Hi,

I have created a 3 chip instrumentation amplifier using an LM324 Op-Amp for a project. I have simulated the circuit in Circuit Lab using a differential DC input voltage of 3.5mV. The gain is approx. 982 and I have split this up providing roughly 18 in the amplifier and 54 in the first stage.
The circuit works as expected in simulation, however I have built this on the breadboard and the only way I can get it to work is by connecting the virtual ground of the 9.1V Zener split rail supply to the common or zero volts on a separate split rail bench power supply.

This poses an issue, as the idea of the Zener split rail supply was to provide stable voltages from two 9 Volt batteries connected in series, so the transducer could be placed elsewhere and powered from a separate supply.
The only connection between the simulated transducer circuit and the instrumentation amplifier was to be the two sense wires from the transducer to the IA.

I have provided a link to my Circuit Lab model for further information, and attached a picture of the circuit.

I am hoping to get some information as to what I could try, or perhaps some direction and guidance in relation to what is not immediately obvious to some my problem.

https://www.circuitlab.com/circuit/7jn33336vqnk/ia_with-zenner_circuit/

Also, there is a slight difference between the finished build and the simulated, which is the inclusion of an output offset adjustment circuit, which uses a non-inverting buffer and 10Kohm potentiometer. The output of the buffer is connected to the other side of R13, so I have removed the ground reference. However, I don't believe this contributes to the issue I have described.

Regards,
Dan

 

hi dan,
The INA LM324 inputs must have a DC path to 0V.
E

 

the only way I can get it to work is by connecting the virtual ground of the 9.1V Zener split rail supply to the common or zero volts on a separate split rail bench power supply.

As Eric noted, that circuit cannot operate with two separate grounds.
There must be a ground path from the signal ground to the circuit ground.
An instrumentation amp can detect a differential voltage that has a common-mode voltage above/below ground, but that voltage must still be reverenced to a common ground.

For true ground isolation you would need a circuit designed for that, which typically uses optical or magnetic coupling isolation.

 

Hi Eric, Crutschow
Thanks for the reply.

Yes, I just looked trough some books and have come across some information which mentions what you both have.

It appears I was trying to do something which cannot be done with my current arrangement. That's ok, as I thought by connecting the reference I was cheating.

 

The LM324 is very noisy, produces crossover distortion and has a poor slew rate that cuts signals above 3kHz.
Use an audio quad opamp instead.

 

hi dan,
This is what LTSpice shows for your version of the LM324 INA.

E

 

The LM324 can have an offset voltage of 3mV, so the output can have an error of (3mV*982+3mV*18) which is 3 Volts.

 

Hi,
As I have repeatedly posted, in the real World this is not a problem, it is simply trimmed out, as is the Span adjustment to correct for component tolerances.
E

 

Hi,
As I have repeatedly posted, in the real World this is not a problem, it is simply trimmed out, as is the Span adjustment to correct for component tolerances.
E

Indeed you have, but I suspect that @Danlar81 hasn't come across the previous posts where you have mentioned it. Otherwise there would be a preset in his circuit diagram for exactly that purpose.
The real world doesn't have time to be messing about with presets - the real world uses a better op-amp.
Also, it was well worth @Audioguru again mentioning (as he always does) that the LM324 is very noisy

 

hi Ian,
It is the same 'Dan' with whom we discussed this point some time ago.
E
https://forum.allaboutcircuits.com/threads/instrumentation-amplifier.195354/post-1840767


PS: As this a Homework assignment, I suspect it one of those cases where the TS has to use the devices specified by his Tutor?

 

Hi All,
Yes it is the same Dan and as Eric has pointed out, the LM324 is specified by the lecturer.

The purpose of this task is to identify the errors associated with instrumentation amplifiers and account/adjust for them and verify limitations specified in the data sheet. These are important parameters to gain an understanding of, so the IA can be characterized.

"The real world doesn't have time to be messing about with presets - the real world uses a better op-amp." Ian I understand your point, but isn't it a better approach to see and experience these issues and learn how to fix or, look at a datasheet and understand why you choose some devices as opposed to others? What would I or others learn by using a device first up that accounts for these errors?

Using specific instrumentation amplifier IC's have parameters that mask errors and limitations of the LM324, and does not provide any avenues for further calculation or correction, and hence does not allow us to understand and appreciate the errors associated with the device and how to account for them.

In Circuit Lab there was no issue in relation to the output offset due to it being so small, approx. 45.16 micoVolts. However as you have mentioned Ian, it is a problem in the real world which I have accounted for in my build circuit with offset adjustment. My real word measured offset with the inputs shorted was -1.126 Volts.

With all this being said, the simulation gives and output of 3.773 Volts, and the built circuit output is 3.805 Volts, however temperature is certainly an influence on the delta.

Please see the attached, as it is an updated schematic which accounts for the very small Circuit Lab output offset.

I have also attached a link to the updated simulation.

https://www.circuitlab.com/circuit/vzyj5ghs73pg/ia_offset_adj/

Also, I decided to decrease RG from 10Kohm to 9.1Kohm so I could have a gain which is slightly over 1000 as opposed to slightly under.

The next task is to introduce various noise sources such as magnetic, capacitive, non-electric to gain an appreciation of EMI and how to protect against it.

Ultimately, this is a learning exercise. The long term intent is to churn out engineers that understand this phenomena and to construct robust designs that account and minimize the effects.

 

Hi Eric,
Based on the response it looks ok until about 3kHz.
I am going to do the same in Circuit Lab with the updated version. Thanks for showing that.

Should get interesting when I introduce my noise sources such as a 230VAC power drill, blender, PC with case removed and mobile phone. I will be using a long sense cable to put in parallel with these sources. Then I have to hand make twisted pair cables, shielded cables and verify the AC results.

I will post my findings once I have performed them.

 

"The real world doesn't have time to be messing about with presets - the real world uses a better op-amp." Ian I understand your point,

Hi Dan,
Consider that you were producing a quantity of these INA's for for different applications, you will find that variations in component parameters and whatever sensor the client connects to the INA inputs will produce different output results, also, if the INA's need circuit repairs during service, some method of Zero and Span adjustment is essential.

I am talking from practical experience.

E

 

Based on the response it looks ok until about 3kHz.

Hi @Danlar81
Recheck the -3dB point on post #6, the Red solid line.
E
Added:

 

It would have been useful to have known from the outset whether this was a question about the best way of designing an instrumentation amplifier with high dc gain, or a question about evaluating the performance of an LM324.

 

Hi Eric,
I have a question for you if you don't mind which is related to the EMI testing I have been doing.

I have performed testing with two single core cables as the sense wires first. I then had to hand twist those wires to create a twisted pair cable and re-test. I tried to make sure the twist was every 100mm or so, but I am not a machine. After this test the twisted pair cable needed a shield, so I wrapped it in aluminum foil and re-tested. The length is approx. 2 meters.

The tests required, were to measure the AC ambient noise at the amplifier output, and then measure the the output once more when a noise source was introduced, such as a drill, radio etc.

In any case, I have obtained the following ambient values.
Two Single Core Cables - 17mV to 22.1mV
Hand Made Twisted Pair - 8.2mV to 12.7mV
Hand Made Shielded Twisted Pair - 6mV to >100mv

I was expecting a much smaller and consistent noise value on the Hand Made Shielded Twisted Pair cable, but what I have observed is the value moves all around depending on where I am, so it's as if it's been affected by my bodies capacitance?

The configuration I have used is attached.

I have read for an isolated sensor the shield can be connected to the IA differential pair (-ve input). I have tried connecting the (-ve input) to the shield, which is connected to common zero volts, considering I have a +/- 9 Volt split rail power supply, and not connecting the (-ve input) to the shield. Connected or not, I don't see a noticeable difference in the AC ambient noise output.



I have also read that SCIN (Shield Current Induced Noise) can occur due to an imbalance in the magnetic coupling between the shield and the two signal conductors.

Considering this is a hand made shielded twisted pair cable, I was wondering if the AC noise value I am observing is directly related to my cable? If it is, then perhaps I just have to live with the ambient AC values, even though they should be lower than the other two cable types assuming it was a manufactured shielded twisted pair cable?

I wish I had a manufactured shielded twisted pair cable to compare, but I do not.

Anyway, I'm interested to hear your, or anyone else's thoughts and opinions on what I have described.

Regards.

 

Hi dan,
Redo the twisted/screened pair test, with the screen connected only at the OPA end

BTW:
To make twisted pairs, cut the single wires [ over length].
Put the end of the wire pair in a bench vice, and the other end in the chuck of a handle drill.
Turn the drill drill slowly to give you about 5 turns per inch.

You can also add resistive/capacitive coupling of the twisted pair at the OPA inputs.

What is the maximum frequency rate that you want to take actual measurements of the bridge signal?
eg: every second , minute etc ..?

Will the final OPA output voltage be connected to an ADC???

Is there any electrical noise on the bridge voltage supply?
E


Updated: Added link.
https://appmeas.co.uk/resources/pressure-measurement-notes/how-do-you-avoid-noise-in-sensor-cables/

 

Hi Eric,
I won't be able to get that twist rate with the cable I am using as the conductor is 1mm^2 and the insulation diameter is 4mm. I would be lucky to get 2 turns per inch. Do you think I should try smaller gauge/diameter cable?

With the circuit, it is all on a breadboard, simulated transducer also, however it is powered from two 9 Volt batteries to provide my split rail and the OPA is powered by the bench power supply. I only connected the virtual ground of my sensor circuit to the shield so it would have a return path to the bench power supply common reference zero volts.
Do you think it's ok to do that? I would have done it that way if the transducer circuit was removed from the breadboard, so I'm just trying to simulate that.
Originally I did have the transducer shield floating as you mentioned, and the virtual ground of the transducer circuit was connected directly to the common reference of the OPA. At the OPA end the shield was connected to common zero volts.
That probably sounded confusing so I have attached a mock up.

When you say resistive/capacitive coupling could you describe that further? Do you mean adding either a resistor or capacitor from each of the sense inputs to common zero volts?

The measurement rate is not very specific in the sense that I connect the circuit wait for the values to settle a little and take the measurement. I just do the same when the noise source is introduced.

No the OPA will not be connected to an ADC.

There is 0.3mV of AC noise on both bench supply rails.

Regards

 

Hi Dan.
I use 7/0.2 stranded copper wire, it' more flexible.
Look at this PDF Section: 5-21 Figs 5-24 and 5-27

Do you have a simplified up-to-date diagram you could post?

E

 

Hi Eric,
I just put it up, Sorry I forgot to attach it before.

The cable is not ideal, but it's what I had lying around.

I will have a look at the pdf.