Opamp noise

comp.arch.embedded.piclist wrote: ... > What type of caps are the best to use for this. Ceramic, mylar, mica, > tantalum ...?

Hi Jim

Low-Voltage Measurement Techniques:

maj 25, 2007, Using three-point current reversal to reduce error in low resistance/power measurements:

Analysis And Measurement Of Intrinsic Noise In Op Amp Circuits:

For your own sake - build test circuits and measure noise (with and without circuit-like current/voltage) over:

It might be only one problem, that need to be addressed.

regards,

Glenn

... snip ...

I have read only some of the replies you have received, and I suggest you look for parasitic oscillations. These can be low amplitude high frequency, get rectified, and cause the sort of effects you are seeing. Just a possibility. Try putting 100 ohm resistors in series with the opamp input pins.

My transducer inputs are 250 ohm. I placed a 4.7K ohm resistor in series and used a 1nF cap to common. This has helped with the DC level drift, which is now only a few mV. I got a response back saying that I should use 2.2uF instead if I wanted a 15Hz lowpass filter, so I tried that and it made the drift worse. If fact, with the 2.2 uF cap to ground on the inputs, the DC level would shift over 100mV when you moved around the circuit. So I placed the 1nF cap back and got it back to a couple of mV DC level drift. I am content with that, so long as when I commit to an actual circuit board, this won't change. I.E. everything is working fine now, then if I commit to a circuit board, I get the same effect as before, then I am at a lost. I liked the idea of a phase sensitive rectifier and injecting a AC excitation, but I am a bit sketchy on how do to the lock-in circuit.

I think, I may leave the 1nF and 4.7K lowpass on the input and also place a 2.2uF in parrallel with the 1nF and just commit to an actual circuit board.

Also, the protoboard that I have been using is a high quality one, 3M and it does have gold placed slots, so it is the best that one can have for a proto board.

Well, I guess we'll see.

Thanks for all input. All the more reason why I stay digital.

James

Probably the 2.2uF had poor RF performance. For example, an electrolytic or tantalum does almost nothing to short out RF such as FM stations nearby. Or resonates and makes it even worse.

Yep, and probably you'll find out that when you have a good ground plane your noise drops to less than a millivolt.

Wow, we must have really scared you here ;-)

and got it

No, I find this stuff interesting, but I don't have the time to dig deap into the pains of the analog world.

I do like working with sensors, though most already produce above 10mV signals, so I don't need such an agressive gain as with the force transducer.

In the past, I had worked with a K-type thermocoupler and used Analog Devices AD595 and interfaced that to a microchip PIC 18 controller. It worked fine, but one in awhile I would get a readout that was outside of the possible temperature rise profile. I ended up just doing a 128 point averge to keep these out. But I would not be surprised if I had the same problem as I am having with this transducer amplifier.

However it dispoints me that AD would not have better docs on IC specific to that type of app in which that include RF and RFI ellimination techneques.

James

As for the 2.2uF I used mylar, which is the same as I used for the

When you say a force transducer, is this a complete commercial device (if so which one), a strain gauge or a capacitive device? With each of these there are a number of reasons why you may get the sort of variations you indicate with people walking around the set-up (yes I have read your other response to a question).

One of the problems about force measurements is the mechanical mounting itself. This sometimes takes very expert placement and design and the device manufacturer is often the best help here.

The next issue will almost certainly be the configuration of the elements in the measurement circuit. Are you using the sensing elements in a Wheatstone Bridge formation?

Then you need to consider the wiring. I hope you are using twisted pair shielded cable with low noise factors. Also remember to terminate these decently in your system and pay attention to eliminating the earthing loops while maintaining decent signal grounding (things you get to appreciate when your sensors are 150 metres from the electronics and in noisy environments).

It'll be a long learning process. But well worth it. Once you master most of this you'll become a hot commodity because the majority of analog engineers is already retired or soon will.

Most likely someone had a cell phone and it was connecting to a new cell tower at that instant. Or maybe it's a Radar. Sometimes it is best not to average but delete data where the jump is technically impossible and can't be explained. Better yet, ruggedize it against outside influences.

