1. Home
  2. Electronics Design
  3. Upper limit on power dissipation in NTC thermistors

Upper limit on power dissipation in NTC thermistors

I've just been reading

"A versatile thermoelectric temeprature controller with 10mK reproducibility and 100mK absolute accuracy" by K.G.Libbrecth and A.W. Libbrecht, Rev. Sci. Instrum. 80 page 126107 (December 2009).

The authors don't mention self-heating in their thermistor, and dissipate 2.5mW in the device. It isn't obvious what the self-heating coefficient of their thermistor probe might be, but since it seems likely to be a packaged YSI 46000, which does 10mW/C in a well-stirred oil-bath, they are probably seeing 250mK of self-heating, whcih makes nonsense of their claimed 100mK absolute accuracy.

Since I've seen a bunch of smaller glass-encapsulated thermistors provoked into showing unstable resistances in 2002 by dissipating 1mW in them, 2.5mW sounds much too high for comfort.

The 2002 event involved getting calibration data by measuring the resistances of a bunch of thermistors in a well-stirred water-bath over a range of tmperatures. The resistances were being measured by a big Thurlby-Thandar multimeter, set to autoranging, and the resistances read went unstable - in the last few digits - at highish temperatures when the multimeter had moved down to the 1k or less range.

By disabling autoranging and moving up to the 10k or less range, we went back to seeing stable resistances.

I've poked around since then looking for discussions of "hot channel" formation in NTC thermistors, but what analysis I've seen of self- heating and its consequences seems to assume that if 10mW/C is the self-heating at zero power dissipation, it is the right figure at any dissipation.

In fact, even if NTC thermistors were homogenous objects (which they aren't) self-heating is going to move the current paths in towards the axis of the thermistors, so that the heat is going to have to diffuse through more thermistor to get out, and the heat flux per unit area around the core is going to rise as the current gets moved in towards the axis, so the self-heating per watt is going to rise with increasing dissipation.

In reality, thermistors are inhomogenous lumps of incompletely sintered metal oxides, and the conduction paths are going to be a chaotic network; as the dissipation in the thermistor increases, the shorter inner paths are going to get to carry more of the current and the self-heating per watt is going to rise, but it is going to be a somewhat random process, which could explain the unstable resistance readings that I saw back in 2002.

There's probably a paper in this if anybody has access to a useful batch of themistors, a well-stirred water-bath and a six-digit multimeter or resistance bridge.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman
Loading thread data ...

Hi Bill, I know Ken Libbrecht and his son. The circuit is very similar to what is used in the diode laser that we built 'together'*. I would guess that you are right. The accuracy is most likely that stated by YSI. They could run these at a lower bias voltage and reduce the self heating. Most likely they forgot to think about self heating. (I run the same circuit from a 5V reference.)

The transient step response in figure 2 is less than stellar. Looks like a little less gain or more integration might be in order.

But does anyone ever want absolute accuracy from a thermistor? (Mostly I just want stabilty and I hardly care about the absolute temperature.) I'm frankly amazed that they can make them repeatable to 0.1 C.

George H.

*well there were a few rough spots in the collaboration... but what else is new.
Reply to
George Herold

I'm partial to thinfilm platinum RTDs for stuff like TC reference junction sensing.

John

Reply to
John Larkin

y

ds

ul

Me too. Mainly because you can dissipate more power in a thin film platinum RTD than you can in an NTC thermistor.

Sloman, A.W. "On microdegree thermostats", Journal of Physics E: Scientific Instruments, 11, 967-968 (1978).

The signal voltage tends to quite a bit less than you can get out of a thermistor, which can make it necessary to go over to AC excitation. Joerg would hate what that does to the budget.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

s
l

The circuit is a tolerably competent reworking of Bradley, Chen and Hulet's circuit from 1990, as they acknowledge - the paper is in their list of references. Bradley et al achieved 300uK, which is rather better than the Libbrecht family seems to have managed.

The paper does talk about modelling the settling behaviour, but the proportional and integral gain amplifiers use 33k and 10M resistors and a 1uF capacitor, which doesn't suggest that they did much in the way of fine tuning.

