Page 3 of 3 FirstFirst 123
Results 21 to 25 of 25

Thread: Data logger project

  1. #21
    Join Date
    Mar 2006
    Posts
    2,478

    Default

    Are those LM335CZ or whatever the number is? Small temperature sensors based on temperature coeffecient, but redesigned to exploit that, and calibrated to 10mV per degree C? Anyway, I like them. But it's been too long since I used one, I even forgot the exact number.

    If you need fast reaction time and very consistent swappable sensors for pre-calibrated sicruits, PT100 is nice. Bare elements about 15 quid, likely a lot less in quantity. They can stand the condensation that build up on coolers better than most stuff, but they're originally intended for heat. I like them for fast response and very low thermal mass, and very efficient thermal coupling.

    Once you have bored the guests, don't forget to insert little thermistors into the holes.

  2. #22
    Join Date
    Jun 2010
    Location
    Australia
    Posts
    3,734

    Default

    Quote Originally Posted by The_Doctor View Post
    Are those LM335CZ or whatever the number is? Small temperature sensors based on temperature coeffecient, but redesigned to exploit that, and calibrated to 10mV per degree C? Anyway, I like them. But it's been too long since I used one, I even forgot the exact number.

    If you need fast reaction time and very consistent swappable sensors for pre-calibrated sicruits, PT100 is nice. Bare elements about 15 quid, likely a lot less in quantity. They can stand the condensation that build up on coolers better than most stuff, but they're originally intended for heat. I like them for fast response and very low thermal mass, and very efficient thermal coupling.

    Once you have bored the guests, don't forget to insert little thermistors into the holes.
    Nope. Much smarter, they are Dallas single-wire DS18B20 digital programmable sensors with 12 bit resolution (0.06c). Slow, but fine for a logger logging at 5, 10, 30 or 60 seconds. Benefit? Accurate and no calibration.

    I have used LM35 before, 10mV/C requires an amplifier so for fast analog applications I still prefer NTC resistors. I do a little trick with the Atmel MCU's, I set the ADC ref to 1V1 and operate the NTC divider from 5V0 with no amplifier required (simple, low cost, no drift). Provides "more gain" giving 32 values per degree celsius which is usable with PID (given correct cycle time, derivative can operate correctly).
    This space for rent.

  3. #23
    Join Date
    Mar 2006
    Posts
    2,478

    Default

    Ok. I just thought the Dallas things would be bigger, not sure why...

    In the last post I managed to miss out that it was the tempco of voltage references that led to the LM35. I like weird inversions or transpositions of an idea like that. (A similar one occurred to me just two days ago: that 'noiseprint' based noise reduction seems to be based on an inversion of the old 'Vocoder'. Maybe not exactly, but just pondering stuff like that is totally interesting at times.) And things like your trick of resetting a different voltage ref in a ADC, I like that one too. (And have been known to use asymmetrical supply rails in op-amps for getting zero-setting and wide voltage range to increase resolution).

    One thing: I only have a vague memory, but isn't temperature control one of those things that will work just as well if you omit the I or the D (I forget which) from PID? At least, I remember that there was something I was told worked just as well with PD when I asked about PID sometime back on usenet.

  4. #24
    Join Date
    Jun 2010
    Location
    Australia
    Posts
    3,734

    Default

    It depends what your trying to achieve.

    P gets you in the ballpark, although alone a steady state error remains.
    I removes said error.
    D is the prediction of what is going to happen based on rate of change. It reduces overshoot and undershoot.

    On some systems P is enough, and for fan cooling with slow rates of change is passable. I want to regulate within +/-0.3c so I is required. Without D I have too much under/overshoot.

    Some systems work fine with PI. That's all part of the fun. Model the system with just P and determine what is best to achieve your DEFINED requirement.
    This space for rent.

  5. #25
    Join Date
    Mar 2006
    Posts
    2,478

    Default

    Quote Originally Posted by dnar View Post
    Model the system with just P and determine what is best to achieve your DEFINED requirement.
    Well, that's what I do.. Even when not for a control by closed loop servo, it has a use, like the thermopile meter provides the P in the circuit, the I in the slow response of the pile, but the rate of change combined with a differentiator gets the D (which leaves me wondering how interchangeable 'derivative' and 'differentiator' actually are). The result of boosting the D signal and mixing with the main signal is a fast response (it twicthes strongly if you pass a hand fast across the input beam), in a meter that otherwise takes at least 40 seconds to settle. Mine settles in a few milliseconds and does not deviate more than a very few percent when transiting from D signal to main. Which is good because not many laser power meters are rated for better than 5% anyway. With coaxing, I think it might do 2% or better.


    Btw, what's the steady state error? I thought the main purpose of I was to integrate, i.e. low pass filter, to remove noise on some defined time scale. Noise isn't steady state, as such.. One characteristic of my meter circuit is that some noise is inherent to the acceleration, but it's a small price to pay as it reduces on a rapidly declining exponential curve after any step change in input. This is an open loop thing though, nothing's being controlled, only observed.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •