Print 3D

Edit:19 janv. 2018, Cre:14 janv. 2018

TronXY X5S

Introduction

After having designed and built Delta printers, the next step was CoreXY.
I had looked at many existing CoreXY design and was not very happy with them, so was intending to design mine, enclosed as are all my printers. I now know the difficulties associated with enclosed printers and am prepared for that.
However, recently a new CoreXY appeared on the Chinese market, the TronXY X5S, at a very competitive price. Review were underlining some defaults, but that was seemingly easy to correct, especially when these default are known. So I ordered it to GearBest. It took 6 weeks to arrive home, with a somewhat sketchy tracking, but there was no custom nor taxes.
Awaiting the printer, in order to ease modifications, I made a, OpenSCAD model of this printer (note that it is not the Standard X5S, but the model as I was intending to built it, more on that later). In this simulation process, I discovered that all I was wanting made a lot of modifications and at the end of the process, I will use about 25 to 30% of the supplied part, so it might be preferable to build a printer from scratch.

When you buy the machine

You have to make adaptations and improvement to this printer. It will take you a lot less time to make the modifications at the building phase, and so you have to buy some stuff, So I created a page referencing stuff you shall buy, modification pages, which often propose printed parts. You probably have to wait a few weeks to get your machine, so as you shall already own a printer to buy this one, you can print stuff. For miscellaneous reasons, one of the mains being that I want to enclose the machine, I prefer doing structural stuff in aluminium (only with hand tools) than print part, but you may have another opinion. One possible solution is to use PLA specialised for heat treatment, which is actually the best plastic you can print in term of stiffness and heat resistance, but you have to add a heat treatment phase to your process (in a domestic oven)

Assembly

No paper manual, just a pdf file, that I printed because I find using a tablet for assembly very unpractical.
The manual is reasonably clear, however I want to underline some faults in it:

  • The LCD wiring shown is wrong, the cables are reverted. When you start the board, nothing display, so it looks as a dead board. Hopefully, there was comments in Youtube video on the topic, so I was aware of the problem, exchanged the plugs and then the LCD lit and display starting board. This is a minor fault, but if you are not aware of the problem, that could be a showstopper and you may think your control board or LCD is dead. Note that it seems that on new boards plugs are now colored to identify the connexions. What I don’t understand is that it is a known problem, there is no paper manual, so TronXY just have to modify a file to clear the point, which will cost them near to nothing and could be done quickly. This should have been done immediately. Also, even if their was no error on the manual, this point shall have been underlined, as it is an easy mistake.
  • When installing the Z rods, I had problems of the rods pulling and bending the top Y rail beam. In fact, you have to add washers between the rods and the beam, but while this is written in the manual, this is not clearly visible on the image and you are at risk to definitely bend the rail beam (for other reasons, that I will explain later).
  • The principle of belt installation is clearly defined but a very important point is forgotten: When installing the belts, you have to mechanically lock the X beam against the stepper plate to have it parallel to the frame, because tensioning the belts rotate the X beam. The X beam shall only be released when the two belts are installed and equally tensioned, which shall be checked by testing their resonance. This clamping could be done with big zip-ties, moderately tightened screw clamps or pipe clamp. If the X rail is not perpendicular to the Y rail, your parts won’t be either, squares will be diamonds and circles will be ellipses.

Design faults

First, safety

  • No fuse whatsoever on the 12V. This is particularly critical as there is no separated MOSFET board and the Chinese 3D printer boards are notorious to start fires with the onboard MOSFET. You shall at least add a fuse on the 12V power supply line. Note: separated MOSFET board is NOT a good thing in itself, it is a required remedy to insufficiently rated board MOSFET and screw terminals on cheap Chinese boards, in order to avoid fire hazard.
  • No terminals/ferrules on stranded wires (power supply, bed supply, hotend heater and board fan). This is dangerous and terminals are required by law in all Europe, for a reason. I am equipped, so no problem for me to add terminals, but I understand that people may be reluctant to expense 10% of the printer cost to buy the required tools to install ferrules.
  • No box for the electronic, just plastic side shields. Boxes does not only have a much better look, but they can contain an electronic fire (again, this does occur). Acrylic plastic shield also burn easily, spreading an eventual fire.

