Not so Hot; temp sender fault?


I’ve highlighted the small corks that cover the adjustment ports on the rear of the gauges on my '69 S2. If your gauges don’t have them it might indicate that the gauge has been replaced, and that may be related to your problem. The two adjustments control the zeroing of the gauge and the range in a rather crude manner. If you have a bench power supply there are some small dots on the face of the gauge that can be used to calibrate the gauge at certain input voltages. Let me know if you need the details. My memory (not a good source) says there are dots for 4.8v and 7.2v

1 Like

Do yourself a favor and don’t muck with gauge adjustment. Smiths had a special tool for this. Although you can do it with a screwdriver, the wrong tool and inexperience can make a mess of it. It was calibrated at he factory, and it’s unlikely to have gone out of adjustment, so let it be for now.

The most likely problem is the sender. Smiths had terminology for thermistors which doesn’t translate well to modern art, so the odds are very good that the sender is incompatible with the gauge. I have four different senders on the way, from four different manufacturers. By this time next week, we’ll know which ones really work. At least we won’t have to rely on sloppy measurements.

If you have your heart set on keeping this sender, I will also work out procedures for calibration using external resistors, so you won’t have to dig into the guts of the gauge.

All I ask is to be patient, good info ieill be forthcoming.

Thanks david, I do have the two corked holes but was looking for a single hole with sealing goop. If you have any details as to which is the zero set and which is the range set and any further parameters are used to set them

Those adjustments may not actually work. I was able to adjust one gauge, but in several others the rivets were completely tight and would not allow any movement. I came to the conclusion that the gauges were adjusted at the factory and then the rivets tightened so the adjustment would not drift.

Yup. Adjustment is more likely to result in a ruined gauge than an accurate one.

The first of my test units arrived today. It’s the repro part c168695, sourced from SNG. I suspect the manufacturer is Caerbont, which is the latest company claiming the Smiths name. By all appearances, a well made part, but I’ll have more to say about it when the other senders arrive. For now, let’s look at the numbers:

Temp C Ohms
20 620
25 481
30 381
35 313
40 249
45 202
50 165
55 134
60 110
65 92
70 76
75 65
80 53
85 45.6
90 39
95 34.5
100 33.4


Thanks Mike and Michael for the timely responses. I’ll wait for results from the sender side of things and leave the gauge alone. I relocated my heat sensor to the rear hose off the intake manifold as it closely matched the area around the temp sensor in results using a IR thermometer (the manifold to head stud was much hotter) So I seem to run about 70 to 80 coolant temp on a short run. At 75C my sender is at about 50 to 40 Ohms which matches the findings in the sensor curve above that 50 Ohms is approaching max on the gauge. It seems it should be about 100 Ohms at 75C to put the needle in the mid range according to the above curve graph
When the pressure builds (7 lb cap) I get some leaking at the water pump (new pump with the engine rebuild 20 years ago but mostly just sat. I assume it is somewhat due to the Evans coolant as seeping thru is one of the downsides to the water-less stuff. Got much improvement from re torquing the W pump bolts. Evans recommends Barr leak stop with their fluid. has anyone been impressed with another leak stop product?

A few responses:
I became interested in this topic because the sender in my own car needs replacement, and now it has my full attention. Eventually, I’ll be writing up a page on this so that the data is permanently available, with a full discussion. There’s more to this than I expected. But to cut to the chase: the Caerbont sensor is the only modern product available with a screw terminal. If it isn’t a good match to your gauge, you’re going to be stuck with either switching to a blade terminal sender (assuming any of those are better) or applying an external resistor to correct the sensor. You may have to experiment a bit, but 30 ohm resistor in series with the sender should get you close. Install it in the fold down panel area, so it’s out of sight.

I’m not a fan of waterless coolants. But one thing you shouldn’t be seeing is higher pressure under standard conditions. I think you have a weak pump seal, which should be addressed. I’d suggest you replace the seal, and use a light coat of gasket sealant on the new one. I don’t suggest chemical leak stoppers under any circumstances, not ever, not even if you have a bleeding hole in the system. They’re more likely to plug your radiator than to plug a leak.

Finally, an IR thermometer is a terrible thing if you need precision readings. If you have an ongoing need to measure your cooling performance, use a self adhesive plastic thermometer on your thermostat housing.


Thanks Michael for the good advise, I’ll try to see whether the leak is from the WPump gasket or the impeller seal and fix it. I know tightening the WPump bolts greatly reduced a dripping of fluid but not completely and only when hot under pressure. So it may be I just need a new gasket ; do you recommend any particular gasket seal for the WPumps ? I usually use Halomar Blue for most gaskets but hear RTV silicon type is better for water-less coolant.

I do not seem to have high pressure as the Radiator cap removes (carefully and slowly) with out boil over, vapor, steam or trouble after a run up to temperature.

I found a variable resistor made to adjust temp senders on a “Vanagon” and will try that behind the dash as you suggest.

I attached a heat probe w/ a clamp to the rearmost coolant hose fitting on the manifold to give me a more accurate temp reading (Wire not long enough to reach the thermostat housing) . It’s called an “Engine Guard” made is Australia and it gives me a digital read out placed near the dash. I formerly had it on a Inlet manifold stud but that was usually about 30 C hotter than the coolant (I say “about”, cuz as you say, the IR readings bounce around a lot; plus/minus about 10C) .

Thanks for doing the research on the temp senders avail.; I will wait for your findings.

I’ve turned up a total of six sensors since Friday. Testing so many was a strain, the numbers may be off a bit. The 20,000 foot preliminary view is that Smiths seems to have used two different thermistors, and modern versions try to copy one or the other, but don’t necessarily distinguish correctly. Unfortunately, this means that I’ll have to do some testing on gauges as well.

