Buffered voltage divider: how to improve noise behaviour?

[Bigman]

Hi,
Out of a normal battery supply (~3.8V nom.), I want to produce a voltage of 70mV. Since those 70mV are intended as a reference voltage for an inverting-OpAmp amplifier, a normal voltage divider with two resistors alone wouldn't do the job and a buffer was needed. As a buffer a choose the OPA347. My target is to have those 70mV as "clean" as possible, without any noise. From the internet, I also got some "inspirations" about additing some capacitors as noise-filters (see attached circuit diagram).

To check the noise generate by this circuit, first I connected the battery which drives this circuit directly to my oscilloscope (Rigol 1102E). The measured Vpp was 4.8mV (which is more or less the noise of the oscillsocope itself).

Next, I measured the desired output-voltage of my circuit Vap. Now Vpp is 6.4mV. My assumption is the, circuit itself generators a noise of Vpp=1.6mV (=6.4mV-4.8mV).
I am wondering: what can I do, to further reduce the noise. The thermal noise from the resistors must be already pretty low in the 10-6V range. Same for the OPA347's input noise. One reason, why I choose OPA347 was also it's small Bandwidth of 350kHz (I assume the larger the BWP, the more noise in higher frequency areas I wlil get).

[dom0]

Noise is added geometrically (a = sqrt(b²+c²+..)), not algebraically.

Use AC coupling to remove the DC offset and be able to use a higher sensitivity on the scope, since the 4.8 mVpp (at 20 mV/div) are the scopes noise floor, not that of the battery.

With the LPF at the output you will get a DC error if loaded by DC currents.

[Bigman]

Thanks a million time for you kind answer!

Noise is added geometrically (a = sqrt(b²+c²+..)), not algebraically.

so funny ... I read this a million times before, but when I finally need this knowledge, I am not remembering it.

Use AC coupling to remove the DC offset and be able to use a higher sensitivity on the scope, since the 4.8 mVpp (at 20 mV/div) are the scopes noise floor, not that of the battery.

Yep ... this reduces the indicated noise/Vpp significant. Since I am not that proficient pls. allow me a comprehension question: all what I know regarding DC and AC coupling is, that AC blocks in a signal the DC component. Is it always the case, that with an AC coupling the probes' sensitivity will be improved and why is that so?

With the LPF at the output you will get a DC error if loaded by DC currents.

hmmm ... as a beginner I am always curious: if I have a DC current running through my LFP, I understand that this yields into a voltage drop at the LFP-resistor, which I assume you mean with the DC error. Anyhow, the resistor is small and so it's noise also, hence how should it affect my noise/Vpp?

[Dave]

Would you mind taking a picture of your probing setup? I have a strong feeling that the culprit lies there.
Please also provide your whole schematic (including your inverting amplifier). You might be unnecessarily complicating things.

[Bigman]

Would you mind taking a picture of your probing setup? I have a strong feeling that the culprit lies there.
Please also provide your whole schematic (including your inverting amplifier). You might be unnecessarily complicating things.

Attached a picture of the setup.
The blue/black wires are going directly to the battery supply (+3.6V and 0V)
The grey wires are connected with the output of the circuit and with 0V (~70mV and 0V)

The whole setup is within a selfmade tinplate/copper box to shield it from external influences (in the picture the box is missing the upper lid).
The connection from the tinplate/copper box to the oscilloscope is done via BNC-connectors and double shielded coax-cables.
The 0V from the battery is also connected with the tinplate/copper box and so also with the coax-cable cladding.
While testing the circuit is NOT connected to any loads.

[Dave]

The probing setup looks good, now let's see the schematic of the circuit that this thing will be attached to.

[Bigman]

The schematic is already attached in the very first post above.
A picture from the real circuit with it's components is attached to this post. Since all traces are covered with solder-mask, I have also attached the artwork.
Don't get confused by the three unconnected pins on the right side. I just use them for to fix the pcb on a bread-board (mechanical stability reasons).

