JP86SS

iTrader: (1)

So what are you guys using for EGT sensors and signal conditioning?
Probes look inexpensive enough but the conditioning to get a 5V output is a bit much ($150-$200)
Is there an existing method to use resistors or something to obtain a good reading device inexpensivly?
Part numbers etc, greatly appreciated.
TIA
John

RednGold86Z

iTrader: (1)

You'll need a thermocouple amplifier basically.

I found one a few years ago that converted it to a 0-5V signal, but the range was limited by the supply voltage, so we had to use a 12 to 24 V converter (cheap) in order to measure cat temparatures. Just can't remember the name on it though.

But, if you want a serial connection, you can use a PICO thermocouple amplifier. That has 8 inputs, and requires no power supply (other than the serial port). I think it was pretty cheap, like $100 or so. edit: looks like it's a lot more than that now but it's USB. There's a cheaper one for about $200 with fewer inputs.

Building your own would not be a good use of time. Thermocouples generate very tiny voltages, and you'll need a cold junction reference.

Do you need it to output a voltage, or do you just need something that you can log somehow, like through a USB or serial port?

JP86SS

iTrader: (1)

Would like to log to the ECM if possible but be able to figure out the actual scaling needed. I don't think resolution needs to be insanely accurate for the 8 bit input. Just some way to verify tyou are in the ballpark of 70*F from actual. (better than 5% which should work)Thermocouple may not be the best way to go because of the very low voltage output.Since there is already a 5V source, couldn't it use a resistor to reduce the current/voltage supply and then read it in like a CTS?Something along those lines. Maybe an RTD would be easier to use for that. Don't know, still fishing for info.

RednGold86Z

iTrader: (1)

I don't know of any thermistors that are used for EGT measurement. Thermocouples seem to be the industry standard (Type K for most applications - has a wide range).

Out of curiosity, I went to Ebay Motors and typed EGT in the search. Plenty of gauges available. Maybe one of them has an output.

I bet Grumpy has something that'll do what you want.

JPrevost

Thermocouple Amplifier Boards

I asked him to make me a simple board and he did... apparently (just checked his site) he added an upgrade to what I had him make for me. $35 is reasonable don't you think? It works, it's cheap, and it's output is linear making for an easy digital temp gauge. Hook up a K-type thermocouple and be good to go. When hooking up, remember to use the k-type wire ALL the way to the TC terminals on the board. So plan on buying enough wire to get to where you'll be placing the board. I don't recommend placing the board in the engine compartment.
Mine:

JP86SS

iTrader: (1)

Matter of fact thats the chip I've been looking at using to make one.
There's another guy that has one that is like $24 too but now I lost the site and have to search again rather than make my own.

Edit: (Duh!, the site IS the one listed on your board, now I don't have to search)
The output on the specs show it will be up to Vs-2 at 1000*C (10 volts)
Were you able to knock down the output with a voltage divider or something?

JPrevost

Take another look at the attached image and see if you can't figure it out. Next hint is the website pdf file spec sheet. Okay, I won't make you look . The resistors bring the voltage down into 0-5v.

LCD with built in thermocouple controllers - $100

junkcltr

iTrader: (1)

Sort of true. It appears that the ckt does a divide by two. Watch out though. Here is what can happen:

If the EGT goes above 960 *C, then the 9.8 volt output divide is approx. 4.65 Volts. If the EGT gets to 980 *C, then the 10 volt output divide is approx. 5 volts. I am guessing the voltage regulator is a 10V device. That way the max output voltage is 10/2 = 5 volts at 1000*C. Make sure that the voltage divider resistors are within spec to give a voltage divide of 2 or greater (i.e., think about resistor tolerances). I would measure the two resistors and make it so the there was some margin for the 10V input voltage reference. Something like a divide by 2.1 and calibrate the 2D lookup table accordingly.

The stock ECM A/D can't handle inputs greater than 5 volts as far as I know. You risk blowing out the ECM A/D if using an input greater than 5 volts. BTW, the ckt designer calls it a DAC which is incorrect.

Also, the ckt has a NON-LINEAR output and requires a 2D lookup table. The author says it is linear in the datasheet, but that is incorrect. That is why he offers a calibration service for $15. He will do the calibration and give you the 2D lookup table values.

Seems worth it for $35. The AD595 is about $10 - $15 in low quantities. A custom "real fab" PCB in low quantities will be about 20$ with a min order of three ($60 total).

Junk

junkcltr

iTrader: (1)

If you do end up building your own then use the AD595. It is a simple PCB that you could wire onto a RadioShack mini board. Digikey has the AD595 devices for $10+. You would also need a 10 V linear voltage regulator. If you did a divide by 2.1 on the output that would put at 4.76 volts max to the ECM A/D. ECM 730 device U6 has all kinds of inputs waiting to be used. The best part is that the EGT would be correlated with the other ECM ALDL values.

