Rigol DS2102 - grounding the probe doesn't flatline the trace

[agdr]

When I ground the probe tip on my Rigol DS2102 on a high vertical gain setting I don't get the flatline trace I would expect.  Instead I get a fast, complex, waveform.  Any input on what might be going on here is appreciated!

This is a DS2102 100 MHz scope with the 200MHz software upgrade.  The probe is a 200MHz probe set to 10x and compensated.  Not only does this happen with the probe tip grounded,  it happens with the probe entirely disconnected and even with a BNC shorting cap applied to the input.  The only thing that properly flatlines the trace is setting the scope input coupling to ground.

The scope is set to AC coupling, 10x ratio to match the 10x probe, and BW limit off which should give the full 200MHz.

The first photo shows the problem.  Probe tip grounded.  This is on channel 2. Moving the probe to channel 1 results in exactly the same thing.  10mV/div vertical and 2nS time base.  Trigger is set to single sweep to freeze the waveform.

The second photo shows the probe removed entirely!    I have another photo (won't fit here, my photos are too big) with a shorting cap attached that results in exactly the same waveform.  But setting the input coupling to "GND" will finally flatline the display.

On lower vertical gain settings the scope works just fine, such as viewing a 2Vp-p waveform at 500mV/div, or the probe compensation square wave from the scope's compensation ring.

[technogeeky]

Is the frontend busted?

Check the resistance of the center pin to the outer ring; when the channel is on and off. Is it 1.000M in both cases?

[agdr]

Looks like 1.2 meg on both channels, the entire scope is powered off here.

About the only thing I can think of - other than what must be operator error on my part here, lol! - is just a noisy vertical amplifier in the thing.  But i can't imagine a company would design a scope such that the highest couple of gain settings show up significant noise in their amplifier.  That "gnd" coupling setting that flatlines the display must either be after the vertical amplifier (given that a BNC shorting cap didn't flatline it) or simply a software trick that draws a flat line on the display. 

[agdr]

Here is the BNC shorting cap...

5mV/div and 2nS time base

[agdr]

...and the flat line (finally!) with the coupling set to ground.

[rs20]

The real question is, does anyone else have the same scope who can confirm/deny whether this is working as intended? Given that you're horizontally zooming right into the shortest timescale where all the high-frequency noise would be evident, I'm not 100% convinced that this is actually a problem? If you zoom out to a more normal timescale it'll just appear as the classic fuzziness which is inevitable when dealing with a 200 MHz scope collecting 200 MHz BW worth of noise.  I'll admit though, it does look a bit more than I'd have expected.

Good on you for trying with the shorting cap, that was going to be my next question! Do you have probe compensation still set to 10x when using the grounding cap? Because then 5mV/div is actually 0.5mV/div on the frontend, which would be indicated correctly if you configured your scope to read 1x.

Also, sorry to be pedantic, but nS is nanosiemens. ns is nanoseconds.

[technogeeky]

Other things to try:

• Turn on the 20 MHz bandwidth limiter. Does the noise go away?
• Make sure you are testing on calibrated timescales, not the vernier timescales. This shouldn't make that much of a difference, but at least we can rule things out.
• Use a probe-tip wire (the little coil spring that fits on the ground on the tip of the probe) and then connect that to the tip. This is a much shorter wire, so any real high frequency noise (via inductive pickup) will be eliminated.
• Turn off any nearby switching power supplies; especially LED ones.
• Similarly, turn on AC triggering and see if the signal correlates with that trigger or not.

[A Hellene]

This is something to contemplate on:
Rigol DS1052E nasty surprise!

In a few words, "I guess that we have both been, amongst others, the lucky winners of this specific (non-)QC market..."


-George

[agdr]

A Hellene - thanks for the pointer!  I figured there must be a thread on this already by now. I tried searching for "grounded probe noise" and even just "Rigol" but I must not have gone back far enough in dates.  Well wow!  Apparently the Rigol is "noise challenged".    Skimming through that thread I see you even tried the shorting BNC caps and the issue of coupling "gnd" working came up.

I've had the scope for a year or two, bought new from Tequipment.com.  I had noticed the trace wasn't particularly sharp, but hadn't much need to go down to super-low voltage readings until recently.  Then I became aware that "fuzz" was actually a small waveform.  Then tried grounding the probe... and still there.

Interesting about your 95MHz measurements.  I hadn't measured the period yet, but the frequency must be around there since it is showing up as a few cycles at 2ns (rs20 - you are right! - thanks).  Must be a clock signal in the thing somewhere for the digital logic that is leaking.

