So now the project is complete.
Return to the shop
The last few weeks were taken up with vacation and conferences. So there has been no progress on the crib. Teal and I had brought in the last of the wood from the shed, but it was not thick enough for this project and has to wait for another project. This was the last of the Wisconsin Woodworker’s Guild Logfest hauls that I had set up to air dry. At its peak, it was over 1300 board foot of lumber, which has now been reconstituted into many pieces of furniture for the family and friends.
So yesterday we set off for Kettle Moraine Hardwoods which is our local lumber mill. They had a nice selection of thick Red Oak (5/4 and 8/4 – 1.25-2″ thick) and we picked up some Hard Maple and an Elm slab for future projects, as well.
Once home, it was time to surface the boards on the jointer and planer and then cut them to rough length in preparation for sending them to the CNC router. At this point, there was a minor design change as the thick stock for the legs was completely cleaned up at 1.86″ vs the 1.5″ I had in the design. So we decided to go with thicker legs. However, this leads to more work as I did not have a router bit that would cut deep enough, which will be detailed later below.
I fired up the computer for the CNC router and some problems arose. I had not used it for ~2 months and only had the windows 10 logo showing for 45 min. At this point, I rebooted again and it came up in a few minutes. However, Mach 4 which is the CNC controller software for the router had a whole series of errors when starting, and was unusable. Most of the plugins would not work. So I restored it from the backup copy, restarted the PC again and it started to work. However, in testing, many of the configuration parameters were missing including “little things” like the home switches and control for the spindle. Digging through my notes for the configuration values, I got it running again. Now, thoroughly annoyed, it was time for a reward of our home brew Imperial Stout which is now ready for consumption.
After re-zeroing the CNC router it was time to set up the first part and make a test run. The crib end top rails were chosen as they are the smallest parts and least costly in case of problems. The 8/4 stock for the leg pieces was over $150. Making wach leg approximately $35-50, so I was not going to try those first.
So the stock is clamped on the CNC router and you can see my “cheat sheet” where I have printed out the outline and marked the distances from the near end for the various clamps.
Next you can see the cut under way. At this point I have left off the dust shoe, so there are chips EVERYWHERE.
Now the top rail is completed and vacuumed off.
Another shot in progress, looking down with the dust shoe in place.
With the top rails successfully completed, now I move on to making the legs. One of the front legs being cut. In order to prevent the part moving I used not only the 5 clamps shown but also some small strips of double stick tape which help reduce the part sliding under load immensely. With this CNC router, the limitation on cutting speed is not the machine, but the ability to clamp the work and avoid it slipping under the cutting forces. Cutting speed was 100 inches per minute, 18,000 RPM at 1/4″ depth of cut with a 1/2″ 2 flute carbide end mill.
Set up for one of the rear legs – 43″, 110 cm long. Not your ordinary tabletop CNC router. . Note the beautiful curl figure in the stock. Later, you will see how I make this “pop” when finishing.
The other back leg ready to cut
This is why I use wooden clamps. Just a minor nick this time. These are shop made on the CNC router.
Another look at the scale of these cuts and the finish off the CNC router. This was without a reverse last pass as I don’t (yet) have a router bit long enough to do so and that would look even better.
Here are the legs off the CNC router. As you can see I was not able to cut all of the way through. The depth of cut was limited to 1.5″ based on my largest end mill / router bit. So now they need to be run through the band saws and then flush trim routed. The end curves are too tight for my big band saw (24″ with 1/2″ resaw blade ) and need to be run through my small one (12″ with 1/4″ skip tooth blade).
Bandsawing the excess off the legs on the 24″ band saw.
Flush trimming off the excess. The holdfast works great to clamp irregular stock like this. Teal also assisted in taming the work. The Oak is a bit unruly. I often had to reverse directions to minimize tear out. This means taking climb cuts which try to throw the work and router around.
First dry fit test. Not bad. The cross pieces will be flush with the upper / inside edges of the legs in the final assembly.
The CNC router made this work feasible in a few hours. Otherwise I would have had to make templates, band saw to size and flush trim through several steps. I had done a similar project with curved legs – Elyse’s Sleigh Bed. This is MUCH easier and with less chip out to fix.
Movies of the CNC router at work
The first part is without the dust shoe (chips Everywhere) and the second part is with the dust shoe in place (much neater) .
