Solving OP Amp & EMI Problems - Joe's Experience

[chemnut220]

I have a temperature probe that I amplify with an OP amp and the circuit has to sit next to a video transmitter operating at 1280 MHz outputting ~ 800 mW of power.

I can get the circuit about a meter from the transmitter if I absolutely have to but it is going to be in somewhat close proximity.

I am not well versed in this sort of thing so my first design was with the good ol LM741. Well - the stupid thing worked really well and had no EMI problems (none that I could detect).

Then I wanted to operate on a single supply and a lower voltage (rail to rail) so I bought some TL972 opamps (was planning on using 2 stages on the same piece of IC - I can do that with a dual op amp right?).

Well I built a single stage inverting amp with a gain of 10 and it all looked nice until I turned the video transmitter on... not so good... the output voltage swung wildly all over the place. I tried adding capacitors between the in +, in - and virtual ground (everything I had in my box from 10pf to 1000 uf) which helped very little.

So I started reading about op amps and emi which brought me to stuff about EMIRR rating. I am about to order a few TLV2372 which have around -20db rejection in the band I am operating. That is actually kind of poor compared to some of the better EMI hardened opamps that TI has, but I need a pdip package so I can breadboard without strange modification to SMD packages (soldering jumper wires - i just can't bring myself to do it...) and the TLV2372 was the only one offered in this package with any sort of EMIRR rating (that I could find).

Does anybody have any suggestions? I want to operate on voltages between 6-8 (+/- 3-4) and would like to have pDIP packages if possible.

Thanks,


- Joe

 

[berkeman]

I have a temperature probe that I amplify with an OP amp and the circuit has to sit next to a video transmitter operating at 1280 MHz outputting ~ 800 mW of power.

I can get the circuit about a meter from the transmitter if I absolutely have to but it is going to be in somewhat close proximity.

I am not well versed in this sort of thing so my first design was with the good ol LM741. Well - the stupid thing worked really well and had no EMI problems (none that I could detect).

Then I wanted to operate on a single supply and a lower voltage (rail to rail) so I bought some TL972 opamps (was planning on using 2 stages on the same piece of IC - I can do that with a dual op amp right?).

Well I built a single stage inverting amp with a gain of 10 and it all looked nice until I turned the video transmitter on... not so good... the output voltage swung wildly all over the place. I tried adding capacitors between the in +, in - and virtual ground (everything I had in my box from 10pf to 1000 uf) which helped very little.

So I started reading about op amps and emi which brought me to stuff about EMIRR rating. I am about to order a few TLV2372 which have around -20db rejection in the band I am operating. That is actually kind of poor compared to some of the better EMI hardened opamps that TI has, but I need a pdip package so I can breadboard without strange modification to SMD packages (soldering jumper wires - i just can't bring myself to do it...) and the TLV2372 was the only one offered in this package with any sort of EMIRR rating (that I could find).

Does anybody have any suggestions? I want to operate on voltages between 6-8 (+/- 3-4) and would like to have pDIP packages if possible.

Thanks,


- Joe

It sounds like you will need to use some shielding and filtering of the opamp circuit to get this to work. Can you post a schematic of the opamp + sensor circuit? For sensing temperature, you have pretty low bandwidth requirements, right?

It sounds like you are getting rectification of the RF signal in the opamp or its associated circuitry...

 

[chemnut220]

So shielding would probably help, but the only thing I know about shielding is what I have observed in cat6 cable (foil wrapped around the wires with a drain wire going to ground). How would I implement RF shielding on a small opamp circuit? This probably sounds really naive, but the only thing that comes to mind is putting heat shrink around the final circuit, then wrapping the whole thing in aluminum foil and connecting that to the negative battery terminal?

I can try that I guess on breadboard and see what happens.

Also, you mentioned rectification of the RF signal. I will include those terms in my google searching and do a mind dump later today.

I am still puzzled why the LM741 worked so well. There were several key differences. For starters, the 741 was operating at +/- 12 volts and well those to my knowledge are not cmos op amps; and from my reading, cmos op amps are the most susceptible to RF (the TL2372 and most rail to rail op amps I could find on digikey were cmos - even the emi hardened ones).

I attached a schematic of the circuit I am using. In the attached schematic, I show a voltage divider creating a virtual ground, but I have also set it up to work on +/- 3.75 volts as well and found the same sensitivity to RF. Also that circuit is completely isolated (for the time being) from other circuits in my system, I just use a mulitmeter across the virtual ground and Vout to observe what is happening.

Lastly, I replaced the thermocouple with a little voltage divider network that generates a small DC signal similar in amplitude to the thermocouple. I thought maybe the long thermocouple wire was behaving as an antenna, but it did not solve anything...

 

[Enthalpy]

At 1280MHz, it is not the op amp's role to provide rejection. Shielding is what you need, both at the cable and the reading circuit. Then, proximity is not a worry.

More: a thermocouple delivers a ridiculous signal, need a cold reference, and has essentially drawbacks in this use. Why shouldn't you have a semiconductor junction as a temperature sensor, which would give you 2mV/K instead of 20µV/K? Better, this sensor can be embedded in an integrated circuit like the AD592 or a modernized version. It fits in a TO92 that you can glue on the monitored part. Add a few ceramic SMD capacitors locally, and you get 1µA/K ready to be carried away.