I'll read through your thread in more detail tomorrow.  I'm wondering if the higher priced 4000 series 500MHz Rigol scopes have the problem.   I've been thinking of dropping some money into an upgraded scope this summer.  I'm *really* glad I've found out about this problem now!  It may already be in your thread that I haven't seen yet, but if anyone out there reading this has a Rigol 4000 series, please try grounding the probe tip (to probe ground / bnc ground / calibration ground), turning the vertical sensitivity and time base to maximum, and post what you get.  I would be most curious.

[tautech]

There's some history with design problems, search the forum for Project Yaigol.

[A Hellene]

You 're welcome, agdr!

By the way, do not miss the reverse engineered DS1000E schematics that I've drawn and shared with the community while I was investigating the problem in my device.


-George

[agdr]

tautech - thanks!  Found it:

https://www.eevblog.com/forum/projects/project-yaigol-fixing-rigol-scope-design-problems/

Well geezzzz.....    Just that first paragraph doesn't look encouraging in terms of scope quality.  Some great detective work by the forum folks here.  I'll read through this week.

[agdr]

A Hellene - I just scanned through all 6.  I can see a tremendous amount of work went into ringing out the circuitry and transcribing it all.  What a great resource.  Thank you for the contribution!

[rs20]

Then I became aware that "fuzz" was actually a small waveform.  Then tried grounding the probe... and still there.

What did you think it was?

Anyway, my DS2202 produces a similar-looking waveform with a grounding cap and all the dials turned to the extremes like you have (that's 500uV/div actually, not 5mv/div.) I don't want to be "that guy" desperately defending his purchase, but the set of problems that can't be solved by a) setting bandwidth limit, b) using higher level signals, or c) using more gentle timescales appears to be a small one; certainly one I haven't encountered after using the scope for many years now.

One thing that is disappointing, though, is that Rigol don't seem to specify a noise value or figure in their datasheet. Not sure if it's standard practice to omit such a figure from the specs, but it certainly makes it harder to defend the performance of a product when they don't make a specific claim.

Noise is an inevitability of using wide bandwidths, although my attempts to calculate a figure have given an unexpectedly high result. A plain old resistor, such as the termination resistor in the scope, of 1 megaohm at 20 degrees C produces 1.8mV RMS white noise when observing over a 200 MHz bandwidth. This is over 3 divisions on your scale, RMS, which would mean that your scope is performing better than a theoretically perfect scope, which is clearly wrong. I'm not sure how the 16pF capacitance of the frontend factors in, or whether I'm missing something else?

TLDR; I suspect you should watch the EEVBlog episode debunking the claim that digital scopes are noisier than analog. It does a good job of conveying that noise is an intrinsic part of life, not some design mistake that can be arbitrarily quashed. Clearly there is some sort of design imperfection in the Rigol scope, but still cruicially important to understand what a theoretically perfect scope looks like -- because it's not dead flat, ever.

But i can't imagine a company would design a scope such that the highest couple of gain settings show up significant noise in their amplifier.

The high gain settings are 100% useful on more sensible timescales, when using acquisition averaging, or when using the bandwidth limit. So your expectation is unreasonable here.

[rs20]

Correcting myself: turns out the noise (across the entire bandwidth) of a parallel RC circuit is dependent on the C alone, so we can calculate that we should expect a theoretically perfect scope to have 16uV, or about 1/3 of a division. So this confirms what has already been pointed out that this scope is far from theoretically perfect (let alone, apparently, a top-of-the-line modern scope), but that the "flatline" that the OP was hoping for was also unreasonable.

[tautech]

Correcting myself: turns out the noise (across the entire bandwidth) of a parallel RC circuit is dependent on the C alone, so we can calculate that we should expect a theoretically perfect scope to have 16uV, or about 1/3 of a division. So this confirms what has already been pointed out that this scope is far from theoretically perfect (let alone, apparently, a top-of-the-line modern scope), but that the "flatline" that the OP was hoping for was also unreasonable.

Yep, noise on a DSO is one thing and can be managed in a # of ways as Dave outlined in his 2 vids but periodic noise is entirely another thing, it originates from the immediate environment or from within the DSO itself.
On a cramped bench it is not unusual to pick up the noise from the display.....dangle an unterminated probe around a DSO and you'll see. 

[DimitriP]

With the input shorted at the BNC jack , expecting a flat line is not unreasonable.

I'll accept it as a posssible side effect of a performance vs price design compromise but that's as far as I wll go

Someone reading this with a megakilobucks scope should be able to confirm if there is a flat line displayed with the input shorted ....

