Noise from NTC connection (temperature measurement) 4

[schnell]

Hello,

Please could you help me in making a noise-free (as far as possible) connection to a NTC Thermistor?

We are continuously detecting temperature by using the ADC of a microcontroller to read the voltage across a divider ……one resistor of which is the NTC Thermistor.



-before reading the voltage with the ADC , we use an op-amp (LMC6035) to buffer the thermistor signal, and also to expand the signal range, so that we read the temperature with a higher resolution.

CIRCUIT DIAGRAM:

http://i56.tinypic.com/125qqs9.jpg

Anyway, the thermistor is not on the PCB that contains the opamp and the microcontroller, but it is about 10cms away from this pcb, -and in fact the thermistor is actually mounted on a water pipe.

-so the thermistor connects to the pcb via a 10cm pair of wires, which are not twisted together, but do run pretty closely to each other.

-these wires are connected to the pcb by a push-in connector (i.e. the typical connector with shrouded-housing soldered to the pcb, and the crimp-housing connected to the thermistor wires).

Anyway, unfortunately we could not get the op-amp to be located on the part of the pcb which is near the thermistor wire connector.
-in fact, the opamp is about 10cms away from this connector.

-therefore we had to run a 10cm pcb track from the pcb thermistor connector to the opamp.

…I am worried that such long runs of wire, and also of pcb track, will cause noise to be picked up, and corrupt the thermistor signal. ?

Do you think that this is possible?

Anyway, I have brought the thermistor signal to the opamp via a pair of tracks which run alongside each other…from the pcb thermistor connector..
……I needn’t have done this since one side of the thermistor is connected to Vcc (5V), and there is a Vcc node available at the op-amp (Vcc actually powers the op-amp and the microcontroller)

…..however, I just assumed that running the thermistor signal as a pair of pcb tracks , side-by-side, would mean less noise corruption………do you know if this is true?

Do you know of any other measures that I can take to reduce noise corrupting this temperature measurement.?

The thermistor is around 10K at 25C.

 

[Skogsgurra]

Considering the rather slow temperature variation, you could filter the signal heavily. I am right now using a thermistor with the same spec to measure temperatures using a cirka ten m long unshielded cable and have no problems at all with noise. But then again, I am using a 1 uF capacitor instead of 10 nF.

An NTC thermistor is a rather high-impedance device - especially when it gets cold - and noise pick up is inevitable.

I think that can simplify your circuitry a lot, BTW.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[schnell]

hi
its the temperature of water going to a shower, so it needs to read at a high rate so as to avoid scalding.

regarding filtering...........

would it be best to put bypassing caps across the thermistor tracks? (i.e. across the thermistor)

-or is it best to put bypassing caps from each thermistor track to the pcb ground?

the thermistor itself is the high-side divider resistor, so is not connected to pcb ground.

is it right to route the pcb tracks as a pair, as i mentioned above?

 

[Skogsgurra]

The time constant will not be more than 10 milliseconds at all temperatures and you will never need faster response than around 100 ms to control water temperature in a shower application. So filtering is your friend.

Yes, of course, making a loop of the conductors is always a bad idea.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[itsmoked]

You should twist those leads. It borders on trivial effort and helps a lot. You should be running a ground plane below those two traces too.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[potteryshard]

Schnell...

You stated: -and in fact the thermistor is actually mounted on a water pipe.

I would be concerned that the thermal mass of the pipe itself, especially if that pipe as is usual in shower installations is of copper, is going to add significant delay in sensing the water temperature.

I would tend to think that the thermistor would work best if placed in a well which directly contacts the flowing water mix and is itself thermally isolated from the piping material. I would think that the thermal mass of the thermistor well itself would provide first-stage filtering of the water temperature. Would interposing the piping material between the water and the thermistor create too much time lag?

 

[schnell]

Hi Potteryshard,

sorry i was cutting a long story short....the thermistor is actually mounted sticking into the flow of water.

