The design of the fisher, while inheriting from the first personal delta printer Rostock by Johann C. Rocholl and other deltas, does have some specificities in its design.
RepRapPro is using for all its machines a board equipped with a 32 bit arm processor, the Duet (same processor as an Arduino Due). A delta printer requires much more calculation power than a Cartesian and 8 bit boards are struggling to operate deltas. So a 32 bit board is for a delta a major improvement. The configuration is also more simple than 8 bit boards, as the parameters are simply defined in a configuration file. All parameters are defined in files which are placed in an SD card. This SD card is also used for printing, which allow the printer to run in full autonomy.
Stepper drivers are on the board and motor current is digitally set by the board itself and can be modified during operation.
A web server is installed on the board and you operate the printer with this web server, through an Ethernet link, or if you don’t have a router, through a Wi-Fi bridge. The printer is autonomous so if you loose the link, you cannot see the parameters but this does not stop the print. There is no local panel. A very complete touchscreen panel the ‘PanelDue’ is sold by David Crocker and Think3DPrint3D. No simple panel is available nor seems forecasted.
Provided your network runs DHCP, the connection to the printer is automatic, simply calling its name (modifiable) in your web browser. Alternatively, you can configure a fixed IP adress.
The availability of a very capable web server in the main printer board is unique and makes it the only real ‘all in one’ board. Others board are using an external computer, frequently a RaspBerry Pi to run the Octoprint webserver.
Having such a capable brain in a low cost machine is unique on the 3D printer market.
It shall be noted that the main developments for the Duet used in the Fisher were not done by RepRapPro nor the board designer, but by independent programmers, the Delta Geometry and automatic calibration by David Crocker (DC42) and the Web server by Christian Hammacher. These people are now involved in the new duets boards
The firmware have an automatic calibration procedure which try to detect faulty geometry and compensate for it. It does indeed ease printing, but the software cannot really detect faulty geometry and creates a fictive geometry which drive to inaccurate dimensions.
As example of the problems associated with wrong geometry:
‘bowled’ beds (common on Fisher) is compensated by the automatic calibration procedure, but modify significantly the part dimensions. To recover proper dimensions, some have successfully tried to cheat the arm length to obtain accurate dimensions. typically, arms lengths set at 163mm were imposed (while real length is 160mm) to recover proper dimensions.
Non level bed (bed levelling is not adjustable on the Fisher) compensation drive to significant differences between the X and Y dimensions and squares may not be square. To check the bed levelling, you shall measure against the bottom plate. The level default is amplified, say 0.1mm levelling error will drive to larger dimensional error.
The probing is done by pushing the bed on its springs, which make the machine flexing. Due to load difference, there is some difference of deflection between center probing and near column probing. The Z-offset shall be adjusted for different probes point by measuring these differences. This deflection and associated offset depends from the mechanical construction and belt tension and may vary with time.
The effector (mobile part maintaining the hotend) is not in itself the smallest you can find, as by example the Kossel effector is smaller, but as it does have the fan within the effector area and as the hotend is above the effector, this is practically one of the smallest effector ever devised for a delta printer, the smallest being the Tiko one, which does not have any forced cooling and is for a much smaller printer.
Its total weight including fan and duct, for hotend cooling and part cooling, is approximately 100g. Again, the Kossel effector weight is only 40g, but without any fan and with a J-head hotend (not full metal).
The hotend of the Fisher is an ‘all metal’ type, as the PTFE tube does not go to the melting zone and is not submitted to high temperature. It is ‘integral’ as the nozzle is integrated with the melting zone and, so cannot be replaced. This non-removable nozzle does offer less flexibility, but have significant operational advantages :
no risk of leaks as there is no joint in the hot area
no need to tighten anything while hot
It is not the first of this type, as we could cite the Prometheus, but is a bit different, having a fixed heating area and a real heat break.
So, how is it possible that an all metal hotend have a so small finned area compared to others? First, it is designed for PLA mainly (which does not mean it is not capable of higher temperatures). But the design key is that the installed fan is a radial fan, with higher pressure capability than axial fans. As such, the real flow which can go through a small area will be higher. So, the better cooling allow for a smaller finned area.
And this also allow a very unconventional design, the flow which goes through the hotend fins is then used to cool the part. This is usually not recommended, as this flow is heated. But a high flow rate does reduce temperature elevation and this work properly.
As for any design decision, there are trade-offs:
It is not possible to stop the part cooling while printing the first layer, which tend to be delicate to manage.
This radial fan runs at relatively high speed (for a radial) and is really noisy.
The bed is of what seems an original design, and is really a strength of this printer.
On top of the bed plate, there are 3 balls (diam 6mm) positioned at the external bed radius at 120°.
There are three springs which push up the plate and each ball then contact two horizontal cylinders, which makes sort of a ‘V’ and positions sideways the ball.
The ensemble of three make a kinematic positioning where only one position is possible for the bed plate whatever its thermal expansion or warping.
It is also used as the probe system, as the cylinders are electrically chained to the probe connector and as soon as one ball no longer touch any cylinder, the probe circuit is open.
But there is more than that, the spring push are moderate (~200 g/spring) and the bed is easily displaced sideways as soon as there is some side load. This is a remarkable safety as after evading the parasitic load, it reposition EXACTLY at the same place. When you do have contact between the hotend and the part, as is frequent for the first layer due to overcooling, this is really good.
My experience show that it is extremely reliable as while I got many side move during my first prints, the print quality and accuracy was anyway constant. Particularly, while printing PETG, I got a lot of hotend scratching the part (due to insufficient retract, overextrusion and fast cooling of PETG) and this never fails to reposition properly.
