EEVblog
ฝัง
- เผยแพร่เมื่อ 26 เม.ย. 2014
- Why do digital oscilloscopes appear noisier than traditional analog oscilloscopes? PART 2
Learn how and why your digital storage oscilloscope display the noise that it does.
In this 2nd installment Dave shows how waveform update rate affects the apparent displayed noise of a waveform, and alters the waveform update rate on an Agilent DSOX3000 to show this.
Also, how sample rate and timebase settings affects the boxcar averaging high-resolution display function.
PART 1 is here: • EEVblog #601 - Why Dig...
Forum: www.eevblog.com/forum/blog/eev...
EEVblog Main Web Site:
www.eevblog.com
EEVblog Amazon Store:
astore.amazon.com/eevblogstore-20
Donations:
www.eevblog.com/donations/
Projects:
www.eevblog.com/projects/
Electronics Info Wiki:
www.eevblog.com/wiki/ - วิทยาศาสตร์และเทคโนโลยี
The reason they appear noiser is because the new scopes aren't made with oxygen-free copper like the Monster cables, hence noise is so high. As well, they don't have cryogenically treated capacitors which stabilizes the electron spin, reducing noise and adding "airiness" and ambience to the signal without losing detail. Also are those knobs balanced? Ideally they should be using wooden knobs balanced to 1/100 of a gram and made from redwood - these introduce less jitter when you turn them and hence less noise.
Power cord - What are you using for a power cord into the scope? Should be using one with cryo-treated LiniPure monocrystal conductors and at least 6 die-electric layers in order to avoid quantom tunnelling effects cause by atoms jumping across impurities in the copper. This opens up the "stage" of the signal without adding excessive cohesiveness or sharpness to it.
Next time, know the actual science of what is going on, instead of quoting meaningless non-sense terms like "frequency".
what you should talk is what noise you will get at 8b resolution, and how that noise is related to sampling rate and snr of adc itself . (like Nyquist limitation and so on)
always informative and entertaining. i love to watch and rewatch at a later time
i love getting a response from dave and i love the vlog i wake up every morning and check for a new video! so much learning!
In technical terms, because the sheer bandwidth is so high, the noise is being folded back on itself. Effectively, the sampling process is an RF mixer which converts the fundamental, and every harmonic, down to baseband, whether you wanted it or not (the trick is, you do want it, since that's why equivalent time sampling works). Thus, with a million updates per second, and supposing a 60Hz display refresh, every 1MHz-wide band from 0Hz to 500MHz is being aliased into the ~16k waveforms being displayed -- bringing roughly sqrt(16k) ~= 127 times more noise to the party.
Those numbers are probably a load of crap -- the bandlimited noise at 2mV/div hardly looks like it's 0.008 div tall -- but the general idea applies, if probably with different parameters. That kind of dynamic range would be hard to observe with an 8 bit ADC anyway.
Whereas, the high res mode (or the intensity graded mode -- which does histogram weighting of the samples, for a similar effect) limits bandwidth to what's visible on the screen: basically, update rate times number of pixels. Suppose the visible waveform is 512 pixels long: the FFT of exactly one single-pass waveform (as visible on screen) might cover, say, 1kHz to 250kHz (at 100us/div). Any lower frequency content looks like gradual curvature or DC (it can't be resolved into a complete cycle), and any higher frequency content is aliased, and either spreads the waveform (aliased), or is filtered out (in high-res mode).
Finally, supposing the front end (which particularly contributes noise on the lowest mV/div settings) or ADC itself (noise, quantization) has a given noise density (some constant nV/rtHz), the sheer bandwidth means HF noise dominates: there's a hell of a lot more MHz above 250kHz than there are below, so you should expect that high bandwidth settings, ranges, or scopes in general, *should* be noisier!
Excellent explanations... Part 1 and Part 2
It definitely enlightened me on my digital and analog scopes.
Thanks for the informative easy to understand videos.
Very good explanation. I agree with it completely.
Other way of explaining it is that if it is gaussian noise, then looking at less samples, you will rarely go over over few sigma. But if you increase number of samples at each point (proportional to the waveforms updates, even with analogish persistance), say 10 times, then you basically add about 1 sigma to the width of your sample, making signal appear noisier. It is just statistics.
Actually if you run infinite persistence, and leave DSO running for few days, the width of the signal will increase even more, due very rear noise events.
Digital just make it possible to see signal much better, and make it possible to show very rear events, and high sigma deviations from average, which on normal CRT would be too rare to light up phosphor (without using long time exposure photos).
Really good video!
Great example. I have argued with many on the differences between old analog scopes and new digital such as the Agilent. It might be easier to just point then to this video when I am asked about it! :)
Good information, especially for a non-engineer (me). Thanks for another informative video!
Important lesson, when looking at lower frequency signals turn the persistence on, otherwise the signal will appear much better then it really is.
Instead of using the internal wavegen, you could've also reduced the trigger rate by increasing the trigger holdoff.
Thanks for that, I could not think of the trigger holdoff.
That won't work in free running auto trigger mode. And in normal mode there is nothing to trigger off for that delay to work.
Ah, I had my analog scope brain engaged. On the analog scopes, the trigger holdoff will also adjust the free running / self-trigger rearm rate. Not so on the digitals....
