EGT sensors

madjak · Post by **madjak** » Mon Jul 20, 2015 8:06 am

I've just placed an order for an el-cheapo EGT sensor setup. At this stage I've ordered just a single thermocouple and amplifier for testing but will eventually get 4.

The thermocouples are the K-type sensor with exposed end. You can buy these on ebay for around $15. http://www.ebay.com/itm/EGT-Thermocoupl ... 3aac01c0f9

For the amplifier I've ordered a single channel unit from adafruit for $12. These units will output a 0-5v analog to hook up directly to the RCP. https://www.adafruit.com/products/1778

This makes it possible to set up 4 channels with amps for less than $120. I have no idea how good it is at this stage but I'll report back once/if it's working.

Edit: oops... I linked to the wrong amplifier. Fixed

acarrano · Post by **acarrano** » Thu Jul 23, 2015 10:07 am

I have bought the similar sensors from the same place. Looking at the amplifier though, it shows a different temp range. How easy will it be to calibrate voltage v temp? When are you looking at testing this?

acarrano · Post by **acarrano** » Thu Jul 23, 2015 11:57 am

http://www.ebay.com/itm/Quad-Channel-J- ... 35e2d1bb3c

I found this that may keep things tidy. Just waiting for a response to see if this is viable.

Alex

toga94m · Post by **toga94m** » Thu Jul 23, 2015 4:50 pm

Not sure where you're planning to mount the quad amplifier board, but the chips are rated for 0-50C so probably not under the hood. The chip is rated to +125C operating temperature, but it's designed for better accuracy if the sensor connects to the board at 0-50 (cold junction compensation range). There's also an AD8497 chip that does type K in an ambient temperature of 25-100C, same pinout, would fit the same board.

For calibration - thermocouple output voltage isn't linear with temperature, but RCP has a nonlinear mapping option. The type K AD8495 applies a gain of 122.4 to the TC millivolt signal to create the (roughly) 5mV/C output voltage. Depending on how accurate you want it and what range you care about, you can use this scale factor combined with a type K chart to create your linearization map in RCP setup.

http://www.omega.com/prodinfo/thermocouples.html overall TC info
http://www.omega.com/temperature/Z/pdf/z204-206.pdf type K in degrees C

With 5 bins available, you might want to see temperatures from 0C to 900C, but not require good accuracy at the cold end. So you could have a larger gap from 0C to the next datapoint, like this :

0C = 0mV * 122.4 = 0.000V
300C = 12.209mV * 122.4 = 1.494V
500C = 20.644mV * 122.4 = 2.527V
700C = 29.129mV * 122.4 = 3.565V
900C = 37.326mV * 122.4 = 4.569V

You can see the temperature and voltage track pretty closely at 5mV/C, so you could skip the linearization table if you don't mind a little inaccuracy at higher temps (+/-15 degrees C). And if all you're really concerned about is EGT balance between the four cylinders (and watching for extreme high temperatures), then you probably don't need to know absolute temperature within 1 degree.

Disclaimer - I do analog instrumentation for a living, though not in cars. I play with cars after work.

toga94m · Post by **toga94m** » Thu Jul 23, 2015 5:43 pm

I had to take a look at how non-linear the type K thermocouple really is.. Here's a chart I made showing error between actual temperature and indicated temperature for a non-linearized thermocouple, at 5mV/C right from that chip/board. Worst case is indicating about 815C at 800C actual.

Seeing this curve, I'd pick the five RaceCapture datapoints at 0C (min), 350C, 680C, 790C, and 900C (max). This 5-point (4-segment linear) fit will keep you within a degree or two from 0-900C, with tighter accuracy from 350-900C.

C V
0 0
350 1.7494
680 3.4625
790 4.0226
900 4.5687

tom

toga94m · Post by **toga94m** » Thu Jul 23, 2015 7:10 pm

Just a little more refining and tuning.. these five linearizing points keep you within +/-1 degree error from 0-988C, and 1000C is only 1.3C low.

