The effect of fan quantity and fan shrouds on ‘short’ layers

Created on Saturday, February 13, 2016. I last modified it on Sunday, February 21, 2016.

‘Short’ layers are layers of a 3D print that finish so quickly that the plastic may not cool down before the next layer is printed on top of it. Good cooling is essential when printing these with PLA. I investigated how the number of cooling fans, and the shrouds fitted on them, affected short layer success.



Short layers are regions of a print that may not have time to cool before the next layer is deposited on top of them. Poorly-cooled short layers leave tapered sections of the print with an undesirable ‘melted’ look. In this article, I investigated what cooling set-up (number of fans and the type of fan shrouds used with them) produced the best-looking short layers with the least warping. I concluded that printer operators and designers should try to use two fans, one fitted with a shroud that distributes air broadly and the other fitted with a narrow-airflow shroud that focuses air at the plastic deposition area. If only one fan is available, a narrow-airflow shroud should be used instead.


In my first article on 3D print cooling, I tested how different types of fans and fan shrouds — accessories that shape and direct a fan’s airflow — affected the success, appearance, and warping of unsupported overhangs. My conclusions were:

  1. Use a blower fan because it can supply higher pressure, and
  2. Use a shroud that blows air over a large area, because this reduces warping.

In this article, I look at a different design feature in 3D printing.

Figure 1. A real-world example of a challenging print. This unicorn bust by kellesabelle has two difficult features: A challenging bridge in the unicorn’s muzzle, and a thin horn.

The unicorn bust in Fig. 1 has a thin horn. As it prints towards the top, the horn has less material in it, so the time required to complete each new layer decreases. I call these 'short’ layers because they print so quickly that the plastic may not have time to cool down. If new layers are placed onto still-hot layers, the shape will sag and warp and the print will fail. It’s therefore important to cool the new layers as quickly as possible. Here I continue the work of my first article, testing how the number of cooling fans and the fan shrouds fitted on them affect the success and appearance of a print that has lots of short layers.

I want to answer two questions:

  1. Should I use one or two fans? My 3D printer, the Printrbot Simple, comes with only one cooling fan at a time when many other consumer printers come with two. Is having two fans an improvement? Is it enough of an improvement to justify installing two permanently?
  2. What shrouds should I use to get the nicest short layers and the least warping? In my previous article I recommended shrouds that blew air over a large area because they reduced the warping of thin suspended planes. Are they also the best to use for short layers?

Cooling conditions being tested

One or two cooling fans?

I designed a dual-fan mount for my Printrbot Simple, and designed a custom blower fan mount to fit two blower fans of the same type used in my first article (Fig. 2). For tests that used only one fan, I swivelled the blower fan out of its mount so that it pointed out of the enclosure. I did not differentiate between which shroud was fitted to which fan; I designed the dual-fan mount to eliminate this as a factor.

The fans are connected in parallel to the Printrboard’s fan input via a Y-splitter extension. You might be wondering if this is safe, so I asked Brook Drumm, founder of Printrbot, about it beforehand:

Desi: Can the Printrboard handle two 12V 0.10A fans wired in parallel? I’m using a 6A power supply.

Brook: Yes, no problem. I’ve run three before.

Figure 2. Dual-fan setup. The dual-fan mount holds the two blower fans equidistant from the hotend. This means that shrouds can be fitted interchangeably on either fan mount.

Fan shroud types tested

I tested the same fan shroud types that appeared in my first article (Fig. 3), plus two additional shroud conditions:

  • None — No fan.
  • Naked — The fan blows onto the print without any shroud to direct the airflow.
Figure 3. Shroud types tested in this article. The shroud types encompass a range of approaches from no airflow shaping to very specific airflow shaping.

Treatment matrix

I tested all possible combinations of the fan shrouds above, except for the shrouds that would not fit while the Surrounding-type shroud was installed (Table 1).

Table 1. The combinations of fan quantity and shroud type tested here (23 in total), marked with white squares. Shrouds paired with the None option were tested with one fan only. Blacked-out combinations are duplicated elsewhere in the matrix. Most of the combinations for the Surround shroud type are blacked-out because the other shrouds would not physically fit around it.
None Naked Open Funnel Wide Nozzle
Sharp Nozzle
Wide Nozzle

Test prints

All of the models for this study are available for download here.