No further interpretation tonight, but here’s the raw data for anyone who cares to pick at the problem:


Mike, when the sender died on my S3 E-Type I simply took one off my MK2 parts car. Unfortunately the readings were completely wrong, so I also installed the meter from the MK2 which worked correctly with the MK2 sender. This experience agrees with your observation of two different temperature measuring systems.

This indeed the case. But first a little background. Thermistors have an exponentially decreasing characteristic curve. So it’s not easy to specify one with a shorthand notation. If you go to Newark and Digikey looking for a thermistor, you would search two numbers: Ohms@25C and Beta. Beta is a calculated number:

B= ln(R1/R2) / ((1/T1)- (1/T2))

Sort of like the slope of a line. The combination of a point on the curve and the Beta is a fairly good description of the performance curve of the thermistor. I’ve calculated these numbers for each of my samples, you can find them in the spreadsheet. That’s now, but not then.

Smiths used various shorthands for specifying thermistors, none of which lines up with Beta. In some places you might see them specify two points: the resistance at 30C and at 110C. Which not only can’t be readily compared to a modern component, it’s a PITA to measure something above the boiling point of water. Today, I found that they had an even less precise specification method: simply specifying not one design resistance, but a range@100C. Which is also hard to measure, and tells you nothing about the shape of the curve below that temperature.

This is why aftermarket parts (probably including those from the latest Smiths company) are crap. There’s no modern analogy to these old specification systems. I suspect that every one of the modern parts I’m seeing was reverse engineered from the scrapyard.

Getting back to the two groups of senders in my collection, I was able to get some time with an old Smiths catalog today, and here are the listings for the relevant senders. Each one has a range of temps @100C specified, and there are two main ranges. There are a couple of oddballs with very different ranges, as well. Not only does this give credence to the “two sender” theory, it suggests that the parts books may have confused the parts, due to Jaguar’s habit of assigning their own part numbers. So more research, and I’m going to have to dig through my box of gauges.

1 Like

Not that it would help the overall problem but would it work to use antifreeze mix to reach 110C?

I suppose, but the point isn’t that I can’t measure the part. The point is that the stated spec can’t be matched to a current component without experimentation.

I haven’t forgotten about this, testing has been time consuming. I came up with three gauges, two from my spares collection and one from my S2. I have to confess, I can’t say for certain what the provenance of any of these gauges is. I’m assuming that the two spares are S1 gauges, based on having blue lens filters. The one in my S2 may or may not be original. So there’s that.

What I found was that the two S1 gauges behave identically. Considering that they’ve not been well treated through the years, it reinforces the idea that the calibration isn’t worth messing with. The S2 gauge reads slightly lower than the S1 gauges with the same input, which reinforces Mike’s theory that there were two gauge systems. I used John’s S2 mapping to estimate the readings on my gauge.

But the big surprise was plotting the measured response of the gauges to the measured response of the sender, in this case, the “C16895” from SNG. Here are the results, without further ado:

Temp C S1 Gauge S2 Gauge C16985
58 50
68 40
70 76
75 30 65
80 50 53
85 40 45.6
87 20
90 30 39
96 20
100 10 10 33.4


A few points: notice that both gauge curves are straight, while the sender curve is a curve. The S2 gauge is a better match to the sender than the S1. None of this helps the OP, because his readings were consistently high. I think the simple solution for him is to add a resistor in series with the sender. I Think the best I can do here is document how to “shift” the sender curve up or down by using resistance.It seems as if the gauge systems were designed to read either 75 or 85C dead center (depending on generation). But above or below normal operating temperature, accuracy goes to the wind. I’m not sure any of the senders, old or new, can produce accurate readings.

1 Like

Ok, first the bad news. I made some sort of error with the gauge measurements, I’m going to have to go back and recheck those numbers. Disregard the above post for now.

Today, I received a C15472 from SNG. This is another repro part, similar to C16995, except it has a 1/4" connector instead of a screw terminal. Measurements were not consistent between the two senders:

Temp C Ohms
25 495
30 393
35 325
40 270
45 220
50 181
55 154
60 130
65 110
70 96
75 83
80 72
85 63
90 54
95 47
100 42.9


C15472 has somewhat higher resistance than C16995, so it will read lower on the gauge at any given temperature. Among the seven senders I’ve tested so far, it’s about the middle of the range. Among the seven, there should be a sender to match any requirement. More about this after I retest the gauges.

Fascinating research… thanks!

I don’t like where this is going: the gauges seem pretty consistent, but no two senders are the same.

I recognize there are good gauges and there are poor gauges, I just have trouble thinking theses gauges were ever intended to be overly accurate. And many gauges, in particular analog gauges , are most accurate at mid range. Which means they are less accurate at the high and low. I think it is great that Michael is testing these components to determine the better and the not so better, but I just do not see the accuracy happening with these components that some would like., and although accuracy sounds nice, I also do not see that it is needed.

It’s the senders, not the gauges. A sender is just a thermistor. There’s no reason that they can’t be spec’d and manufactured accurately, but they aren’t. Even the two Caerbont repros behave differently. And no two of the seven senders are a matched pair. They do tend to fall into two groups: a “reads high” ground and a “reads low” group. There’s no ‘reads right’ group.

As for the gauges, they seems to be consistent. But since any thermistor will be nonlinear, it must have been a comedian who divided the gauge scale into linear divisions. The later gauges with no temperature markings are more honest. That said, I need to rerun the gauge test, I think I made the error of not giving the gauges long enough to react. Their action is slooooow.