AD8429 as thermocouple amplifier

Trying to build a thermocouple amplifier with AD8421/8429 in-amp.

The thermocouple is 0.0005" diameter type K which will be working in non-linear range, and that's why we could not use any existing thermocouple amp. It has a resistance around 1300ohm: 800ohm on positive arm, and 500ohm on the negative side. Only the joint potion of the thermocouple will be subjected to temperature change.

The proposed solution is to use AD8421/8429 in-amp with a gain of 100. then the output is captured by NI X-6366 simutaneous DAQ differential analog inputs, which have another set of in-amps. Desired sampling rate will be 50kHz+. For mobility reason the amp is battery-powered, and the reference voltage is set as the mid point of the battery packs (two independent regulated battery packs, +/-5V).

While let the thermocouple floating or connect the IN- pole directly to the reference, the amp is working but accuracy might be questionable considering the bias current. Multisim said so and I have not tested it out. By connecting either one or both of the IN-/+ poles to reference with 10Mohm resistor (as stated in the data sheet and other sources) will disable the amp. The output does not change from zero.

Q1\ Considering the amp part, what's the right way to create current return parth in this case? Or, may the error appear to be a constant zero-drift? Then it can be processed later in software.

Q2\ For RFI suppression, I would use the resistance of the thermocouple as part of the single stage low pass filtering. Thus only capacitors are needed. Is this correct? Is there any guidelines I should take in choosing the type of these capacitors (~200pF)?

Q3\ Considering the DAQ side, AI+ connects to AD8429 OUT, AI- connects to AD8429 REF. Then AI+ and AI- are connected to AGND with a 10Mohm resistor, respectively. Is this setup correct? Or should I ground the AD8429 REF? Or should I take the AD8429 as a referenced source? I tried all setup and observed no difference in output.

Q4\ For the focus of tracking fast temperature change, which one among the AD8421/AD8429 appears to be a more suitable unit in this case? May the low bias current of AD8421 minimize the problem in Q1?

Thanks for your help!

  • Hello neeviour,

    For Q1, can you post a screenshot of your schematic? I am not sure I follow your description...

    Q2. You can't use the resistance of the thermocouple, you need a real RFI filter, especially if the thermocouple is long (say, more than 30cm). That means, you need to place some real resistors.

    Q3. The reference pin MUST be connected to a low-impedance return. In this case, I think you want to connect it to AGND. By the way, the NI DAQ boards can implement different input configurations, such as NRSE, RSE, Differential, etc... which one are you trying to use? Please refer to their documentation.

    Q4. How fast are the temperature changes you are trying to track? What accuracy are you trying to achieve? For now, I am going to say that you really want to use AD8421, because it has better DC performance.

    If you can provide answers to some of my questions above, that should make it easier to understand your issue.

    Regards,

    Gustavo

  • Thank you so much for your reply. And sorry that I did not attach the schematic at beginning. I thought it was quite simple.

    Design 1.

    Taking V3 as the thermocouple, the above shematic is what suggested by the data sheet of AD8429.

    Design 2

    Design 3

    Above two are some other solutions.

    Initially, I found none of them working. The only working scenario is for the first design, set R4=0, which is another commonly seen scenario. But, I do observe signal drifting. So part of Q1, is this drifting a constant value?

    I did some investigation and found that the values of R3/R4 do matter. With +/-5V supply, R3/R4 must not exceed around 2.2Mohm. If the supply increases to +/-6V, that marginal value would turn to around 3Mohm. On the other side, an in-amp (TC913A, which is not capable of high gain and high frequency) with 150pA bias current can work fine with same setup at any given supply voltage. Thus I am suspecting the AD8429 will work fine with R3/R4=10Mohm only under +/-18V for the design 1 and design 2. And the critical issue IS the relatively large bias current of AD8429. In general, if I set R3/R4 to 1Mohm, all designs worked fine.

    And I would prefer design 3 as it rails the amp when the thermocouple breaks, a nice feature for really thin thermocouple. With design 3, the marginal value of R3/R4 under +/-5V supply would be around 4-5Mohm.

    That's all what I found for Q1. If this is true, I would suggest some notification in the data sheet of AD8429.

    Q2, The thermocouple is around 20cm in length. It is very difficult to shield it or coat it (surface tension of most coating/paint at such diameter will form bubbles instead of a layer). So I guess we do need to use RFI filter. Thanks for your info. Does the type of capacitor matters? Stacked film or ceramic?

    Q3. I can use the DAQ under any configuration. Taking the thermocouple and the battery powered amp as a floating source, the REF return is to the battery, not the DAQ side, right?

    Q4. Thermocouple signal will be above 10kHz but not to 100kHz (to be tested now). Accuracy is not a major issue, +/-5C or something like that will be fine enough. I may try AD8421 later.

  • Hello Neeviour,

    From your description, your circuit sounded very simple indeed, but sometimes there is a little detail that is overlooked...

    - For your simulation to work, you need to provide a reference for your circuit. SPICE needs a "node zero" which we usually refer to as ground. Add a ground symbol to your supply common or label it as "0".

    - You are correct to point out that the bias current may induce offsets in your case, this is why I recommended AD8421 over AD8429 for your application.

    - Most people designing thermocouple front-ends like something like your third option for the same reasons you mentioned.

    - Q2. The type of capacitor does matter if you want to be really accurate. The dielectric matters a lot. Stay away from high dielectric types, and pick the largest voltage raiting you can. As a rule of thumb, COG/NP0 are good choices. Plastic films vary in characteristics, and I have come across COG caps that are better than some plastic films. Look for polypropilene if you go for a plastic film.

    - Q3. REF is the terminal to which your output voltage is referenced, and it needs to be connected to your ground. If you don't have anything else on your board but the batteries and the amplifier, you should be safe to do what you are saying. And in that case, I would use RSE mode.

    - Q4. Just be midnful of the RFI filter choices then. You don't want to limit your bandwidth before you reach the amplifier.

    For more information on how to design thermocouple interfaces, take a look at a series of videos that Matt Duff recorded on the topic, I think you will find them quite useful.

    http://www.youtube.com/watch?v=Xp7ZNAc9Fis

    Good luck!

    Gustavo

  • Thank you so much!

    Just to explain, the schematics are done with multisim just to explain the circuit. The descriptions above for Q1 are all from real circuit. When simulating, I do refer to ground.

  • A further question here, about the resistance of thermocouple.

    In normal cases, designers would take the thermocouple as short, taht is fine for TCs with relatively large diameter.

    The potential generated by a TC is between the hot joint and the cold point - which is not necessarily the "cold junction". The cold point might be just a few millimeters away from the hot joint. As long as there is no temperature difference between the rest part of the wire, there will be almost no potential on it.

    Like the case above, the long leads will act merely as a resistor (a unfolded wirewoud wirewound resistor?). In my case, the resistors will be as large as a few hundred ohms.

    Now I need to set the RC filter parameters, and it matters a lot. Why I should not consider these resistance? Or maybe I should make the TC as short as possible, and use discrete resistors. But for the filter part, what's the difference?