A call for help/ideas to develop the Heated Bed

With the meters and meters of left over 30 gauge nichrome wire I have, plus the left over Kapton tape, I need only pay for the sheet of aluminum. I bought one at remnants pricing that is 0.25" (6mm) thick. Now I need to layout the PCB and get a bunch of extra MOSFETs. It looks like Digikey has a better price on logic level input MOSFETs.

I had originally expected to use nichrome wire for the heating elements, and that has a resistance that changes with temperature. By measuring the voltage across a current sense resistor, or maybe even the MOSFET itself, I could measure the resistance of the heating element and know how hot it is to regulate how much more power to send to it. But the copper on the PCB has a very low electrical conductivity change with temperature, so it will be difficult to get the temperature and power the coil with the same 2 leads. I may have to go back to adding a thermistor for every single zone just to keep track of each zone's temp.

Mike

Mike,

I think you'll find these various articles/documents/etc. of interest.

[focus.ti.com.cn]
[www.ami.ac.uk]
[www.ultracad.com]
[www.ami.ac.uk]
[www.ultracad.com]

I haven't sat down and carefully gone through these yet but I got the impression after an initial read that it may be possible to measure temperature of PCB etch without adding thermistors, particularly if you can make and store reference measurements at a known temperature (i.e. 25 degrees C) for each etch segment.

TC

Mike,
Actually the resistance of copper changes about 20 times more with temperature than nichrome. [www.allaboutcircuits.com]

---

[www.hydraraptor.blogspot.com]

Funnily enough looking at the figures Aluminum more so.....

Shame there is not an easy way to etch aluminum. Otherwise it would be possible to laminate cooking foil onto plate glass and then etch a heating track into it. Perhaps using the cheap toner transfer method.

Voila heated perfectly flat build table.

I agree it does open the way to cheap evenly distributed heating elements though.

Following the figures it becomes apparent that the "Cold" resistance of a copper or aluminum heating element needs to be somewhat lower such that enough power can still be introduced once the resistance increases with temperature. At higher temperatures the dissipation will be greater, but the increased resistance will limit reduce the amount of available power for heating.

There will be a sort of self regulating effect at or around an equilibrium temperature. This will change as material is printed. due to the material either conducting or insulating the plate.

Given the above and that the cold resistance will be lower (ergo more current drawn when cold) a current regulating driver circuit might be a needed to regulate inrush and keep the temperature at a set point through out the build.

---

Necessity hopefully becomes the absentee parent of successfully invented children.

Cold/hot air blow to help remove the printed piece

Yesterday I let a piece cooling on Heated Bed. This morning it was impossible to remove. Then I remember to try quick hot it with blowing air, using hairdryer. It quickly almost jumped alone from Heated Bed :-)

Also I already did the contrary, blowing cold air from my mouth when the Heated Bed was hot. It worked for small pieces.

Does anyone have the same experiencies?


PTFE tape

DealExtreme have now PTFE tape, maybe it can be good so insulate some parts of Heated Bed?

---

---
New cutting edge RepRap electronics, ARM 32 bits @ 100MHz runs RepRap @ 725mm/s:

[www.3dprinting-r2c2.com]

TC ->
Great research work there! Thanks for find all the good formulas for me!

nophead -> Amazing! The fact that nichrome has an execptionally LOW thermal coefficient of conductivity blew me away! I really thought that copper would be flatter, and resistance wire steeper. Especially since Nichrome is mostly nickle and chrome and iron, it has the lowest of the ones listed. I would have expected the alloy to have a greater coefficient, not a lower one.

The 0.0039 (or 0.00404 for pure copper) make sense in that if you go in the opposite direction, 0 degrees Kelvin, absolute zero, is around -275C. And if you multiply 0.00404 by -275, you get -1.111, which makes the conductivity infinite at absolute zero, which is correct, as copper approaches superconducting at absolute zero (prevented form quite getting there by impurities)

This means that all copper traces for heating and temperature sensing is back on. The difference between the resistance (and therefor current flow) of the copper heating coil changes by only 4.38% as the copper goes from 200C to 220C. So to get even a 1 degree C resolution, the analog to digital conversion must be sensitive, and accurate, to 0.21% changes. Over the full 5V range, that means accurate to 11 bits, which is a bit more than an Atmel ATMega can handle. They can do a differential mode down to 200X, but the precision drops at the same time.

