Oscilloscope 100x Pre-Amp for noise measurement?

[RogerThat]

Hi all,

I have a project with a 24-bit ADC reading an 0-50mV air pressure sensor. The thing needs noise free 5V power to be able to take fast measurement with the accuracy I want.

My oscilloscope (1054z) reads down to 500uV (actually less since there is noise)  and I need( maybe 'want' is a better word) to scope noise down to around 1uV. So, I'm thinking to build a 100x gain amplifier front end to the scope.

My basic idea is to use a 1ghz OP AMP with dual supply (two 3v lithium coin cells and two 2.048v references). This will give me a +- 2V output nicely match to the high gain mode on the 1054z. The thing needs to be battery powered due to noise...and yes, it's going in a metal can to shield off radio noise.

I've been playing with the idea since this morning so maybe the details are a bit off....but the idea. Have anyone done this before here? Maybe someone already have a schematic? 

[HB9EVI]

Just some thought

My guess would be, that a 1GHz bw opamp doesn't necessarily have the best noise performance, and is quite obsolete, since you obviously work with a DC source.

In a setup with such an ADC, you probably have to care about mains frequencies, so 50/60Hz; some ADCs have filters builtin, additionally sample rate will be low to reach 24bits, what reduces noise of higher spectrum as well; so in the end we are mostly talking about 1/f noise

[Kleinstein]

It is very difficult to get very low noise at a high BW. This is the reason for the usual 20 MHz filter setting with most modern scopes.

There is another thread about AC coupled amplifiers to measure low frequency nose.

[In Vacuo Veritas]

Why are you yelling "op amp"? It's not an acronym. It's puzzling to me that so many people write "OP AMP". Why?

[Conrad Hoffman]

The old Tek plug-ins go down to 10 uV/division. You'll be limiting the bandwidth to a few kHz max at that level if you expect to see much.

[DaJMasta]

First, find out what sort of bandwidth you need, then try to start looking at the rest of the system... it sounds like it's worth figuring out how much gain you really need to see what you want, since a 24 bit ADC is going to have <20 effective bits before massive averaging and your scope can probably already see trends in the 1-50mV output.


That said, if you need significant gain, it's almost always best to do it in stages to use lower noise/GBP amps.  You may need a higher input impedance if the thing you're measuring can't supply much current so you don't effect the reading with the loading of the probe.  You want low pass filters at every stage to keep noise outside of the band of interest suppressed, as this will keep overall system noise to a minimum.  Also, don't use a voltage reference as your supply, just use the batteries directly.  Most opamps have some margin they need from the rails for the signal anyways, so two 3V batteries would likely get you something like ~4V worth of usable signal level.  Then let the gain determine the output.... and switching to a wider supply (like 2 9V batteries instead of 3V), could give you the option to pick lower noise amps that require higher supply voltage and could get you onto ranges of your scope that have less inherent noise, something like the 1V/div range.

In any case, single microvolts resolution with real bandwidth is a challenge, but if you can constrain your design somewhat, you can probably make something that will be useful to the project.

[Conrad Hoffman]

The above is true, concerning using multiple stages, but remember that the signal to noise of an amplifier is established right at the input stage. Nothing that happens after that can improve it, so don't blow it at the front end!

[RogerThat]

Thanks for all the good input.

For this specific application I don't need much bandwidth but if I build it I would like it to be a tool that I can use for other projects also. So, for constraints let's say it should not limit my scope. 

I've looked at some circuits and my idea is (this is my first time building an AC thing) like this:

First an AC couple amp like the picture below. Keep the resistors at 100k and 10k to achieve a gain of 10 . What value would you recommend for the capacitor? I guess R3 form an RC filter with C1 so it would preferably be quite low?

Second feed this to a normal voltage follower with a gain of 10 also.

Power from two coin cells (3V) giving +3 to -3, 6V span. On-Off switch.

BNC in both ends for connections.

I'm thinking a dual op amp in a dip-8 package. Recommendations which?

[EEVblog]

I've looked into this, and actually have a prototype amp (called the uFrontEnd) with selectable gain up to x50. It's not easy to do even for the standard 20MHz bandwidth noise measurement. Few chips on the market are capable of the noise floor vs bandwidth required.

[iMo]

Long time back I got these opamps:
https://datasheets.maximintegrated.com/en/ds/MAX4223-MAX4228.pdf

You may use 2 opamps with 10x gain each. BW should be ok, noise is not bad as well. 6mA per amplifier.

[RogerThat]

I'm looking at this: https://www.analog.com/media/en/technical-documentation/data-sheets/AD8011.pdf
2000v/us slew rate, should be good?