Then, questionable design points

  • Hotend fan box
The fan box (cannot realistically be called ‘duct’) not only have no ducting or air flow guiding to the hotend heatsink, but in addition, send a big chunk of the air flow directly to the hotend block and the bed. This have two effects:
  • The hotend does have some difficulties to heat up, as it is quite efficiently cooled. This is quite aggravated by the fact there is no heat block insulation, just one layer of Kapton. You need to add insulation, finding on the market insulation sets or better, the famous ‘silicon sock’. As a temporary measure, I made a rough insulation with many Kapton layers.
  • When the hotend is above the bed, the flow prevent it to heat, as its power is marginal (~100W for a 33×33cm). First trial was ok because the homing send the head outside the bed area and lower the bed, but if you try to print after a bed adjustment (which is quite the normal process at the beginning), You will never reach your target temperature. On my first test, I was capable to reach 55°C in a bit more than ten minutes. On my first print, after twenty minutes, I still did not reach the 50°C target, when I finally understood the problem, you shall lower the bed and install a temporary shield between the hotend and the bed to prevent cooling by the hotend fan.
This fan support weight 77 g, add 9 g for the two fan grids, which is insane. This support have no structural purpose and weight many times the hotend and even more than the carriage plate (which is quite heavy).
  • Heavy X carriage printer
The X carriage is in steel (2mm thickness), so it is quite heavy. I replaced it by an aluminium plate of 3mm thickness, reducing its weight by half (so ~38 g gain)
  • Lost travel
There is a lot of travel lost. when the Y carriage is stopped at the bed side, there is still 40mm remaining between the fan box (which is quite bulky and could be reduced with good design) and the frame. On the X carriage, if the twin wheels were placed on the bottom instead of the top (with the added benefit of better stability), it will give approximately 40mm more travel on each side, so 80mm total.
If the wheels were replaced by ‘Mini’ openbuilds wheel (as is done on deltas or other CoreXY printer like the D-Bot), you can still add some travel. So, a good design may allow to reduce the width of this printer by 100mm and its depth by 40mm, which is quite something. That will be lower cost, stiffen frame and will use a lot less space.
  • Too high screw lead
The screw lead is 8mm, so the motion is reversible, which means that when you push the bed, it moves. The problem is that this is a two motors bed, so when it moves, this is never the same amount and the bed is tilted. A unique driver control the two Z steppers, so there is no possibility to adjust each side with another Z switch.
The game is to not push the bed when you remove the part, It is not easy and not quite funny, so you end up having to adjust the bed for almost each print. Using a 2mm lead screw should have costed exactly the same. Sure it makes the Z move quite slow, but adjusting the bed at every print is not nice and you rarely move the entire height.
  • Wrong attach of the Z rods
As explained in first paragraph, when you install the Z shaft, you have to add washers to compensate the space between the top of the rods and the rail beam. A washer is 1mm thick, so the adjustment is not very precise, and you will end up at best, with a bending of 0.5mm, which is a lot when you depose your first layer. If your beam is bent to the bottom of 0.5mm, this is equivalent to a bed bulge of same value. Remember that this machine does not have a sensor for bed height and no automatic calibration. To aggravate the problem,if you tight firmly the rods at the bottom (you shall as this is structural), the rods bury themselves in the profile slot. I did not noticed first and I ended to bend the rail beam by more than 5mm, hopefully it seems to have recovered its original shape. The final space was around 8mm, so I had to add many washers and change the top screws for longer. There was a benefit, anyway, having many washers, I was able to play with different washer thicknesses to make a better adjustment than possible with the original design (3 washers max). This is a bad design and the rods should have been attached by supports screwed sideways to the profile. It shall be noted that such twin lead don’t need 4 rods, only two rods of larger diameters (12mm) installed just aside the screw and properly installed with rod supports. This shall not be more costly than the actual setup and be better. You can find this Z setup by example on Eusthatios, Herculien or other printers. It shall be noted that in term of stiffness, the actual setup is quite stiff and sufficient. When you push the bed on one side, it lower easily rotating both screw without visible tilting (but too much to print properly anyway).
  • Insufficient bed power