[ebastler]

With the input shorted at the BNC jack , expecting a flat line is not unreasonable.

I beg to differ. Of course it is unreasonable to expect a flat line (if you crank the vertical amplification up to the maximum). I am flabbergasted by this whole discussion, in a TME forum! The only valid questions one should ask here are: "What level of noise should one expect? What level of noise do other scopes produce, under comparable circumstances -- same bandwidth, same vertical amplification?"

[rs20]

Someone reading this with a megakilobucks scope should be able to confirm if there is a flat line displayed with the input shorted ....

Wasn't aware that the laws of physics could be bribed with millions of dollars. Vrms = sqrt(k_BT/C) = 16uV in this case. End of story, even before the noise figure of the frontend is considered.

(Unless I'm misapplying the formula)

[tggzzz]

When I ground the probe tip on my Rigol DS2102 on a high vertical gain setting I don't get the flatline trace I would expect.  Instead I get a fast, complex, waveform.  Any input on what might be going on here is appreciated!

Do some experiments to characterise the signal...

What's the frequency? (Looks like ~100MHz - that ought to remind you of something )

Is it constant with time, or pulsed?

What makes it change? Loop orientation, position relative to animate and inanimate objects...

What happens if you get rid of that 100nH inductor (the ground lead) and directly connect the tip to ground?

Welcome to the world of EMI/EMC.

[rs20]

...

I think you missed the post where the OP used a BNC shorting cap, and ended up with essentially the same outcome. This is partly an imperfection with the Rigol, as explored in other threads.

[DimitriP]

Someone reading this with a megakilobucks scope should be able to confirm if there is a flat line displayed with the input shorted ....

Wasn't aware that the laws of physics could be bribed with millions of dollars. Vrms = sqrt(k_BT/C) = 16uV in this case. End of story, even before the noise figure of the frontend is considered.

(Unless I'm misapplying the formula)

More expensive scopes have better lawyers.

[capt bullshot]

That's just noise, as some other pointed out.
But look at my "good 'ol" TDS224 here: this waveform is result of some defect (first picture) no matter if the input is open or shorted

second picture: same input settings, just maximum speed - see the noise

third picture: this scope's also cheating on you when you set the input coupling to GND

[David Hess]

The first photo shows the problem.  Probe tip grounded.  This is on channel 2. Moving the probe to channel 1 results in exactly the same thing.  10mV/div vertical and 2nS time base.  Trigger is set to single sweep to freeze the waveform.

This is a common problem when the probe tip is shorted to the ground clip; the loop created is large enough to pick up a lot of EMI.  Grounding the probe tip with a coaxial ground should produce a clean trace.

The only thing that properly flatlines the trace is setting the scope input coupling to ground.

Ground coupling does not actually ground the input on these oscilloscopes; instead it disconnects a later stage of amplification so the display does not represent the actual input noise.

If you zoom out to a more normal timescale it'll just appear as the classic fuzziness which is inevitable when dealing with a 200 MHz scope collecting 200 MHz BW worth of noise.  I'll admit though, it does look a bit more than I'd have expected.

The fuzziness is inevitable but that much?

We can measure the amount of noise to make a quantitative comparison.  The photograph showing the shorting cap and 500uV/div has about 0.7mV peak-to-peak noise.  Dividing by 5 produces a good estimate of RMS noise so about 140uV RMS noise.  My 200 MHz analog oscilloscopes under the same conditions have an accurately measured RMS noise of about 20uV.  A 7 times difference could be reasonable for a poor design.

One thing that is disappointing, though, is that Rigol don't seem to specify a noise value or figure in their datasheet. Not sure if it's standard practice to omit such a figure from the specs, but it certainly makes it harder to defend the performance of a product when they don't make a specific claim.

Any good marketing department knows not to advertise things which show their product in a poor light.  Rigol has a history of misrepresenting things.

Noise is an inevitability of using wide bandwidths, although my attempts to calculate a figure have given an unexpectedly high result. A plain old resistor, such as the termination resistor in the scope, of 1 megaohm at 20 degrees C produces 1.8mV RMS white noise when observing over a 200 MHz bandwidth. This is over 3 divisions on your scale, RMS, which would mean that your scope is performing better than a theoretically perfect scope, which is clearly wrong. I'm not sure how the 16pF capacitance of the frontend factors in, or whether I'm missing something else?

