AVOR – a steel-framed medium-size router

[Neale]

The Electrickery

The early design decision which drove much of the design in this area was the selection of the CSMIO-IP/M motion controller. This choice was driven mainly by discussion on this forum. Not the cheapest option, but I was attracted by a number of features:

• Ethernet communication to PC (noise resistance)
• 24V supply and external digital I/O signalling (noise resistance)
• Differential signalling to stepper drivers (noise resistance)
• No BOB needed – built-in to unit
• Built-in analogue output (and input, although I’m not using this)
• Good reputation for build quality and reliability

It required a move to Mach3 (Mach4 would have been possible but I preferred to go Mach3) from LinuxCNC which I had been using with the Mk1 router. LCNC does have motion control hardware available, but this seems rather more difficult to specify (and understand!) Not sure if I would go the same way today, but would need to do more research on the LCNC added-hardware options first. Anyway, the IP/M was available off-the-shelf from Zapp and I bought during one of their “special offer” periods where (at that time, I believe) the LCNC-supported hardware was only available from the US. Again, this might have changed – I’m talking about decisions made around 3 years ago.

Having chosen dual X motors I wanted to go digital for the stepper drivers, replacing the older analogue drivers I had used with the Mk1. Generally described as giving smoother motion with various anti-resonance features, for me, the most important feature was stall detection. With a dual motor setup, if one X motor stalls for any reason, you want the other to stop immediately before you potentially break something by driving the gantry on one side with the other locked. These motors are surprisingly powerful! I bought EM806 drivers which are a later updated version of the popular AM882. At that time, I thought that the AM882 would become obsolete and I wanted something that was going to stay current for a while; in the end it seems that the AM882 is still going strong. However, I only use EM806 for X and Y; the Z axis uses an old M752 salvaged from my old router. Saved money and I’m not sure that the Z axis needs the advantage of the digital driver or stall detection. Seems to work fine in practice. I use a 68V linear power supply for the steppers. The original 500VA toroidal transformer failed and I replaced it with a 650VA item but it doesn’t even get warm with my normal machine use.

To keep the control box contents cool, I have a couple of 12V 120mm fans on one side with foam dust filters that get cleaned from time to time. However, they are run in series off 12V so just give a gentle waft of air over the drivers and then through the box.

I was able to pick up a cheap “new but shop-soiled and obsolete” Pilz safety relay on eBay. 24V so it worked well with the PSU needed for the CSMIO. I’m still not quite sure exactly how the thing works internally, but the effect is like having two relays with their coils wired in parallel, but their switch contacts wired in series, and with a latching capability so the thing is reset with a momentary-operation switch. Both coils have to be activated so that all their switch contacts are closed for the machine to work. Even if one relay fails to release when the e-stop is operated, or a switch contact welds together and does not open, the other relay will still act as a safety device - it is very unlikely that both would fail together. If I were doing this again and couldn’t buy a cheap Pilz device, I might well consider doing something similar with two ordinary relays. Not quite the same level of reliability, perhaps, but pretty good if you don’t have to satisfy the safety mafia who want everything documented. Anyway, my safety relay has (the equivalent of) three n/o contacts plus one n/c contact. One n/o switches the enable signal to the CSMIO, one n/o switches mains to the stepper driver PSU (via a 24V relay), and the n/c switches the enable signal to the drivers. Ideally, this last signal should also go via a n/o contact but the wiring was slightly easier this way. Belt, braces, and electric trouser hoist – any one of these three actions should stop the machine. The safety relay can be tripped by any one of the three e-stop switches mounted around the machine and there is a single "reset" button on the front panel.

The fault signals from the digital drivers are wired to the “driver fault” input on the CSMIO. The hardware documentation tells me that the response time is very quick like this, although clearly I am relying on a single safety mechanism via the CSMIO firmware here. However, a motor stall might damage the machine but it is not the same as a “personal safety” issue which is handled by e-stop switches and the safety relay. Limit switches are wired to the CSMIO; a limit switch event does not cause an e-stop event and I have to trust the CSMIO firmware to do the right thing.

So, I have to rely on the CSMIO for limit switch events. I do have limit switches at max X and Y travel as well as min; the min limit switches are also used for homing as usual. Each axis has both limit switches wired in series. There is a single min limit switch on the slave axis but this is not wired to the motion controller. Just a single limit switch at top of Z travel, also as usual. I’m not happy about the reliability of the cheap proximity switches that I have bought. It was a cheap box of ten from eBay; one was the wrong type, one never did work, and the rest seem to have virtually no hysteresis between on and off switching points. This was a pain when first setting up the machine. Because I have the home switches for each axis wired to separate CSMIO inputs, I can home more than one axis at a time. I have the machine configured to home Z (to raise the tool clear of any work) then X and Y together. What happened was that Z homed fine, then X and Y started moving, and almost immediately I had a limit switch trip. This could only be on the Z axis, as the X and Y switches were being used for homing at this point. Eventually I discovered that Z homed correctly by moving until the switch tripped “off”, then moved a tiny amount in the opposite direction until it tripped “on” and Mach3 used this position as “home”. This is normal Mach3 homing action. However, because the on and off trip points were so close, once the Z axis had homed any slight vibration was enough to cause the Z switch to trip “off” again – which was now being interpreted as a limit switch event. I am very grateful to Dean/JazzCNC for pointing out that the CSMIO allows you to define a small offset for the home position; now, my Z axis goes through the usual homing process, then drops 0.5mm and sets “home” there. Far enough that the switch does not trip accidentally.

