Constant heat hot-wire. How to accurate measure resistance?

[moreil]

temperature resistance feedback

Full documentation to date at
www.homecnc.org/index.php/Constant_Heat_hot_wire_cutter

I'm trying to design a circuit to maintain a wire at a constant heat (for a hot-wire foam cutter). The wire I'm using is 0.5mm stainless steel wire with a temp coefficient of resistance of around 0.003 ohms/degree.

So I was planning to use something like the circuit below as a circuit to measure the voltage and current across the wire, then use the AVR to caculate the resistance and modulate the current on the wire to suit.

My problem is that the circuit above has a temperature resolution of about 10 degrees celcius best case. This isn't ideal.. The kerf that the wire produces is fairly sensitive to temperature and so maintaining the temp to +/- 10 degrees is pretty rough.

I can obviously use high precision ADCs (the ATtiny13 is 10-bit) but that involves switching microcontrollers (with knock-on effects on prototyping and development environment support). I can use an external ADC, which drives significant complexity and cost.

Question is: Is there some way of amplifying a selectable range so I can get better precision on the current measurement?? i.e. The circuit currently measures current from 0-5A. Is there some simple way of changing it to measure (say) 2.75 - 3.25 A?

Maybe use one of the spare outputs on the AVR as a 1-bit DAC to move the voltage reference for IC2A to something above ground?[/img][/url]

 

[moreil]

measuring resistance of a wire

Updated circuit. This one has a PWM output on the AVR into a 2nd order low-pass filter to form a crude DAC, which is then used to implement a 'coarse' and 'fine' current measurements.

**broken link removed**

If the 'PWM' line is feeding a 5V, 1Mhz PWM signal into the low-pass filter consisting of R8,R9,C4, and C5, will the ripple at R10 be less than 500uV ?

( full detail updated at **broken link removed**)[/url]

 

[moreil]

pwm hot wire

Small modification. I've switched to the LMP2014 op-amp as I've worked out I need to detect a 20uV signal, which is tough to do on a 2mV offset op-amp.

Now I need to worry about getting the noise low enough...

 

[Siniša]

calculating hot wire temperature and foam cutter

You should use Kelvin 4 wire method, separate connections for feedback of voltage on the wire. Contact resistance will change much more than wire tempco as you apply pressure. Also I would go with premade diff.amp because poor resistance tolerance will screw your CMRR. You will have fun also of getting this thing to work stable, as system will change thermal inertia a lot depending on material , feeding speed ... or wire itself. One set of PID parameters that will work with wire alone will not work when you apply material.

 

[moreil]

avr measure resistance

You should use Kelvin 4 wire method

That sounds nice. What would that sort of circuit look like?? Is that akin to a bridge? One of the complication is that the wire is run at relatively high voltage. Would that make an impact?

editNever mind. I just found it with google. Isn't that exactly what I've done? Measuring the current through the wire, the voltage drop across the wire and then calculating the ratio to accurately measure the resistance?

Also I would go with premade diff.amp because poor resistance tolerance will screw your CMRR

I'm assuming that a diff.amp is something different than the op-amp I'm using above? Is this akin to the circuits you get at the front end of instrument amplifiers like this?

I'm sorry, I don't understand what 'poor resistance tolerance' means. The CMRR shouldn't be too important as the values in question change fairly slowly (10s of degrees per second rather than 100s).

You will have fun also of getting this thing to work stable, as system will change thermal inertia a lot

That is indeed true! On the bright side though, the wire never actually touches the foam (it's pure radiant heat transfer) and as it's a CNC wire cutter, the cutting speed can be held pretty constant which reduces that impact. I'm fairly hopeful that it won't be too difficult to stablise.

 

[Siniša]

4 wire resistance measurement op amp

No, you didn't do 4 wire method because you attach your electronics with only two connections on J1. That way you are measuring all resistances between your electronics and wire including contact resistance of J1. You need to have another two contact connector that will carry only feedback voltage from the wire.
CMRR refers to DC signal as well AC, so it is very important that you have this paraeter high as well. Because your measurement is DC, you will have to watch out for things like thermoelectric voltages of contacts, temperature offset drift of opamps, both voltage and current, and probably bunch of other stuff. I would suggest using wire that has as high tempco as possible, because it makes it so much easier to measure. Ni would be good material I presume. You should also work backwards, set your desired tolerances, than from that you can calculate what is your error budget and see if you fall within those parameters. My gut tells me that you will quickly run into problems. Running PWM and trying to do measurement at such low signal will quickly turn into NASA project. You did not specify what is ppm/K of that wire? Work in steps, don't try to come up with solution right away, because you will likely run into corner otherwise.

 

[moreil]

avr current measure

No, you didn't do 4 wire method because you attach your electronics with only two connections on J1.

Ahh, I see what you mean. I've ignored the cables from J1 to the wire as the resistance is negligable. (~ 0.1 ohms versus 15 ohms for the hot wire).

I would suggest using wire that has as high tempco as possible, because it makes it so much easier to measure. Ni would be good material I presume.

I'm constained to using something fairly strong to survive the tension on the wire. I'm currently using 0.5mm stainless steel wire, which is a bit over 0.003 ohms per degree.

You should also work backwards, set your desired tolerances, than from that you can calculate what is your error budget and see if you fall within those parameters.

I've shown my calculations at **broken link removed**

The key number is that the voltage over the current sensing resistor will vary by about 40uV per degree.

I've never had to do design around these sort of numbers and I'm worried about the noise levels. As you can see, I've put 20Khz low pass filters on both the voltage and current measurements and I'm now using precision op-amps (~ 0.8uV offset, ~ 130dB CMRR) but I suspect there's a bunch of things I should be worried about that I'm missing.

The critical path is obviously around I2C which comparing the output of the low-pass filtered PWM from the CPU with the current sense voltage and then amplifying by 200x.

I'm worried that the 3 stage low pass filter won't have low enough ripple, and I'm worried that there'll be noise ingress that that I haven't acounted for.

My gut tells me that you will quickly run into problems. Running PWM and trying to do measurement at such low signal will quickly turn into NASA project. You did not specify what is ppm/K of that wire? Work in steps, don't try to come up with solution right away, because you will likely run into corner otherwise.

More details of my working out are on the wiki above. I'm planning to PWM fairly slowly (the thermal inertia of the wire is relatively high, so a 50Hz or even 25Hz PWM should be ok).

I was intending to run a fairly simple control algorithm: Turn on current, and sample at 5Khz until the wire resistance reaches a given value, then turn off the current for a fixed duration. This is actually variable frequency modulation rather than pulse width of course.

Current circuit is:

**broken link removed**

 

[Siniša]

measurement of a wire resistance

Well... If you want to measure 15Ω resistor with 1.5mΩ resolution or better you need A/D converters with at least 14 bit resolution. That is for both voltage and current. 0.02%/ºC really isn't much of tempco to work with, and much more difficult as this is done as DC measurement. That's more reason to abandon those resistors and regular opamp for measuring voltage and go for instrumentation amp. Or switch load from high side, low side of load connect to 0 volts, and sense current as negative voltage with inverting amp on sense resistor connected bellow your 0V. Setting reference higher for current will solve only half of the problem if you could get it to work.