AD caters to analog engineers. They (usually but not always) assume that their customers know all this stuff. I believe they've got some good application notes about EMI/ESD though. But take these with a grain of salt. For example, their frequent recommendation to split ground planes underneath an AD converter is in my opinion wrong. I would never do that.

Whoops. Mylar caps are usually coiled up inside, not good for RF. 1nF requires much less coiling and maybe that's why it worked, sort of. Mylar caps are great for low frequency AC-coupling because they aren't plagued so much by microphonics and other nasty side effects of ceramics. But for shorting out RF you've got to have ceramic caps. Just get yourself a nice quantity of 0.1uF SMT caps. 0603 size, or if your eyes aren't that good anymore 0805. I use coin envelopes for all those SMT parts, then it's always at hand and occupies much less space than the tape snippets they often come in.

For further information on opamp circuits I would suggest reading

Regards

James

e quoted text -

This transducer is a wheat stone bridge configuration. It is from Grass Technologies. The link is

For testing perposes, I have it mounted on a rubber pad so that vibrations in the floor don't interfer with the testing. I do believe that the problem is EMI and RF because if you were to hold your hand over the circuit, and leave it there, the level shift would remain until you move your hand away. I am using twisted pair wire to the transducer, but it is not shielded. The cable is only about 5 inches from the circuit for testing.

I am a bit new to this. So bear with me. One question I have allows wondered, when I would as an electrician years ago, we had to deal with ground loops, but I never understood why we only grounded the shielded cable at one end. Except that maybe the ground potential were different at the two ends. But then the question would be, which end do you ground, the instrument end or the circuit end and second how does having a grounded wire at one end not effect the signal in the wire.

Thanks

quoted text -

Since you wrote in another post that your want to commit to layout I suggest to provide a shield connection. You don't have to use it but I bet later you'll be glad it's there. In really tough environments this will not work well without a shield.

That is because the ground at one side of a large building can be several volts different from the other side. It shouldn't but it happens anyways. During thunderstorms it can get even worse. If you connect a shield from one machine to another this can create a low impedance connection and you might find tens of amps on that shield. Not a safe situation.

I always opt for the side with the sensitive amps in there or where power electronics such as transmitters are. Usually you can use a Y-rated capacitor at the far end. There should also be bleeder resistors. In really tough cases transformer-coupled isolated transmission is going to be the ticket. This is often done with RS485.

I have a question about Ferrite beads. Are they placed in series with the signal lines? I.E. from my transducer, I feed the wire to a single bead and then from the output of the bead to my lowpass filter? Or should the wire wrap around a ferite core. And then, should both wires from the transducer wrap around the same core, in the latter case? In the first case, I assume then that there is a bead for each sensor line. Correct if this area.

Thanks,

James

Typical is one bead per wire (not one per pair), right at the entrance into the enclosure of your ADC unit and not that the far end. No wrap around, just one end in and the other out. This is because a multi-loop will have some inter-winding capacitance and that makes it leaky again.

If you have low frequency noise like from VHF-I band stations multiple loops through a bead can be beneficial though. When using beads on power wires mind the saturation limit. More turns -> saturation at less current. When a ferrite saturates it becomes inefficient, almost as if it isn't there at all. The limit can be found in the datasheet.

Put a dab of soft glue on the beads because otherwise they rattle.

Ok, so that rules out PCM mounted pass ferrite beads. Where can I get these beads were you just pull the wire through it. The ones at DigiKey seem to be of the type where you break you wire and run input to one side of bead and the output comes from the second pin of the bead. I.E. they are like in circuit inductors. However, from what I understand you don't want any wraps, you just want the wire to go through the center of the bead, in which case my original question is where I can find these small beads like this.

James

You can use those, I do that all the time. Just place them really close to the edge of the board and make sure that's near the metal enclosure.

