Searching \ for '[EE]: Vbe temperature measurement again' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/io/sensors.htm?key=temperature
Search entire site for: 'Vbe temperature measurement again'.

Exact match. Not showing close matches.
PICList Thread
'[EE]: Vbe temperature measurement again'
2002\01\06@185217 by Peter L. Peres

picon face
Hi all,

I made a small test using a thermostat, a thermistor, and a BC547C with C
and B shorted. The test resistors (to put current into the BE junction)
were 4k7 and 47k respectively. They were switched alternatedly (when one
was in the circuit the other was not - not by paralleling 47k with 4k7).

The measurement was done by hand in a block of foam used as thermostat
(jury rigged). The reference was a glass lab thermometer that should be
better than +/- 3C at 100C (I tested it in ice water so the 100C point
should be ok too).

All errors are almost certainly due to measurement error, parallax reading
the glass etc etc. Anyway the trends are clear imho:

- By not using a current source the instrument constant is not clean (last
column).
- It is probably not worth the effort to do this without a proper current
source. The error induced by the changing current is too big. The effort
required to correct this could go into a table lookup for a thermistor.
- I did not try to change devices and see what gives. I will try this with
a proper thermostat and automated measurement. Even so the thermistor
creeps badly (see zero repeat at 20C).

The application note (the one with the baby bottles) was very helpful,
but they CLEARLY used a current source and other parts that made it
somewhat expensive. Any comments are welcome.

Here is a table (it was clobbered together so no comments on quality
please) (YOU WANT TO SEE THIS ON A WIDE SCREEN WITH FIXED FONT):

Thermostat measurements  BC547C with C and B shorted, 5k6 thermistor

vtest=4.99V Rlow=4k7 Rhigh=47k
spam_OUTplpTakeThisOuTspamactcom.co.il 2002

          Temperature Ilow    Vlow     Ihigh    Vhigh    Therm      deltaV     Vt-V(t-10) R(t-10)/Rt mV/K
          Deg. C      uA      mV       uA       mV       Ohms       mV         TEST mV       TEST    273.15

Meas.:     120.0       98      408      924      526      0.156      118        18.00         1.37     0.3001
          110.0       97      429      920      544      0.214      115        22.00         1.38     0.3001
          100.0       97      456      915      566      0.296      110        19.00         1.39     0.2948
           90.0       96      477      911      585      0.410      108        17.00         1.34     0.2974
           80.0       96      498      908      602      0.550      104        19.00         1.41     0.2945
           70.0       95      519      904      621      0.774      102        18.00         1.47     0.2972
           60.0       95      542      900      639      1.140       97        19.00         1.43     0.2912
           50.0       94      563      896      658      1.630       95        17.00         1.47     0.2940
           40.0       94      584      892      675      2.390       91        19.00         1.55     0.2906
           30.0       94      605      889      694      3.700       89        16.00         1.51     0.2936
           20.0       93      626      885      710      5.600       84        16.00         1.57     0.2865
           10.0       93      645      881      726      8.800       81        19.00         1.64     0.2861
            0.0       92      673      876      745     14.400       72        16.00         1.47     0.2636
          -10.0       92      686      874      761     21.200       75

Zero rep.:  20.0       93      623      885      708      5.640


Note  Test columns are to check monotonicity etc
Note1  The dV value at -10C is not in line with the others - could be meas. error

--
http://www.piclist.com hint: To leave the PICList
.....piclist-unsubscribe-requestKILLspamspam@spam@mitvma.mit.edu


2002\01\06@215328 by Tom Messenger

flavicon
face
Hi Peter

I was just looking at the Linear Tech apnote recently about this so your
data was of interest to me. Your conclusion about using a good current
source is certainly on the mark.  The deltas are so small that they are
easily lost if attention to details is not paid. I would think that the
current source needed should not present a problem. Of course, if you are
looking for rock bottom parts cost, it might be.

The theory does look interesting; Jim Williams writes interesting ap notes.
I do believe as he claims that you can do away with calibration from unit
to unit if you do a good implementation here.

Good luck and please let us know how any further work goes on this. Thanks
for taking the time to fill us in.

Best regards,
Tom M.

--
http://www.piclist.com hint: To leave the PICList
piclist-unsubscribe-requestspamKILLspammitvma.mit.edu


2002\01\07@014909 by Robert A. LaBudde

flavicon
face
At 06:50 PM 1/6/02 -0800, Tom wrote:
>Hi Peter
>
>I was just looking at the Linear Tech apnote recently about this so your
>data was of interest to me. Your conclusion about using a good current
>source is certainly on the mark.  The deltas are so small that they are
>easily lost if attention to details is not paid. I would think that the
>current source needed should not present a problem. Of course, if you are
>looking for rock bottom parts cost, it might be.

When using a diode-wired transistor as a sensor, the principal error arises
from variations in current. The current should be low enough to avoid
voltage drops across the base-spreading resistance and high enough to avoid
shot noise issues. Typically the current used is 50-100 uA.

The current source must be temperature stabilized or a large error will
result. Typically the current source used is a matched transistor pair
wired as a current mirror.

Given the difficulties associated with obtaining high accuracy from
transistor sensors, it's usually better and cheap enough to simply buy a
solid-state sensor package with the details buried internally.

In principle, solid-state sensors are capable of accuracies on the order of
0.01 C if used correctly.

================================================================
Robert A. LaBudde, PhD, PAS, Dpl. ACAFS  e-mail: .....ralKILLspamspam.....lcfltd.com
Least Cost Formulations, Ltd.            URL: http://lcfltd.com/
824 Timberlake Drive                     Tel: 757-467-0954
Virginia Beach, VA 23464-3239            Fax: 757-467-2947

"Vere scire est per causas scire"
================================================================

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads


2002\01\07@163812 by Dave Dilatush

picon face
"Peter L. Peres" <EraseMEplpspam_OUTspamTakeThisOuTACTCOM.CO.IL> wrote...

