op amp buffered voltage reference question.

[EnzoF]

Hello!

Sorry for a newbie question.

I need to make 7 separated voltage references to use as virtual grounds to provide a center point for ac signals from piezo sensors. I have limited board space on the PCB and I did this:



Will this work or is there some kind of problem with this?

I need reasonably clean virtual grounds and I can't use only one because the circuits have to be separated from one another.

Thanks

[razvan784]

You shouldn't need to use two separate dividers. You also shouldn't need two consecutive buffers.
What you need is a capacitor (start with 100 nF) in parallel with R2. R1 and R2 can be increased to 10k.
If you want to be "extra sure" no crosstalk happens between the channels, add a 10k resistor between the divider output and each opamp + input.
Also, if crosstalk is an issue, the opamps should be individual (not 2 or 4 in the same IC) and each should have a decoupling capacitor (100nF) between its supply pins, physically close to the chip, and maybe a low (100 ohms) resistor in series with the positive supply.

[EnzoF]

You shouldn't need to use two separate dividers. You also shouldn't need two consecutive buffers.
What you need is a capacitor (start with 100 nF) in parallel with R2. R1 and R2 can be increased to 10k.
If you want to be "extra sure" no crosstalk happens between the channels, add a 10k resistor between the divider output and each opamp + input.
Also, if crosstalk is an issue, the opamps should be individual (not 2 or 4 in the same IC) and each should have a decoupling capacitor (100nF) between its supply pins, physically close to the chip, and maybe a low (100 ohms) resistor in series with the positive supply.

This schematic was just to show how it ended up on my PCB. The PCB is quite busy and that was the most convenient way to do it. I want the dividers close to the op amps and the op amps are not close to each other so that's why I used two. May not make a difference but I noticed that higher impedance traces picked up 60hz hum so that why I used two dividers.

I already have 100nF capacitors on the dividers. 10k was too much noise, 1k - 5k worked best on my breadboard circuit so went with 5k.

Anyway I discovered that my circuit can be improved to avoid having to use separate voltage references.

The points you made about the 100ohm on the op amp supply and the resistor before the positive input is something I will have to try. Will a smaller resistor work on the non inverting inputs? Say 5k? I'm a bit concerned about noise.

[razvan784]

The points you made about the 100ohm on the op amp supply and the resistor before the positive input is something I will have to try. Will a smaller resistor work on the non inverting inputs? Say 5k? I'm a bit concerned about noise.

Sure, 5k should be fine, even 1k. If the PCB is already built and works fine, none of these should be needed. Same with the 100 ohm supply filters.

[Jan Audio]

Please show why it has to be isolated, thanks.

[Terry Bites]

Its an odd way to go about it. Generating Vref from the power supply using a divider will be noisy. You only want a cap on the lower resistor or you'll throw away the low pass filtering.
Go for better Vref source with low output impedance. Re route your pcb and make vref a priority signal- are you setting ref for a heap of comparators I wonder?
You guys and your DARPA projects, ha!

[EnzoF]

I'm making a PCB for 10 piezo sensors. I need a voltage reference of 2.3V to serve as the virtual ground. Didn't find one, used a resistor divider which works adequate. I'm using a teensy microcontroller and the ADC is not the cleanest anyway so a voltage divider works well enough.

I made the circuit on a breadboard and it worked really well but on the PCB it didn't. It was because the circuits interacted with one another through the ground plane and voltage reference.

I changed my circuit, I don't think I will need 10 separate voltage references for this one.

Will this work the same if I make 10 of them and connect them all to the same voltage reference?

[EnzoF]

It's a balanced charge amplifier that will cancel out common mode noise, especially 60hz hum which is far bigger than any resistor divider noise after some digital filtering.

[TimFox]

Your circuit to generate multiple ground-replacing reference voltages is rather complex.
Would it not be simpler to use an appropriate split power supply?
"Real ground" tends to be cleaner than buffer outputs.

[EnzoF]

Your circuit to generate multiple ground-replacing reference voltages is rather complex.
Would it not be simpler to use an appropriate split power supply?
"Real ground" tends to be cleaner than buffer outputs.

It's not by choice. The microcontroller only has +3.3V and ground.

Also Did you read the post above? I don't think I need multiple ground references with the new circuit posted.

[TimFox]

There is also a possibility to generate a low-current negative supply from a single supply, such as  https://www.ti.com/lit/ds/symlink/tl7660.pdf?ts=1634980912985
I realize the appeal of single supplies, but they do "confuse" signal ground and negative rail.

[Benta]

Nah.
Use an LM4041-ADJ (from TI) set to 2.3 V and buffer it with a single rail-to-rail opamp. Job done.

[EnzoF]

There is also a possibility to generate a low-current negative supply from a single supply, such as  https://www.ti.com/lit/ds/symlink/tl7660.pdf?ts=1634980912985
I realize the appeal of single supplies, but they do "confuse" signal ground and negative rail.

Thing is, it works really well for my needs and the advantage is I can set the centerpoint to 2.3V which is optimal to capture the whole signal as it's not symmetrical. That way I get a lot of resolution.