Assuming an average peak heating rate of 200W for fast warm up with all zones on, with each zone only getting 1/8 of the total power that it 200W/8 = 25W per zone. So at temperature, I want the resistance each zone to be 12v²/25W = 5.76 ohms. That gives the current at 12V/5.76 ohms = 2.083 Amps. Now, with the 0.0039 temperature coefficient, that means that 1 degree C will be (1.0 + 0.0039*201)/(1.0 + 0.0039*200) = 1.0022 times the resistance, and 1/1.0022 times the current. Now assuming that I use the Fairchild FDU8878 MOSFETs I just bought with an on resistance 0.012 ohms at 4.5V gate voltage, that means that the voltage drop across the MOSFET should be 0.014*2.083 = 0.02916V, and when the heater is 1 degree C hotter it will be 0.02916V/1.0022 = 0.02910V. This means a difference of 0.02916 - 0.02910 = 0.00006V per degree C around the 220C working temp. Even with the times 200 differential input of an Atmel ATMega, that is 0.0124V. The datasheet says that the differential ADC input is only good to 6 bits at the times 200 scale, so it looks like the ATMega can not measure that small a change.

The plan B for this is to use a couple of quad op-amps set for a gain of 150, which will make the voltage across the MOSFET 4.374V, and the difference per degree 0.009V. With a 10 bit resolution over 5 volts gives the minimum resovable step size of 0.0049V, or about 2 steps per degree.

Plan C is to assume that the heated bed only starts at 200W. With the change in resistance going up 1.78 times its room temperature value, the current will already be 1/1.78 = 0.562 times the starting current. 2.083A * 0.562 = 1.170A at temperature. Multiply that by the On resistance of 0.014 and the voltage drop should be 0.0164V. Setting the op-amps for a gain of 250, the 220C voltage would then be 4.096V, and the one degree C difference 4.096v / 1.0022 = 4.087V or 0.009V. No improvement over plan B, which means plan B was probably wrong and optimistic.

It is fortunate that the copper's electrical conductivity increases with temperature to form a negative feedback loop and stabilize a bit around the desired temperature, because it looks like it will be tough to use the change in resistance of the heating element to keep the temperature of each zone precise.

Mike

---

Team Open Air
Blog Team Open Air
rocket scientists think LIGHTYEARS outside the box!

RE: A/D resolution check out this article about improving A/D resolution using a "running average filter".

[www.planetanalog.com]


Have you figured out that you can achieve a resistance as high as 5-6 ohms using serpentine etch?

I was thinking that the approach would be to have a fixed current source that can be pulse-width modulated to control temperature. Measure the voltage across the resitance at a fixed current and you can calculate the resistance (R = V/I, of course).

I'd bet that a differential OpAmp will be needed to scale the voltage across the resistance up to a range that can be measured easily by your typical A/D converter. A fixed offset voltage could be applied to so that the output of the OpAmp is zero (or near zero) at 25 degrees C, and nearly full-scale at maximum temperature (220?).

I'm sure that many references for low resistance measurement methods can be found via a web search.

TC

aka47 - good observation re: aluminum.

I'm going to think about that!

TC

Surely if you're going to all that bother it'd be much easier to use thermocouples/thermistors/i2c temperature sensors? Or even:

[www.rapidonline.com]

in series with the heating element. Trying to auto-regulate using a negative feedback loop based on resistance of the element is going to take forever to reach a stable temperature. I'm using pla which is easy to remove, but for anyone with kapton/abs could you leave spaces between strips of tape? Pla doesn't stick well at all to hot aluminium, but i've had no problems leaving 3mm gaps between strips of masking tape. It might be a way to reduce contact area, and hence difficulty to remove, without letting the part warp any more.

Certainly, any "solution" should be compared with alternatives.

However, I don't think a PWM modulated current source, a differential OpAmp, and an A/D converter is very complex to construct. Sure, careful analysis for the PCB serpentine etch approach is required, but construction "could be" relatively simple. Simple construction is the thing that interests me about this approach.