[Dave]

... those 70mV are intended as a reference voltage for an inverting-OpAmp amplifier ...

This part.

[Bigman]

... those 70mV are intended as a reference voltage for an inverting-OpAmp amplifier ...

This part.

this part is just an example and I don't see the relevance to the noise-question - especially since the above attached results are measured without that part.
Anyhow. in the 1st intented setup it will be a damn simple inverting-OpAmp amplifier:
inverting-input connected via Rf=250kOhm with an OpAmp output. Parallel to Rf is a capacitor (approx. 1pf, since OpAmps f_c is around 44MHz). The non-inverting input is connected will be connected with the 70mV provided by the device discussed above. Additionally, there maybe some additional capacitors at the OpAmps supply input (40uF).

[Dave]

Out of a normal battery supply (~3.8V nom.), I want to produce a voltage of 70mV. Since those 70mV are intended as a reference voltage for an inverting-OpAmp amplifier, a normal voltage divider with two resistors alone wouldn't do the job and a buffer was needed.

Well, if you are connecting this to the non-inverting input of an opamp alone, you don't actually need buffering (unless that opamp has an awful input bias current, in which case you should pick another one). Pick a corner frequency for the desired LPF and select the right capacitance to put in parallel with Rst2.

[T3sl4co1l]

It's sometimes better to use a separate divider for each op-amp in a signal chain, rather than use a common reference that must be buffered.

The standard topologies can be modified with resistors pulling important nodes up or down, as needed.  Consider the differential amplifier configuration, with resistors added, one to each input pin, pulling them up towards +V or down towards -V or GND.

Tim

[Mechatrommer]

Out of a normal battery supply (~3.8V nom.), I want to produce a voltage of 70mV. Since those 70mV are intended as a reference voltage for an inverting-OpAmp amplifier, a normal voltage divider with two resistors alone wouldn't do the job and a buffer was needed.

why do you want to combat noise from the joke "voltage divider voltage reference" setup? or probably is that 70mV is relative to 3.8V ie 1.84% relative to Vs? or is there another circuit you didnt show to ensure the Vs is always 3.8V?

[Bigman]

Out of a normal battery supply (~3.8V nom.), I want to produce a voltage of 70mV. Since those 70mV are intended as a reference voltage for an inverting-OpAmp amplifier, a normal voltage divider with two resistors alone wouldn't do the job and a buffer was needed.

why do you want to combat noise from the joke "voltage divider voltage reference" setup? [...]

why do you call this circuit (discourteously) a "joke"

[Mechatrommer]

because you wont maintain 3.8V from the circuit you showed.

[Bigman]

because you wont maintain 3.8V from the circuit you showed.

as you can see from the post above:
battery voltage: 3.75V
Voltage Vap at the output of the circuit 70mV

of course, both voltages are SOC dependend. However, the exact absolute voltages are not the key.
I was most interested to understand where the noise is coming from ... (btw: totally independent which other circuit this voltage-divider will finally supply).

(... I would better have posted this in the "Beginners" section. If a moderator can move it there, it would be appreciated, so the pro-contributors can better understand where this belongs to)

[Mechatrommer]

you dont have much noise there, only resistors, capacitors and opamp noise.... maybe your bet is high priced of resistors, capacitors and opamp. and just maybe a light reading... http://en.wikipedia.org/wiki/Johnson–Nyquist_noise

[Bigman]

you dont have much noise there, only resistors, capacitors and opamp noise.... maybe your bet is high priced of resistors, capacitors and opamp. and just maybe a light reading... http://en.wikipedia.org/wiki/Johnson–Nyquist_noise

  .... this we already discovered in the 3rd post (https://www.eevblog.com/forum/projects/buffered-voltage-divider-how-to-improve-noise-behaviour/msg688893/#msg688893) and the noise-question was already cleared (that there is indeed not so much noise and that it's more or less the "Rigol floor") ....