This is a good setup, with the only limit that you may not have a part weight exceeding ~300g. Considering the Fisher size, this is highly improbable.
Cylinder screws tend to became loose with time and shall periodically checked.
To my knowledge, while similar kinematic coupling were already devised for effectors, this is the first time that this is used for a bed, and using it also as a calibration switch is clever.
Due to the different load while calibrating the center of the bed (which need the force of 3 springs) and calibrating the bed side (load of one spring), the level effectively reported varies depending the position. The variation is linked to construction quality and belt tension and varies from 0.05 to 0.2 mm. This is compensated in software configuration.
The arms are simply acrylic laser cut parts.
While I was dubious at start, this was working quite well for me till they break (Aware of the fragility of acrylic, I was careful while tightening the nuts in construction). There were multiple reports of arms broken during construction. Others reported too large holes, which may be related to laser cut tolerances. I was also careful to use ‘plastic compatible’ lubricant, as most ordinary lubricant initiate cracks in plastic or chemically destroy it. This kind of lubricant could be found in garden train shops (and is relatively costly).
Note that in above photo, taken during a print, the duct is removed to print PETG
Acrylic is widely used in low cost printer as it is very easy to cut with a CO2 laser, its cost is very low and it have a good looking. The problem is that the laser cutting process creates crack initiations and so there are significant risk of cracking. Indeed, some users broke miscellaneous parts (I broke one effector plate). But nothing is always certain, while having some stability problems, I seriously overtigthened the side panels, yet I got no crack on them. As others, I broke arms in operation, which I replaced with printed arms.
Also, the thermal expansion of acrylic is high and so the stability if submitted to variable ambient temperature is not very good.
One of the main problem is that the acrylic bed is often non flat (mine had a curve of 0.5mm) which drive to difficulties in printing and poor dimensional accuracy. Non flat beds shall imperatively replaced. The replacement shall be non metallic if you do not install a bed heater (metal bed cool too fast and drive to warping).
The overall stability of this printer shall be much better if side panel were in wood. But wood is more costly and gives an ‘artisanal’ look to printers.
A few years ago, there was a trend to build 3D printers in acrylic, because of the low cost and good aspect. We could cite as an example for a delta printer the Kossel clear. Most of these printers disappeared or changed their construction method (from laser cut to moulded parts). Kossel clear seems to be still on sale but did not had much success compared to other deltas.
Anyway, some designers still use acrylic part. RepRapPro, the Fisher manufacturer is using acrylic on its others printers. We could also cite the micro-delta, of external dimensions similar to the Fisher, but with smaller usable area. It extensively use acrylic, which raised a flame war in forum (that is in French), but while there was indeed problems, the micro-delta lasted better than some expected.
The extruder of the Beta version was direct drive and using the same motor as for movement. the torque of this motor is low (2.2 kg.cm) and the current is already set to the usable maximum (1.2 A). So the flow-rate capability of this extruder is limited. For comparison, the other basic printer of RepRapPro is using the same motor, but with a gear reduction of three.
Does it work ? It works well for layers of 0.2mm, but for layers of 0.25mm, you shall take care to have limited width to height ratio and it is better to impose the width than left the slicer makes own choices. For this thickness, you cannot increase the speed of infill and in fact, practically, your flow-rate is limited. Going for a thicker layer don’t make sense, as you will not improve the overall printing time. This extruder was replaced by a geared extruder (ratio 2:1) for the final Fisher version. I designed a geared extruder (ratio 3.1:1) to replace the original with the same stepper.
The specification tells that the usable space is diameter 150mm per height 180mm. In practice, the limit is lower.
The diameter of 150mm is only usable if you are very cautious with the cable routing (which are along the panel), as the cable clearance is insufficient. Also, as the opening in the panel is not sufficiently high, the 150mm diameter is only usable on approximately 83mm height, then the arms conflict with the panel.
The 180mm height is only available the central point. If panel opening were sufficient, the height at the periphery allowed by the geometry could be 160mm.
In addition, the 150mm diameter is obtained with arm angle of 12.5°, much lower than other deltas, which makes high carriage travels for small effector move. The small motors of the Fisher cannot cope with the required speed and acceleration and this diameter could only be reached with lowered speed (especially travel speed) and may need reduced acceleration. The lack of bed flatness also can make printing large part quite difficult. While changing my arms, I lengthened them of 10mm to improve large diameter prints, which make me losing 11mm height capability.
When issued, this printer was one of the lowest cost printer kit available on the market, and was quite capable. Some components were of an unusual quality for a printer of that class (the Board and the hotend). A clever design gives it the largest usable area for its overall size. The bed design is brilliant but the material used for it was inadequate.
It was relatively easy to operate (as could be a 3D printer) and the configuration have been improved over time.
Since, you can find much lower cost kit, but with sometimes quality problems. The low cost printer market have changed and moved to Asia, which explains why RepRapPro abandoned the market before facing unavoidable troubles. Though, the board lives now an independent life, new versions were created and its capabilities maintains a market against low cost Asian boards.
The research of economies show mostly in the mechanical area and can be improved relatively easily as your needs expands.
An important weakness of this printer, its high noise, may, however, be more complex to correct. I moved mine away from computer because of this noise, which underlines the lack of a local panel. However, without installing a panel, a Wi-Fi bridge will allow operation from a low cost tablet or from a smartphone.
Other design details
There are other points of lesser importance:
Instead of GT2−2mm belts commonly used on printers, the belts are T2.5. As T2.5 don’t run well on their teeth on idlers, the belt is twisted to have its back running on the idler.
Power supply is an external block of 19Vx 6.32A. Accessories (fan and hotend heater) have their voltage matched to 19V.
The hotend is maintained by the bottom fin, which is larger than other fins, by two small screws.