.....as you imply, we could not cope with the poor signal that we would obtain if the thermistor was mounted to the outside of the pipe.

in fact, we need this thermistor signal to be pretty unfiltered......this is because our showers must allow the customer to virtually instantaneously halve or double the water flow rate.....without the water temoerature wandering too far from the demanded value..............

...since we do not know the individual temperatures of the incoming hot and cold water flows, we just have to initially best guess the flow settings on the individual hot and cold valves (stepper valves).....and then let our control algorithm trim the temperature to the demanded value....

(....dare i say we sometimes get temporary over-temperature during these flow change times............and are currently trying to improve the situation so that we keep the water temperature pretty much the same through this sudden flow-change event.)

 

[IRstuff]

What's the response time of the mechanical portions of the valve control? Your valving response time is already a low-pass filter, so noise faster than the response time of the valving won't affect the temperature much at all.

What limits are there on the maximum temperature of the water in the hot supply? Bear in mind that 154°F water will "instantaneously" cause 1st degree burns, but 149°F water will take about 1 second to do the same.

And what does your "instantaneously" mean? a nanosecond? one second? Seems to me that 1/2 to 1 second is pretty close to "istantaneous" for most consumer applications.

Why wouldn't you measure inlet temperature as well, at least, the hot side? Seems to me that as a product liabiliy risk mitigation, you should measure the hot side temperature to ensure that your product is not being used outside of its safe-operating-curve.

Are you measuring flow rate as well? Seems to me that just using mechanical settings to vary flow without knowing the actual inlet pressures or flow rates is somewhat speculative.

My personal experience is that flow rate greatly affects the apparent, and possibly, the actual temperature mixing. This is possibly due to the difference in line pressures between hot and cold.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[potteryshard]

As I understand the way the old mechanical safety mixing valves work, there is first a pressure balancing spool between the hot and cold inlets, and then a thermostatic control system.

The pressure balancing valve is the customer protection against sudden system upsets; usually a sudden drop in pressure due to toilet flushing or dishwasher valve operation for example. The thermostatic action merely allows the user to select the desired mix; it is the pressure balance functionality which would seem to preclude the temperature spikes in the final mix.

To accomplish this electronically via a single set of flow control valves would seem to require very fast control response. It might take pressure sensing on each input upstream of the mixing valves as an additional inputs to the control algorithms, or possibly looking at the temperature rate-of-change rather than temperature for priority control inputs.

Or, a mechanical pressure balancing valve upstream of the electronic may be more cost effective.

 

[schnell]

we take scalding as 45 degree c....you cant shower above 45c


we have pots for flow rate and max water temp that we pre-set for the customer. max temp of hot is indicated to micro with pot.......the highest it could be is 55c

 

[VEBill]

If you're designing the entire system, then perhaps the plumbing side of the system could include enough post-mixing water volume to slow things down a bit.

 

[DPolak]

Hello.

C23 and C25 - can be easy 10 or 100 time larger.
Thermistor response time is in 100mseconds range, no sense to use low pass filtering with KHz range (10nF, and 6kohms).

Second problem - signal should be measured by instrumentation amp (High common mode rejection), with 2 wires.
This schematic looks like single ended method (U4A),
Yes, 2 wires go to thermistor, but only one wire is "hot".
60Hz (on pin 3) will be amplified by U4A.

Advice: if you too lazy to redesign, just move the filter below 60 Hz (say 10Hz).

thx

 

[schnell]

DPolak:
Thankyou...i have in fact now suggested that we change to using an instrumentation amplifier.

However, my boss and colleagues feel that i am being silly.

They point out that the schematic that we have above (first post) works absolutely fine.

Also, this application note from Texas seems to support our non-instrumentation amplifier method. (fig 1 first page is pretty much identical to our own set-up)

http://focus.ti.com/lit/an/slyt156/slyt156.pdf