Wednesday, November 12, 2008


History Repeats Itself

Smug in the knowledge that I'd outsmarted this particular pitfall, I still managed to get caught by the very problem I thought I'd fixed.

The other night I noticed my Darwin had shut down just after laying down the raft for the X carriage. The error log showed that my temperature sanity check had been triggered, so I went down to the lab to check on things.

At first glance it looked like the PTFE insulator simply failed...

However, once I peeled off the insulation, the thermocouple popped right off the heater, as did most of the JB weld. Oops. So, similar to what happened to Nophead's extruder, my thermocouple came loose, causing my heater to overheat, which in my case caused the heater barrel to melt its way out of the PTFE insulator.

The sanity check did catch it, and prevented any further damage from occurring, but it didn't catch it quite in time to prevent wrecking the PTFE insulator. Which is too bad; I've managed to print over 80 parts so far, and only had 4 more large parts to go (plus a handfull of small bits) for a complete set of printed Darwin parts.

Ah well, back to the drawing board. Next up is fire cement!


Well I know this info might be kind of late but on my Darwin the PTFE has always been kind of weak. I have been more afraid of the heater barrel getting ripped out/knocked sidewyas more than of it droping out of the PTFE. So I decided to drill a small 3mm hole about 1 cm up the PTFE Barrel and sunk a set screw in there to help hold the barrel in place. Now I realize in you case this would only maybe give your code a bit more time to realize something wrong and delay the inevitable of the heater barrel and now set screw to melt through the heater barrel. But this still will not solve the problem of the JB weld giving out. Also when I drilled the hole I mad sure that the hole went deep eough to take some of the threads off the heater barrels so that there was a place for the set screw to sit.
The metalab group is using PEEK, I believe glass filled, succesfully, as mentioned at:
The latest lasercut extruder design is worth looking at. That traps the PTFE and takes the load on a big washer. Have a look on the Wiki:

Oh, I'd suggest sticking that thermocouple on a tab...

Vik :v)
Vik, how much of a temperature differential do you get between the bore and the tab? That does look like it might withstand a few temperature accidents. And what do you stick it on with; I'm a little leery of using more JB weld at the moment... :)

I'm actually building up one of Ian's laser cut extruders right now; about to lay down the cement on the heater. I'd been putting off using it before as I was afraid the gear ratio was going to slow down my flow rate, but at this point it's all I've got.

I've got spare parts for 2 more Mark 2 extruders, except for the PTFE bits. I'll have to find a Canadian source for PEEK. And a new space heater, can barely type it's so cold right now!
Would it be possible to monitor the electrical load characteristics of the heater itself, rather than an external thermocouple, to get slightly quicker detection of faults?
You guys are aware that thermocouples are not fragile, aren't you? In fact, I believe that in this case, you should spark weld the TC wires directly onto the barrel, and thus do away with adhesives & all.

That's right: the two thermocouple wires don't even have to touch. All you have to do is make sure there's an electrical path of reasonably low resistance between the two, and you will get a reading indicating the mean temperature between the two wire attach points.

I would strongly recommend to anyone using thermocouples to read carefully the wealth of engineering info in the Omega Engineering handbooks. There are, of course, many other fine publications in which you can learn about TCs, but those are free.
There's an idea - spot welding. I don't have anything that'll weld brass right at the moment though, but I could probably come up with something.

There is the issue of the heater coils shorting to the barrel. With a welded thermocouple, I think that would mean I would now have a blown thermocouple chip, if not a blown Arduino too, as well as a burned out heater. Will have to look into that more closely.

Most people are using thermistors though, thus the interest in tabs and mounting ideas.
You are right about the risk of shorting the heater, but damage can be easily prevented with a pair of resistors and a few diodes (provided the input impedance of the amplifier is high enough). Adequate protection of high-impedance inputs is always a good idea, no matter what.

As far as spark welding goes, you could be surprised of the results obtainable with only a large capacitor (1000uF @ 20V comes to mind). Some experimentation may be required (I've never done this on brass), but using small diameter TC wire makes welding fairly easy.

But my point was more to the fact that you can generally abuse a TC quite a bit without loss of precision. Whether it is welded, jammed in a hole by a tapered pin or held down by a saddle clamp, pretty much anything goes. Heck, just separate the TC leads, insert the barrel between the two, and twist the (insulated) leads together to lock in place. Quick and efficient, as long as the measuring point touches the barrel.
Hm, sounds good. I suppose I could just pinch the thermocouple in a tab, no glue no nothing. Or even better, I could just jam it in between the nozzle and the barrel as I screw it in.

Wish I had some more thermocouples to play with... It's rather amazing how many different things can go wrong on a Reprap; I'm a little worried about destroying my only thermocouple at the moment.

D'oh! It came with really long leads, which I cut and most likely threw out; if they were the two metals all the way up, then I could have made a pile of thermocouples out of it...

I'm off to rummage in my spare parts bin...
Ah, yes, very cool. Jamming the wires in the threads between the nozzle and the barrel works great. I now have three thermocouples too, when I thought I only had one. Funny thing is I thought to order extra thermocouple boards and chips, but not the thermocouples themselves.

We'll see how it holds up over time, but you really couldn't ask for a better temp sensor placement - I should be getting a reading something like the average temperature between the inner nozzle surface and the outer barrel.

Thanks Bert! I'm off to do some thermocouple reading.
Funny thing is I was taught this in school too; I just didn't make the connection. :)
Glad I could be of some help. Sometimes, it is the simplest things that get overlooked.