With an open input, the 16pF input capacitance has an overwhelming effect on the 1 megohm input resistance limiting the bandwidth to about 10 kHz.  With the input shorted, noise is dominated by the high impedance input buffer.

TLDR; I suspect you should watch the EEVBlog episode debunking the claim that digital scopes are noisier than analog. It does a good job of conveying that noise is an intrinsic part of life, not some design mistake that can be arbitrarily quashed. Clearly there is some sort of design imperfection in the Rigol scope, but still cruicially important to understand what a theoretically perfect scope looks like -- because it's not dead flat, ever.

Unfortunately Dave did not make any actual noise measurements although this would have been difficult on the analog oscilloscope he showed (Tektronix 2225?) because its input noise is about on the same level as its trace width.

There is a difference in how the noise is displayed which makes DSOs look worse but many DSOs are noisier than the analog oscilloscopes they replaced.

But i can't imagine a company would design a scope such that the highest couple of gain settings show up significant noise in their amplifier.

The high gain settings are 100% useful on more sensible timescales, when using acquisition averaging, or when using the bandwidth limit. So your expectation is unreasonable here.

Sample rate and time/div should have practically no effect on noise except changing the low frequency cutoff.

The highest gain settings are more useful when bandwidth limiting, averaging, or high resolution mode are used.  It was not uncommon for analog oscilloscopes to deliberately limit input bandwidth at higher vertical sensitivities.

[A Hellene]

- Everything that has been stated about EMI is more than valid.
- Even the suggestive hint about the FM radio band interference intercepted by the pick-up head being formed by the ~100nH inductor of the probe ground lead when it is attatched to the probe tip is also valid.
- Not to mention that I, myself, was really hesitant to scare our fellow EEVBlogger by my mention of a probable hardware (or poor design) fault of his electronic test equipment device.

BUT,

..what kind of everyday electromagnetic interference (meaning that we are not residing within the Van Allen Belts area!) can find its way through a BNC Shorting Cap, like the one depicted by the OP in the following picture?




-George

[agdr]

Great comments!    Some good insights in there. 

I was poking around Rigol's support site today and found this, "noise performance of the DS1000Z"

https://rigol.desk.com/customer/en/portal/articles/2285691-ds1000z-noise-performance

10mV/div with the input open.  They also have a link to that video of Dave's about the digital vs. analog scope noise:

https://rigol.desk.com/customer/en/portal/articles/2286590-why-do-digital-scopes-appear-noisy-

I had taken these two yesterday with the input open.

[tautech]

They also have a link to that video of Dave's about the digital vs. analog scope noise:

https://rigol.desk.com/customer/en/portal/articles/2286590-why-do-digital-scopes-appear-noisy-

For the full picture you can't watch one and not the other.
Episodes 601 and 610
https://www.eevblog.com/episodes/

[agdr]

For the full picture you can't watch one and not the other.

Ah, a part 2!  Here is a direct link:

[tggzzz]

..what kind of everyday electromagnetic interference (meaning that we are not residing within the Van Allen Belts area!) can find its way through a BNC Shorting Cap...

I missed the OP's subsequent reference to the shorting cap - but is it really a shorting cap or just a blanking cap? I've seen far more blanking caps (or 50ohm terminations) than shorting caps.

In my defence I'll note that some of my other suggestions about "what makes it change" indirectly encompass that possibility That a cap doesn't change it is obviously significant. From an EMI/EMC PoV, tying an input line to ground directly/50ohms could have a significantly different effect than enclosing it at one end.

However, some of the other posts indicate there may be more fundamental issues in this low-end scope. For example, "Ground coupling does not actually ground the input on these oscilloscopes; instead it disconnects a later stage of amplification so the display does not represent the actual input noise" would mean that offsets earlier in the input chain would "disappear" when GND is selected.

Personally I have nothing against the Rigol scopes, but they are a reminder why professional engineers always preferred HP and Tektronix scopes: there were few "surprises" waiting to be discovered.

[A Hellene]

There is no need to be defensive! This is a discussion about exchanging our knowledge and our educated opinions on a problem introduced by one of us to their probably more knowledgeable counterparts of this community.

In the picture above, there is a (copper, maybe, by its colour) pin machined in the center of the cap that suggests that this pin holds something that might be connected to the opposite (female) BNC input. Ideally, that pin should introduce a 50.0 ohm resistance to the BNC input.

Now, in my opinion, the Ground Coupling option has nothing to do with hardware (since there is no input shunting hardware on board in the front-end analogue section, according to my schematics of DS1000E) but it merely is a firmware gig that blocks any actual input feed to the video processor for the channel in question --thus, the flawless depicted flat line that lacks even the ADC stage digitisation error/jitter (the quantisation noise), which is inevitable.


