Index Home About Blog
From: John De Armond
Subject: Furnace Temperature measurements (was Re: small furnace?)
Date: Wed, 09 Jun 1999 04:03:33 EDT

MikeFirth wrote:

>   I turn to my favorite source of information, the Omega Temperature Handbook
> (free from 1-800-TC-OMEGA, where I find that a high temperature
> infrared thermometer (display on the back of a gun shaped unit) can be had for
> $1995 or more and bare units requiring housings and connections can be had for
> about $1,000.  All of the units listed claim +/-1% of reading plus 1 or 2
> degrees F.
>   Now, 1% of reading at glass temperatures is +/-20 degrees or more.  That
> means a temperature measurement of 2040F might be between 2020F and
> 2060F which makes a difference in the handling of the glass even in my
> limited experience.
>   On the other hand, for $79 I can get a single input Type K digital
> thermometer that is stated to have +/-0.3%+2F of reading and for $140 I
> can step to 0.2%+2F and for $295, I can step to 0.1%+1.8F.  That last
> one translates to about +/-4F at 2050F.

(takes off his glass hat and puts on his metrologist's one)

This is an apples & oranges comparison.  

First, it is important to understand that Omega makes almost nothing
and simply private labels other peoples' products and sells them at
high markups.  (hey, someone has to pay for that 100 lbs of paper
they send you when you request a catalog set.)  Those instruments
are either Raytek or Wahl units and are available for significantly
less directly from them.

The 1% of reading specification quoted for the pyrometer is the
overall system accuracy and is typically derived by root-sum-squared
combination of:

	*	sensor absolute error. (the error inherent in the sensor)
	*	indicator absolute error. (the error inherent in the indicator.)
	*	Error in precision of indication. (how precise the indication is
for identical input)
	*	error in resolution of indication.
	*	error of repeatability (the spread of indication for subsequent
identical inputs.)
	*	errors of long term drift.

Note also that the error specifies how far the indication will
disagree with a calibrated standard.  It does not represent how
precise or repeatable the instrument is.  Both these figures are
typically an order of magnitude better than the absolute accuracy. 
This matters because while we don't care so much what the absolute
temperature is, we do care that the temperature at which we find the
glass to work properly be repeated consistently.  In other words, it
is important that the instrument hit that sweet mark you
grease-penciled on the display dial.

It should also be noted that as reputable brand names who sell to
the military and GSA, Raytek and Wahl derive their error
specifications using standard procedures.  This is not the case for
inexpensive instruments.

The cheap T/C indicator's specification is only for certain
laboratory conditions and even then only applies to the A/D
conversion.  I've looked at a LOT of these instruments in my former
professional capacity.  Here are some problems:

*	Cold junction compensation - top of the list.  A precision
instrument connects to the T/C through heavy metal blocks with long
thermal time constants and have cold junction temperature sensors
embedded in the metal.  Cheap instruments have stamped metal
connections with some sort of low precision temperature sensor
located somewhere inside the case.  Typically a cheap forward-biased

*	Linearization - the temperature vs EMF curve is non-linear. 
Analog instruments handle this with a non-linear dial.  Quality
instruments handle linearization with either calculated (low
accuracy) or table lookup (high accuracy) linearization.  Cheap
instruments do either a) nothing, b) analog approximation using the
transfer curve of a diode to approximate the T/C curve or c) table
linearization using only a few segments, sometimes as few as 2. 
Such an instrument can be repeatable but not accurate.

*	Reference stability.  This refers to the internal voltage
reference that the T/C signal is compared to.  Cheap instruments use
either simple external references or worse, the reference built into
many A/D chips.  These show both poor long term stability and poor
temperature compensation.  The latter is important in a glass shop
because of the extremes in ambient temperature encountered.

*	Power supply stability - cheap instruments are typically poor in
this area.  As the batteries run down, the reading changes.  This
affects both the accuracy and the repeatability (ability to hit your

Without individual characterization I'd never rely on a cheap meter
to be better than perhaps 5% of the indication.

>   Of course there is that much mentioned drift of thermocouples that can affect
> temperature measurements.  But what is the drift we are talking about?
> In the back of the Omega book is an article on the advantages of the
> Type N thermocouple which has about the same range as the Type K and is
> proclaimed to have better durability and less drift.  But when we look
> carefully at the data presented we find that while the wonderful N is
> much better than a K, exhibiting, for example under 2C (about 3.6F) at
> 1077C in 1200 hours (16 gauge wire) and well under 3C at 1202C for 1200
> hours (8 gauge wire) the K type (8 gauge) is only showing on the order
> of 6-8C drift over the same time.

There are more error terms than that article indicates.  There was a
comprehensive article a few months back in the Instrument Society
(ISA) Journal about T/C error terms.  One term we have to deal with
in the temperature range we're interested in is the long term drift
cause by crystalization of the TC metal.  This article reported
errors as large as 20 deg at the high end of K/N range.  This error
can be annealed out using the proper procedure but most users don't
know to look for this error.

Other error terms:

*	Unless selected for the furnace environment, the ceramic
insulators can go conductive at the temperature of interest to glass
workers.  This shunts off part of the signal, introducing an error
term.  This is worse with precision instruments that read the T/C
voltage than with traditional analog pyrometers that actually
operate off the generated current.

*	degradation caused by the T/C connector.  Especially significant
for cheap connectors that are made from brass instead of actual T/C

*	Errors from the T/C lead wire.  lead wire has a much lower error
specification than T/C wire.  this is not a problem when the
transition from the T/C to lead wire is close to the same
temperature as the reference junction but in the case of glass
T/C's, the transition point (connector) is likely to be much hotter
than the readout reference junction. One can run T/C wire all the
way to the indicator but most people don't because of the expense.

*	Contamination of the lead wire.  Typical lead wire for furnace use
is either silicone impregnated fiberglass or ceramic fiber.  Either
jacket can be easily contaminated which alters its electrical
properties.  This is a significant consideration for batch
operations where metal oxide dust can settle on the wires

In addition to all that, there are the other uncontrolled factors
that can't be generalized.  While the optical pyrometer is a closed
system where error terms can be carefully controlled, a T/C system
is distributed with different elements exposed to different
influences.  This is a major consideration if one expects to rely on
a T/C indicator for much better than about 5% overall accuracy.

The reason we can do so well with a relatively inaccurate system is
that we learn to compensate, usually unconciously.  The apparent
sweet spot temperature may slowly drift with time but we just
mentally note that "the best place is now half a division below my
mark" or the equivalent with a digital instrument.  The problem with
this is that when you have to replace a component or even just
change something, you're back to square one.  You erase your mark
and experiment to find it again.

Because the optical pyrometer is self-contained and because it is
remote from the furnace, it minimized these problems.  Even the
cheaper sub-$500 instruments are still self-contained and thus
exhibit better long term stability than T/C systems.  And one can
easily convert even a cheap instrument to a glass-specific
instrument by buying a cheap optical bandpass filter from Edmund
Scientific to grab only the infrared from the actual glass.

The other significant benefit from an optical pyrometer is that
since it sees down into the melt, it naturally averages the
temperature over the cone of vision.  The T/C, on the other hand,
measures only the spot where it is located.

Please don't interpret this article as some sort of condemnation of
T/Cs.  It is not, though it is a condemnation of cheap instruments. 
People, myself included, do quite well with T/Cs.  But an optical
pyrometer simply does it better at moderate cost.  I wouldn't look
at it as pyrometer vs blowpipes.  I'd look at it as pyrometer vs
lost time trying to get the melt right and keep the color correct.


Index Home About Blog