I mentioned proximity switch reliability – I have had two of them fail while fitted to the machine. This is one reason why I also use soft limits set slightly inside the physical limit switch trip points. I find this very useful as I often jog an axis out of the way, and you can just run the axis up to its soft limit without any ill-effect or physical limit switch event. The only down side that I have found with soft limits is that they prevent use of jogging unless the machine has been homed. If you want to rehome the machine and one or more axes have a long way to go, it can take a bit of time as homing speed is rather slower than full rapid speed. However, homing is a bit faster by having X and Y home simultaneously so I don’t usually bother to turn off soft limits to jog in this situation.




For completeness, here are a couple of pictures of my control box. External controls are mains on/off and pilot LED, reset button and associated LED, e-stop, and spindle coolant pump switch. This has two positions and centre-off; turn to left gives 12V on pump (only used for filling system and purging bubbles) and turn to right gives 5V – slow trickle of water through system which is more than adequate and should extend the life of the cheap water pump I am using. I often don’t bother with the spindle coolant if I’m only doing a smallish job – I just check the spindle case temperature from time to time. Everything else is driven through Mach3, using a wireless PC keyboard with integrated touchpad and a wireless MPG “pendant”. You will see warnings against using wireless keyboards like this because of potential electrical interference generating false keystrokes but I have never seen this problem.

External connections are made via cabinet-mounted sockets, with the exception of the hard-wired mains lead. Stepper motor connections are via 4-pin XLR connectors. They lock in place, so minimal chance of one falling out, and the contacts, on paper at least, are rated for the motor current. I started by hard-wiring motor connections to DIN rail terminals but found that I was continually needing to remove the control box while setting up and testing so went back to the same connectors that I had used on my previous machine, and where they had worked successfully for a number of years. Control and similar connectors use what were described as “avionics connectors” or GX16. I use 4-pin versions which have a screw-lock to keep them in place. There are 5 of these altogether and carry e-stop, limit switch, spindle coolant pump, and VFD connections. The remaining connector looks like an XLR connector but inside the shell is an RJ45 ethernet connection. The actual plug and socket are pretty much standard RJ45 but the cabinet-mounted socket has an XLR-compatible fitting around it and the plug is an XLR shell that fits over the RJ45 plug for protection and locking. These are available off-the-shelf from people like Rapid Online, but make sure that you get the one that fits over an existing connector if you want to use a commercial Ethernet cable. Otherwise, you are going to have to fit the RJ45 connector yourself AFTER putting the shell on the wire. Guess how I know this… The RJ45 bulkhead connector has a short length of Cat5 cable inside the cabinet connecting it to the CSMIO.

All cables on the machine run through cable chains; motor, limit switch, spindle power, and spindle coolant pipes all run through the same guide. I have tried to keep limit switch and motor cables (all CY) on opposite sides of the cable chain and I have not seen any interference or false triggering effects. I suspect that the 24V signalling voltages are a great help here – much better noise rejection. I have also run ground wires through each section of cable chain between ground points close to each end of the cable chain. This removes any reliance on ground connections via the Hiwin rails. Certainly the touchplate mechanism is very reliable – no need for crocodile clips to the tool bit or anything like that.

As well as a movable touchplate, I have also installed a fixed touchplate set into the bed, just below bed level, which works with my Mach3 macros for tool-height setting for either single tools or multiple tools per job.

[Clive S]

Bloody hell Neale you are putting me to shame good write up.

[Doddy]

Well... that's me nicking a few ideas, designs and pointers.

In particular - very interesting about the all-steel construction, well argued case.

[Neale]

Well... that's me nicking a few ideas, designs and pointers.

Please do - that's exactly what I did!

It's taken me nearly as long to write this as I did building the machine, but I know that I have left out a lot of detail. For example, I'm very happy with the way I was able to run the Y ballscrew through the gantry uprights because wherever else I put it seemed to have disadvantages. I've made a few mistakes as well - believe or not, I completely forgot to allow for the bed supports and bed thickness when I was calculating Z clearance under the gantry. Fortunately, it's fine for the kind of work I normally do.

I'm happy to provide more information on any points I have not mentioned - or any of the ones I have but where I haven't gone into detail.

[Neale]

Bloody hell Neale you are putting me to shame good write up.

Not really, Clive - I just wanted to demonstrate that I do really build stuff and not just talk about it!

[Boyan Silyavski]

I like the gantry very much

[routercnc]

It's taken me a while to post because it took so long to get around to reading it all !
Nice build and great write up. Will be very useful for someone starting out to get up to speed with it all.

Epoxy levelling comments made me think a bit and yes levelling without the bridge is possible if there is adjustment on the gantry to ensure the x bearings run smooth and the y axis is epoxy levelled afterward. Got me thinking which to choose for my mk4 machine. Still like the bridge but agree it can work without

Nice control box also and full of nice kit

[Zeeflyboy]

Looks like a really nice design... Will be following with interest!

[magicniner]

You could have called it RWAIN - Router Without an Interesting Name ;-)

[Neale]

Thanks, Nick - I'll save that one for the Mk3!