-George

[A Hellene]

By the way, agdr, speaking of 50 ohm input termination methods,

Does your DS2102 have the 50 ohm input impedance option? I think that all the later DS2000 models have that option (I am not sure, though, if that option is standard or purchasable).

If it does, try enabling it without anything else connected to the BNC and watch the trace: It should be as clean as possible, with noise levels of less than 1/5 of a division typically, even at the highest gain settings! That is because the standard 1.0 Mohm input impedance in parallel with a merely less than 15pF capacitance introduces very higher levels of thermal noise than the 50 ohm one.

If it does not, get a 51 or a 47 ohm (1/8W or 1/4W) resistor, trim its leads as short as possible and insert it to the BNC input connector without touching the center BNC line. Watch the trace again.

Then, you will know with a fair amount of certainty if you device is problematic or not.


-George

[Gixy]

I have an MSO2072A transformed into MSO2302A (300 MHz). In the same operating configuration, the noise is much more lower than shown by the OP. I'm wandering if his scope is self-calibrated. I had some strange behaviours which have been solved after this self-calibration function (utilities/system menu).

[Fungus]

With the input shorted at the BNC jack , expecting a flat line is not unreasonable.

I beg to differ. Of course it is unreasonable to expect a flat line (if you crank the vertical amplification up to the maximum). I am flabbergasted by this whole discussion, in a TME forum! The only valid questions one should ask here are: "What level of noise should one expect? What level of noise do other scopes produce, under comparable circumstances"

This.

[agdr]

I have an MSO2072A transformed into MSO2302A (300 MHz). In the same operating configuration, the noise is much more lower than shown by the OP. I'm wandering if his scope is self-calibrated. I had some strange behaviours which have been solved after this self-calibration function (utilities/system menu).

Good thought!  I just gave it a try, photo below with the cal in progress.  lol - "Last Calibration Time: 2015-9-4".  9/2015 was at least one Rigol software update ago, probably two.   But unfortunately no effect on the problem.  I put the BNC shorting cap back on (inputs unconnected for calibrate), cranked it up to 500uV/div and 2ns, and the same waveform is there.

BUT.. I did seem to hit a nerve on something.  The suggestion was made to try limiting the  bandwidth.  I set the 20MHz bandwidth limit and the result is the second photo.  Looks like normal noise now!  The amount and type of noise that Dave describes in his videos (still on high-gain 500uV/div here).  Can't get it to trigger either, scanning the trigger voltage level through the whole thing, which may mean random noise (?)  as opposed to something periodic.  The p-p amplitude appears to be about 1/3 of a division, so maybe 500uV/3 = 166uV.  Using the 16uV number calculated for thermal noise earlier in the thread that would be roughly 10x, so probably more going on there than just thermal/Johnson. 

Going up to 100MHz with the bandwdith limit makes the amplitude about 30% larger, but still appears to be mostly random noise with just the occasional flash of a defined waveform.  Then removing the BW limit entirely and I'm back to that original large(r), complex, waveform.

So a few of quick theories come to mind.    One is EMI entry at the higher frequencies, as has been suggested, but keep in mind a BNC shorting cap is on the input.  The EMI would have to be getting into the internals of the scope, past any shielding.  At such a high level of gain though it is probably quite possible.  Another idea is something to do with the 200MHz software hack.  Keep in mind this is a 100MHz scope, DS2102, software hacked to 200MHz, That might explain why it seems to be more-or-less behaving up to 100MHz.  Yet another idea is that PLL and PLL instability described in the Yaigol thread affecting the ADC.  This is the "older" scope hardware, the non-A (DS2102 not DS2102A).  It may be more likely to have that PLL problem.  I'm just assuming fixing the PLL may be one thing Rigol did with the change to the "A" version, more stuff than just adding the 50R input option. That might explain why yours doesn't appear to have the noise waveform, being the "A" version.   This one is running the latest software though, or at least what was latest last I checked, 00.03.05,  HW = 2.0, from the system info screen.

[tggzzz]

There is no need to be defensive! This is a discussion about exchanging our knowledge and our educated opinions on a problem introduced by one of us to their probably more knowledgeable counterparts of this community.

Ah, I don't think anyone has ever accused me of being defensive before In this case I'm irritated that I didn't spot the shorting cap, and barged into the middle of a conversation!

More seriously, your attitude exactly mirrors mine; it is a shame more people don't think that.