Depending the bed you received, the bed power is 80 to 100W, which is way insufficient to heat up a bed of 33×33cm (~0.085 W/cm2). At best, you can increase the temperature from ambiant by 30°C for some, only by 25°C for others. This needs time and patience and allow you to only print PLA. It is quite marginal to print PETG (while I succeeded to do with caution and slow speed) and it is unthinkable to print ABS. But honestly, I think printing ABS without a filtered enclosure is unreasonable, the odour is unbearable. In addition to its low power, there is no bottom insulation on the bed.
The power supply is 360W, so you can at maximum use 300W, take into account 70 to 80W for the steppers and heater (40W), approximately 220W was available for the bed, which is still a fairly low value (that makes a power density of 0.2W/cm2, while a rule of thumb estimate requirement to 0.37W/cm2 and a comfortable power - fast heating - at 0.5W/cm2), but more usable. however, this is a 12V machine, and the board used is not 24V capable, 220W makes 220/12 = 18.33 A . Board Mosfet did not accept more than 10A, and you shall have an external MOSFET board to handle that, with a safety added benefit, as the onboard MOSFET frequently cause problems.
  • Wheels details
    • The wheels are installed on cantilevered bolts, and the plates are way above the beam (~5mm), while they should be just above the beams (0.5 to 1mm) to reduce the lever value in order to minimize bending.
    • OpenBuilds Wheels were designed for CNC and they are overkill for printers. It exist ‘mini’ wheels which are largely sufficient for printer loads, are at least equally stiff or better, are lighter, need less space and allow carriage size reduction, hence further weight benefit. You can find on Internet some Chinese Openbuilds wheels with abnormal wear, but this is probably related to poor material selection or poor assembly (see the eccentric paragraph below)
  • Poorly positioned LCD screen
The vertical position of the screen needs you to be kneeled on ground or be crouched to operate it. It became quickly very boring.
This screen should have been reclined to be usable or installed on the top beam. Installing it on the top needs to raise hands, which may be tiring, but much more practical. In addition, a position on top will shorten significantly the wires and reduce the huge bulk of the wire harness. A nice solution can be the control board near the top for good wiring and a reclined screen on the bottom but the board/LCD screen cables are rather large.
  • Too small pulleys for steel core belts
The belts supplied are in shiny material not in rubber and are reinforced with steel chords. This makes them very stiff, which is good for quality printing, but they have difficulties to run over idlers and especially over the motor pulleys (20 teeth). In principle, steel armed belts shall run on pulleys with a minimum of 36 to 40 teeth. People who have experienced this kind of belts tell that when running on too low diameter, the belts tend to disintegrate over time and loosing wires. Using 40 teeth pulley is doable on this board but may need a bit more powerful steppers, which may be 0.9° (400 steps/rotation) to maintain precision.
  • Flexible frame
The frame is not bad in itself, bolting the aluminium profile is a good and strong solution, but a CoreXY do impulse its loads on the top of the frame, which is very demanding for a printer of that size. So the frame tend to vibrate at high speed and acceleration. Rippling remains moderate, but reinforcing the structure, especially in the YZ plane may improve the situation.
  • Minor design faults
    • Wrong position of Z chain anchor
    The Z chain anchoring is not properly designed and shall be moved by 20~25mm to have reasonable radius and limit stress on the anchors. The Z chain move a lot less than the XY one, but it moves anyway (and there are a lot of small movements with the ‘Z-hops’), so minimizing stress on chain and cable is required. This end chain support is a 3D printed part, so that will not be a big problem to redesign it correctly. While other users have reclined the anchor point, I used long screws found as spares in the kit to space the chain end from its anchor.
    • Misplaced Spool support and extruder
    The spool is installed on the back of the printer and it protrudes a lot on the back, increasing the space required for an already very large printer. The extruder is also on the back, which makes it very difficult to access, check and use given the printer depth. I designed printed parts? to install these parts at the front of the machine.
    • Return idlers wrongly aligned
    Return idlers are not aligned with the Y carriage pulleys. This may not have a lot of consequences because on this side, the lost travel makes the Y carriage stopping far from the pulleys, so belt misalignment angle is limited. But this sort of very basic mistake underline the fact that the design wasn’t properly checked.

