How Thick is Your Baselne on Your Digital Oscilloscope?

[rf-loop]

It is very nice of someone do tests and show images about:

Scope (DS2000: 70, 100 or 200MHz model) settings 1 or 2 channel in use.
Full BW
No average, no any kind of filtering, no high res mode.
Horizontal speed 1ms/div
Acquire (sampling) mode peak (or how it is named in Rigol) and normal 8 bit mode (no high res)
(or if not peac acquire mode, with 14M memory selected and then scope stopped and zoomed out so that more data is displayed)
Inputs open. (or external 50 ohm terminators on channels BNC. specially if envinroment is noisy)

(Just all settings so that pure maximum level of front end peak-peak noise is visible on the screen.)
Probe factor 1x

Vertical:
500uV/div
1mV/div
2mV/div
5mV/div
10mV/div

[Paul Price]

I have discovered that if I select Hi-Res mode and AUTO trigger on my DS4024 and set the time/div to 100uS/div or faster, then the display is very quick to update and the line is a flat, but dim, as can be possibly be displayed. The only problem here is that if there is any repetitive signal in the D.C. waveform the scope goes nuts.

Thanks again for everyone for taking the time to post pictures!

From what you all have seen here, which scope do you think is best in having the best thin baseline display?
Of course I must consider all the pros and cons of a scope besides this one behavior and there are many things I like about the DS4024 and I think overall, I like the cost v. performance of the DS4024.

I am going to waste a fortune of inkjet ink and print some of these pictures out and compare them to my baseline.

[ejeffrey]

Generally if you want to accurately get the DC value you would just use the scope measure functionality either on the whole trace or with cursors.  That isn't to say it isn't worth figuring out how to get a cleaner signal display on your DSO, but you should be able to quickly measure voltages without it.

[JuanPC]

From what you all have seen here, which scope do you think is best in having the best thin baseline display?

under $10kusd., the only scope that can go to 0.5mV/div is the Rigol DS2202,
no scope can do all things,

but then again, when i had Instek gds-1102A-U,
the baseline when connected directly to the wall was higher & unstable vs. connected to Ps Audio Power Plant Premiere, small & stable, if i made a short on the tip of the probe with the ground, was even smaller, i could see a very thin base line way under <1mV/div.
Better Power first.

All DSO/MSO have a Switch Mode PSU, and those are noisy Meanwell are around +/-100mV ripple & noise, i´m surprised to see that a scope with a switch mode psu has a 1mV/div sensitivity.
A switch mode psu is affected by external power coming from the wall, increasing noise.

If you want better than Ps Audio Power Plant Premiere, the New P5 or P10, and/or the Accuphase PS-  series.
http://www.accuphase.com/model/ps-1220.html
http://www.accuphase.com/cat/ps-1220_e.pdf

i don´t know how stable is the regulation of the Accuphase, but the PPP is flawless 120.5v.
the Accuphase noise is near -84.375dB.
The Signal Generator in the Accuphase is Analog, in the PS Audio is DDS.

[Galaxyrise]

Thanks for the pinup's JBP. This looks twice as flat as my Rigol which shoes about 1-minor div noise bar regardless of the volts/div setting when viewing a top or baseline bottom of a clean square wave.

That just sounds like ADC jitter to me.  It's an 8-bit ADC and each value corresponds to one pixel, so a 2 pixel line is as good as you can get without some form of averaging.

The DS2000 provides three forms of averaging: Average mode, which averages the results of multiple captures.  As you noted, this is not what you want.  Anti-Alias, which weights the intensity of each pixel based on neighboring pixels, and Hi-Res, which averages multiple sample points to generate each pixel.  (I find it best not to mix these different averaging methods as having multiple active can produce some ugly results.) I use Anti-Alias most of the time, the main effect is just to enhance the trace.  I've had Hi-Res severely distort the signal, so I don't use it any more. Both Anti-Alias and Hi-Res can slow down the capture rate depending on the other settings, as you noticed.

Another tip is to go into Dots display instead of Vectors.  The sin(x)/x interpolation can cause the trace to go above the measured values. 

Turning the intensity down can help minimize the appearance of noise. Since you're just looking to measure DC, you can turn down the memory depth, too.

This is the thinnest trace I can get out of my DS2102, and I would guess you can achieve similar results:

[JuanPC]

This is the thinnest trace I can get out of my DS2102, and I would guess you can achieve similar results:

Please Make one with 0.5mV/div.
one with 1mV/div
other with 2mV/div.

shorted probes.

[Galaxyrise]

Please Make one with 0.5mV/div.
one with 1mV/div
other with 2mV/div.

shorted probes.