 

[Carl Pugh]

Add 10 to 100 uf capacitor from V1 + to ground.
Add 0.1 uf capacitor from V1 + to ground. Place capacitor close to OA1.
Add 10 to 100 uf capacitor from V1 - to ground.
Add 0.1 uf capacitor from V1 - to ground. Place capacitor close to OA1.

I agree you may be getting rectification from the RF signal.
Add 0.1 uf in parallel with R2.
Add 1 Kohm resistor in series with OA1 +input.
Add 0.1 uf from OA1 + input to ground.

I also agree that if lower temperatures are being measured, a intergrated-circuit temperature sensor would be prefered. Have used the LM35DT with good results.

 

[Carl Pugh]

If this is battery operated, you might try wrapping the whole thing in aluminum foil. The aluminum foil that you buy in a grocery store works well.

 

[Carl Pugh]

After thinking about it there is also another problem.
The ground tap for V1 is two 10 kohm resistors. R2 is also 10 kohm. The ground will bounce around when there is a thermocouple input.
R2 should have much higher resistance than R3 & R4, maybe 1000 times higher.
I believe that TI has an IC that provides a ground rather than using R3 and R4. There is also a circuit using an operational amplifier IC that will provide a ground. The advantage of these circuits is that they use much less power than R3 and R4.

 

[AlephZero]

More: a thermocouple delivers a ridiculous signal, need a cold reference, and has essentially drawbacks in this use. Why shouldn't you have a semiconductor junction as a temperature sensor, which would give you 2mV/K instead of 20µV/K?

Semiconductor sensors only cover a small temperature range (about -50C to +150C). We don't know what the temperatures the OP is measuring or what environment the sensor is in.

 

[chemnut220]

Ok guys, a lot of good input and suggestions so thanks a bunch for that.

I really do need a thermocouple. It is an exhaust gas temp (EGT) sensor. I actually don't know the exact EGT value, its just being converted to an analog voltage for either the end user to have as a relative tuning parameter, or for an analog input into a micro controller. (I anticipate actual values in the 500-800+ deg C range).

I actually can and will get rid of R3 and R4 completely because the amp is being powered off a 2 cell lithium battery (LiFe). These batteries have balance plugs that give me the option of +/- 3.xx volts and an actual 0v reference. I just drew the resistor network so I could show someone in another forum how to wire an opamp on a battery.

I tried wrapping the circuit on breadboard in foil but that did not help. I have sense gotten rid of the breadboard and just soldered stuff to a 8 pin socket. I will try the aluminum shield again - this time I will try some ground wires from the shield.

I will put the stuff back on breadboard and try the specific suggestion by Carl.

Thanks a bunch guys,

- I will report back if I have success. I am also throwing money at the problem too lol. I just ordered a few different op amps to try; one of them is supposed to be RF resistant. I can't help myself, I am a chemist by training and often times we just throw a bunch of crap to see what sticks...

 

[yungman]

It is not likely to be the circuit design or opamp you choose that is the problem. Very likely is the layout and the grounding that is causing you the problem. So take a picture of your circuit.

I use CMOS, bjt opamp, I mix high speed digital signal with low level analog signal all the time. It's all about layout. Your circuit do miss some filter cap.

1) Depend on the impedance of the thermal couple and the frequency that you are looking for ( I take it is very slow as you use thermal couple.). I would put a cap from +ve input to ground right at the input.

2) You ABSOLUTELY need to have a ground plane for the circuit. Or else, all bets are off.

3) I don't see 0.01uF cap from supply input of the opamp to ground. You are going to need that.

4) If you use R3 and R4 to set up the ground reference, you better have bypass cap at the junction between R3 and R4 to ground.

5) If your thermal couple is connected to the pcb with wires, you better use coax and have the shield solder to the ground.

This is RF, you have to have all the RF precaution technique. Take pictures at few angle so we can look at before we talk. There is no further point of talking electronics until there is some pictures.

With good RF layout, simple circuit like this works. Devil is all in the pesty details. Never have to worry about RF susceptibility of the opamp itself yet. AND I use TL082 all the time.

 

[Enthalpy]

Exit the semiconductor sensor, OK. Thermocouple or platinum resistor.

Layout more than schematic, sure. And to filter out 1280MHz the capacitance shouldn't be too big: what you need is a low inductance, so take SMD capacitors. Cable on a ground plane.

Since good shielding of a thermocouple cable is unlikely, you should filter brutally the signal as it enters the board. In addition to SMD ceramic capacitors (1nF suffices) I'd put series inductors in a C-L-C pi arrangement. Just buy some SMD inductors that resonate over 1280MHz; they exist for cell phones. Or use resistors, in a C-R-C pi arrangement, but this will increase the thermal noise.

 

[yungman]

We need pictures, one picture say a 1000 words. I always trouble shoot EMI by eye-balling than talk out of thin air.:rofl:

Case in point. I was doing CE test one time and the system failed. We trouble shoot down to a cable that had plastic connector. It was a shielded cable that the shield connected to a pin and the pin was grounded right onto metal panel inside the metal box. Everything looked good, we went around and around and spent 3 hours with no luck. Finally I told the engineer to open the plastic connector of the cable. We found the shield was connected to the pin, BUT the assembler was lazy, he/she left a 3" wire connecting from the shield to the pin! #$%&! Shorting the wire, passed the test right on the spot!

One picture speak a 1000 words.