Now that we have a starting point for the design, it is time to start gathering dimensions from the photos. Traditionally, I have done this by printing photos at half or full sheet and then scaling it manually. My goal when borrowing from an existing design is not an exact replica but rather using the original as a model to start for overall proportions. I am not into doing reproductions. Additionally, in this case I am making improvements in the design.
First, I will note the major known dimensions. In this case it is overall height width and depth. This is done on each photo. Do not assume (or forget) that the scaling will be different in each image. Next the actual size in the photo for these is measured and the scaling amount calculated. I will normally do the actual measurements in mm to make the math easier (avoiding fractions). You can use a caliper or ruler – I like using a caliper.
I double check that the X&Y dimensions yield similar results. If all you have is an angled or isometric view then this gets more complicated. Now I take measurements of measured features and do the math in a spreadsheet so I have a listing of my measurements and can easily double check for errors in case of a conflict. I will also do the actual photo measurements in one color and the final scaled in another to avoid mix-ups
I will typically start on paper and make sure that the dimensions make some sense before transferring to Sketchup.
You can also import a photo into sketchup and then start drawing your components on top of it. For rectilinear pieces it is better to use it to set a few lines and then draw as you normally would. Trying to match the photo underneath will otherwise lead to errors and out of parallel edges.
Here I have started with the curved end pieces. With the rudimentary curve drawing tools in Sketchup (arcs, lines and bezier curves) it took a lot of tries to get something that looked decent. I wish I could have gotten the NURBS lines to work which offer much more control. At this point I will lay down a couple of lines to mark the lower rails and then hide the photo. So now I can move into 3D space. each of the parts show becomes a component so I can mirror them for the other end and all of the modifications transfer as I work back and forth.
The top and bottom rails are then fit to the end posts. Now I mirror the ends, and set them approximately the right distance apart. I am still not sure if the end top and bottom rails will fully or partially overlap the legs.
Next come the bottom rails and I then draw guide lines up the leg posts to show where the top rail should line up vertically and the offset for the spindles from the inside edge. The inside spacing is an important dimension as you are only allowed 28 +-5/8″ by CPSC guidelines so you don’t have too big a gap around a standard crib mattress.
Now with the the back rail placed, and the frame colorized you can see the shapes better. I think it looks better with the taller top rails, but lest see what the girls say tomorrow.
Baby Crib Project
My daughters, Jessie and Elyse, are both due at the end of March 2018. Jessie still needs a crib. So the question of: “What is the next project?” has been decided, and I have a rather short timeline. However, if push comes to shove, she can use the same cradle that she slept in as a baby that my dad made, while I finish the crib. This is another project where I am following in my Father’s footsteps making things for the grandkids.
- Solid back – avoid little fingers finding outlets and easier cleanup after dinner gets launched. Flat panel rather than raised panel per Jessie’s preference.
- Adjustable mattress height
- Convertible to a bed. This will make it last, rather than be a 2 year item. Appearance as a bed will take precedence over appearance as a crib.
- Follow CPSC guidelines for safety. This includes no cut outs – so much for Mackintosh style slats I had wanted to do on the CNC router (besides Teal and Jessie are not fans).
- Design needs to complement the other furniture I have made for Jessie which is of a craftsman style and have the same finish
So we have been digging through many Google Images, Pinterest, LumberJocks and many other websites looking for ideas as a launch point. There are a lot of ugly cribs out there! Besides, many designs that look OK as a crib do not look very good as a bed. On top of it you have the predominance of MDF and particle board based junk that is on most of the web sites.
Given that Jessie is expecting a girl, I was hoping for a sleigh crib / bed sort of design. However, many “sleigh cribs” have weird lumps for curves that seem to have been tacked onto an otherwise square leg. I want something that will be smooth and flowing. Plus, this will give me another excuse to play with the CNC router. Although, a band saw and spokeshave or template and router would probably work well, too. For Elyse’s queen size bed, the curves were laid out on a template and then pattern cut with a band saw and router. Her bed (ca. 2009) is shown below.
Now, I want to do something that is more “organic and flowing” for the end posts rather than ending squarely at the floor. This is a tall order for an engineering mind.
Much has been said about the changes in safety standards for cribs. The slat spacing requirements have changed, no more drop fronts (yeah!). However I was still worried about safety and started more research. The primary resources I used are:
- CPSC website: Full-Size Baby Cribs Business Guidance & Small Entity Compliance Guide which provides the inside dimensions but is silent on spindle spacing.
- International Association on Child Safety: iafcs.org/docs/Docs_Head_Entrapment_Presentation_Revised_2013.pdf which does provide spindle spacing and other useful information.