I'm not sure what you're after; there's so many settings to mess with!  With just a 50R terminator attached and no BW limit, the trace is about a division tall and the scope reports 880uVpp at 100us/div.  Turning on the 20MHz bandwidth limit cuts the noise in half, the trace is about half a division tall, and the scope reports 400uVpp.  Turning on Anti-Alias while displaying vectors makes the center of the noise much brighter than the outside, making the apparent trace thinner. With dots, I didn't notice a difference (though it looks like 500uV/div is done in software, zooming in on 1mV/div.) Using High Res instead of Anti-Alias results in a one-pixel trace and a reported Vpp of 100uV.

Not counting HighRes (which always shows a thin line), going to larger V/div makes the trace thinner until it's like in my screenshot.

[Uup]

@ Paul, is the following what you see on your DS4024 at 10mV/d?

10mV/d no bandwidth limit
10mV/d 20MHz bandwidth limited
10mV/d 100MHz bandwidth limited
10mV/d Averaging on (2 ave)
10mV/d Hires mode

I also tried with a 50ohm terminator attached and it made no noticeable difference.

[Galaxyrise]

It is very nice of someone do tests and show images about:

Scope (DS2000: 70, 100 or 200MHz model) settings 1 or 2 channel in use.
Full BW
(Just all settings so that pure maximum level of front end peak-peak noise is visible on the screen.)

As I have a 2102, I implicitly some bandwidth limiting.  Did you want dots or vectors?
Dave did show what the noise floor looks like in his DS2202 playing around video (#369, around 10 minutes in.)

[rf-loop]

It is very nice of someone do tests and show images about:

Scope (DS2000: 70, 100 or 200MHz model) settings 1 or 2 channel in use.
Full BW
(Just all settings so that pure maximum level of front end peak-peak noise is visible on the screen.)

As I have a 2102, I implicitly some bandwidth limiting.  Did you want dots or vectors?

2102 have of course some amount less noise as 2202.

the purpose is to see the actual analog front-end peak to peak noise level (including ADC).

Peak-to-peak instead of rms noise level is what we meet with the measurements if can not use any methods for hide this noise. Of course rms numbers, depends of method, give more nice values and manufacturers like show these, if they show any number about noise levels.

Agilent: http://cp.literature.agilent.com/litweb/pdf/5989-3020EN.pdf
"Peak-to-peak Noise:  the peak-to-peak range of noise in an oscilloscope based
on a particular criterion such as time, number of acquisitions, and/or acquisition
memory depth.

RMS Noise random noise measured as one standard deviation"

It is just the same as the pulse generators, for example, where the manufacturer like to tell us  the rms jitter value, but the user will have to live, however, with peak to peak values. And more fun come when they do not tell us how they have measured and calculated rms. If now need sure know what is real peak to peak example so that 99.99% of samples are all inside some value (IF distribution is gaussian true random)... well... with what number you multiply this rms value. Is it ten or more...

Sure, I know that the noise images are "ugly", but it is a reality.

Of course, in practice, it is possible depending on the situation, "filter" produce noise out of sight. For example, using the BW limit, or waveform average (usual average) or High-Res (other kind of averaging with low speeds as long as true to the ADC sample rate is much higher than displayed sample rate)

No average, no any kind of filtering, no high res mode.
Horizontal speed: 1ms/div
Memory: 14M  (memory depth not auto but 14M selected. Becouse acquisition peak detect there can also use example 1.4M or even less, no markable difference?)
Acquisition: Peak Detect
Sampling: Real time
Antialising: Off
Coupling: DC
BW limit: Off
Probe 1x (Edit: in oscilloscope menu! Probe is not at all connected, open or terminated inputs)
Display: Vectors.
Sin(x)/x imitation or what ever it is and all other possible after ADC user adjustable filters off (if there is any).

Scope Input BNC: Just open, nothing connected (if have 50 ohm terminator and if it makes difference, then with terminator and without)   (envinroment need be quite silent about EMI)
Also 50ohm terminators can use so that center pin do not make connection but still terminator works as emi shielding cap.

Vertical:
500uV/div   
1mV/div
2mV/div
5mV/div
10mV/div

(scope internal TFT images)

[JuanPC]

Probe 1x

Inputs: Just open, nothing connected (if have 50 ohm terminator and if it makes difference, then with terminator and without)   (envinroment need be quite silent about EMI)
Also 50ohm terminators can use so that center pin do not make connection but still terminator works as emi shielding cap.

X1 is debatable, better to include X10 too.


Probe Shorted is better.

[Paul Price]

Thanks to UUp for clearly showing the problem, especially when trying to measure mV DC signals. What good is 10mV/div sensitivity if it is swamped out by noise.

But a Big Thanks again to everyone who posted pictures and comments.

Ejeffrey, the idea here is to quickly visually make DC measurements. Setting up cursors is not going in that direction.

Well, things are drifting a little off topic  into ways to make some unusual noise on a DPO/DSO oscilloscope. No thanks, I have enough!

