Opamp noise

There are some electronics groups on USENET, you might try those as well.

In general, we'd need a lot more information to even begin to help.

How about:

a)A description of the transducer.

b)Several schematic diagrams, a baseline plus a drawing of what you tried to do to remedy the problem, one drawing for each attempt.

c)A photo of the physical circuit, the transducer, the wiring, the power supply, etc.

d)More information about what happens when someone comes near the arrangement (does the output level go up and stay up indefinitely?).

What most people would do with a problem of this complexity is devote a web page to it somewhere and post a link.

I can't even begin to help until I see the actual op-amp circuit ...

The Lizard.

to

eb

I appreciate your response. I will try to get this information sometime today. Is it possible to post info on this newsgroup? I thought you couldn't do that.

Also, when someone walks and just stands there, the level remains. If I put my hand on the oscope wire, and hold it, the level stays. Normally, I would think this is just my electric voltage, but I would think that on a diff configuration that would be common mode which would not get amplified. I guess one more piece of info is that this is on a proto board at the moment. I pref to do this before I commit to an actual circuit board.

James

to

eb

Also, the transducer is a Grass Instruments force transducer # FT03.

Most people have access to some sort of website where they can put the information up. Then they put the link here on the newsgroup.

For example, let's assume I messed up a motorcycle valve adjustment and wanted to shame myself:

That is why I'd want to get a look at the schematic. That is a very odd effect.

When op-amp circuits are too responsive, there are standard approaches to numb down the frequency response. But the effect you've described ... I've not seen it before.

The Lizard.

That has to be a RF field effect, in other words, a RF susceptibility. You do have a gain of 1000, so 50mv becomes 50uv of IP level equiv.

Short the pins, right at the opamp, and remove any other leads, and run some tests, then slowly add more lead-length. RF susceptibility is mostly poorly documented, and varies widely with Opamp families.

-jg

This is what I am thinking, but how can I shield my circuit against this. Keep in mind, that I am using 2 9V batteries for my power source so I know it isn't a power supply problem. Also remembering that this is currently on a protoboard, I have earth grounded the protoboard with the oscope with the reference rail of the op amp. I have also tried to put a sheet of aluminum foil over the circuit and also grounded that, but I still get the same effect. If you move around the room, the mean level changes 50mV or so. I will put together some pics from my scope and circuit and post them by tomorrow. I have never had to work with transducers that work in the uV range before, so I have not seen this kind of problem before.

James

One possibility is that the op-amp is oscillating and you are getting "theremin" type effects. Micropower chips are more prone to that sort of thing. Do you have your 'scope set to BW limit?

It's also possible that the front end of the amplifier is acting as a detector for RF signals.

You need proper filtering and construction techniques to get useful results with G=1000, a plugboard will not cut it. It's possible to prototype with a ground plane and some PCB material but you have to be a bit careful.

Best regards, Spehro Pefhany

"The Journey is the reward"

eff.com- Hide quoted text -

Filtering on the front end will elliminate my DC gain. I have thought of filtering, but then I lost my DC steady state force and only get the force changes. Is it possible to use a two or three stage amp to elliminate the unwanted signal. Also, I don't see how this is a bigger problem when I get a DC level increase. I and going to post my scope reading in a little bit. Also, I do have the scope on bandwidth limit of 20MHz.

Not if you do it right :)

Google Low Pass filter. or

You could also look at ferrite beads and series resistors. Normally in such sensor application, you filter as agressively as your system response / reading rates allow.

-jg

I'll second that, since you mentioned that the same problem happens with shorted inputs.

Aha, here we get closer to the matter. Circuits this sensitive will hardly work without a proper layout that contains a full ground plane. Or at least a prototype board with a full ground plane and built by someone who knows EMI really well.

If you can find one, try a pure CMOS amplifier. The RF susceptibility that Jim mentioned above has its roots mainly in the fact that most opamp inputs are bipolar. There is always a BE junction and that rectifies RF from cell phones, cell towers, strong AM/FM/TV stations, pager transmitters and so on. The rectified RF is basically DC until something moves, such as a person walking around. A human body that moves can make the EMI amplitude vary.

If you must use a BJT opamp things will get tough. I had to deal with such as case earlier this year and that client really needed a relayout, we couldn't avoid that. We placed capacitors between the opamp inputs and to ground, set a few good LP filters up front, bypassed the rails super-good, and at the end the noise was gone. Shielding will help but only if you have a solid ground plane, a fully grounded metal chassis (no anodized aluminum) and filter anything that enters or leaves this shielded cavity.

BTW, you cannot post schematic attachments and most binary groups have shriveled up because some overzealous AG has ... well, you know. Got to use a real web page or some free (but honest) upload service.

So, back to the Weber barbeque now, Thanksgiving dinner is cooking in there :-)

Hide quoted text -

Low pass filtering will not affect the DC gain.

Sounds like high pass filtering. Not what you need.

Not really.