[ebastler]

I set the 20MHz bandwidth limit and the result is the second photo.  Looks like normal noise now!  The amount and type of noise that Dave describes in his videos (still on high-gain 500uV/div here).  Can't get it to trigger either, scanning the trigger voltage level through the whole thing, which may mean random noise (?)  as opposed to something periodic.

Going up to 100MHz with the bandwdith limit makes the amplitude about 30% larger, but still appears to be mostly random noise with just the occasional flash of a defined waveform.  Then removing the BW limit entirely and I'm back to that original large(r), complex, waveform.

I would say that the main difference between the 20 MHz bandwidth screenshot and the ones you shared earlier is that for the 20 MHz capture, you did not have the scope set to single shot mode ?

I am not convinced that the single shot screenshots really show a significant periodic signal. It's easy for our eye/brain to interpret some structure into a more or less random curve. Could you try capturing an FFT on the "noise" signal, to see if there is a pronounced periodic component in it?

[MrFox]

I have a DS2072A unlocked to 300MHz and I'm not seeing anything out of the ordinary, at least nothing that isn't explained in Dave's wonderful two parts video above.

When measuring wideband noise I would think the trigger must disabled. Otherwise it will trigger on random peaks with no indication of how frequent or rare these are. I mean it can't be called periodic unless you see other stable peak at a fixed distance from the trigger point in run mode. Each trigger is a brand new time reference.

I noticed the triggering seems to have a minimum cutoff point, it will only trigger on steps higher than maybe 200uV (at 1x). So when you place your trigger level in the center, it will only show the peaks that are above that threshold.

This becomes obvious when measuring a real signal like a 2mV sine wave,  it ends up looking much clearer than you'd expect. It's a relatively fat trace but nothing like the above noise screenshot.

[Andreas]

I had taken these two yesterday with the input open.

Holy crap, what are you doing?

The facts:
- you are sampling with 2ns/div (I did not believe that since it could not be seen on the fuzzy photos before)
- The ADC can deliver at maximum 2 GS/s (single channel) so in every div only 4 ADC measurements
- some "snake oil software" (on automotive you would say defeat device) then interpolates the 700 points for the display from that

So don´t believe that what you are seeing in this case.
There cannot be some waveform below 1 ns rise time on a 200 MHz bandwith signal.

With best regards

Andreas

[agdr]

I am not convinced that the single shot screenshots really show a significant periodic signal. It's easy for our eye/brain to interpret some structure into a more or less random curve. Could you try capturing an FFT on the "noise" signal, to see if there is a pronounced periodic component in it?

Good idea!  I know the FFT ability in this scope is meager, but it should be enough for the purposes here. 

The first photo here is the FFT with the input set to the 20MHz bandwidth limit.  Hanning window.  The second photo is the same thing but with the input bandwidth limit removed.  Interesting!  There is a small peak of some kind between the middle frequency reticle and the one of the left.  Also appears to be a low(er) frequency peak around the first reticle.

[mstoer]

I have the DS2102 as well, bought about 4 years ago now and optioned up to 300 MHz.  I have the same noise with the probe tip (antenna style) grounded at full bandwith,  but I expect to see it like that.

If I put a 50 Ohm ground cap right on the scope it goes flat.  There is no option to change the input to 50 Ohm on mine, the menu shows it as a greyed out option set to 1 MOhm.

[agdr]

I have the DS2102 as well, bought about 4 years ago now and optioned up to 300 MHz.  I have the same noise with the probe tip (antenna style) grounded at full bandwith,  but I expect to see it like that.

If I put a 50 Ohm ground cap right on the scope it goes flat.  There is no option to change the input to 50 Ohm on mine, the menu shows it as a greyed out option set to 1 MOhm.

Hey thank you for the input!  It sounds like you save the same pre-"A" hardware as mine (no 50R input option on mine either, greyed out the same way).   I've tried the shorting cap, but I haven't tried a 50R cap.  I actually have been thinking about that. I don't have any here but I'm sending an order out to Mouser in a couple of days and will get a couple.    Who knows, maybe something in the vertical amp goes unstable with zero ohms on the input.  Certainly worth a try to see if I can get similar results with mine with a 50R cap.   

[agdr]

I noticed the triggering seems to have a minimum cutoff point, it will only trigger on steps higher than maybe 200uV (at 1x). So when you place your trigger level in the center, it will only show the peaks that are above that threshold.

Interesting!  I'll have to ponder that triggering information a bit and mess with it on the scope.   Thanks!