- National Institutes of Health: Age, side height, and spindle shape of the crib in climbing over the side.
Where to start?
After several evenings of web searches and IM messages back and forth with Jessie, a leading candidate for the basis of the design emerged. It is the “Franklin and Ben Mayfair Crib”. For an example see: Mayfair Crib This was the first site that it popped up on. Others sites list it as discontinued.
This is just a starting point. As on many projects, I will look at multiple designs and then take the pieces I like and modify for more robust construction and the techniques that I prefer.
Things I like in the design:
- The outside sweep of the posts – especially the inward curving feet and no “lumps”
- Curved top rails (teething deterrent). These will end up looking more like Elyse’s sleigh bed
- Looks nice as a bed
- Nearly solid back
- Openings at the ends of the back panel
Things that must be improved:
- Flat panel in back looks like it was tacked on
- Lots of fasteners and holes showing
- Flat inside edges of the legs when viewed from the side
- Extra side and bottom parts hanging on when converted to a bed
- Front bottom rail – likely will be removed as this will not be a day bed but rather converted to a full size directly (pending approval)
So now I need to gather basic dimensions and start the new design.
With the first batch of the top drawers out of the way, it was now time to start on the rest of them. These are bigger, and I had learned a few lessons from the first set of top drawers:
Consistency and repeatability are paramount. This lead to fixture board end stop additions and minor redesign to remove most human error. I had some sides on the first batch which did not line up perfectly and I was lucky to be able to fix the fit by sanding the backs of the drawer fronts.
Drawer and side thickness MUST be consistent. This time all of the boards were drum sanded to the same thickness. It is surprising how much variation even a large well tuned planer has (mine is a 20″ Jet). This made a big difference in the fit for the 18 drawers vs the earlier batch of 9. Planing single handedly does not help.
Scoring the faces of the sides to prevent tear-out is a must. The roughing cut sequencing of Joint Cam causes a lot of tear out. It would be better if the roughing cut was the same sequence as the finish / dovetail pass. However , scoring the face heavily with a marking gauge, largely mitigates the problem .
Fixture additions must account for bit profile differences between the roughing and finish passes. It would pay to run the finish pass air cutting to make sure the fixture fingers are not in the way or you could end up with a snapped bit or missed steps. I had to trim back the corners by hand to allow room for the dovetail shaft.
The new LED light ring around the front of the motor nicely illuminates the work .
Now for cutting the dadoes and assembling the 18 drawers.
One of the goals when making the CNC router was to be able to do general woodworking joinery including dovetails, sliding dovetails and mortise and tenon joints. The mortise and tenon capability was proven out when I did the blanket ladders. The dovetails needed another project and the dresser project with 27 drawers is a perfect fit.
The CNC router was set up with a special base for the dovetails and the horizontal fences were CNC routed to final spacing in place (20″ apart) to match the vertical fences.
Over the last week I had prototype the joints using JointCam and using scrap pieces of plywood that were of the approximate thicknesses of the drawer fronts and sides. I tested both equally spaced and equal spaced dovetails and both turned out well after a bit of initial tuning. Teal liked the equal spaced better and these could be done with both pins and tails in one pass so this was chosen.
Wood for the drawers
The drawer faces are white or red oak depending on the dresser . At this point I am doing the top row of smaller drawers. The 3 top drawers for each dresser are cut from a single board so that the grain runs across the face of the dresser. If one is botched then all 3 are scrap. So there is a bit of tension here with some nice pieces of quarter sawn white oak (big flakes) and near quarter sawn red oak for the drawers.
The drawer faces are approximately 3/4″ thick. The drawer sides are 1/2″ nominal (0.47″ actual) baltic birch. I like the baltic birch plywood for drawer components. The even coloring and lack of voids make for nice side pieces.
Now for the “production”
The off cuts were again tested and everything seemed to work although the joints were a bit looser than expected from the earlier testing . Jointcam has a setting for adjusting the fit and this was used. New pieces tested OK so on to the real stuff.
At this point things fell apart. The router bit was progressively creeping out of the collet during the cuts. This was apparently exacerbated by the oak drawer fronts. I reduced the feed rates and increased the spindle speed but it was still happening. I tried a new dovetail router bit with a longer shank and really cranked down on the collet but it was still creeping out and then the shank of the router bit snapped just past the collet. Time for a beer (or 3) and think through the alternatives.
