Pipe heater, induction or element?!

THE_RB

Joined Feb 11, 2008
5,438
... and find that the LM34's response time is very fast, almost instantaneous. I don't foresee a problem in your application.
...
Agreed, it won't take 3 seconds to respond to the small change in temperature required for setpoint control.

...
It is also my opinion that, despite the preference of the scientific community for Centigrade measurements, the LM34 is inherently more precise in absolute terms than the LM35, simply because one degree C is so much larger than one degree F.
...
The LM34 might be more precise when used as a thermometer, but in closed loop apps the reading should vary between two consecutive readings only which is an infinitely small threshold. The granularity of the sensor (ie; 'C or 'F) will not affect that and both LM35 and LM34 will regulate the temperature identically around that very small threshold change.
 

#12

Joined Nov 30, 2010
18,224
I guess I should clarify. The 15 second number is the time from a cold start where the LM35 has to change by several degrees to get to the temperature of the subject. You know, 5 time constants at 63% of the difference per 3 seconds? (or was that 67%? darn these senior moments!) Anyway, once it gets stabilized, it can track really close, depending on how fast the subject's temperature is changing. When the intent is to get stable at a certain temperature, the LM35 gets so close it's hard to tell the difference between the reading and perfect.
 

wayneh

Joined Sep 9, 2010
17,496
The time response of the electronics will be trivial compared to whatever is in between the bulk milk and the LM35. I mean, you can't immerse the LM35 in milk and therefore you'll need some arrangement like a thermal well in the flow stream. That thermal well will have a boundary layer separating it from the bulk. The well needs to be some distance (often specified in number of pipe diameters) away from the heat exchanger, to allow for bulk mixing. Hopefully the flow regime is turbulent rather than laminar, which will cause a lot of headaches. The flow rate through a pipe is not uniform at all positions, it depends on the velocity profile across the pipe.

I really think that, unless you have experience designing such things, that you'd be better off with the standard tank. Even a modest amount of residence time - a fat spot in the pipe - solves so many problems you otherwise have. If you DO have experience with inline heating (as I do), you'd likely decide not to do it. There's a reason that a big, insulated hot tank is the de facto standard for delivering a heated liquid.
 

tracecom

Joined Apr 16, 2010
3,944
The LM34 might be more precise when used as a thermometer, but in closed loop apps the reading should vary between two consecutive readings only which is an infinitely small threshold. The granularity of the sensor (ie; 'C or 'F) will not affect that and both LM35 and LM34 will regulate the temperature identically around that very small threshold change.
Maybe I don't understand, but here is my rationale.

Both the LM34 and LM35 output voltage that is directly proportional to the temperature in .1V steps. Then, those steps are converted A/D by the micro-controller. Because one degree C is larger than one degree F, a .1V step from the LM35 represents a larger absolute temperature change than a .1V step from the LM34. Thus, finer resolution is possible with an LM34. Or am I missing something?
 

THE_RB

Joined Feb 11, 2008
5,438
Hi Tracecom, yes you are correct about generally reading the temperature.

But in a standard closed loop setpoint control the heater will switch based on one setpoint which is between two readings.

So let's say it's set to 40'C;
<=39'C = heater ON
>=40'C = heater OFF

So the sensor only needs to change a VERY small amount to toggle between 39 and 40, the actual temp at the core of the sensor might go from 39.95 to 40.05 degrees when operating.

The total voltage change only needs to be enough to trigger the ADC to toggle between one adc value and the next;
<=ADC 71 = heater ON
>=ADC 72 = heater OFF

If the sensor and ADC were perfect models the threshold would be almost infinitely small, so that only an incredibly tiny temp change is needed to toggle the heater. In reality the sensor and ADC are not perfect and have some very small hysteresis etc, but allowing for normal noise levels on the signal etc the difference between a LM35 and LM34 in setpoint control is practically nothing, smaller than noise levels, much smaller than the base resolution of the sensor and much much smaller than the thermal lag etc of the entire system mass which is the main issue in regulation.
 

Thread Starter

geoffers

Joined Oct 25, 2010
475
Thanks for the replies, I think I'm going to stick with the tank! It seems fairly clear now that it could be a bit risky using a pipe heater. I think I might look at altering the heater I use in the tank though, at the moment its a fish tank heater but by the time the heater has decided to start, the water may all ready have got bit cold. I think a sensor on the input and a water heater that starts when the milk starts to flow maybe the way forward? It could then stop when or if the water gets 0.5-1 degree warmer than needed?

Should give a good safety margin and be a improvement on what I've got at the moment! If I use a off the shelf immersion heater is a ssr the best way to switch it or would I be better using a ordinary relay? I've only used normal relays before, are ssr's worth the extra £'s?

Cheers Geoff

Ps #12, it would be a sort of bllaaawwwt noise rarther than a squeak or moo! I don't want to hear it:eek:
 

tracecom

Joined Apr 16, 2010
3,944
μ
Hi Tracecom, yes you are correct about generally reading the temperature.

But in a standard closed loop setpoint control the heater will switch based on one setpoint which is between two readings.

So let's say it's set to 40'C;
<=39'C = heater ON
>=40'C = heater OFF

So the sensor only needs to change a VERY small amount to toggle between 39 and 40, the actual temp at the core of the sensor might go from 39.95 to 40.05 degrees when operating.

The total voltage change only needs to be enough to trigger the ADC to toggle between one adc value and the next;
<=ADC 71 = heater ON
>=ADC 72 = heater OFF

If the sensor and ADC were perfect models the threshold would be almost infinitely small, so that only an incredibly tiny temp change is needed to toggle the heater. In reality the sensor and ADC are not perfect and have some very small hysteresis etc, but allowing for normal noise levels on the signal etc the difference between a LM35 and LM34 in setpoint control is practically nothing, smaller than noise levels, much smaller than the base resolution of the sensor and much much smaller than the thermal lag etc of the entire system mass which is the main issue in regulation.
I am at a public computer and will study your reply in detail later.

