around 30% less extrusion to get a decent print. There must be something else wrong that is masked by under extruding. I've done repeated tests on tall cubes with different temperatures every 10mm or so and keep coming back to 190 to 200 deg C as being the best temperature with this particular hot end/filament combination. So I don't think it's temperature and I've lowered the speed and acceleration to less than I was using in my old Mendel.
There has to be an answer so all suggestions, however obscure will be greatly appreciated.
I'll keep digging.................
Edit. Sorry Tony, we must have been typing at the same time - anyway, I think I've done what you were suggesting.
Last edited by deckingman (21 October 2016 14:11)
To be honest whilst the difference in how we are each getting good results from our titan extruders is of academic curiosity I think you use the value which works for you. Maybe its to do with hobbed drive gear differences, maybe there have been revisions to the titan over time, maybe its pinch wheel pressure. But if you are printing okay then set it up however it makes most sense to you.
I printed this two days ago on my T3P3 kossel mini with flying titan extruder at 250 steps/mm. Those fan cowls are single walled.
So can be in any doubt its dialled in correctly its just around 60% of the steps/mm that the calculation for a titan suggests.
(Nice prints by the way). I agree with what you say but it's bugging me.
Yes posts crossing. If after your adjustments you get to the point that what you ask the extruder to extrude in gcode is what happens then you need to focus on the slicer settings to get good results.
Yes, Slic3r settings or machine mechanics. I'm using more or less what I've always used with Slic3r but maybe for some reason a CoreXY needs something completely different. Or maybe, it's the actual nozzle diameter that's incorrect - the micro drills I ordered haven't arrived yet so I can't measure it. I guess I could try another slicer but I know my around Slic3R
Either way, I'm now reasonably confident that the extruders are doing what they should do - it's just that I have to under extrude to mask some other defect.
I'm not going to give up - there has to be reason.
I seriously think that assuming its a corexy issue is a red herring - I'd be interested if you prove me wrong but corexy is just a kinematic scheme I can't see how it requires any different extrusion settings. The two titans I have to "under extrude" with if you follow that logic are both installed on deltas.
I will when my current print finishes closely inspect and measure the drive gears on each titan.
I think you have proved the issue lies at the slicer end because you did a print and there was only ~4% variation in the gcode commanded extrusion and what actually happened. Modifying your steps/mm so that error is reduced to a minimum will confirm this.
I recommend trying another slicer like Cura (I use an older version as AFAIK the newer versions don't support as many printer types).
OK folks. Here is the latest. I changed the steps per mm from 424 to 405 as indicated by the ratio of filament used compared to filament expected. Re-printed the 1cm^3 cube and measured the filament used. This time I got 410 actual vs 417 expected - not too bad and it shows that changing the steps per mm has the expected effect. However, the finish on the cube itself was pretty dire. One corner showed distinct bulging and the surface finish on the top was very "ribbed", although the sides look good. Then I printed another cube with an extrusion factor of 80%. So I was expecting to see 410mm (the previous value) x 0.8 = 328 mm. The result was..............330mm. Spot on. So that proves to me that everything is working "extruder wise" as it should.
What is freaking me out though, is that the finish on this last cube is PERFECT. It's silky smooth all round, including the top. No lumps or bumps, no bulging corners. When I measure it, it is a little small. Height wise it is 9.87 mm and width wise (both X and Y are the same) it's 9.71. So around 1.3% in height and 2.9% in X and Y.
How can this be with genuinely 20% less plastic extruded?
I seriously think that assuming its a corexy issue is a red herring - I'd be interested if you prove me wrong but corexy is just a kinematic scheme I can't see how it requires any different extrusion settings. The two titans I have to "under extrude" with if you follow that logic are both installed on deltas.I will when my current print finishes closely inspect and measure the drive gears on each titan.
Yes, it could well be Red Herring - I admit I'm clutching at straws.
Hi Ian.I think you have proved the issue lies at the slicer end because you did a print and there was only ~4% variation in the gcode commanded extrusion and what actually happened. Modifying your steps/mm so that error is reduced to a minimum will confirm this.I recommend trying another slicer like Cura (I use an older version as AFAIK the newer versions don't support as many printer types).
I don't see how it's the slicer. The numbers it came up were perfect for a 1cm^3 cube. I've modified the steps per mm and re-printed the cube and it used the correct (theoretical) amount of filament. So slic3r does all it's little calculations then at the end, adds them up (at least I'm assuming that's how it works) and comes up with the exact amount of filament and print volume.
............But I'll try Cura at some point.
I have never tried to check a slicer's volumetric output , vs what you should expect.
My point is simply that if the entire printer system (ie the mechanics, firmware, electronics) is doing precisely what it was commanded to do by gcode (ie extrude exactly N amount of filament) then the issue must logically be what it it being told to do by the slicer is not correct. I know that is not particularly helpful in saying what is wrong, which is why I suggested another slicer.
I hear you and a different slicer has to be worth a try - if only eliminate the possibility that Slic3r is at fault.
What I'm trying to get my head around is that I have proven that the entire print system is doing what it's supposed to do and extruding exactly the right amount of filament, as dictated by both manually extruding a fixed amount and by dynamically extruding an amount dictated by the gcode generated from a slicer. Also, it's not specific to any particular object. It's the same for a tiny 10mm cube and a giant 320mm tall x 220mm diameter "crinkly shaped" vase. Yet there is something wrong with my machine which means that, in order to get a good quality print, I have to extrude at least 20% less filament than is optimal.
My current line of thought is that 1mm^3 of extruded filament wants to take up more space than 1mm^3. Or alternatively, 20% less extruded filament is required to fill a given volume. Hence the poor surface finish bulging at the corners, and the tendency for the print head to drag on the recently extruded filament when using an extrusion factor of 1.00.
So why could this be? I'm aware that filament expands as it comes out of the nozzle - die expansion or some such I believe it's called. But it should shrink again as it cools. If it didn't shrink as expected for some reason, then that would account for it, but of course that begs the question "why doesn't it shrink back after expanding". Alternatively, if there were gaps between the lines of filament, then a smaller extruded amount would be needed to fill a given volume. That would also account for it, but there are no visible gaps that I can detect with the naked eye - certainly not 20% worth of gaps.
What if the nozzle diameter is (say) 0.6mm ( I won't know until I measure it) and I've told slic3r it's 0.5. In theory, it shouldn't affect the volume extruded but I'd expect the slicer to adjust the width between each line accordingly. If it comes out of the nozzle 0.6mm wide but slic3r spaces the lines 0.5mm apart, would it have the effect of squashing the lines together, making them bulge upwards? If it did, it would like a severe case of over extrusion as subsequent layers would compound the problem.
There's a thought............
The only times I've ever had over extrusion caused by the slicer, was when I accidentally used settings for 1.75mm filament for a 3mm filament printer and when I had the wrong nozzle size, set it for .5mm and it really was .4mm.
My understanding is that die-swell occurs most in free air and is approximately 20% for the materials/temperatures we use routinely in 3d printing. However it does not occur in the same way when you lay filament down onto a build plate or a previous layer, then it is less as the filament is constrained as it is extruded. ABS shrinks by around 0.7% on cooling but this is an order of magnitude less than the "effect" you're measuring.
Try printing some cubes with different nozzle sizes selected in slic3r. I can print with a 0.4 nozzle using slic3r set for a 0.3 nozzle and it works, sometimes the prints even look a little sharper, but the top solid infill isn't always as good, though often slic3r doesn't do such a good job of top solid infill anyway. I think all that happens is the default extrusion width reduces, and the minimum feature size reduces too, but with the larger nozzle tiny features are likely to be fuzzy.
To take your example in the last paragraph, I'm not sure that would happen. Firstly the spacing between the lines isn't usually your nozzle diameter because of the die swell. So Slic3r prints 0.48mm extrusion width (spacing between the lines) for a 0.4mm nozzle as a default, you can change it under the advanced tab. If you're laying down filament with a 0.6mm nozzle but slicer is setup for a 0.5mm nozzle, then the lines are by default 0.6mm apart, however they aren't going to bulge out (might do on first layers but that isnt unusual is it) as it is extruding the volume correct for a 0.5mm nozzle. This is the same as me printing with a 0.4 but using settings for 0.3mm - it works although some details are going to be poor. I suspect for a simple cube you wouldn't even notice the difference. It might explain your issue, but it doesn't explain why I see the same effect as I am using a verified 0.4mm nozzle and settings for 0.4mm and getting great prints but sending around 75% of the steps/mm to the extruder than I (supposedly) should.
@DJ. I'm not using default values. Slic3r is set in my advance settings to 0.5mm thick so this will override the default width. So if my nozzle is actually 0.6 dia, but slic3r is extruding setting the lines to be 0.5mm apart (as set in the advanced tab) then surely that will have an adverse effect.
I'm not saying that is the cause but just maybe........
I appreciate everyone's help but it just seems to me that whatever theory I come up with to try and explain what's happening, it just gets dismissed as not being feasible. I could do with some positive suggestions or ideas.
Sorry, it's been a long day. I've been playing around with this since 8.00 am. It's now 10:15 pm and I'm no further forward.
Have you tried a different roll of filament? Maybe that one expands more than usual.
Duet WiFi hardware designer and firmware engineerhttp://www.escher3d.comhttps://miscsolutions.wordpress.com
Yes, - about 5 rolls and in all 3 extruders too.