[agdr]

I thought that I had found the solution a couple of day ago when I ran across these two YouTube videos from 2013 about noise on a Rigol DS4024 scope:





The type and level of noise he is showing is exactly what I'm getting, with the center peak.  In the comments on the second video the poster states that he found the problem, EMI from his laptop power brick.  Well I have a laptop brick (Sony) plugged into that same outlet the scope is plugged into.  Has to be the problem, right?

I moved the scope out to a room that has a dedicated  power plug (standard NEMA 5-20) fed by 6AWG wire (10AWG pigtails to the plug), is on a dedicated 20A breaker, and the whole thing is just 80ft away from the power drop from the pole and utility electric meter.  The 20A breaker feeding it is in a panel attached to the wiring gutter right off the electric meter.  Doesn't get much more direct than that.  In addition to a 8AWG ground wire the plug is in a metal box attached to a metal EMT conduit that attaches to the gutter, and the whole thing is grounded at that point with a 8ft copper-clad ground rod (electrical entrance ground).  This room used to have a hot tub in it, lol, hence the 6 gauge wire.    Now it makes a great conducted-EMI test room with no other electrical or electronic devices within a 60ft radius, home run wiring to the electrical entrance, and solid ground connection.

THEN I attached an old Waber EMI filter plug-strip I have, back before Tripp Lite bought them, similar to their Isobar filters now.  Cascaded EMI filters, the first plug is rated at -75dB and the second -100dB.

After all that: same problem remains.     Exactly the same waveform with the same test conditions.  So I can say with some confidence the problem isn't powerline-conducted EMI.  I still need to look up the frequency response curve of that powerline filter to verify it is good through 200MHz.

[ebastler]

So what?! You are looking at noise, and triggering on it -- with the trigger threshold set fairly high. And hence, what a surprise, the scope will select and show those traces (out of many random noise samples) which have a positive edge exceeding the trigger level. It will show this positive edge in the center of the screen, where the trigger time point is shown. That's exactly what you have set the scope up to do, isn't it?

I don't think you are seeing anything out of the ordinary. It's just noise. Time to move on, there's nothing to see here... ;-)

[agdr]

The 50 ohm BNC terminators finally arrived from Mouser,  Amp #46650-51RFX:

http://www.mouser.com/Search/ProductDetail.aspx?R=46650-51RFXvirtualkey52310000virtualkey523-46650-51RFX

net result: exact same noise waveform.  It was certainly a good thing to try!

Time to get a real scope.   Luckily I had planned on upgrading scopes this summer anyway.  The Keysight DXO4000's look nice and I trust they won't have this problem. Can any 4000 owners confirm?  The DSOX4034A is what I have my eye on.

In the photos below - first is the 50R terminator on Ch1 and second is with acquisition set to averaging/16.   500uV/div and 2ns time base again.

[TurboTom]

We've got an MSOX4034A at work and guess what - the last "real" sensitivity of the input amplifiers is 5mV, lower than that (i.e. 2mV and 1mV) is just a digital magnification of the 5mV signal. The scope wouldn't trigger on noise at that low level with the input terminated, but in free running mode and in the 1mV range, the noise easily fills a graticule peak-to-peak (which is completely normal).

So I guess regarding the noise / low level signal quality issue, you won't be happier with the Keysight than with the Rigol (I've got a MSO2072A "+" at home and more or less can reproduce your findings, but I don't have any complaints regarding it's low level / noise characteristics). Other than that, of course the Keysight is a completely different machine with a much smoother user interface and better overall performance.

Cheers,
Thomas

[David Hess]

We've got an MSOX4034A at work and guess what - the last "real" sensitivity of the input amplifiers is 5mV, lower than that (i.e. 2mV and 1mV) is just a digital magnification of the 5mV signal. The scope wouldn't trigger on noise at that low level with the input terminated, but in free running mode and in the 1mV range, the noise easily fills a graticule peak-to-peak (whcih is completely normal).

That level of noise is about right.  A noisy 100 MHz input like from a wide common mode range differential input could be about 200uVrms or 1mVpp.  A clean 100 MHz single ended input could be about 20uVrms or 100uVpp.  Taking into account the 350 MHz bandwidth of the MSOX4034A, it should be between these which it is although it is noisier than it could be.

[ebastler]

Time to get a real scope.  
[...]
In the photos below - first is the 50R terminator on Ch1 and second is with acquisition set to averaging/16.   500uV/div and 2ns time base again.

@agdr -- My earlier comment may have been a bit harsh, but please read it anyway. It can save you a few $1000 for buying a new scope which would most likely show the exact same behavior...