Input noise is specified at 2 nV/√Hz.....so that would be for 100mhz: 2 nV / 10000 = (alot lower than 1uV).
While writing this it feels odd that noise goes down when frequency goes up. Am I doing it wrong?

[Karel]

I'm looking at this: https://www.analog.com/media/en/technical-documentation/data-sheets/AD8011.pdf
2000v/us slew rate, should be good?

Input noise is specified at 2 nV/√Hz.....so that would be for 100mhz: 2 nV / 10000 = (alot lower than 1uV).
While writing this it feels odd that noise goes down when frequency goes up. Am I doing it wrong?

Yes. In this case, it's 2nV * √Hz. (20uV @ 100MHz)

[iMo]

@RogerThat: you have to be more clear on what performance you are for.
You indicate you want to measure the "noise of the 5V voltage regulator".
Why are you looking for a high BW then?
People usually are measuring the noise of such stuff in 0.01-10Hz, or 1Hz-10kHz BWs. Why do you want to mess with ~MHz then?

[dzseki]

Also when dealing with noise measurement, mind the resistor noise, like having a super-duper low noise preamp chip with a 100kOhm feedback resistor kills all the efforts (with having 40nV/sqrt(Hz) self noise).
En(resistor)=sqrt(4kTRB)
k - Boltzman's constant (1.38E-23)
T - Temperature in (K)
R - Resistance (Ohm)
B - Bandwidth (Hz)

[Marco]

I'm looking at this: https://www.analog.com/media/en/technical-documentation/data-sheets/AD8011.pdf

Why though? Why not go straight for something like OPA847?

[tggzzz]

There's a lot of useful information about this (and many other topics) in The Art Of Electronics - particularly chapters 4,5,8.

You can see the table of contents in http://artofelectronics.net/wp-content/uploads/2016/02/AoE3_chapter9.pdf

There's subject and parts indexes at https://artofelectronics.net/

Get that book and you'll save yourself a lot of time and effort, leaving you free to concentrate on your USP

[RogerThat]

Aah, realised my formula misstake a few minutes after posting. Thanks Karel.

"you have to be more clear on what performance you are for". Well, as good as possible. And what do I mean by that? In my current project I want to measure voltage noise/rippel down to 1uV. I'm starting to get a grip about what are the limiting factors regarding noise/BW. So what I understand now is that a resonable BW would be up to 10khz for 1uV.

The input sampel cap and resistor on the ADC form a 32khz low pass filter so anything above say 40khz will make little impact on the measurement. So, let's limit to as good as possible with 40khz and components that can be had for thru hole mounting (for one-off things I really avoid SMD stuff).

Ok, so the signal chain would be like this then (maybe just one amp if it can make 100 gain straight away):

BNC->LPF (32khz)->AC couple 10x amp ->10x amp -> BNC

Low resistor values and I guess film/tantal caps?

tggzzz: great links, will read.

[Marco]

It's hardly worth it to go two step for 100x40kHz, OPA847/LMH6624 combine low noise with way more GBW than you need.

[RogerThat]

I hate when someone doesn’t follow up their thread, so…
I built the thing .

100x Gain, 0,36hz-31.8kHz (-3db at 31.8khz that is).
amp is good for 500khz @ 100x.

BOM:
LT1028 (DIP-8)
C1: 2 x 22uF film cap (what I had in my box-of-many-components)
R3: 10k
R1: 10
R2: 1k
C2 (parallel with R2): 5n/5000pF
2x 470nF decouple Film caps
Power from two 9V batteries, +-9V.
RG316 input cable….just put some solder on the measuring ends.
Twinings tea metal box
To-Do: will put BNC chassi contacts at the box input/output.

Probe (@ 1x) measure directly on the protoboard. If I calculated correctly the theoretical noise would be less than 1uV. In reality my scope tells me around 20uVpp and 5uVrms when shorted with a 1k resistor. As you can see(picture two) the the FM Radio 100mhz  is creating the noise floor.

Picture one is the output from my voltage reference (the spike is caused by the ADC sending SPI). Picture two is the measuring ends shorted by a 1k resistor. 

Now off to part 2 of this, how do I reduce the noise from the voltage reference. I probably done something wrong, specs say 1uVpp. I'll create a new thread for that. Have a nice week-end everyone.

[Kleinstein]

The LT1028 has quite some input current noise. With the 10 K at the input this gives quite some noise from the low frequency range, below the input coupling limit. Especially in the low frequency range the LT1028 is good only for low impedance sources (e.g. < 100 Ohms).
So there is quite some low frequency noise - that is expected to come from the LT1028.

So one could reduce this noise quite a bit if C1 is chosen even lager and the lower frequency limit set by another RC filter / AC coupling behind the amplifier.

Another option would be to use an amplifier with less current noise, like AD8675, OPA827 or OPA140.