On the other hand, I'm not convinced that an array of thermocouples/thermistors provides is simpler to construct. I'm also not convinced that a control loop limits the time to get up to temperature. The PWM can be set to 100% when the temperature is not in the desired control range and so the rate at which the system heats up is limited by the peak current.

Of course, if the serpentine etch can only yield a very low resistance then the peak current would have to be very high (but, at the same time, the voltages would be very low). Of course the etch has to be wide/thick enough to carry the required current. This fights against the desire for high resistance. Until someone figures out what is a reasonable/achievable resistance its tough to know. So for now, I'm not convinced that the serpentine etch approach will prove to be practical.

Of course, these discussions are speculative, and so are my opinions.

TC

TC: I didn't mean a control loop would take longer to get up to temperature, I had thought rocket_scientist was trying to use the increased resistance of metals at high temperature to auto-regulate the temperature. As for thermocouples/thermistors, I have a bunch of i2c chips lying around so i'd find it easier, but objectively neither approach is particularly difficult.

I have to say, i'm curious about the approaches people are using, but for me I don't think i'll have much use anywhere between the heated bed i'm using now, and some sort of roller bed that can act as both an axis, and a way to dump parts in a bin.

Thanks for clarifying James. I definitely didn't get your point before, but do now (and I agree).

I don't have a RepRap yet (still learning) and so I'm a long way off from experimenting with any of these ideas. I'll probably build a vanilla machine to start with just to get something working. But I expect a heated bed will be among one of the first modifications I'll want to try.

TC

LOL

I have been contributing here for a couple of years and still don't have a working machine.

I keep getting distracted with new bits of machines to play with.

The serpentine etch is complex to design and a bit mathy, there is going to have to be a certain amount of trial and error too. The big advantage is once you have the parameters sorted and a functional design done, production is very cheap and easy.

I could even do it on my kitchen table (Where I do most of my work) using stuff I have a round in the kitchen.

Has anyone got any ideas for a suitable high temp adhesive to laminate kitchen foil onto glass with ?? Pref cheap.

I seem to remember reading somewhere about self adhesive copper foil for EMF screening etc. Can't think that it would be inexpensive though given current market prices for copper.

---

Necessity hopefully becomes the absentee parent of successfully invented children.

Just thought, removing kapton tape from glass is very easy as you can use a scraper (The sort that you put a stanley knife blade in) due to the glass being very smooth and hard.

A quick google for etch aluminum suggests that it can be done with a number of solutions.

1. Ferric Chloride (same conentration as for PCB making)
2. Salt and Copper Sulphate (50:50 solution, works better as an electro etch)
3. Caustic Soda
4. Hydrochloric Acid and Peroxide (the same starter solution used to kick start copper chloride etching, ie brick acid and peroxide)

Questions:

Which side to put the heating trace on. Do we put a heating trace on the top side sandwiched between the glass plate and the kapton layer. Or do we put the heating trace on the back of the glass plate and the kapton layer on the front.

Certainly putting the heating trace on the back makes it possible to replace the kapton without trashing the heating element. But means you are heating up what is a poor conductor (which could be desirable though as less energy would be needed to keep the base of the print warm)

---

Necessity hopefully becomes the absentee parent of successfully invented children.

James, TC,
I should clarify myself, too. I was planning PWM control of a fixed voltage (12V) to control each of the 8x8 or 64 zones. I was not planning on using solely negative feedback from the resistance change, even with thermistors or thermal switches added because that would limit the heated bed to only one temperature. I am not sure how often I will work with PLA because it melts when exposed to very hot water, but the Nylon would likely take an even higher temperature.


Using a separate thermistor or thermocouple for each zone is going to up the cost and complexity. I would probably have to run a second sense line for each column, and mount all 64 thermocouples or thermistors in the middle of it's zone, directly connected to the Row Power Line, and also connected the column sense line. Doable, but I think I would rather add 8 op-amps and a reference voltage generator on the card edge than double the number of column lines and add components to each zone.

Mike

---

Team Open Air
Blog Team Open Air
rocket scientists think LIGHTYEARS outside the box!

OK, the feasability study.