I'm a bit peeved that they cite my paper - which does bang on a bit about dissipation in the sensing thermistor - without having read it carefully enough to notice that self-heating can be a problem.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

Thermistors are amazing in narrow temperature ranges, where you don't need the linearization network, and so can take advantage of their high dR/dT without needing to dissipate a lot of power. Their 1/f noise is pretty variable, I would say, and I've never seen a spec for it in a thermistor datasheet.

RTDs are better in a whole lot of ways, but of course they're much more expensive in general.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
Reply to
Phil Hobbs

y

ds

ul

You need to distinguish between the glass-encapsulated "interchangeable" NTC thermistors, and the rest. Yellow Springs Instruments made a big step forward when they introduced the first stable parts, and their data sheets did talk about long term stability.

A couple of the micro-degree controller folk claim stabilities down at the Johnson noise limit of their set-ups, without beign aware that there was a Johnson-noise limit, which does suggest that the 1/f noise isn't too bad.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

y

ds

ul

t- Hide quoted text -

OK next thermal project I'll keep thin film RTD's in mind. I never thought too much abotu the noise in the circuits. Most of the time I've got fairly long integration times in the control loop.... 10's to

100's of seconds.

Say another interesting question is where to put the sense element? I've always found that it's best to put it as close to the thing I'm trying to control as possible. I was testing commercial TEC controllers. I found all of them had too short of a time constant. I hacked some of them by sticking bigger caps in them. When I asked the vendor why they didn't make longer time constants.. The reply was to put my sense element right next to the TEC. Then they said I could control it's temperature really fast. Yeah I said, but I don't care about the temperature of my cold plate I care about the temperature of the diode laser.

George H.

Reply to
George Herold

d

ey

e

rds

ful

Hey you should be happy they cited you at all! At least they used a second thermistor to montior the temperature. I say a paper where they used the variation on the voltage from the thermistor in the loop as a measure of the temperature stability.

George H.

Reply to
George Herold

g

ed

s
W
a
"
y

hey

he

e

ards

d

eful

e
s
y

le

is

net-Hide quoted text -

I

I had the same discussion with my boss - Jim Molloy - at Affinity Sensors when we started the project that is written up in the 1996 paper, and we put our sensing thermistor as close as possible to the TEC.

When checking the performance of the system we glued a separate thermistor on top of the block of silica, at the point where we wanted the temperature to be stable. As you can see from the paper, the approach worked very well.

If you really need to stabilise the temperature at the laser diode, put another thermistor right where you need it on the laser diode, and feed any offset into the integral term. The time lag from the TEC to the primary sensor fixes where you have to roll-off your high frequency gain, and we ended up with insufficient gain at 2.5mHz to suppress the noise coming in from the bang-bang controller on the room air-conditioner (as is mentioned in the paper). That limitation, rather than any temperature gradient from the cold-plate to the block surface - is what limited our performance.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

ng

red

es

mW

a
h
s

l"

ny

they

the

se

wards

e

nd

e

seful

e
s
y

le

I am.

That would be

Barone F, Calloni E, Grado A, de Rosa R ,di Fiore L, Milano L and Russo G 1995 Rev. Sci. Instrum. 66 4051=964

I wrote a very acid comment on that and a couple of other drop-offs in that paper, but Rev. Sci. Instrum. didn't publish it. The comment didn't cite any new references and thus came a little too close to just pointing out that their referees and their editor hadn't done their job right.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

Hide quoted text -

It depends on the source of the thermal forcing, and how complicated you're prepared to make your control system. Thermal conduction is very slow--in fact it slows down quadratically as you move the sensor away from the TEC or heater. That limits the loop BW very badly if the laser isn't right there on top of the TEC.

