View attachment 46790Hi.
I was wondering if someone could help me sort out a minor problem I'm having with a temperature circuit that I've made that is part of a larger project.

It's based on a 7107 A/D converter and an LM35 temperature sensor (no programming). It's running large LED digit displays.

My problem is that the temperature jumps every second say from 19 to 20 to 19 to 20, etc. I'd like it to be more steady - only update every minute or say. Is this an easy fix? Details of the data sheet below as well as my circuit:


http://www.electro-tech-online.com/custompdfs/2012/07/fn3082.pdf

 

Try something like figure 14 in the datasheet for lm35 And also add two decoupling caps on the LM35 like they have done in figure 4. Try a 100nF and something in range of 4.7uF to 10uF. Coupled in parallel. It will cost you little to try this
http://www.ti.com/lit/ds/symlink/lm35.pdf

 

I can try the 100nF and range of 4.7 to 10uF - no problems - I have these on hand. However I'm not sure about what to do with figure 14. Which parts are you suggesting?

 

I can try the 100nF and range of 4.7 to 10uF - no problems - I have these on hand. However I'm not sure about what to do with figure 14. Which parts are you suggesting?

Ah what I meant was the the R-C Damper used in figure 4. That is used in figure 14 also. But it is perhaps more clear in figure 4

 

Also there is always LSD uncertainty which can cause such apparent bobble. (bobble is a technical term - see http://books.google.com/books?id=mwkApRCi-qAC&pg=PA528&lpg=PA528&dq=last+digit+bobble&source=bl&ots=KvSMARX7vp&sig=HDea9mJkHEvrrbewYn_z7CPSoFY&hl=en&sa=X&ei=_e5mUPuxLYSm9ASfrIHICw&ved=0CCQQ6AEwAQ#v=onepage&q=last%20digit%20bobble&f=false)

 

So is it simply a matter of connecting a 0.1uF ceramic across Vs & GND, and a 75Ω resistor and 1uF ceramic series across Vout and GND?

 

measure the voltage at the output of the LM35 and its 0v supply directly at the sensor ,

it should be giving 10mV per degree C, so if its 20 degrees C it will be giving out 200mV,

see if its fluctuating , looks like the supply is not correct, as you have the 0v of the LM35 not connected direct to the supply?

 

So is it simply a matter of connecting a 0.1uF ceramic across Vs & GND, and a 75Ω resistor and 1uF ceramic series across Vout and GND?

It may help. But it could be a other causes also.

 

measure the voltage at the output of the LM35 and its 0v supply directly at the sensor. It should be giving 10mV per degree C, so if its 20 degrees C it will be giving out 200mV,

see if its fluctuating , looks like the supply is not correct, as you have the 0v of the LM35 not connected direct to the supply?

I just measured between Vout to GND of the sensor. It measured 186mV (and the temperature is 18 celcius) and only fluctuating by only say 0.5mV.

 

try to power the LM35 directly across the 5v supply and see what happens then?

Or remove the LM35, and put a variable supply at the input of the 7107 say 0 to 1v and see how it reads then.

 

try to power the LM35 directly across the 5v supply and see what happens then?

Or remove the LM35, and put a variable supply at the input of the 7107 say 0 to 1v and see how it reads then.

I connected Vs of the LM35 directly to 5V - no difference
I connected GND to pin 30 - dropped temperature from 19 to 14.

At the moment the temperature is steady at 19. No fluctuation whatsoever. I think it's when it goes say to 19.6 it jumps maybe to 20 and back and forth.

Maybe it's like Wmodavis said above - it's "bobble"???

 

Uncertainty in one count is to be expected, for example when the temperature is 19.5C.

Since you are already using extra components to drive the display, the only solution is to latch the display data at a slower rate (once every 60 seconds).

Since this complicates the hardware, it would be simpler to move to a MCU design.

 

Uncertainty in one count is to be expected, for example when the temperature is 19.5C.

Since you are already using extra components to drive the display, the only solution is to latch the display data at a slower rate (once every 60 seconds).

Since this complicates the hardware, it would be simpler to move to a MCU design.

You are right, but I've steered away from microcontrollers as it's a bit out of league.

 

You are right, but I've steered away from microcontrollers as it's a bit out of league.

Out of league? Can you elaborate on that. Your scoreboard project would have much less complex by using microcontrollers. And even for a beginner like you. You could have saved a lot in development time and component cost also, as the control unit would more or less be reduced to one chip. It is nothing wrong in using discrete logic by all means. But it ads a lot of complexity to a project, and hence possibilities for errors. And any modifications are also very tedious.

 

I haven't studied all the posts in this thread, so if my suggestion is redundant, just ignore it. The timing of the 7107 seems to be controlled by an RC network consisting of a 100pF cap and a 100k resistor, which gives a reading of 3 times per second. Could you simply change the 100pF cap to say .01μF, which should change the reading to about twice a minute (actually 1.8 times)?

 

Your absolutely right. The scoreboard would have been alot less complex had I gone down the programming line. I've simply headed down this direction because I don't have that specific programming knowledge. It also has forced me to to spend more time learning about discrete components. The problem solving has also been so valuable. Yesterday I made alterations to my count down timer board and it didn't work. I found 3 small mistakes. I actually found them by elimination. So I'm learning.
I think as my next project I'll try to make the board again with this knowledge but go down the MCU path - so be expecting me knocking on your door in the near future!!!

 

100pF cap and a 100k resistor... gives a reading of 3 times per second. Could you simply change the 100pF cap to say .01μF, which should change the reading to about twice a minute (actually 1.8 times)?

That sound promising. I'll try that. But now it's 1am so I better head off. I'll try it in the morning.

 

Changing the 7107 clock frequency is not that simple. This is a dual slope ADC and changing the clock frequency will alter the conversion factor as well as AC line frequency rejection.

 

Changing the 7107 clock frequency is not that simple. This is a dual slope ADC and changing the clock frequency will alter the conversion factor as well as AC line frequency rejection.

You may be right; I have never used one. It seems to be explained on pages 9 and 10 of the datasheet. Unless Chris has already modified it for 50Hz operation (which I think is necessary in Australia), it may still be set for 60Hz. It's worth a read and a check.

 

I think as my next project I'll try to make the board again with this knowledge but go down the MCU path - so be expecting me knocking on your door in the near future!!!

Of course feel free