Re: Temperature sensing technologies

TI make a wide range of pre-calibrated sensors with digital interfaces. I have used a lot of the TNP275 i2c sensors. They seem very reliable. For high accuracy look at the TMP119. For very low cost the TMP1075. John

Getting good measurements of still air is hard, because the sensors have a very strong tendency to measure the temperature of their leads instead.

We usually care much more about speed and millikelvin repeatability than anything else, so we use 0603 thermistors with one end soldered to a copper pour. For our stuff, the pour is in intimate contact with the cold side of a Peltier cooler, producing a time constant of 100 ms or faster.

Something similar would probably work for air as well, though of course much more slowly.

Cheers

Phil Hobbs

cars use a lot of NTCs, comes in all kinds of form factors and are easy to interface

When I was much younger, I told one of my bosses that thermistors were unstable, and so were the people who used them.

He responded by showing me a datasheet for a Yellow Springs glass- encapsulated thermistor - which had only recently been introduced - and I became a convert.

I haven't looked into modern surface mount thermistors, so I don't know how they are encapsulated.

I do know that Betatherm glass encapsulted thermsitors have pretty much the same long term stability as the Yellow Springs parts and the Philips parts weren't as good. If you want good long term stability in a glass encapsulated thermistor you do need to be picky about the supplier.

Semiconductor temperature sensors get standard semiconductor encapsulation because pretty much all semiconductors get that same encapsulation.

Platinum resistance sensors don't need a lot of protection, but they are insensitive.

They are quite useful as temperature detectors on heatsinks.

Yes, it was a hell of a rush job. The organ had been fitted with a transistor amplifier by a previous owner; he had just cobbled-in a budget Japanese hi-fi amplifier with some sort of monobloc output stage and tinplate heatsinks. With two 6-ohm loudspeakers in parallel as a load, it did not like running flat-out sinewave for long periods. Two weeks before the organ was due to be the centrepiece of a national organ conference, one of the output transistors shorted and bunt out the bass loudspeaker coil.

Luckily I had just been reading an advertisment for Goodman loudspeakers from the 1950s, which said the factory could offer a 24-hour turnaround service for repairs. I realised that these Goodmans industrial-quality loudspeakers were intended to be easy to repair, so I dismantled this one, made a new voice coil and a polepiece centring tool and had it reassembled and installed within 36 hours of being told about the fault. (I remember setting the Araldite quickly by leaving it on the front seat of the car in bright sunshine.) This was connected to a temporary amplifier while I set about designing and making a replacement for the original in time for the conference.

There wasn't time to mess about finding a supplier of suitable heatsinks; I needed something which I could make myself quickly. A local metal factors cut all the plates while I waited and sold me the threaded bar and nylon rod to make the spacers.

The electronic design was pretty basic but every possibility of abuse had to be guarded against, so it was robustness that scored over performance at every stage. A lot of it was set by what I had already available - the mains transformer had come out of a fruit machine, the die cast box was a spare one I had on hand. Just to complicate matters, the gain distribution and impedances had to be the same as the original Compton valve amplifiers (which I believe were designed by Williamson), so that they could be interchanged.

A lot of my designs use 2N3055s, they are basic and simple - and easily available. I still use a valve amplifier for some P.A. work, that has EL34s in the output and ECC83s in the other stages.

I recently had to sort out a problem for a friend who uses those while teaching traditional operatic technique to a small group that meets in a large room in his house. I found the stereo imaging didn't seem right as I moved around the room - but, of course, I had stood up to walk around. When I set down, everything was right aghain.

The Quad electrostatics have the high frquency source in a narrow vertical line at the centre, this means that the frequency response in the horizontal plane is good but there are quite severe phasing effects in the vertical plane. You really need to listen at the same level as the loudspeakers to get the optimum quality.

If there's a temperature difference, the air will convect without help from fans, and if there isn't you don't need to stir the air.

The trouble with 1N4148's is that they have a loosely specified forward drop at any temperature and current. The cost of calibrating each one has to be set against the cost of buying a pre-calibrated semi-conductor sensor.