If the thermal forcing is mainly from the heat sink temperature variation, you're far better off with the sensor mounted right at the cooler--even drilled into the cold plate. If the laser dissipation is varying, e.g. due to mode hops or current tuning, you need a second sensor right there, with a second loop driving a local heater. I like to use the monitor photodiode for that--dump some forward bias current into it and look at the change in Vf. It isn't foolproof, especially nowadays, since the monitor photocurrents have got so much bigger...you used to be able to swamp the photocurrent with a milliamp or so of forward bias, but not anymore. Its compelling advantage is that it's brazed to the same header as the laser, so the response time is usually well under 1 second. For a heater, you can use 2520-size resistors--they're made of alumina, so the TC is also fairly fast.

Re sensors:

If you have a look at Figure 20.3 on P. 803 of

formatting link
it pretty well summarizes the reasons why thermistors and RTDs are the right way to go for high accuracy things.

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
Reply to
Phil Hobbs

Hide quoted text -

Another sensor or so, somewhere between the end element and ambient, can do some feed-forward correction.

Here are some 1206 thermal time constants:

ftp://jjlarkin.lmi.net/RTD_in_air.JPG

ftp://jjlarkin.lmi.net/RTD_lotsa_copper.JPG

ftp://jjlarkin.lmi.net/RTD_on_board.JPG

Soldering a surface-mount resistor or polyfuse to a bunch of copper has radical effects on its thermal behavior. In the cases above, theta and tau change by numbers like 10:1. An 0603 resistor can dissipate a watt if you mount it right.

John

Reply to
John Larkin

g

ed

s
W
a
"
y

hey

he

e

ards

d

eful

e
s
y

le

h

is

e

.net-Hide quoted text -

=A0I

ww/beos2e/thermal2.pdf, it pretty well

t- Hide quoted text -

Thanks Phil, I read your Thermal chapter a few years ago. The last thmeral control project was stabilizing the temperature of a permanent magnet used for NMR. The magnet is insulated from the outside, a ~1" diameter copper rod runs from the donut of magnetic material to the outside world. On the end of the rod is a TEC and then heatsink. The biggest path from outside to inside is through the heatsink/TEC/rod. We'd built the copper rod with a long hole drilled down it for the thermistor, almost touching the magnetic material. After reading your chapter, we tried the experiment of mounting the sensor at the magnet side of the TEC. There was no observable improvement in stabiltiy. We had a very sensitive measure of the magnet temperature difference.. We can set the NMR frequency to 10 Hz out of ~20 MHz. The biggest problem was that this increased the time constant for the magnet temperature to stabilize by about a factor of 2.

Re Sensors: Very cute idea of using the monitor diode for a temperature sensor.

George H.

Reply to
George Herold

ing

rred

kes

1mW

y a

sh

ss

e

el"

any

they

the

use

a

owards

he

and

a

ce

useful

y
*.

at

uce

oks

tly

able

gh

e is

re

l.net-Hidequoted text -

=A0I

OK, I've never actually done the comparison. It seems so 'obvious' that I want the sensor close to the control object.

George H.

Reply to
George Herold

Hide quoted text -

Depending on the geometry and the amount of adaptive control you're prepared to put in, that's probably true. Without an MCU to do some tweaking, I'd worry that device-to-device variations in things like epoxy bond line thickness would produce artifacts.

I generally like to put a loop on both ends of the cooling string--one on the TECs to get rid of the environmental forcing, and one at the laser to improve the loop bandwidth. They're liable to fight, of course, if you're not careful--if you let the fast loop dork the slow loop, it can take over the entire job once the slow loop has done the slewing. I'm using that trick in a laser locker for a customer--the slow temperature tuning loop gets into the ballpark, at which point the fast current-tuning loop takes over. I don't know if Tim W. would approve, but it's one pretty effective way of getting decent bandwidth without windup problems.

Yup. That's why you have to be so careful with air temperature measurements--it's really really hard to measure anything but the temperature of the leads.

Hopefully it's made of something reasonably electromigration-resistant!

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
Reply to
Phil Hobbs

net

formatting link
Hide quoted text -

How about 30 watts from a 2010? Or 300 watts from a surface mount resistor 3/8" square?

formatting link

John

Reply to
John Larkin

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.