Best regards, Spehro Pefhany

is

ail

the

=A0"The Journey is the reward"

://

.speff.com-Hide quoted text -

"The Journey is the reward"

eff.com- Hide quoted text -

Ok, here is at link to scope readouts and my circuit. In my circuit, I would need a low pass response of 15 Hz or less should be sufficient, so anything above that should not get amplied even if it is micro Volts signals. I am wanting 0-15Hz response coming from my transducer. Also, then INA 126 that I am using is an instrumentation amp, not an op amp so I can't easily put capacitor in the feedback to attenuate the high frequencies, so I am at a loss here. I have not dealt with instrustrument amps before. Only opamps.

I will take any suggestions at this point. Normally I am a digital person, not analog. But I will certainly try something new, as I have done.

Heres the link.

Thanks

You probably want bypass cap(s) near the power supply input to the amplifier chip. Electrolytics won't do--you need something with good frequency response.

Scan the manufacturer's sheet for a suitable value and multiply it by 5.

The power supply bypass cap(s) should be as physically close to the amplifier as possible.

I didn't realize until I read the data sheet that you were using a chip that is not an op-amp. I would try one of the following two approaches.

a)Use RC filters on both inputs wires (if the transducer will support it) to numb the frequency response down to a few Hz if that will work for your application.

b)Change to a dual op-amp design or perhaps a single op-amp design so you can numb down the frequency response of the amp. Looks like the chip you're using won't allow that.

I'm not an analog guy either. Your design makes me nervous because of the regulator feeding from the same supply as the amplifier ... I get the feeling there might be a feedback path between the transducer and the amp.

I hope someone gives you competent help ...

Hide quoted text -

Seriously, you can't build sensitive circuitry like this on solderless breadboard. I suggest to use a piece of copper-clad (a chunk of single sided FR4 or phenolic) and prototype on that. "Dead bug style" would be to glue the chips inverted onto the copper clad, "living bug" requires little spacers but then the order of pins is not inverted, easier to handle. Place chip 0.1uF caps right at the supply pins.

It looks like you are getting FM stations into your circuit. Theoretically you could cut up a metal cookie can, drill holes and place ferrite beads on every (single) wire that leave, right at the can. But it's much faster to build this prototype correctly, on copper clad. Plus you'd have to empty that can, too many calories ...

Also, shield the cables going to the bridge, for example by using a piece of shielded CAT5 cable.

Welcome to the ghastly world of analog :-)

Anyhow, the prime rib on the barbeque is almost ready. Got to go.

I have to agree with Joerg. These solderless breadboards are okay for uncritical AF circuits. For anything more critical use another prototype method.

Here's a alternative to the dead bug style. It's called manhatten prototyping and works very well if you have to deal with signals in the RF-range.

ier

hat

to

ou're

the

The 5V regulator is there because the transducer does not allow more than 8V max. It is a 350 ohm wheat bridge transducer. I can certainly put another battery in the circuit to provide the the voltage for the transducer.

Ok, I tried a low pass on the Diff input, the bridge resistance is 250 ohm so, I put a 100K in series with each input and a 0.1micro farad mylar cap to ground. I did this for both the - and + inputs to the instrument amp. Now, the level can change as much as 130mV on the DC level when you move around. So it made it worse. I am beginning to think that it would be better to use to unity gain op amps with active lowpass filters and have the output of those feed the instrument op amp with a gain of 1000. Would this be a better approach?

James

s.

ng is

vel

gh

rail

r the

am

es

is

ove

be

=A0 =A0"The Journey is the reward"

tp://

ww.speff.com-Hidequoted text -

my

dth

=A0"The Journey is the reward"

://

.speff.com-Hide quoted text -

The only reason I built this on the protoboard was to make sure I would get what my PSice showed, but as you have mentioned the protoboard is giving me more problems. I guess the bigger question is, with the circuit that I have drawn, if I have a print ciruit board made for it with a copper ground plane on the bottom, will I still get the same problem, or will it go away. I don't need any of the AC and RF, I just want the DC response which is < 15 ro 20 Hz -/+ DC from my force transducer. Or, am I better off, using a 2 op amp unity gain buffer which has lowpass filters on the input and feedback and then have the output feed into the instrument amplifier with the 1000 gain. This is new territory for me, so I don't know.

Thanks for all input.

James

What Jeorg and Spehro said, but:

See if you can AC couple the force transducer. If it's a resistive bridge you may be able to feed it with a square wave or a sine wave, amplify that a bunch through an AC coupled amplifier, rectify it (either synchronously with something like a 4051, with a precision rectifier, or in software after sampling the AC signal), then finally finish up with some DC gain.

The up side is that you sidestep a whole bunch of nasty low-frequency and DC phenomena; the down side is that it's harder to get a reliable gain with an AC circuit.

With more than one opamp, you have more RF detector points :)

What can help, is to split the gain to (say) 30:30, (or 20:50 etc) rather than 1000 in one chunk. Then, any RF effects, which tend to be Vos related on the first detecting stage, are reduced, and you can place more LPF between the amps. Generally, a small PCB right at the pressure sensor is the best, and send either a voltage, or current to a second stage.

-jg