Assume a 12 inch by 12 inch heated bed, with 8x8 separate heating zones. That makes each zone 12 inch/8 = 1.5 inches square. Let me further assume that the traces and spaces are the same thickness, and that 90% utilization of the 'copper' half of the space. That gives 45% copper trace and 55% open space. The area of the copper is length times width = 1.5inch² *0.45 = L*W*1.0125. This is close enough to simplify to L * W = 1 inch² or L = 1/W or L = 1.0E6 mills²/W


Resistance is L*R(t)/(W*thickness). I can easily use the PCB Fab-in-a-Box boards with my laser printer, and they are 1/2 ounce copper, or 0.0007" thick. We want 5.76 = 6.788E-4 ohm mills*L/(W*0.7), or 5.76*0.7*W = 6.788E-4*L. Substituting L = 1.0E6/W we get 4.032W = 6.788E2/W or W² = 678.8/4.032 = 168 mills². W = 13 mills, and L = 1.0E6/13 = 76.9 inches.

To check my math, 0.013 * 76.9 = 0.9997 square inches. Resistance is 76900*6.788E-4/(13*.7) = 5.74 ohms. SO the math checks out. The current carrying capacity from the reference above for 13*0.7 mills square = 9 square mills. This drops completely off the bottom of the tables in [focus.ti.com.cn] table 1, but then their max delta T is much smaller, too. At 25 mills² and 20C for the delta T, the max allowed current is about 1.8 Amps. Multiplying the delta T by 10, dividing the crossectional area by 3, we should be able to handle 3 times the current. This means that the design goal of 2 amps should be safe, assuming the PCB can safely heat to 220C at all.

So, the answer is yes, it is feasible to make a 64 zone PCB heater. IF the PCB can handle working at 220C.

Mike

---

Team Open Air
Blog Team Open Air
rocket scientists think LIGHTYEARS outside the box!

Why are you aiming for 220C? That is the temperature we extrude ABS at, but the bed is about half that.

At 220C I think a PCB will degrade pretty quickly. It will certainly go brown and I think the glue might let go of the tracks. It is not far short of peak reflow temperature which you only hold for a few tens of seconds.

---

[www.hydraraptor.blogspot.com]

Today morning I had another Mendel frame vertex to remove from the Heated Bed. I leave the piece and Heated Bed cooling all the night.

In the morning it was impossible to remove it and so did it again using hot blow air, using the hair dryer. After 20 seconds a sound like "cracking" happen and the piece released itself :-)

I will write this info on the wiki page. I guess that if I use cold blow air when both piece and Heated Bed are hot, the piece will also release itself.

---

---
New cutting edge RepRap electronics, ARM 32 bits @ 100MHz runs RepRap @ 725mm/s:

[www.3dprinting-r2c2.com]

rocket_scientist (Mike)...

Cool! Thanks for doing that feasibility analysis. I walked through the resistance calculations and I agree with your numbers. As a double check I filled your numbers into this calculator:

[circuitcalculator.com]

I got 5.75 ohms at 25 degrees C and 10.1 ohms at 220 degrees C.

I have to admit that it is hard for me to conceptualize fitting 77 inches of etch in a 1 square inch area. It is also hard for me to believe that a 13 mil 0.5 ounce copper trace can carry 2 Amps (I haven't checked those calculations yet).

I do wonder why you are thinking that the heater bed has to get to 220 degrees C. I thought the goals were more like 100-110 degrees C. I think standard FR-4 laminates are good to about 130 degrees C (but I don't know for sure). There are PCB laminates designed specifically for high temperatures and use of one of those may be needed.

Constructing the board from standard FR-4 sure would be the simplest approach. Any change you can fab a prototype of a 1 square inch serpentine board with your laser printer fab method and report some experimental results? Or is more research and calculation needed?

TC

Here's another calculator. You can enter current, length, oz/in, temperature rise, etc. and it calculates trace width, resistance, voltage drop and power dissipation. The temperature range is limited however.

[desmith.net]

TC

How much power is needed for the heated bed in total (watts)?

I would think there should be a reasonable peak power that is needed to heat the bed up at a decent rate and a nominal power level needed to maintain the bed at an even (hot) temperature.

