Why Are Digital Scopes Noisy (Or what appears to be Noisy)

all electronics have an inherent noise, including oscilloscopes that cost more than a home. with any amplifier circuit, as gain increases so does the noise.

it would still be visible if you shorted your input to ground. your only seeing internal noise, with nothing connected.

this is without probe.

same settings with probe attached.

Legacy analogue scopes (and that's pretty much all of them) have one primary source of noise but digital scopes have three:

front end amplifier noise (common to all)

adc sampling noise (digital scopes only)

psu and built in computer noise (modern scopes which means digital in reality)

The ADC noise on my (quite pricey) scope is about 5 bits pk-pk, and doesn't change between 2.5Ms/s and 10Gs/s. When you are looking at slow stuff you can make it about 8x better by letting it combine samples for a lower noise lower sampling rate effect.

The front end noise is almost always masked by the ADC noise.

I've found that several digital scope sinject quite large noise currents into earths and can often upset the thing you are measuring. The latest LeCroy is not bad. I keep an old 20MHz analogue (Gould) scope for looking at the occasional noise issue.

Most scope manufacturers don't quote noise very well.

The latest Pico 5000 series looks interesting  - 16 bit, noise is quoted, but be careful because it will only work plugged into a PC via USB and this can be an earth noise issue.

MK

I guess is the same inherent noise that you find in the high-tech digital multimeter aswell. As you already know, once you use the DC/AC or ohm measure scale, the multimeter oscilates values of tenths of mV or mΩ at high precision setup.

All very good and valid points but still

an oscilloscope with say a 2mV per division range should be able to show a sub 2mV waveform without much noise

take a typical scope with 8 to 10 vertical division, if the noise is 1 division then that's 10% or more of the range, does not this seem excessive for the range

some of the examples I have seen on you-tube  where showing a division of noise even on the 100 - 200mV ranges, that's an awful lot of noise even for a cheap oscilloscope

So yes I get the ADC noise, input front end noise etc etc, but I would never expect it to be more than sub 2% on a low cost scope and probably less than 0.5% on a high end scope

Am i crazy for expecting that ???

Scope manufacturers are very coy about noise specs.

Looking at the 12 bit LeCroy data sheet there is no simple "divisions pk-pk on range x" spec.

Instead it says that it has 55dB SNR - thats a ratio of 562:1 so at 8 divisions full scale you could expect as little as 1.4% of  a division which would be very nice, despite being well short of 12 bits.

Here's an interesting link:

http://www.evaluationengineering.com/special-reports/201307/instruments/Resolving-Finer-Detail.html

MK

That is a very cool article

Thanks

Sorry for digging an old topic back up, I found it interesting ... let me start off by saying, I'm not a high performance scope expert - so please do correct me (nicely) if I'm wrong ...

Well, part of it might be the better bandwidth of modern DSOs as well, thus being able to let through and measure high frequency noise which would have been filtered by less bandwidth capable scopes. The phosphors of old CROs didn't shine that brightly for "random" signals, so any noise would be a very dim, possibly hardly perceptible blur.

The inherent noise contribution is likely the thermal noise (Johnson–Nyquist noise - Wikipedia) and shot noise (Shot noise - Wikipedia) which is a matter of physics and can't be eliminated, all mixed in with other sources of noise already discussed (i.e. A/D, RF, SMPS). I suspect a wider bandwidth means that more thermal noise is admitted which increases the apparent noise floor of your channel. If you have a bandwidth limiter, maybe you can see its effect on the readings (at the cost of bandwidth, of course).

I suspect the method of scaling the horizontal axis also has something to do with it - as scopes are likely to show "bars" which span the lowest to highest value in that pixel bin, rather than the "most common value" or intensity coded value. This is just the way it's "combined" several A/D values together at that particular time interval, and it's not the same as the phosphor of CROs which will have an intensity gradient representing how often the signal is at a particular reading, thus the noise "fades" around the signal and is often not seen.

That's what old topics are for so no worries

My main concern was that on the lowest settings, it is over a full division of noise, with a 8 bit ADC less a bit for overshoot management, you have say 250 count, so 30 count per division , that's well over 10% noise on the range

The thermal noise from a  1k resistor over a 1GHz bandwidth is 127uV, from 1 1M resistor it's about 4mV which are about 0.762mV and 24mV pk-pk respectively. If you are using  a x10 probe it is the 1M case that applies, regardless of the actual source impedance but the scope will multiply the reading by 10 so the indicated noise will be 240mV pk-pk.

I set my scope to 200MHz bandwidth and with a shorted =10 probe I got 9.5mV pk-pk (1Gs, LeCroy 610zi) -  the expected value is 10.8 mV pk-pk.

So I conclude that the scope is not noisy but as Gough suggested is just showing what physics says should be there.

Of course this was only one quick experiment with one scope.

And watch out for those x10 probes - handy but noisy !

(thermal noise data from Thermal Noise Calculator one of many such pages on the web).

MK

Thanks Michael for that handy link ... and the calculations. Very enlightening!