The test object

The test print is a narrow conical spike (Fig. 4) with the top cut off so that it is never narrower than 0.4 mm (the width of my nozzle). A conical shape was chosen to minimise air direction as a factor in the experiment; many prints are large enough to interrupt the flow of air from one direction, so the orientation of these objects relative to the air stream is important.

One side of the spike’s base is marked with a slot which indicates a common side for repeatable measurements between all of the different prints.

Figure 4. The 'spike’ test print. One side of the spike’s base is marked with a slot, which helps me measure each print in a standardised way.

Printer and slicer settings

I used Cura 15.04.2 to slice these models, with these settings:

  • Layer height 0.2 mm
  • 2 Shells
  • 3 Surfaces
  • 10% infill
  • 60 mm/s print speed
  • Minimal layer time 2 seconds
  • Fan always on
  • No cool head lift

Minimal layer time was reduced to 2 seconds (from a default of 15) to represent a “sensible minimum” setting that a printer operator might choose; enough time to cool the plastic a little, but not long enough to significantly increase printing time. Cool head lift was kept off because it ruins many prints by allowing the nozzle to ooze, and then dragging that oozed plastic onto the print where it can interfere with layer deposition. Cool head lift also has a problem with retraction that causes layers to skip.

I used the same PLA filament for these test prints as before. All spikes were printed within a period of 3 days. The temperature inside the printer enclosure during printing was 34–39 °C. The bed temperature was always 1 °C colder than the enclosure temperature.

Quantifying the prints

The printed spike needs to be measured in a repeatable way. The spike is pushed lightly into a measuring jig (Fig. 5), which is a negative with the exact shape and dimensions of the spike as it was designed in CAD. If a spike is melted and warped it will not slide as deeply into the jig as a spike that is perfectly shaped. A caliper jaw is inserted into a cut-out in the measuring jig’s surface, and the opposite jaw slowly closed against the edge of the spike’s base until it makes light contact (Fig. 6). This was repeated for all four sides of the spike (using the flat edges of the base to index each side) to measure the warping of the spike on all sides.

Figure 5. The measuring jig. The hollow shape matches the ideal dimensions of the spike. The square hole at the head of the jig is made to accommodate a caliper jaw.
Figure 6. The measuring method used in this study. The spike is pushed lightly into the jig, and calipers are used to measure the distance from the square hole to the spike’s base. This is repeated on all four sides of the spike.


Visual inspection of test prints

The print quality of the spikes appeared to improve as the number of fans and the specificity of the fan shrouds increased (Fig. 7). Once the fan shrouds became more specific than the Funnel-type, however, the visual difference between the spikes became harder to discern.

Figure 7. Completed spikes arranged on the treatment matrix from Table 1. The appearance of the spikes improves as the matrix progresses towards the narrower fan shrouds, from left to right in the photo.
Click here to look at the prints in closer detail if you like.

The page has a super large table, so use your browser zoom to navigate it.

Cooling schemes for short layer success

Because of the lack of replication here (I only printed one spike for each cooling setup, and that already took 8 hours of printing in total), it would be meaningless to talk about hard numbers because the sample size is so small. Instead, I will use an approximate weighted score calculated as follows:

  1. I measured from the base of each spike to the first printing defect visible, and recorded the distance as the height to first defect.
  2. I subtracted the nominal height of the spike in CAD from the measured length of the spike (minus the distance added by the jig) to find the protrusion distance. This number is directly related to the warping of each print; remember that a perfectly-shaped spike can be pushed more deeply into the measuring jig than a warped and bulging spike, so its protrusion distance will be less.
  3. I found the absolute difference between the length of a given side versus the three others, and then averaged all of those to get the warping.
  4. Finally, I assigned a weighting to each spike, calculated as Height to first defect − (Protrusion distance + Warping).

The weighted score favours the success of a print (ie. how much of the spike could be printed cleanly) and penalises any warping.

Overall ranking

Using no cooling fan produced the worst-looking print and the lowest weighted score (Fig. 8). The print quality and scores improved as the cooling setups began to employ shrouds that focused and targeted the airflow more aggressively.

Figure 8. All tested cooling setups, sorted by weighting (higher is better). It appears that short layers print best when they are cooled by constricted and targeted airflow. The None setup failed so early and warped so much that its weighting was negative and thus omitted.

One fan versus two for short layers?