With a multi-zone bed, the total peak power would get divided by the number of zones to yield peak power per zone (watts/zone). It seems to me that the calculations that rocket_scientist did results in much more power/zone that is really needed. Using this calculator:

[desmith.net]

... and assuming a 25 degrees C ambient + 85 degress C rise (total 110 degrees C), and assuming a 1 Amp peak current, 0.5 oz copper, and 76.9 inch length the calculator reports 15.9 watts per zone. With 64 zones that's over 1 KW (and that is at half the current - 1A instead of 2A) with a peak current of 64 Amps.

That just doesn't seem practical to me.


TC

Noticed this on nophead's blog...

experimenting with CU and PLA

And since I don't have a blog yet (so I can't seem to post)...

In the case of plastic on metal, I think the plastic must be sticking to the fine features of the metal, based on something in some way related to the Coefficients of Linear Expansion for each material. (PS: Useful table, but I can't see PLA on it).

My guess (and I'm just musing here) is that as the plastic and metal cool, they cool at different rates, with the result being that the fine structure of the plastic "grips" the fine structure of the metal, holding it in place as it cools further. If the thermal coefficients are either very similar or very disparate, then they won't grip each other well. If they're close to some magic ratio, they grip. By changing the temp of one part (eg: the metal or the plastic) through blowing hot/cool air, one of these pieces expands slightly and the grip is lost.

Regarding removing the piece: After letting it cool, have you tried turning the bed on again and then occasionally attempting to knock the piece off? If my idea above is correct, the right temp should release the lock between the metal/plastic enough to remove the piece without too much force.

Another option to try might be to cool the metal even further (eg: a burst of cold from a can of air, etc), below the average room temp. If you've got a design where the layer the object is printed on is separate to the actual heated bed (eg: like some of nopheads designs) then this should be easy to test.

As for my musing above: Anyone with a heated bed got a microscope handy?

Just a thought....

Makerbot Industries have recently launched their own version of a heated bed, using an etched copper boardm (I have a Makerbot, but no financial interest!)

[store.makerbot.com]

it's just that some of the questions about heat tolerance of pcb's etc have been answered....

Could we scale up their solution for Mendel?

Ever hopeful,

Sven

nophead's experiments are very interesting indeed.

When I posted my questions about peak power / average power I wasn't thinking about the whole problem, and in particular, how to make multiple zones cost effective.

The point is that to minimize the amount of electronics needed then an approach where the the zones are multiplexed is likely to be necessary (i.e. time-slice and turn on one zone at a time and not all zones simultaneously). So, the peak power of a given zone has to be N-times higher (when N is the number of zones that are multiplexed).

- the instantaneous power of a zone is X
- the average power of a zone is X/N
- the power of all zones is N * X/N = X

For example, if we had an array of 64 zones arranged as an 8 row x 8 column matrix then 2 * 8 FETs would be needed to select one-zone at a time (1 FET for each row + 1 FET for each column). Regardless of the approach, I would think the objective would be to minimize the cost and some sort of zone multiplexing will likely be used.

A two layer PCB seems like it would be sufficient to construct a multi-zone serpentine-etch heater. On the top would be the serpentine etch with interconnections on one axis (i.e. rows). On the bottom would be the interconnection on the other axis (i.e. columns). On the periphery of the board would be the FETs and other heater-bed circuitry.

A bed surface would then have to be layered (sandwiched) on top of the PCB to spread the heat from the individual zones and provide some thermal mass (to average out the peak heating due to multiplexing of the zones). This layer would then be the heater bed on top of which parts are extruded.

It seems to me that this surface layer has to conduct heat well but be an electrical insulator. Would something like alumina ceramic work? Or are there better ideas?

TC

Folks, sorry for the delay in replying. I was fighting getting this years taxes done, Only US residents will understand .

Nopkead,, you are quite right. I got my extruder and heated bed temps mixed up. 110C, maybe 120C is the most that should be needed, and that 'feels' more realistic for a FRP-4 board. I have made a test layout, but have not yet etched the board. Here is a look.