>...Anyway the trends are clear imho:
>
>- By not using a current source the instrument constant is not clean (last
>column).

Even with just your pair of resistors it's not bad, though.  These are
nice results.

>- It is probably not worth the effort to do this without a proper current
>source. The error induced by the changing current is too big. The effort
>required to correct this could go into a table lookup for a thermistor.

Also, this method doesn't product much of an output change per Deg. C;
that's one of its drawbacks.  It isn't necessary to keep the currents
themselves constant, but it is important to have a constant ratio
between the two currents.

>- I did not try to change devices and see what gives. I will try this with
>a proper thermostat and automated measurement. Even so the thermistor
>creeps badly (see zero repeat at 20C).
>
>The application note (the one with the baby bottles) was very helpful,
>but they CLEARLY used a current source and other parts that made it
>somewhat expensive. Any comments are welcome.

I think where this "Delta-Vbe" sensing method shines best is in
situations where the ability to interchange sensors without having to
re-calibrate is extremely important, along with extremely low sensor
cost, but where the cost of the measurement circuitry is relatively
unimportant.  It's ideal for monitoring die temperatures on CPU chips.

Dave

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads


2002\01\08@024949 by Vasile Surducan

flavicon
face
On Mon, 7 Jan 2002, Dave Dilatush wrote:

>
> I think where this "Delta-Vbe" sensing method shines best is in
> situations where the ability to interchange sensors without having to
> re-calibrate is extremely important, along with extremely low sensor
> cost, but where the cost of the measurement circuitry is relatively
> unimportant.  It's ideal for monitoring die temperatures on CPU chips.

 I wasn't follow this subject but if delta means diference I have some
doubts about the interchangeability of the sensors. Maybe a ratio.
There are five basic methodes of measuring temperature with a transistor:
1. Ib = constant, measuring Vbe, only be junction is used
2. Vbe = constant, measuring Ib, only be junction is used
3. Ib = constant, measuring Ic
4. Ib = constant, measuring Vce
5. Ib = constant, measuring both Vce and Vbe

But all methodes are too complicated if we think to the LM35 family.

regards, Vasile

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics


2002\01\08@063109 by Dave Dilatush

picon face
Vasile Surducan wrote...

>I wasn't follow this subject but if delta means diference I have some
>doubts about the interchangeability of the sensors. Maybe a ratio.

This "delta Vbe" technique is a well established and documented
technique.  Linear Technology, Inc. application note AN45, by Jim
Williams,  discusses this method and shows an example design on page 7,
Figure 11.
Circuits of this type produce a voltage output which is proportional to
the log of the ratio of the two currents times the absolute temperature.

The output shows very little dependency on transistor characteristics or
device type; Williams obtained less than 0.4 degrees C spread over 25
randomly selected 2N3904s and 2N2222s from various manufacturers.

>There are five basic methodes of measuring temperature with a transistor:
>1. Ib = constant, measuring Vbe, only be junction is used
>2. Vbe = constant, measuring Ib, only be junction is used
>3. Ib = constant, measuring Ic
>4. Ib = constant, measuring Vce
>5. Ib = constant, measuring both Vce and Vbe

>But all methodes are too complicated if we think to the LM35 family.

I think that last statement oversimplifies just a bit: there are many
ways to measure temperature, and each method has its strengths,
weaknesses, and a set of applications where its use is appropriate.

The "delta Vbe" method of sensing temperature has two application areas
where it is especially attractive:

1.  When the temperature sensing element MUST be a semiconductor
   junction, such as measuring the die temperature of a CPU chip
   for purposes of cooling control.  The delta-Vbe technique allows
   highly accurate die temperature measurement without any need for
   calibration.

2.  When there is a need for EXTREMELY low sensor cost (but NOT total
   circuit cost), together with a requirement that sensors be
   interchangeable without re-calibration.  For example, I can bond
   a small, cheap transistor to an assembly to monitor its
   temperature during testing or some other process (e.g., potting)
   and then discard it when I'm done, at little expense.

If you'd like to go back and follow the original thread to see how we
got to this discussion, look in the archives for "[PIC]: current
generator with a PIC" by Ashly Dearden on 27 December; that thread will
provide some context.

Dave Dilatush

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics


2002\01\08@063748 by Vasile Surducan

flavicon
face
Dave, thank you for the tips !
Vasile

On Tue, 8 Jan 2002, Dave Dilatush wrote:

{Quote hidden}

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics


2002\01\08@150921 by Peter L. Peres

picon face
> very stable current source to drive transistor sensor

I was half hoping that the math that promises one-point calibration when
using the two-current method would also help if the current would change a
little bit together for the h and l value (it does not help). This was the
point of setting up the experiment as I did.

It is already known from at least two references that the system works
when done properly (and with lots of parts).

Peter

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics


2002\01\08@150952 by Peter L. Peres

picon face
I understand that the constant ratio is very important, I will concentrate
on this point. Also I am trying to make a sensor with gain (f.ex. by
leaving the collector unshorted to B and using it as a current sink on a
timing capacitor - or to drive a second transistor in current mirror style
with 'bad' thermal contact between them).

This is the next planned step. I think that I'll use a single JFET+R
current source, with a split R shunted by a PIC pin when necessary. I have
had good results with those before (<30ppm/C without even trying hard -
maybe I was lucky with the R tempco ;-).

I found very few pointers on beta change with temperature. I'd be grateful
for some ideas ;-).

thanks,

Peter

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics


More... (looser matching)
- Last day of these posts
- In 2002 , 2003 only
- Today
- New search...