You would have a range of 0*C(32*F) to 980*C (1796*F). The voltage out range is 0v to 4.76v. A voltage of 4.76 is 243 A/D counts. So the resolution is (1796/243) = 7.4 *F per bit. Way more than you need. This gives a safe margin of 5-4.76 = .24mV for the ECM A/D with little loss in resolution.

Nice thing with the ECM is you can LAG Filter it too. You need a 2D lookup table. Look at the AD595 datasheet for the table values (or equation). Overall, I would calibrate it also.

EDIT: When you do the 2D lookup tables, you can make more than one so that it is very accurate in certain ranges that you think are more important than others.

JP86SS

iTrader: (1)

That was my plan.
Still figuring what I want to do. Build or buy then tweak ?
Buying then tweaking would get me closer sooner but may leave me with cal issues. Still thinking on it.

JP86SS

iTrader: (1)

Wimped out and just bought one for $35. I'll never get around to building it until winter when I can't use it anyway
He has added them to his site now. Neat stuff.
Will just do my own cal though so I know it will be close.
Jp

JPrevost

Wouldn't the voltage out for 960*C /2 be 4.91v and not 4.65v? Or is there something going on that I don't understand about those resistors.
Are you sure you can fry an ECM A/D with a littler higher than 5v? I've put more than that into the TPS and never had it do any damage (8746/7747, and a 7730). I would think that the pin on the ecm/pcm would be able to handle it's own regulated 5v give or take a little. You know, just incase a sensor like the MAP or TPS where to go bad.
When you say the circuit isn't linear, I agree. It isn't perfectly linear but considering the resolution of an EGT thermocouple location and the A/D itself I think it's safe to say the linear approach would work rather well. At 980*C it would be off by 215mV, 21.5*C, 70.7*F which does seem like a lot, but it isn't at those high temps. A better formula is 10.25mv/*C . The other added bonus of having it report 10mv/*C is that it errors on the cold side. The circuit reports 1000*C but it's really only 980 . woot, engine saved.
The resistors do need to match. The closer the better.

HaulnA$$

Just FYI. This is from the LM-1 website. It is an AD595 circuit for a thermocouple. Seems simple enough. I thought about building it but just have'nt got around to it. HTH

JP86SS

iTrader: (1)

Back from the dead because I need to fix this.
I am using the above board and a TC from Cole-Parmer (Type K)
Ran signal into 730 "B3" or "E10" read on U6 IIRC.

Problem: All works well until the Thermocouple is inserted into the header.
Then signal goes to crap (Zero/Nada)
Assuming it must be some kind of grounding issue.
I used a "grounded TC" which I thought would be best for fast reaction of the output. I thought hat may be my downfall and require a different TC.
[Edit] If I lay the TC on the floor and hit it with a lighter flame it reads and changes like it should.

Although... If I put the TC in the header and connect my multi-meter to it for TC reading it works fine. Reads correctly just can log that way.
Just my analog out goes away when using the board and connected to the ECM.
I tried adding an OP Amp buffer on the output and it made no difference.
I do have the signal common tied to the power common, figured it would eventually be the same as the ECM common. Operates fine on the bench when setup the same way so I'm at a loss.

Any suggestions welcome.
TIA,
Jp

RBob

iTrader: (1)

Using a grounded probe I would leave the ground at the AD595 to the chromel connection open. This is the thin blade on the TC connector. Not sure how easy it is to check, but it is that side that should be grounded by the probe.

Push comes to shove using an ungrounded probe and grounding the chromel wire at the AD595 would be a way to go. This avoids the ground loop issues that may be an issue.

Or, thought of this: have the AD595 grounds all go to the probe. Tie the chromel wire to the AD595 grounds, with no other grounds going else where. That too will provide a single point ground.

RBob.

JP86SS

iTrader: (1)

Thanks,
That is one I haven't tried.
Will draw that out and experiment to see what happens. Hopefully the signal won't just be "absorbed" somewhere when grounded then because the mV output from the TC will go to the board input and not have a ground referenced at the TC input.

RBob

iTrader: (1)

I just checked out the Cole-Palmer TC probes. When they say it is a grounded probe, it is the junction that is grounded at the tip of the probe. This is done for a faster response. I don't believe that type of probe is going to work.

Need to use a probe with an isolated junction.

RBob.

JP86SS

iTrader: (1)

Update:
Seems that if I do not connect the signal common at all it reads.
There is some noise on the channel I believe because of this and am adding lag filtering routines to the extra inputs I use.

Rbob or anyone who can answer,
Q: If calling the Lag filter routine and it returns a "D" register value, can I just use STAA with an 8 bit address to store the high end integer value?

I don't have free space directly after the existing address and would need to move things around if I can't do that.

RBob

iTrader: (1)

Yes, you can do that. The LSB is the fractional portion of the result. With the MSB being the integer portion of the result.

Can also round up the MSB by testing bit 7 of the LSB.

The only issue is that when you pass the old value to the filter precision is lost. You may want to use 128 (0x80) as the LSB when passing in the old value.

RBob.