My question is really if I can make the piezo circuit above work as well as on the breadboard when I have 10 of them connected to the same voltage reference? Is it possible?

I had a different circuit before where I used a voltage amplifier as piezo buffer circuit. It used a single voltage divider and worked pretty well without any crosstalk.

[EnzoF]

Nah.
Use an LM4041-ADJ (from TI) set to 2.3 V and buffer it with a single rail-to-rail opamp. Job done.

I also need something that is in stock on jlcpcb so I can get the whole PCB done with SMT service. Who wants to solder stuff by hand yuck 

[Benta]

Nah.
Use an LM4041-ADJ (from TI) set to 2.3 V and buffer it with a single rail-to-rail opamp. Job done.

I also need something that is in stock on jlcpcb so I can get the whole PCB done with SMT service. Who wants to solder stuff by hand yuck 

Just an example from me. The TI TLV431 is probably easier to get (careful: NOT TL431, the "V" is important)

[EnzoF]

Nah.
Use an LM4041-ADJ (from TI) set to 2.3 V and buffer it with a single rail-to-rail opamp. Job done.

I also need something that is in stock on jlcpcb so I can get the whole PCB done with SMT service. Who wants to solder stuff by hand yuck 

Just an example from me. The TI TLV431 is probably easier to get (careful: NOT TL431, the "V" is important)

Yep, the TLV431 is available from many anufacturers, On semiconductor, TI and diodes incorporated. I guess any of them should be equivalent.

[EnzoF]

Ok, just did a spice simulation of the circuit with two of these circuits connected to the same voltage reference.

As soon as I add a second circuit my differential circuit goes unbalanced. The negative wire from the sensor lowers 50% in amplitude compared to the positive wire. It doesn't matter if the references are separated with op amps. It seems as long as they share the same ground, the circuits affect each other.



Am I doing something wrong in spice, I'm not very good at it?

[EnzoF]

Here is the difference, the measurements are taken before R3 and R5. The first circuit is outputting a sine wave at 60V, 500Hz and the second is set to zero. If I set both circuits at the same voltage and frequency the problem disappears. Clearly there is huge crosstalk. (the left one is how it should be).



I have no idea what is going on.

[EnzoF]

Ah, it seems when I copied the circuit I forgot to rename the "in" label. Stupid me. I renamed them to "in1" and "in2" and the problem disappeared.

[Jan Audio]

Enzo if you try increase the resistor it can work with 1 buffer.
Dont forget : the half volt rail is sucking all the bad crosstalk out, it neutralizes it.

*disclaimer i dont know about pizza things.

[EnzoF]

Which resistor should I increase?

[Jan Audio]

I thought the 1K2 resistor, i did not saw the 1M resistor.
Nevermind, just try it, what is the 10n cap for ?

[EnzoF]

I thought the 1K2 resistor, i did not saw the 1M resistor.
Nevermind, just try it, what is the 10n cap for ?

This is a charge amplifier, the cap is part of the charge amplifier. You can see that there is a similar capacitor/1meg resistor on the iverting input as well so it amplifies the signal but not the common mode noise/60Hz noise. Although for some reason I'm getting a littl 60Hz noise. Haven't figured out why. I assumed I needed separate voltage references for every one of these circuits.

The 1.2k resistor is only there to help the voltage reference op amp dealing with the capacitive load. If I increase it the frequency response will not be good. I can decrease it bot not increase.

[EnzoF]

Ok I believe I've solved my problem with the PCB.

I had a bit of 60Hz hum on my circuit, not much but since the circuit is designed to completely reject common mode noise I assumed that something was wrong and that separate voltage references could solve it.

Now I connected all the piezos to the PCB in stead of just one and the 60hz hum disappeared.

For some reason this solves the problem and the noise floor looks like on my breadboard but much less noise so better than expected.

I'm not sure why this happens. Can someone explain it to me perhaps?

Here is the original circuit (only two piezo circuits are shown but there are 7):



The problem with piezo conditioning is they produce maybe 90V or so. I used a 700 ohm resistor to ground on both the positive and negative cable to lower the voltage before entering the charge amplifier (on the updated circuit I used two 390pF capacitors in series in stead to avoid the connection to ground.

As you can see the charge amplifier has a 1meg/10nF on both the inverting and non inverting inputs (highlighted in blue) to get common mode rejection. The 700 ohm and the 70kOhm resistors form a band pass filter with the 10nF capacitors to give me a nice clean signal.

The 1meg resistors need to be matched to get the best common mode rejection so I used 0.1% tolerance. However I noticed that in the circuit the 1meg resistor on the non inverting input is connected in parrallel to the 1meg resistors from the other piezo circuits through the ground plane and the voltage reference. I assumed that this made the circuits unbalanced.

That's why I wanted a separate reference for all the circuits.

However now when I connected all piezos the noise sort of cancelled out.

I have not tested the PCB yet with my application and I will see if they respond as they did on the breadboard with only one circuit.