Interesting design points

The Z motors are installed in an unusual way, bolted to their bottom on a quite flexible plate. This have an interesting effect, the flexibility creates sort of a ‘virtual articulation’ at the motor bottom, while maintaining its torque stiffness. Combined with the flexible coupler, that reduced significantly the side loads which may be introduced by the Z screw and limit significantly the Z wobbling. Some people displeased by this weak installation have designed and installed reinforced ones, but this is a wrong idea.

Cheap stuff

  • Cheap electronic components
My SD card reader lock stopped working after 10 uses,so the reader became unusable. I should have added an extension micro-SD to SD extension from day one, but I did not had one at hand. I will try to glue an extension to repair it.
  • No eccentric to adjust wheels
There is no eccentric on the wheels, so it is quite difficult to adjust them and even if well adjusted, they will became loose with time. The 4 wheels on the Y carriage aggravates the situation. 3 wheels will be better and lighter and if we want to improve rotational stability on X axis, we have room to increase the space (in Y) between the twin wheels of Y carriage.
  • Very flexible plates
The plates are in laser cut acrylic. They are too thin and flex a lot, this apply especially to the Y carriage and motor plates. Some retainer shall be installed on the motors because the belt tightening makes them to bend significantly. In addition, the belt is armed with steel wires, which is good for precision, but need quite high tension. It already exist on the market stepper metallic bracket, so it won’t be really more costly to use an appropriate material.
  • Laser cut acrylic parts
Acrylic is not a bad plastic in itself, but when laser cut, the thermal impact creates micro-cracks, so acrylic parts submitted to high loads (which is the case here) always end up cracking. Acrylic is used because it is the easiest plastic to laser cut but is is known since years to fail at short or medium term.
  • Plastic spacers on cantilevered wheel bolts
The wheels are installed cantilevered on plastic spacers, which make them quite flexible in bending. A lot of users are replacing the plastic spacers with washers, which is much better but weights a bit more. Standard installation are aluminium spacer. Though, it may be acceptable to use plastic spacers when wheels are between two plates.
TronXY had already been firmly criticized for the wheels installations on their other printers, some reviewers even telling that makes these printers unusable, but it seems this company do not listen very well.
  • Laser cut acrylic shields on electronic, no box
All electronic is nearly fully exposed and only protected on its sides by acrylic cut plates. This is not very durable and , as already underlined, is dangerous in case of an electronic fire.
This gives an unfinished look to the printer.
  • Low torque stepper on extruder
Extruder is a traditional MK8 in aluminium, with a gear 10mm diameter and a 40mm stepper. Direct drive extruders are simple and economic, but they need high torque steppers, especially because here we have a gear of 10mm (hobbed bolts have a diameter of ~7.5 mm). The long Bowden (80cm) add loss and increase the force needed from an extruder. This extruder is really weak and will face troubles as soon as there will be difficulties with the nozzle. Hopefully, the hotend is PTFE lined and may requires less pressure than an all metal one. This seriously limit the flow-rate you can have and the possibility to use thick layers to reduce printing time.
  • Not very good design for enclosure
What appeal people for a static frame as the one of a CoreXY is the easyness of enclosure.
However, on the X5S, the Z motion system protrudes from the frame and so do the wheels. It should have been quite easy and at least equally stiff to install the vertical beams on the sides of the Y rails, instead of below, and combined with ‘mini’ wheels, you could have bolted directly a panel on the frame enclosing all the motion system.
Yet, you have to add beam spacers on the sides to bolt side panels. This is a missed opportunity (added to the fact that the Z-motion between the beams is a wrong design as underlined above).

Improvable software configuration

The Marlin version used is hopefully recent and have activated thermal runaway safety.
However, it seems the configuration as delivered have some problems:

  • There is lag and overrun permanently when you turn the button and you shall try three or more time to reach a target
  • There is a problem with the extrusion from the LCD panel. When starting my machine, I experienced my extruder clicking, so I thought my hotend was clogged. I disassembled (unscrew) the ‘Bowden’ on the hotend and try successfully to manually extruden while I found it a bit hard and not totally regular. I try to clean up with the filament removing method, drill with a 2mm bit, with no avail. Then I tested a print, which worked without problem. In fact, there is a software problem in the LCD control which makes the extruder inoperative in manual control. This stuff costed me a pair of hours.
  • This printer have a long ‘Bowden’ tube and as such you tend to have bulges on corners and transitions. This can be helped by using the ‘advance pressure’ algorithm, however, this option is not activated in the Marlin configuration.
  • There is no ‘Baby stepping’ Again, this is an option you just need to activates, but it is not activated in the default configuration
Practically, I found that on this printer, the couplers are easily accessed and when you print your first layer, you can adjust the height by manually rotating the couplers and so, makes manually the ‘baby-stepping’. The steps jump are well noticeable and that is easy to do. 1 step is 8/200 = 0.04 mm, so the adjustment can be relatively precise.
  • Inappropriate accelerations and speed configuration
One of the expected quality of a CoreXY is light mobile mass, so possible high acceleration. High acceleration have two benefits:
  • They reduce bulging at corners and transition, which are especially high for a long Bowden setup
  • High acceleration makes the full speed reached more quickly, so significantly reduce printing time, especially for dense infill
The default acceleration are set at 1000, which is a value for heavy Cartesian directly driven extruders. Tested values (again, for my modified printer) were the best around 3000 for X and 2500 for Y. This can be modified by setting your own G-code start-up sequence in your slicer or modifing the configuration default if you are connected with USB (I am not, my computer being remote). Please find my G-code at start-up for the X5S using slicer Slic3R (I Ignore if other slicers use variables like the first layer temperature here). However, these larger acceleration, aside making a bit more noise, are quite demanding for the machine structure, which is a bit weak due to its large size and lack of reinforcement brackets. Adding reinforcement or cross-bracing (especially on the Y planes) will help.

Note: The Marlin file configuration could be found in the files supplied by GearBest for this printer. You shall have an account to access these files. You can find a copy of the files I get at the end of November 2017

Note 2: Marlin is using the very last resources of the poor 8 bit processors of this kind of boards and often exceed available memory or computing power, notably when you are using LCD which consumes resources, so it is possible that some options were not activated for lack of processor resources. I am new to Marlin and cannot confirms such details.
I have found on Thingiverse a modified firmware which supposedly add many improvements, notably adding baby stepping and removing the knob lag, but I did not tested it (Yet).

The good stuff

Is it a good design

Globally the design is sound but is undermined by abusive cost-saving, poot quality control and absence of tests. Also, the printer shall be much smaller than it is for its usable space.

Was it easy to assemble

Yes, aside the time required by my modifications, while relatively long, this was easy and pleasant to assemble. Cables look messy, but are supplied with the appropriate length and use locked connectors with mechanical coding (and not the infamous and unreliable ‘Dupont’ connectors). There was some spare for screws and other fittings. Some precisions are lacking in the manual, notably for the belt installation.

Does it print well

At least for my modified version, it does print quite well PLA, when you have found a good configuration and modified the default values. One of the reasons is the use of steel core belts, which makes them quite stiff and compensate for the large size of the printer. However, others materials are out of reach with the given configuration.

Is it easy to operate

Yes, aside the quirky behaviour of the knob, it is relatively easy to operate from the LCD screen and SD card (I have not configured supplied computer software because my printers are yet quite far from my computer and are supposed to be accessed by Ethernet or Wi-Fi). There is a curiosity in the operation, the rotating knob is moving ‘the wrong direction’ for parameters adjustment.
I did not had to have a look to documentation for the LCD (except for the option ‘restore failsafe’):
I am used to a high end board hardware and firmware (Duet/RRP), an excellent if not the best firmware/Hardware combination, so I was a bit afraid to go ‘back in the history’ of 3D printing with Marlin on 8 bit boards (I never used it before). That is for sure less practical that what I am used to, and a lot less flexible, but once you found the good parameters, it is usable. I intend to install OctoPrint on a RaspBerry PI 3, which I expect will help a lot to recover a comparable ease of use as my Duets.

The cable chains

Once the Z chain anchor point position is corrected, these chains are neat and work well. The links are not openable, so adding a new cable will not be very easy. I removed 2 links on the X/Y chain (cutting them, as my wiring was already done).

What about the bed

The bed is in aluminium, with a fibreglass plate on top of it and what looks to be a ‘BuildTak’ surface clone. I haven’t glued the surface because I have bad experiences of the BuildTak that are consumables. For a printer that size, changing frequently such surface will be painful (they are not easy to unglue) and costly. So I used hair spray on the fiber plate, with success with PLA. The adhesion of PETG was a bit marginal though, because of the too low temperature. I know from experience that PETG do adhere well on ‘Buildtak’ without heated bed, so using the surface supplied could be a solution for PETG. If sticking to the original Bed heater, bottom insulation is not an option. What I also use for small printed parts is I add a shield atop the bed on areas which are not accessed for the given printing job to have slightly faster heating.

