Hello.

I'm trying to build a hakko 907 (24V 50W) controller, but I don't have much experience with op amps, and I have concerns about the circuit I came up with. Here it is: http://i.imgur.com/ZalB8RG.png

It consists of Wheatstone bridge and a single rail differential amplifier with a gain of 3.6. The 220 Ohm resistor was chosen with assumption that the resistance won't go over 220 Ohms, based on some of the charts I found online. The gain was chosen so that the output voltage don't go near the positive rail, and 3.6V can be divided into ~720 steps with 10 bit ADC, which is more than enough.

The choice of op amps in the local shops is pretty scarce, and I wasn't able to find any rail to rail op amps, just the common ones like UA741, LM324 and LM358. Could probably order them from china, but then I'd have to wait for a few months, so I'm keeping that as a last resort, in case I can't use the available ones.

So, my concerns now:

I wasn't able to find a tutorial written by a "professional" that uses a differential amplifier in single rail mode without voltage splitter circuit, and yet I found some people use it like that (example). Is this a bad way to go, and if yes, why?

Also, I am not really good at reading op amp datasheets. If I understood correctly, the output voltage quality will start to deteriorate with lower difference at the input, as it approaches the negative rail. Where in the datasheet should I look to find out at what voltage should I expect that to happen? If this is correct of course.

Thanks in advance ,
Miloš

 

Hi Milosh,
what are the specs for the temperature sensor? Total temperature range?

Here is the reason I ask: This shows output voltage of the circuit you posted as a function of Rtd using an idealized opamp.



Using an LM358 would not work because you are exceeding its common-mode input. There is a fix. Also, it has a big offset, so you would need a balancing trimpot to trim out the initial offset. Also, the voltage follower on the output just makes the circuit worse; get rid of it...

ps: some reading for you...

 

Hello Mike, thanks for your reply.

These are the specs measured by someone else on this handle, until my thermometer arrives. And for temperature range, I intended the range to be 180-420 degree C, but that's probably too ambitious for this cheapo heh.

Just read about that input common mode thingie, so, if I put two more resistors, one above PTC, and one above 220R, and bring the input voltages around the "middle", it should be ok?

EDIT: I missed your link about linearity on the first read. I will read it after studying, but I was planing to correct the y=k*x+n formula digitally into something more precise after my thermometer arrives.

 

...These are the specs measured by someone else on this handle, until my thermometer arrives. And for temperature range, I intended the range to be 180-420 degree C, but that's probably too ambitious for this cheapo

Here is a straight-line fit of your data:



I changed the circuit so that a LM358 can handle it. The common-mode inputs are now within the allowed range. The LM358 is good at pulling low, and can get quite close to 0V but not all the way, so I biased V(out) up a few mV at 180 degC. The LM358 can only pull to within 1.3V of the Vcc rail, so I set the gain so that the output is ~3.6V at 425 degC.

R5 primarily sets the gain. R3 primarily sets the offset, but the two interact a bit. If you need to trim out opamp offset, make R3 a 20K fixed in series with a 10K trimpot.

Note how I use LTSpice's Behavioral voltage source to create an artificial voltage which is a function of parameter degC. Note how I use LTSpice's resistor-as-a-function-of-node-voltage to create a resistor that obeys the curve fit above.



Make sure that your PTC sensor will stand 425 degC (most wont)! That is getting to the temperature range where you might have to use a thermo-couple.