Vendors are seldom keen in telling you that all you need is the wire, they'd rather sell you small fused beads for a hefty premium! ;-)

Perhaps you should take some time to document this for others...
Will do, a perfect candidate for the Builders Wiki! But first, some testing. My new heater assembly is curing; will test it out more thouroughly tomorrow. Thanks!
Tom, my sanity check does monitor the load in a simplistic way; this error was fairly subtle though. As the thermocouple slowly comes loose, the power input slowly increases for the same indicated temperature, while the actual temperature increases. Difficult to separate that from ambient changes in temperature, air flow, head feedrate or plastic feedrate. And every false alarm results in a ruined part. It's a tricky problem.

I really like having the thermocouple separate from the heater coil; that seems like an elegant solution. No glue is even better!
More by good fortune than anything else my current extruder maxes out at 290C when the heater is on 100% and it is extruding. It is about 8 Ohm and gets the full 12V (no Darlington) but I have no thermal insulation.

The PTFE seems to survive that and I have been experimenting extruding ABS at 280C with some very odd results, which I will blog soon.

Perhaps a good safety solution is to insert a small series resistor to limit your extruder to 290C. You could make it from a bit of spare nichrome and adjust the length. Obviously it will get hot.
About connecting the thermocouple directly to the heater barrel - there's a new problem.

Right now, as the fire cement dries, I've got essentially a wet cell - the dielectric being the wet fire cement, and the two electrodes are the the steel/brass/copper heater assembly, and the copper/nichrome wires. I'm getting point three volts between the barrel and the heater coil with nothing else attached; and since the heater coil and the thermocouple circuit share a ground, that voltage is drowning out any voltage the thermocouple might produce.

It might work once I cure the fire cement completely, but I'm leaning back towards using a tab and some adhesive to insulate the thermocouple; it's easier than redesigning the thermocouple circuit (for me at least).
Assuming you're using the reprap AD595 thermocouple sensor:

From the "Alarm Circuit" section on page 7 of the AD595 specification sheet:

accessed from the AD595 page at:

It seems that if the thermocouple is shorted, it will trigger an alarm, rather than blow the thermocouple chip.

I haven't touched electronics in years, so it would be good if someone who has could read over the alarm specifications and see if you really can put the thermocouple to the metal, safely.
FWIW, I've been using the same thermocouple + AD595 throughout the process of refining my heater assembly, and even though I've abused the thermocouple a great deal, it still works well. The alarm light has come on a lot, but I finally resolved that by wrapping the thermocouple with PTFE tape.
Hm, reading the data sheet a bit more, I see that the absolute maximum voltages on the input is Vs - 0.15V. That means if the nichrome shorts on the positive side, it'll likely kill the AD595, because it's only running at 5V. However, it might work nicely if I run the AD595 at 12 V instead of 5; it still gives the same output, although I suppose a really high erraneous measurement could cause problems - IE, 12 V to the Arduino a-d pin, which is only good for 5.5V. Only 15 mA though, might be OK.

The + input is currently grounded to signal an open thermocouple with an extreme negative reading; that'l also cause problems with a heater short. Think I might cut that trace, and run the alarm circuit back to the Arduino.

My heater is behaving now that the fire cement has cured; I'd like to set it up so that the two thermocouple wires are not connected but clamped in place between the nozzle and barrel; the idea being that any problem, be it shorted heater coil or thermocouple coming loose causes an error condition and shuts down the Reprap.

I think it may work, but I will have to do a bit more testing.

I've been meaning to comment here for the last few days, but I haven't gotten around to it until now. While I, personally, am not an expert on Thermocouples, I work in the technical support and calibrations section of a major, multi-national, safety testing laboratory and one of our most vital testing tools is TCs. We go through 10s of thousands of feet of TC wire every month at this lab site alone, most of it type J. While I'm not a TC expert, per say, I have picked up a decent amount in the process of developing automated test fixtures.

There are a few things I wanted to comment on:

- TCs should respond pretty quickly to temperature changes.

- Omega Engineering is a great resource for TC info

- You can weld or solder TCs to the device, but be aware that the presence of additional dissimilar metals (including the solder) adjacent to the TC joint may cause your temperature measurements to be off. The amount it affects it will depend on the metal being soldered/welded to and the type of TC you're using. Different TC types are made from different metal pairs and have different electrical properties over their temperature ranges. The relationship between voltage and temperature tends to be non-linear, so even if it seems to not have much effect at room temperature, it may have a much larger effect at your working temperature. Ideally, you want a TC type that needs a high change in voltage to have a significant effect on the temperature reading, I’ve been told type T might be good for that.

- I, seriously, doubt you will get anything resembling accurate readings if you put the metal object between the two sides of TC wire. The whole idea behind a TC is measuring the voltage across that joint of two, specific, materials and converting that value to a temperature based on a known lookup table. If you put a third metal between the two known metals I would think you’d totally mess up the assumptions that go into that calculation.

- Usually, here at work, we use epoxies and glues, like Loctite, to attach most TCs to the item being tested. It works pretty well but, obviously, you have to pair the adhesive to the expected temperature range of the item under test.

Hope this helps.
Cool. Thanks for the input!

To follow up, I ended up not welding the thermocouple wires, but mechanically trapping them between the the nozzle and barrel. It was harder than you'd think, but it allowed me to build the rest of the parts without problem. I did have to increase all my temperature setpoints by 10 deg C, but I expected that.

Measured temperatures are rock solid now; previously I had several deg C of noise in the readings, but now the noise level is below the resolution of the output (1 deg C). The noise level on the Arduino power supply also effects this, probably via the A-D converter reference voltage.

The big advantage with not welding the thermocouple for me is that if there is any mechanical failure, and one or both wires come loose, I immediately get a temperature reading of 0 deg C, which triggers the shut down.
Post a Comment

Links to this post:

Create a Link

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to
Posts [Atom]