For that particular one-off, where they were all from the same batch, it was easy to just check that they were similar and adjust the resistor networks to switch at about the right temperature. I didn't have time to go looking up specifications and placing orders, I just grabbed some components from a drawer.

LM35 is (again!) available in TO-220. That can be mounted on some metal surface, or can be bolted to some little winged heat sink.

Devices such as the Analog Device's MAX6698

Diodes such as the 1n4148 don't follow the Shockley diode equation very well whereas a diode connected transistors follow it fairly accurately.

Robert Pease has a chart of a variety of diodes in Appendix E of "Troubleshooting Analog Circuits" showing the wide variation of Vf vs current between devices.

The temperature can vary by several C degrees, depending on where you measure it. The temperature of the inner surface of the south-facing outside wall of my kitchen is always a couple of degrees below that of one of the inside walls in Winter - but a couple of degrees above it in Summer when the sun is shining on the outside of it.

When I first fitted a diesel heater to my van and ran it up to what should have been a comfortable temperature, the air felt cold and clammy and most unpleasant. After several cycles of opening the doors to clear the moisture and re-heating the air, the van interior began to feel cozy as the fabric dried out, the humidity dropped and the wall surfaces warmed up so they were emitting long-wavelength I.R..

There is no point in having a room thermostat accurate to fractions of a degree when there are such large variations in the measurements and the subjective warmth they engender.

I think you are missing something here. You want consistent performance in your temperature sensors. Manufacturers of sensors that are consistent will be confident enough in them to guarantee the calibration. Buying a sensor that has traceable calibration and a long- term drift specification gets you what you want. It is not expensive.

John

Am 12.09.25 um 08:56 schrieb Jan Panteltje:

I still use that old Stax SR-44. I just replaced the cables.

• Krell + BW803

Gerhard

I had a couple of bandoliers of 100 x 1N4148 in the drawer - so they were what I used. With a fortnight to design, build install and test an amplifier that must be reliable enough for 'industrial' use, I couldn't afford to order special parts whose performance and delivery times were unknown quantities.

These had 18" cones and 2.5" voice coils, so the work wasn't particularly delicate. I carefully recovered all the burnt wire to calculate the number of turns and, after measuring the wire diameter, found I had a suitable replacement size in stock.

I machined up a former from aluminium bar and wrapped thin alloy sheet from a lithographic printing plate around it to form a tube (with a gap to avoid a short-circuited turn), then wound the coil onto the alloy tube on a bed of Araldite. That was then Araldited to the cone and 'cooked' in a hot car until it was hard.

The magnet assembly had to be held in alignment while the bolts were tightened, so I had to machine up an aluminium jig to maintain the correct air gap. The cost of doing the job was mainly the alloy bar which was needed for the formers and jigs. If I had been doing the job commercially, I could have spread that cost over several repairs.

I have dismantled a KEF B139 in the past; it was making 'voice-coil-rubbing' noises. I could find nothing wrong so I re-assembled it. Then I checked the amplifier and found bursts of oscillation on peaks. The data sheet had said there was no need for a Zobel network but I later discovered that it only applied if you ran the amplifier at full voltage. I was running mine at reduced power (to protect the loudspeaker) and that must have changed the internal time constants of the amplifier and led to instability. Adding a Zobel network corrected it.

The diode drops and tempcos would wreck the temperature signals.

If you have four sensors, yes.

The math is messy there. Good enopugh for extreme temp sensing at bad accuracy.

There are cases where thermistors in parallel could make sense too.

For drive over it with a truck and it still works reliability along with zero measurable drift over centuries (within your stated temp range) it is hard to beat a grounded tip stainless steel clad type J thermocouple. (K has better sensitivity but the iron wire is prone to rusting out.)

Cheap sensor in a wide variety of packages, some of which are fairly reliable, cheap interface electronics.

We've been using the two terminal AD590 for our temperature sensors: (App Note: Use of the AD590 Temperature Transducer in a Remote Sensing Application)