TC,
You figured it out. I am planning on powering on row at a time with 8 high side MOSFETs, and PWM controlling each zone with 8 low side MOSFETs. This means that each zone gets a maximum of 1/8 of the time on, so it must power 8 times faster. On the other hand, with 8 zones per column, each zone gets only 1/8 of the total average power. But I think a total power consumption of 200W IS a bit high! Of a single 12V supply, that is around 16 amps, which is too much for the 400W computer power supplies I bought. So I think I will scale down to 100W or less to start with, and see if better insulation will make that practical.

By the way, McMasters has high temperature adhesives and putties that go all the way up to 4000C! I am thinking that silicone pot holders/trivets, glued onto the bottom side to handle the high temperature side, followed by 1/2 or thinker polyurethane foam will really help keep the heat in.




Testing will come later, after I run some wiring in my garage to handle all the new power tools I am getting to help me make robots, repraps, and under water ROVs.


EDIT:




I etched the board, and on visual inspection it looked good. But when I scanned it at 32000 by 32000, I could clearly see pits, voids, and breaks. And not surprisingly, the own meter shows an open circuit until past the last of the breaks. I do not know how to repair anything this fine. The design is easy, but the fab it artsy. Might be best for the actual heater board to have it prototyped by a fab house, so that the lines will be solid with no thin spots to over heat and burn out.

I also counted the lines. Only 39 lines, not the 79 I was expecting. Adding an inch for the connecting pieces at each end gives only 40 inches by 13 mills. So when the serpentine path is perfectly etched it will have only half the expected resistance. Perhaps I was doing the length calculation assuming the entire square inch would be copper, not less than half. Still, it LOOKS quite pretty!

Maybe nichrome wire will turn out to be easiest after all!


Mike

Edited 5 time(s). Last edit at 04/05/2010 01:54PM by rocket_scientist.

I'm personally not quite ready to start evaluating heated beds, but is the point of discussion avoidance of purchasing heating elements (power resistors, nichrome wire, etc?)

On that subject, graphite is generally an electrical conductor. It seems if you mounted a thin enough sheet onto an tough insulator (both electrical and thermal), it would be interesting to see if plastic would stick to it. The bed itself could be the heating device.

I think there are a few benefits of the serpentine etch approach:

- simple fabrication
- easy to obtain materials
- multi-zone temperature measurement and control
- low cost

The idea is to use the serpentine etch to function as a heating element and also for temperature measurement (resistance varies with temperature).

I'd guess that insulating the bottom side would help a fair amount. However, I'd be looking for the following attributes in the insulating layer(s):

- simple construction
- easy to obtain materials
- thin
- light weight

I'd sacrifice insulating value (to a point) to improve these attributes. A thin sheet of rigid fiberglass board might be a reasonable approach. The edges could easily be sealed with foil tape (assuming that its adhesives handle the high temperarture).

TC

I wonder if Nop's observations re a copper surface are applicable to a heated glass surface.....

Anyone had any ideas re a spray appliable high temp adhesive.

I have'nt found much with a working temp beyond 85 Deg C. Other than silicone oven glass adhesive. WHich given it's viscosity would be a pain to use.

---

Necessity hopefully becomes the absentee parent of successfully invented children.

I've seen adhesives that go much higher than needed but I'd have to dig again to find them. However, I wonder about the use of glass as a heater bed since it doesn't conduct heat very well.

TC

Your right it does'nt.

This could be advantageous though (rather counter intuitive, I know)

Consider we are attempting to heat/warm something that is adhered to it and conducts heat more poorly than the glass. ie it may kep the temp about right without wasting too much

The surface temp of non covered areas will be lower, Purely because glass is a relatively poor conductor so less power will be wasted uslessly heating air.

The big problem with glass is you cant use a point heat source or you just get hot spots, because as you say it is a poor conductor.

However using a serpentine element (I am thinking of aluminium here, a poorer electrical conductor, but good thermal conduction and much cheaper) The heat is evenly distributed.

I was playing with my outlaws heater plate on their serving trolley recently. This uses a glass plate wiht a serpentine heating elemnt bonded to the glass to keep food dishes hot/warm.

The funny thing is it uses surprisingly little electricity, does'nt over heat the room, but does the job.

Oh and more specificaly for us plate glass is flatter than most machinery can acheive, but is cheap.

---

Necessity hopefully becomes the absentee parent of successfully invented children.