A quick comparison with the Ender-4

Creality sells the printer Ender-4, telling it is a CoreXY, but of a much smaller size than the TronXY X5S. Creality having get a reputation since the huge success of the CR-10, a quick comparison is interesting.

  • First, Creality sells the Ender-4 as being a CoreXY printer, while it is in fact an H-Bot. H-bot cinematic have some problems which are corrected by CoreXY cinematic. It is possible to transform it in CoreXY by adding two pulleys and changing belts, with the same setup as I have done on my X5S. However, promoting a printer with a wrong cinematic designation does not really look serious, as if they are not knowing at all what they sell.
  • The wheels are properly installed, with aluminium spacers and eccentrics, allowing proper adjustment. It shall be noted that the Y carriage semms to have have ‘mini’ wheels, while the X carriage use full size wheels. The Y carriage do have 4 wheels but for a H-Bot, Z axis stiffness on Y carriages is really important and you cannot do a simpler 3 wheel carriage as is possible on a true coreXY.
  • They also use the infamous ‘fan box’ however it is probably the same as the CR-10 and weight less than the X5S (44g instead of 77+4g)
  • The extruder is the same model, just in plastic instead of aluminium, also with a short motor, so with same weakness.
  • Idlers are made with flanged bearings, much better than the infamous plain bearings and fixed washer of the X5S. However, they are of smaller diameter (F623 or F624), which does only work with standard belts (without steel chords). It may have been better to use F625 of larger diameter at same cost for better durability.
  • They also use laser cut acrylic plate. Stepper support plate design is better than on the X5S. The belts are run lower, so with less cantilever. The belts are not steel reinforced, so needing less tension. The stress on motor plate is much lower than the X5S and you can expect lower deformation. However, you still see users replacing these plates with reinforced 3D printer supports, so there may be problems.
  • There is no part cooling fan, which is unacceptable for a printer dedicated to PLA.
  • LCD panel is vertical, so you also need to be kneeled or crouched to operate it.
  • The electronic board does have a cover, but LCD is also open, so the electronic protection is slightly better than on the X5S. Shield looks to be in aluminium instead of laser cut and fittings are more neatly installed.
  • Z smooth rods are properly installed on aluminium supports
  • The hotend heat block is insulated
  • They also use a fragile micro-SD card slot
  • There is no cable chain, hot-end cables are bundled with the Bowden tube, but this can work on a much smaller printer.
  • There are options for a filament detector and a level sensor for automatic calibration, which are both nice to have equipment.
  • It is reputed to be silent, so there is probably a better choice of fans, but a part cooling fan shall be added.

So, the quality of the Ender-4 appear to be better than the X5S, but is not very good either and is far from what is proposed on the CR-10, especially regarding the safety of electronic and wiring.

In conclusion

As all Chinese printers, this one need many adaptations to be safe, reliable and durable. A CoreXY can potentially be a faster machine than other designs, so it is appealing for experienced users which thinks that the X5S can be the base for a good machine. My initial intent was indeed to makes improvements to get a better machine. But at the end, it will need so many modifications that I finally decided to use this machine as a test bed and sell it, to finally design and build the machine I want, using the experience I had with it (and with my former designs).
There is some missed target, as many problems cited here are already known on other machines from the same manufacturer (wheel assembly, insufficient bed power, poor hotend cooling, poor wiring and electronic boxing).

It shall be noted that while electronic and wiring in Chinese printer is known to cause fires, there is at least a positive in this new generation of machine, which is the hotend support in metal, which cannot melt in case of hotend fan failure. This is a vast improvement against printed hotend supports. Also, the use of locked connectors is safer than the infamous ‘dupont’ connectors we were using just a few years ago.
It shall also be understood that such machine could have not been sold pre-assembled in regulated countries because it does not comply with basic safety regulations, especially in EU. Being a kit and sold directly from China, you shall modify the design yourself to ensure your own safety, which is a harsh job, even for engineers specialized in safety. So the gold rule still apply: NEVER run a printer unattended (and a camera is not a fire alarm and won’t help fighting a fire). Please have a look at this page about safety Disclaimer: I authored this page.

(c) Pierre ROUZEAU
Éditer - Historique - Imprimable - Changements récents - Rechercher
Page mise à jour le 19/01/2018 18:39