***** So did 8 other people :-> Today is one of those rare days I have my brain engaged at al!
All good stuff.
Damn I love that scope!
If you work it out, the RMS thermal noise into 50 ohms with a 4GHz bandwidth is about 60uV. Ignoring any internal or external noise, this would make the trace appear about 500uV to 600uV wide.
For a gaussian white noise signal, the peak-peak value over so many sweeps as this scope shows per screen display will be MUCH larger than the RMS value. I have the same scope and I just tried it - I get 230uV RMS. Bear in mind, too, that the scope input is not a true RF low noise amp because it is optimized for other things (like not being damaged by high voltage inputs).
It should have a noisy blanker as standard when you switch it on....i you want to inspect the microworld of micro noises..then you should switch it to see..not as a standard to show all that mess on the screen
this makes me wonder what the noise is like on those 12-bit oscilloscopes from tektronix.. the noise you're seeing on the display is mostly noise the 8-bit limitation, not so much real noise from interferences
they are good for digital type on/off binary signals
Actually, if you LISTENED to his video, you'd understand they're good for ALL signals. They're BETTER than analog scopes in all cases!
+EEVblog, Am I right in thinking that these scopes actually _are_ this noisy, it's just that with a CRO and low wf/s scope you can't see the noise... rather than as you're suggesting, that they _aren't_ this noisy? If you take a single sample, and look at the peaks and troughs, or look at the section of this video where you use infinite persistence, isn't that the noise?
Have a watch of part one. He covers how to see the same noise on an analogue scope.
I know, but he keeps saying 'it's not actually that noisy' or words to that effect... it is that noisy, it's just that you can't see the fact easily on a CRO.
Sorry, I misinterpreted your question originally. I agree with you. Bear in mind that in at least Australia/New Zealand, when something "is not that ___", it's usually a figure of speech rather than an explicit declaration.
I believe the point he is making is that the digital scopes are not worse than the analogue scopes and that they are just displaying roughly the same results differently due to the inherent differences in how they are rendered on the display. He demonstrates this nicely in the first video.
Interesting tip about the subtleties of language, thanks!
Yep, I agree.. but to a beginner I think the wording could be misleading. :-)
Attie Grande Correct, all wide bandwidth scopes are noisy. What I'm saying is in comparison to an analog scope, because that's the myth that is being busted here.
10:25 40 GSa/s!? I want one! (I know you meant 4 :D)
...and you really don't think that scope's 8bit ADC quantization noise should be taken into account? For 8bit ADC it is FAR HIGHER than a possible signal or amplifier noise floor.
I think the problem comes from an assumption made by some people that "good old stuff" was better. Maybe they enjoyed electronics more back then, and it is only a matter of opinion. Anyways, this is very informative. Thank you :)
Well, in this case the analog scope trace does certainly look like it has less noise than a digital one, and from a visual standpoint it does. So it's easy to think that and not quite understand what the digital is doing and why.
EEVblog Then they should have the option to switch noise off.....
This entire series is about how to switch the noise off :D Even my low end Rigol has adjustable low pass filters. You certainly have that option :)
Too much smoke and mirrors, an adjustable LPF is the answer. I chose what I want to see when and where...
what you demostrate? nothing, What I believe, if you has any osciloscope and shortcut the imput to ground (not with the full prove, but with the BNC grounding itselft), and if the osciloscope is not capable to show a 0v noiseless line, so IT IS INTERNAL NOISE from the osciloscope!, is not the same to test it with the input unplugged than short-grounded.
I said the input termination makes no difference. I demonstrated a practical reality of wide bandwidth oscilloscopes. Of course it is internal noise.
ok it is internal, it is a shielding problem, or internal generated by the same digital circuits from the osciloscope?. Internal, so, it isn't noise from the circuit that is being measured, that is quite a badass problem, don't you thinks that?
flyguille I don't known about others, but I don't like to see in the screen noise that is not from the board that I am testing. So, with these videos I concluded that depending on the task , like for low frecuency is prefferable a good analog osc, and maybe for high frecuency digital signals is prefferable a digital osciloscope because its better capabilities about hold/record /multichannel the digital signal to analyze (like in an FPGA develop task)..
What a mess. It was easier in the old days :)
Hay Dave,
In these videos I see a lot of DSO’s just setting around collecting dust. Why don’t ya come off of some that you don’t use and seen a poor Yank like me one f them.
DLMinton
Because you didn't deserve for one ;)
!
Better not to show the noise...they should filter that....if you need to see the system's noise to compare with another noise you should switch to see this noise.... this looks noisy because it is noisy to our eyes...and this is not good to display.
test from where the hell is that noise!, if internal from osciloscope, or external, if is the prove acting as antenna or not, or if it pick up signals from an OPEN input, what if you SHORT-GROUND that input....!, ?, the worst thing that an osciloscope can do, is to introduce internal noise.... , wrap the osciloscope in grounded aluminiun several layers, test everything. Don't show, the product "the noise" as all the show, we already know that it measure noise, !
You are COMPLETELY missing the point here.
That apple is ugly because it is fat and filled with juice... a juicy fruit...but no one will buy...this is the problem
sorry, but digital scopes are rubbish at these frequencies because they are sensitive to the high frequency noise, and you have to compensate for that. for audio, i will always use a high-end analogue scope.
LOL