You are displaying a rather artificial signal there: Via your trigger threshold, you explicitly select the sporadic high noise peaks, and have these displayed in the center of the screen. There is nothing systematically wrong in your scope's input stage, the input signal the scope sees is completely random But you explicitly set up the scope to filter out (in the time domain) these higher peaks which occur sporadically, and display them in the center of the screen.

Switching your scope to averaging mode probably makes this distortion worse: The peak at the trigger position will show up in each averaged trace, and will be as pronounced as before (because each triggered scan picks it out). On the other hand, the random noise before and after the peak will differ between scans (because it is random), and will hence be reduced by the averaging. This makes the "artificial" peak in the center even more pronounced. I still think that what you see is absolutely normal; the scope does what you tell it to do.

By the way -- could you please try saving a screenshot to USB, or transferring it to the PC, and post that instead of a photo of the screen? It is hard to make out any of the printed information on the blurry photographs.

[TurboTom]

FYI and just as a comparison to "agdr's" FFT results (https://www.eevblog.com/forum/testgear/rigol-ds2102-grounding-the-probe-doesn't-flatline-the-trace/msg1149251/#msg1149251) where he configured his DS2102 a little awkward ( -- no offense intended) so only the one and a half leftmost graticules of the FFT contain any relevant information, I decided to take some similar readings with my MSO2072A ("optimized" to full 300MHz bandwidth). The first screenshot is taken with the input directly terminated with 50 ohms. Unfortunately, there's no averaging available on the FFT trace so the screenshot is just a momentary thing. Actually, the FFT fluctuates with no visible local peaks or the like, as one would expect it. The two horizontal cursor lines approximately indicate the window within which the level lies. The maximum of -80.6dBm at 50 ohms equals an RMS noise voltage of approx. 22µV. But without proper FFT averaging and exactly knowing the resolution bandwidth, it's not possible to calculate the RMS noise voltage in the time domain from the FFT. Yet, the average measurement value of round about 135µV RMS could be approximately correct.

If the stock 10:1 probe is attached to the scope, the ground lead directly shorted to the tip clamp, the next screenshot is what one will get. Some averaging and a little tweaking of the FFT parameters clearly shows the six strongest FM radio stations and a few weaker ones as well. And I'm living in a rural area with no strong FM transmitters in the direct neighbourhood, the next one being located maybe some 20km from my place (behind a hill). This is a nice indication how sensitive the unshielded sections of a standard probe is to radiated RF. And it may also shine a different light on what's supposed to be "input noise". Another quite interesting find is that I've got difficulties visualizing the FM stations with the same probe on a DS1054Z ("improved") despite its better FFT functionality. So the analog frontend of the DS2000(A) series of scopes appears to be quite something.

Cheers,
Thomas

[David Hess]

FYI and just as a comparison to "agdr's" FFT results (https://www.eevblog.com/forum/testgear/rigol-ds2102-grounding-the-probe-doesn't-flatline-the-trace/msg1149251/#msg1149251) where he configured his DS2102 a little awkward ( -- no offense intended) so only the one and a half leftmost graticules of the FFT contain any relevant information, I decided to take some similar readings with my MSO2072A ("optimized" to full 300MHz bandwidth). The first screenshot is taken with the input directly terminated with 50 ohms. Unfortunately, there's no averaging available on the FFT trace so the screenshot is just a momentary thing. Actually, the FFT fluctuates with no visible local peaks or the like, as one would expect it. The two horizontal cursor lines approximately indicate the window within which the level lies. The maximum of -80.6dBm at 50 ohms equals an RMS noise voltage of approx. 22µV. But without proper FFT averaging and exactly knowing the resolution bandwidth, it's not possible to calculate the RMS noise voltage in the time domain from the FFT. Yet, the average measurement value of round about 135µV RMS could be approximately correct.

Won't the automatic RMS measurement function find the noise?  Or if that does not work, what about the standard deviation of the voltage which should produce the same result?

I know this does not work on the DS1000Z series but hoped that the DS2000A series was different.

[ebastler]

Won't the automatic RMS measurement function find the noise?  Or if that does not work, what about the standard deviation of the voltage which should produce the same result?

I think the point was to check agdr's assumption that the noise was not random, but that there were some systematic frequency components to it. (Based on over-interpreting the interpolation artefacts he had seen at maximum time resolution.) Hence the FFT to check for the frequency spectrum. If one only wants to quantify the overall noise level, an RMS measurement should be fine.