Code: Select all

C 	K TC mV	AD8494
0	0	0.0
320	13.040	1.596V
680	28.289	3.462V
790	32.865	4.022V
950	39.314	4.812V

New chart - blue is unlinearized error; green is the five points with four line segments; orange is the corrected error (staying within +/- 1C). If you're in the 900's your engine has bigger problems than one degree of error..

madjak · Post by **madjak** » Fri Jul 24, 2015 6:47 am

Nice work there! For me I'm not worried so much about accuracy of the sensor, although having it accurate is nice. I would like to track the exhaust temp through time to see how much it fluctuates and pick up any spikes. Also compare the 4 different cylinders to detect any issues. I think the thermocouples themselves can vary a little in their calibration so do get all 4 accurate you'll probably need some sort of way to measure the temp accurately and map each curve.

Of course for what I'm planning on doing I don't think it matters that much, as long as I know cylinder 4 always reads 10 degrees higher than 1-3.

Also that 4 channel board looks ideal. Saves me from having to stack 4 chips on top of each other. I was planning on making a small 3D printed enclosure for the 4 amps that could be mounted under the dash.

hopefully the thermocouple rocks up in a week or so and I can test.

acarrano · Post by **acarrano** » Fri Jul 24, 2015 8:46 am

Wow you guys certainatley know your stuff. I have the ability to mount the amplifier in cool air in the cockpit next to me as sensors are 2m long (rear enginge single seater). My main concern is also the same, i want to be able to detect 1 cylinder running hotter than the others. Its generally valves that go in my engine and when they do its a new engine!! (not cheap as i've found).

With this in mind do you think it should be easy enough to use the quad board with these 4 sensors?

How would I go about calibrating temps v volts for this? Simply use the figures toga94m has given?

toga94m · Post by **toga94m** » Fri Jul 24, 2015 11:29 am

Sensor config info is here - http://www.autosportlabs.net/RaceCaptur ... ate_Sensor you're going to use the mapped option. I don't have the EGT board, so this isn't tested, but here are my numbers stuck into RCP config. On the chart it looks like a straight line, but it's correcting for about 15 degrees of nonlinearity up around 800C.

Question for you real-world guys.. how are you installing the EGT sensors on the engine? Our Miata has a tubular header, so there's not really enough metal to drill & tap for the probe. I guess if you had a cast manifold it'd be easier, or performance aftermarket with the holes already in place?

acarrano · Post by **acarrano** » Fri Jul 24, 2015 11:38 am

I have a complete custom exhaust so i will have to drill holes in at the angle i want and weld 1/8th NPT threads into it. Then make the exact in spares!

Thanks for the help toga94m. I will revert back with results.

madjak · Post by **madjak** » Fri Jul 24, 2015 12:30 pm

Yes same for me. Drill and weld in 1/8 npt bungs. I have a N/A NA8 Mx5 (miata for you US guys) so I have 4-1 race headers with lots of room. For a log manifold you'd just have to squeeze them in on the top. The bungs are pretty small so it should fit.

If you buy the 90 degree thermocouples then they will fit in better. I'll take a pic of mine when I have the first one mounted.

acarrano · Post by **acarrano** » Sat Jul 25, 2015 6:58 am

I have bought the quad board and hopefully all shall arrive in the next few weeks. Speaking with the guy who makes them and taking what toga94m has said it should come together quite well. Will take pics and bring back results.

acarrano · Post by **acarrano** » Fri Jul 31, 2015 4:14 pm

So all the gear has arrived and will be testing shortly, however will i need the 2.2k pull up resistors in the sensor break out board for the EGT's?

Cheers

Alex

toga94m · Post by **toga94m** » Sat Aug 01, 2015 12:04 am

Shouldn't. The thermocouple chip has a strong drive output, compared to a resistor-type sensor like a coolant temp sensor. Just wire the 5V VRef and common from RCP over to the new board, and the 4 output voltages to your four analog channels.