The two-fan employments of the Wide Nozzle, Open, and Funnel types had similar weighted scores. This suggests that the volume of air reaching the just-deposited plastic can make up for any shortcomings in the shroud design. However, the higher weighting of the Sharp Nozzle shroud in both single-fan and dual-fan configurations speaks to the importance of targeted cooling for printing short layers.

Figure 9. Comparison of one- and two-fan cooling setups for each fan shroud type. The *Surround* type is excluded because two cannot be installed at the same time. In all cases, driving two shrouds with two fans produced better results.

Cooling schemes for reducing warping

Overall ranking

Uneven cooling can lead to shrinkage and warping on the poorly-cooled leeward sides of a print. I took the height measurements of all four sides of the spike, averaged the front/back and left/right height (oriented parallel and perpendicular to the flow of air, respectively), and then found the difference between the two heights to get a measure of the warping that occurred.

I found that wide-airflow shrouds, and the Open shroud in particular, produced the least warping (Fig. 10). This replicated my results from the first article and validated my suggestion re. warping and shroud choice. If uneven cooling is a cause of warping, then it is interesting that the Surround shroud performed so poorly. My hunch is that it is not enough to simply blow air at a print from all sides, but you also need to get the hot air away. I’m guessing that the Surround shroud is collecting warm air from around the hot-end, and localising it around the print because it’s blowing straight down towards the bed. This is in contrast to the other shroud types, which blow air through the printing area so that it can collect heat and then leave.

Figure 10. Amount of warping in the test print, presented as a height difference between sides that were parallel to and perpendicular from the direction of air flow. Less is better. Wide-airflow shrouds are well-represented at the top of the graph, replicating my previous results.

One fan versus two for warping prevention?

If warping is caused by uneven cooling, and a single blower fan cools less evenly than two fans blowing from different directions, then warping should be reduced by adding a second fan. I compared the per-side height difference for paired and unpaired shrouds and found inconsistent results (Fig. 11). When looking at only the one-fan results there is a clear improvement in warping as the airflow becomes broader, but there is no obvious pattern for the improvement (or even worsening) of warping once a second fan and shroud are added. At the very least, this illustrates that you can’t just keep adding fans.

Figure 11. Comparison of one- and two-fan cooling setups in reducing warping. Less is better. When stuck with only one fan, there is a clear advantage in using a shroud that gives broader airflow. When using two fans, then pattern is not so clear.

Conclusion: What cooling scheme should I use?

Table 2 ranks the cooling set-ups based on their print quality and degree of warping (Figures 8 and 10 respectively). Briefly, the best option is to use two fans. Outfit one of them with a narrow-airflow shroud like the Sharp Nozzle type, and the other with a wide-airflow shroud or no shroud at all. If you cannot use two fans, then use the Sharp Nozzle type of shroud.

Table 2. Recommended cooling set-ups for 1- and 2-fan printers. Click on the column headings to re-sort the table. A lower number is better. The best overall result comes from employing both narrow- and wide-airflow shrouds.
Number of fans Shroud 1 Shroud 2 Short layer ranking Warping ranking Sum (Layer + Warping) Rank difference Sum + Difference
2 Naked Sharp Nozzle 2 2 4 0 4
2 Wide Nozzle Sharp Nozzle 5 5 10 0 10
2 Sharp Nozzle Sharp Nozzle 1 8 9 7 16
2 Funnel Sharp Nozzle 7 9 16 2 18
2 Open Open 10 1 11 9 20
1 None Sharp Nozzle 4 11 15 7 22
2 Open Sharp Nozzle 3 12 15 9 24
1 None Funnel 14 10 24 4 28
2 Naked Surround 6 15 21 9 30
2 Naked Wide Nozzle 16 6 22 10 32
2 Naked Funnel 17 14 31 3 34
2 Wide Nozzle Wide Nozzle 11 17 28 6 34
1 None Wide Nozzle 18 16 34 2 36
2 Open Wide Nozzle 15 18 33 3 36
2 Funnel Wide Nozzle 12 19 31 7 38
2 Naked Open 19 4 23 15 38
2 Naked Naked 20 13 33 7 40
2 Open Funnel 13 20 33 7 40
2 Funnel Funnel 9 21 30 12 42
1 None Naked 21 7 28 14 42
1 None Surround 8 22 30 14 44
1 None Open 22 3 25 19 44
0 None None 23 23 46 0 46
That's all there is; there isn't any more.
© Desi Quintans, 2002 – 2016.