Tons of them, for example:

Even better if you've got the space:

See the links above. You can also get them at places like Amidon:

In the end you'll just have to price them out. Even some of the larger hobby electronics stores may have them. Unfortunately most of those stores have become extinct but luckily there is Digikey, Mouser etc.

If you look for loose beads search for ferrite cores, then select the small ones with IDs of less than 100mils and lots of beads will show up.

ith

er

=A0... Where can I get

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ... The ones at

hat I

For my wheatstone bridge wires, I plan on using Belden 4 conductor shielded. I was planning on having that terminate into a mini-DIN 6 pin connect. On the circuit board I was going to use a surface mount femail mini-DIN connector with metal case shielding. That will be right at the metal case. From there I would place the ferrite beads off of the pins of the SMD mini-DIN and the circuit board.

Anyway that was my original thought. Unless you think that this would be a problem. The mini-DIN is designed for RF applications.

James

ith

er

=A0... Where can I get

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ... The ones at

hat I

This is what I was considering. Not quite done yet, but is a start.

example:

space:

Amidon:

Looks good but place a 0.1uF each parallel to C5, C6 and C7. Really close to the amplifier chip. When you find a rotary switch make sure its shaft is grounded if it is metal, to avoid RF leaking in via the RG lines.

Your CAD symbol isn't showing the ground connection for the Mini-DIN, needs to be fixed. It has to show up in the netlist.

What kind of CAD system is that?

with

n,

ce

oop

in.

le

ower

if

=A0 =A0... Where can I get

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 ... The ones at

what I

p.

I use EDWin XP. The symbol was created with the shield connected to the ground plane automatically, thus it does not show up in the schematic, but will be present in the board layout.

Why put 0.1uf next to C5,C6,C7?

comp.arch.embedded.piclist wrote: ...

...

Hi Jim

I have look on your new circuit. I think it might be a good idea to place a ca. 100 ohm at the op-amp output so optionel capacitive load impedance does not make the op-amp to oscillate. Because of the "high" load impedance you do not need a choke in parallel to the 100 ohm resistor:

Function of Output Inductor:

Use star grounding on the analog side:

Star Grounding in Tube Amplifiers:

The following is only meant as a mild [maybe unnecessary] provocation. PS: I am not a native english speaking person. I am aware that your circuit might "just" be a prototype:

Get rid of the big capacitors (C5-C7). Only two 1nF (C6, C5) should be adequate?

I just thought about this - this is the newsgroup: nntp://comp.arch.embedded

How about using some sort of embedded system - a microcontroller for instance?:

Get rid of the DIP switch and variable resistor. Use a microcontroller (connected/internal) ADC (with auto-ranging), and calibrate with SMD buttons or by other means.

"Install" your circuit right next to the transducer. Solder the tranducer wires directly to the microcontroller board.

Output of the controller could be ethernet with IEEE 803.3af power-over-ethernet connection. Why? No expensive wires or connectors are needed - only standard wires and connectors. Ethernet have built-in transformers to insure galvanic isolation. You (and your costumers) do not need "special" power supplies with special plugs, that in a few years might be hard to buy. Many cheap switches has built-in 802.3af support. You might also just settle with USB. This interface can supply

The microcontroller can be programmed to compute measurement using "three-point current reversal". It will minimize 1/f noise - it could be a problem - I do not know. You of course has to design the appropriate analog circuit side, with the possibility of microcontroller (controlled) voltage/current reversal switching:

Low-Voltage Measurement Techniques:

maj 25, 2007, Using three-point current reversal to reduce error in low resistance/power measurements:

Analysis And Measurement Of Intrinsic Noise In Op Amp Circuits:

You could even measure many times and compute truncated mean or trimmed mean:

Want your circuit to be EMC tolerant?

Somebody might place a mobile phone/access point/PC with wireless network card near your transducer circuit or (analog) wires?

Have a look on this site:

E.g.:

EMC and Signal Integrity (I know this is from a digital point of view):

Mitigating EMI Wireless Interference in a Central Office Environment:

EMI Ten Common EMI Problems:

regards,

Glenn