Noise floor

[paulca]

Everytime I get inconsistencies in my circuits I look to noise.  Today it was Arduino ADC values bouncing about by 50mV randomly. (Another topic).

I probe various things and without fail, on everything, I see noise.  The oscilloscope reports it has a Vpp of 60mV or 70mV. 

The frequency isn't stable, the oscilloscope gives frequent values of 8kHz or 80Khz etc.

The waveform is a high freuencies modulated onto a lower one with more random spot spikes sticking out of it.

Even when I probe Vcc from the bench supply or probe ground it is there.

So either it's the oscilloscope probes or ... more likely... this is just my local environmental noise floor.

However is 60-70mVpp not a bit extreme?  With the probe hanging in the air the noise is there but the amplitude rises to several hundred mV.

I even put the sides back on my PC which made no difference.  The fluorescent lamp adds some HF stuff, but I can switch it off.

I suppose I should rule out the PSU by using a battery to power the circuit with the scope (USB) running on the laptop battery.

Any one have any thoughts?  Is it a fools game to chase this?  Is it mostly cause I'm using breadboards in a room with a PC, a laptop, 2 beefy monitors and 2 bench power supplies and at one 30w energy saver light bulb?

Note I have bypassed the hell out of everything with everything ranging from 1-100uF and 1-100nF they don't make the slightest difference (Which might suggest the probes at fault).

[dmills]

Probe technique possibly, but also remember that most scopes only have 8 bit ADCs, so if you are running on a 5V full scale, well, 50mV would be only a couple of LSB at best (And the frequency isplay is almost always a simple minded 'count the zero crossings' thing, not at all good for anything but the simplest of waveforms).

What bandwidth are you measuring over?

Scopes btw make lousy instruments for noise measurements.

regards, Dan.

[Kleinstein]

Computers, compact fluorescent lamps and monitors are good source of EMI. Also instrument with a switched PSU and no ground can be rather bad.

With the background there are 3  mode to couple in your circuit:
1) wire bound / ground loops
2) capacitive  ( avoidable with proper grounding / shields)
3) magnetic  (avoid large loops)

One could probe the source in you you environment to reduce the background.

[paulca]

Scope is 20Mhz I don't set that, I just zoom in.  I could easily see the noise down at 100uS per division and below.  Vertical scale set to 50mV per div.  Probe in X10.

As a random experiment I just took my Extech meter out of the tool box, set it and it's probes on the table and switched it to mV range.  It's happily reading between 5mV and 40mV AC from thin air.

I mean it could just be my environment.  It's always been there.  I only suspect it when things aren't playing like they should be.

In this case I was trying to get good voltage readings on the arduino.  I'd gone through the external reference, but as one of my voltages goes to 4.99V I tried 5V from a regulator as a reference.  It worked but when the USB power was disconnected, leaving the PSU providing 5.00V and not USB providing 5.10V the ADV went nuts and wouldn't read below 1.25V.  I take this to mean that AREF must be < Vcc.  It works if I put a resistor there, but gives my ceiling voltage of 4.61V

Anyway, while it was reading the other end the reading was scattering around.  0.000V  0.029V  0.000V 0.048V and so on.  The actual voltage was 0.042V.  Yet at the other end of the scale it was much better within 10mV.  However at the lower end, both ADCs in use bounced around nowhere near the actual reading.

I was hoping for more than this from it.  I wondered if it's happening to sample some of the noise causing the reading to scatter around.

No matter what I do with it it won't read accurately across it's full 5V (or 4.61V) range.  I'd live with +-10mV but it's out by 100mV or more sometimes.

[David Hess]

Most of the noise is probe related from one of two sources:

1. The loop area of the ground clip picks up everything and this is a serious problem at 100 MHz and above.  Try using a coaxial connection to the probe tip.

2. The ground loop between the oscilloscope ground and the circuit under test also contributes noise.  Differential probing will reject this noise source.

[paulca]

So I can just ignore it basically?

[ez24]

 

[trys]

You have my sympathy. I've been chasing an elusive 1KHz noise source (with harmonics) around my workbench in the last couple of weeks. Switch mode power supplies churn out a heap of noise.

Adding a resistor across signals that have the possibility of floating around too much helps a great deal to kill off those stray capacitances. 

Unclipping the earth lead and earthing directly close to earth of the probe helps a lot too.

Trys

[Andreas]

Any one have any thoughts? 

Perhaps you should switch off your mobile and the WIFI.

with best regards

Andreas

[David Hess]

So I can just ignore it basically?

Whether you can ignore it depends on what you are trying to measure.

To find the potential noise floor of your oscilloscope, set it to AC or DC coupling and either short out the input BNC or leave it disconnected; either should produce roughly the same result.  Attaching a probe and coaxially shorting the input to the probe should also produce the same result.

Many modern DSOs including Rigols lack a true ground coupling mode so that cannot be used; instead they short out the input signal after the first amplifier stages yielding a much lower noise floor than actually possible.  This also means that their "ground coupling" mode ignores input DC offset.

Sometimes people ask about the high noise level displayed by the probe calibrator output.  Some old oscilloscopes used a BNC for the calibrator output so it could be connected to the probe coaxially producing a pristine calibration signal.  Connecting the probe coaxially to a function generator output is almost as good.  In both cases, the noise seen is just from the loop area of the probe's ground strap or lack thereof demonstrating the problem.

[Raj]

try  9v battery via "raw input" arduino pad

[Pack34]

You can improve your low noise measurements with an oscilloscope using a ground spring instead of the alligator clip.

I've been working on some low noise work recently and have taken the approach of "nuking it from orbit". Use a ferrite bead in series with the ADC Vref with a 3 or 4 century bypass capacitor setup to mitigate. Then make sure that all of your routing and analog electronics uses best practices (not crossing return paths, proper spacing, shielded wire, etc).

https://www.digikey.com/products/en?keywords=PK1-5MM-118

[floobydust]

Arduino H/W design is poor for A/D noise. It depends on whose you buy, what the loads are on your 5V contributing to noise there etc.
The Uno is terrible as it connects AVcc directly to +5V 

I cut that trace and add an inductor+filter cap on MCU AVcc (10uH+0.1uF ceramic+10uF tant.) to ensure quiet power for the A/D subsection.
I also add A/D input capacitance like 10-100nF and that gives me low noise A/D results.