What I normally do with a μC thermostat is set the on point and the off point offset from each other by .2 degrees (to avoid switching "chatter.") In other words, if I want to maintain 101 degrees F (body temp of a cow, I think,) I set the heat on point at 100.9 and the off point at 101.1. In order to accomplish the same thing with C, i.e., maintain 38.33 degrees, I would have to set the on point at 38.2 degrees and the off point at 38.4 degrees. Thus, the regulation seems closer with F than C because, in absolute terms, the range from 38.2 C to 38.4 C is slightly larger than the range from 100.9 F to 101.1 F. But, I tend to get fixated on certain things and can't see it any other way. I will think it through as I drive home later today.

Anyway, thanks for your input. I have respect for your expertise and accomplishments.
 

wayneh

Joined Sep 9, 2010
17,496
I also often wish they had made the LM35 to give 20mV per °C instead of 10mV, making it similar in precision to the LM34. The downside would be the less intuitive conversion to temperature from voltage, when you're using a simple voltmeter as a thermometer. I'm sure that's one reason they went with 10mV instead of 20.

There's another reason, too, though. You can cover a wider range of temperature with a given supply voltage if you keep the mV per degree smaller. The higher precision of the LM34 limits the temperature range it can operate in. With a 9V supply, for instance, the LM35 is good to 90°C (194°F) while the LM34 is good to only 90°F (32.2°C).
 

THE_RB

Joined Feb 11, 2008
5,438
...
What I normally do with a ?C thermostat is set the on point and the off point offset from each other by .2 degrees (to avoid switching "chatter.") In other words, if I want to maintain 101 degrees F (body temp of a cow, I think,) I set the heat on point at 100.9 and the off point at 101.1. In order to accomplish the same thing with C, i.e., maintain 38.33 degrees, I would have to set the on point at 38.2 degrees and the off point at 38.4 degrees.
...
Interesting! I generally don't use hysteresis in larger mass temp controllers as the thermal lag of the system means it takes a few seconds for heat and cool cycling, so the thermal delay causes plenty of overshoot by itself and switching is quite clean. even though the setpoint itself is very small (sub 1 ADC count) the ADC may go though 1 ADC count or more of temperature overshoot during cycling.

I also often wish they had made the LM35 to give 20mV per °C instead of 10mV, making it similar in precision to the LM34.
...
Agreed! That would definietly be an improvement.

It's not so bad using LM35 though, for temperature reading apps I normally encourage a bit of noise on the signal line (ie run less filtering) and average a lot of readings. Averaging readings over one full mains cycle (10mS or 8.33mS) both negates the mains freq noise from the signal AND increases resolution as the noise added onto the signal gives a spread of readings (better for averaging).

Another old technique is to deliberately modulate the analogue signal by adding some mV of AC to it, then average lots of readings. That's actually easier than it sounds; you just need a resistor from a PIC digital output pin to the ADC input, and modulate the pin in the interrupt etc.

There's another reason, too, though. You can cover a wider range of temperature with a given supply voltage if you keep the mV per degree smaller. The higher precision of the LM34 limits the temperature range it can operate in. With a 9V supply, for instance, the LM35 is good to 90°C (194°F) while the LM34 is good to only 90°F (32.2°C).
I think you missed a decimal place? An LM35 at 90'C is outputting 90 * 0.010v or 0.9v. :)
 
Last edited:

Thread Starter

geoffers

Joined Oct 25, 2010
475
You're not wrong! The tangents are quite interesting, made me read data sheets anyway!

In my last post I did ask if anyone had opinion on the best way to switch a immersion heater? I will probably use a pic to control it, is there a reason to use a ssr over a normal coil relay? The ssr is a bit more expensive, a previous project was a bulk milk tank washer using a pic, I didn't know about flyback diodes, do now! Took me a while to figure out why the pic kept resetting.

Cheers Geoff
 

tracecom

Joined Apr 16, 2010
3,944
Assuming your immersion heater is powered by mains AC, here are the reasons an SSR might be better than a relay. An SSR should last longer than a relay, should consume less power than a relay, and should make no audible noise when switching. It's possible to build your own SSR using an optoisolator, a triac, and a couple of resistors. Because your load is completely resistive (I think,) you don't have to be concerned with zero crossing switching.
 

wayneh

Joined Sep 9, 2010
17,496
I think you missed a decimal place? An LM35 at 90'C is outputting 90 * 0.010v or 0.9v. :)
Oops, yes indeed. That means that even at the max temp of 150°C, it only outputs 1.5V. They could easily double that voltage and still have room for all but the lowest supply voltages.
 

Thread Starter

geoffers

Joined Oct 25, 2010
475
Thanks guys,

I've been on holiday for a week hence lack of interest from me, a ssr it is then, I've been thinking about it and unless I programme in a fair bit of damping a coil relay could soon get worn out?

While on holiday I was convinced of the wisdom of a water tank, one place we stayed had a power shower, you can guess it, I got a bit of a scalding before it realized the water was to hot!

Cheers Geoff
 

MaxHeadRoom

Joined Jul 18, 2013
28,617
If you roll your own, there is a Triac output opto, MOC3020/23 for one, to switch the main Triac.
Elements are often 'burst switched' to avoid switching noise, zero switch on - zero switch off, or a R/C snubber is placed across the switching triac ~39Ω in series with 0.01μf.
Max.
 
Last edited:
Top