There is not a lot of info on the net about CNC cutting dovetails. There are bits on commercial work where you cut everything flat and add a false front, but little on the trials and tribulations of doing this properly. The JointCam info mentions the use of roughing and dovetail bits and passes but no guidance on when to use this vs single pass. So after the beers and sleeping on it I decide to be conservative and use a roughing pass (straight bit) and finish pass with the dovetail bit but switching to a 1/2 ” shank dovetail in the hope that it would grip better. I had one more of each on hand – 1/4″ straight and 1/2″ 14 degree dovetail with 1/2″ shank. As you can see from the photo below, the pieces are cut 2 at a time as right / left pairs.
Testing went well. However it is tedious as you need to do a bit change for each test (straight and dovetail passes). So by late morning it was a go for this method. Now I just had to make a new set of drawer fronts for the dresser that had them ruined in the first try.
I also found that while JointCam makes a scoring pass for the dovetail / finish bit, one is needed for the roughing pass on the plywood sides. The side faces tend to splinter easily. So I took a marking / cutting gauge and scribed a line across each of the sides before routing. This contained the splintering.
Production part 2
I did the drawer fronts first. These were labeled as to sequence on the dresser, inner face and top edge. The top edges were placed against the fences on the CNC router.
First pass for the stack of pieces was with the straight bit. After all were run, I changed the bit to the dovetail (and it is nice to have a soft mat under the edge of the CNC to catch the dropped bits vs. the concrete floor).
Now that the fronts were done (pins) and they fitted reasonably well against the test sides it was time to do the sides (tails). These were cut and after the first pair I test fitted them to the fronts. They were close but there was still some variation form one front panel board (set of 3) to another. One small adjustment of the fit clearance (-.001 to +.004″) was made in JointCam for the sides to match the fronts. At this point the sides are labeled right and left (blue tape on the rights) and for each dresser set.
Video of the CNC router cutting the joints is here: Dovetail cutting
There was still some tweaking to do. One set was still tight even with the +0.004 correction. These pieces were tight and had a gap at the end of the tails. So I took this set of 3 to the drum sander. I took off about 1/64″ on the inside face and everything fit nicely. So now I have the 9 drawer faces and sides fitted up and ready for the dadoes for the bottom and backs.
The printer is now running It is roughly assembled (as you can see) and now it is time for fine tuning and making parts for itself. I love it when a tool is at the state where it can help finish its own build or rebuild making parts. Cases for the boards, cable retainers, cable covers, cable chains, proper spool support are all on the short list.
There has been a bit of additional grey hair added getting here:
The Rambo control board will not connect at all via USB3. It must be a USB 2 connection . This is the case for both the Arduino downloads and Repetier control software. I did not pay attention when I originally programmed the board and then subsequently rearranged the USB ports…
Along the way the first Rambo board stopped communicating via USB. Many tries to get this going, re-download drivers, try to re-flash the USB chip but to no avail. I ended up returning the first one (still barely within the 30 day amazon return period) and got another. So far so good. I was able to get it programmed and it is running well so far.
As mentioned previously the thermocouple adapter wont work with the 12864 display. So I have jammed a thermistor on top of the heater block as I don’t want to pull out the thermocouple temporarily and risk damaging it and the thermistor is working OK for now. It is still off by about 15 degrees, too hot, and the temperature control overshoots. However the adapters are on their way.
The new nozzles form Micron finally came. It takes almost 4 weeks from Israel with normal shipping. I am currently using the high flow 0.5mm nozzle.
Fan for the print head heater must be directly connected to 12v. Not fan 0 or 1 . It needs to be always on. Not doing this lead to new and novel ways to jam the filament.
It pays to double check the tooth count on the gears. The first test cube was well undersized and over-extruded. The steps per mm settings are now corrected.
12v power supply was just not enough to get the print head up to temp. I needed to switch to 24v.
Still need to find a better way to adhere the cork to the bottom of the heated print bed.
The plain glass rectangular cutting boards form Amazon are working well as build plates. So far printing the PLA to the clean glass with no additives. Clean with denatured alcohol. On the old setup I was getting to decent adhesion with blue tape wiped with lacquer thinner.
Some of the issues were self inflicted as I was time splitting between house / yard projects, this and several other things. At times I hit my multi-tasking limits.
Finished test cube. Size is now dead on. This was with 0.35mm layer height, 60mm/s internal speed 30mm/s perimeters.
Now starting printer parts. Here is the back of the case for the display being printed.
The Repetier control software is great. It is much better than truing to control the printer directly via the twisty-turny knob on the display. Overall, I am rally impressed with the open source software (Marlin and Repetier) of this project.
Completed back panel for the display:
The dresser frames are now all glued up. As you might expect there is an experience curve here with the last one going most efficiently.
I also reached the conclusion that I would not be cut out to be a surgeon. The complex assembly with the time pressure to get everything done within the time window for the glue is stressful. This turns out to be quite vigorous exercise. I was sweating bullets by the time I was done each time (stripped to the waist looking more like Dr Pol pulling a calf than anything you would see in Fine Woodworking) . Lots of moves and also the physical effort of getting the pieces in place and clamped securely in time. With PVA glues you only ave about 15 to maybe 20 minutes to get everything fully clamped or it will be stuck in a BAD position. Even with the ends pre-assembled there are over three dozen pieces to get in position and clamped in that time.
Even with mortise and tenon joinery with nice big shoulders the case will be “out of square” when first assembled. As you can see in the previous post there are pipe clamps o the diagonals to pull things in line. However they are a pain to manage , requiring 2 people and easily falling off. Today I though I would use rope as a sort of “Spanish windlass” to pull things in , but I don’t have much in the garage. So I grabbed a couple of ratcheting tie downs out of the truck and used them first. WOW this is EASY!!! They work great whether attached to the clamps or wrapped through the corners. I am sure others have thought of this before, but I have not seen anything like this posted in woodworking magazines or when I was in the Wisconsin Woodworkers Guild. Here are a few shots of 2 different frames using this technique. My tolerance for these is to get within 1/16″ on the diagonals. If not this close, fitting the flush drawers will be a horrible task. You can easily ratchet them tighter and pull the frames square or back off a notch if you have pulled too hard. Plus you can weave them through the pipe clamps and other clamps without being stuck in an unreachable position as when using clamps to do this.
The printer frame is built out of V-slot linear rail http://openbuildspartstore.com/v-slot-linear-rail/. This is similar to the 80/20 and Bosch rails I have used on the CNC router but with one crucial difference: the slots are V shaped a the top. This allows rollers to run in the grooves to provide the motion without having to add linear rails like I used on the CNC router. This is far lower cost, but also lower precision. However, the test prints I am seeing from a friends unit look great.
The precision of the cuts is critical to rigidity and squareness of the printer. You should use a non-ferrous metal blade in either a chop saw or radial arm saw. I use a Bosch PRO1080NF 10-Inch 80 Tooth TCG (triple chip grind). This has served me well for several years. When using the radial arm saw as I did be careful to prevent over-feeding. The stop system on my saw makes for nice repeatable cuts.
Once cut, the ends of a number of the rails must be tapped for M5 screws. The extrusions already have proper sized holes but this is still tedious if using a regular plug tap. This is is where a “gun” tap or “spiral point” tap which is designed for through hole power tapping comes in very handy. I started using these for the 100’s of tapped holes on the CNC router. I used plenty of cutting fluid and a hand drill with the tap chucked lightly to tap the holes. Practice first and have some spare taps on hand prior to doing this on the v-rail. Snapped taps are darn near impossible to remove from aluminum.
I have not written much about 3D printing. The CNC router had a print head added at the end of the year. This was not what I would call a great success. Mounting hte print head assembly, a Micron EME, was fairly easy, but I also had to add the print nozzle temperature controller, which had to be close to the head as I used a thermocouple for temp sensing de to wanting to use high temperature filaments. I am using Mach4 for the CNC router controller. It has NO 3D specific functions built in despite a stock 3D printing profile. So there is no temperature control for the print head or print bed provided. Both of which are are more important than I would have guessed.
The print head extruder stepper is set up as the A axis. I have the nozzle temperature externally manually controlled as required for each type of filament. Currently I am running PETG from eSun. The print bed is a sheet of glass (cutting board from amazon) with blue painter’s tape to aid adhesion. The CNC router does not have a heated bed, but I can aim a heat gun with diffuser under the raised glass build plate to have a sort of heated bed. The bed needs to be in the vicinity of 80 degrees C to avoid the parts warping and pulling away from the bed. The heat does make a HUGE difference when trying to print anything over about 1-2″ across.
The printing required considerable tuning and testing of the parameters. I am using the open source program Slic3r to do the slicing to prepare for printing. It took about 20 test objects to get things reasonably dialed in from scratch
3D print underway.
Finished product prior to trimming and clean up. This was at 0.35mm layer height.