I like to think my scope as being the universal measurement tool and the idea here is to very quickly visually measure various power supply and other important bias points on a circuit using the scope as a DVM and relying on a thin baseline positioned on the bottom or top graticule line with DC coupling to get an accurate voltage reading just off the graticule.

This was something easy to do with my analog scope and a very useful way to quickly check out every DC voltage quite accurately on whatever circuit I am working on in just a few seconds. Having a 1 minor division or so fuzzy baseline lowers the precision of measurement by an order of magnitude. Having one full division is unconscionable.

[rf-loop]

Probe 1x

Inputs: Just open, nothing connected (if have 50 ohm terminator and if it makes difference, then with terminator and without)   (envinroment need be quite silent about EMI)
Also 50ohm terminators can use so that center pin do not make connection but still terminator works as emi shielding cap.

X1 is debatable, better to include X10 too.

No. If we measure oscilloscope itself front end noise 10x selection in scope do not change anything in front end. If scope inputs are open or terminated there need use 1x setting in scope  probe setting menu so that numbers what scope show are ok.
For this measurement do NOT connect probe.

If test with probe, it is other tests what I ask.

[Galaxyrise]

Display: Vectors.
Sin(x)/x imitation or what ever it is and all other possible after ADC user adjustable filters off (if there is any).

I can't have vectors on but six(x)/x off.  Which would you rather have, sin(x)/x or dots?

[rf-loop]

Display: Vectors.
Sin(x)/x imitation or what ever it is and all other possible after ADC user adjustable filters off (if there is any).

I can't have vectors on but six(x)/x off.  Which would you rather have, sin(x)/x or dots?

Ok, then dots, least 5s persistence.

[PA4TIM]

No filtering etc, 350 Mhz Hameg just standard setup. 1mV/div 10 MHz sinewave from a HP8640 generator

[egonotto]

Hello,

almost no noise at 1mV/DIV.
I am impressed.
Or is it a trick?

Best Regards
egonotto

[Hydrawerk]

Remember that this Hameg has only 2500 waveforms per second update rate.

[rf-loop]

Remember that this Hameg has only 2500 waveforms per second update rate.

How it is related to analog front end noise, can you explain?

-------------------
But also, this was not this kind of noise test what I have asked about Rigol.
Why it is so difficult to do these noise tests. These do not need any equipment. Just oscilloscope.

[rf-loop]

No filtering etc, 350 Mhz Hameg just standard setup. 1mV/div 10 MHz sinewave from a HP8640 generator

Can you show withoiut signal, peak acquire mode, maximum memory, 1ms/div
1Mohm and 50ohm inputs, 1mV, 5mV and 10mV/div (all possible filtering off)
So we get some "reference" before someone show Rigol data.

[ejeffrey]

Ejeffrey, the idea here is to quickly visually make DC measurements. Setting up cursors is not going in that direction.

Once you set up a measurement you can test many values and it will just show you the value.  On most scopes it takes about four button presses to set up a simple DC measurement.  This seems like a really weird thing to bitch about when you are willing to go fiddling with different BW limits, acquisition modes, and averaging to get your clean baseline.

[rf-loop]

Here is tested with very cheap Siglent SDS1102CML
Oscilloscope inputs just open.
Filters off. (user adjustable channel DSP filters off)


Acquire normal.
Note that also Agilent there normal acq but cursors peak.
(agilent insert picture scale is adjusted for compatible with Siglent display)


Acquire peak mode.
Note that Agilent there normal acq but cursors peak.
Here 2mV/div can not compare to other level settings because it have fixed 20M BW.
And even as you see these fat traces, this result is good and acceptable in its class.



Then here is old table for Owon (old version and more new version)





And also same kind of test with new Owon.


Peak mode.
Note that Agilent have normal acquire mode and in image also low samplerate. So image is more nice for Agilent than if it have done with same settings.

There have been lot of these tests available and long time for example Siglent and Owon.
But  example for Rigol what is more expensive and claimed with many words about special low noise front end. But real data is difficult to find.

[PA4TIM]

No filtering etc, 350 Mhz Hameg just standard setup. 1mV/div 10 MHz sinewave from a HP8640 generator

Can you show withoiut signal, peak acquire mode, maximum memory, 1ms/div
1Mohm and 50ohm inputs, 1mV, 5mV and 10mV/div (all possible filtering off)
So we get some "reference" before someone show Rigol data.

Why ? That is a rather useless test, If I have much RFI here then it will show more noise as in a surroundig where no other gear is powered on. A better test is like my picture, a very small signal from a good generator in a 50 Ohm environment. That is what you want to know, how does it performe while measuring small signals. If there was a lot of noise in my trace it could only be the scope or the generator, nothing else ( I use good mil specs cables)

With open inputs I can be everything. Like comparing the readings of multimeters while connected to nothing.

And no this is no trick, look at my website, there are many screenshots from real meaurements made with this scope.

What you want is something like comparing cars idling while waiting for a trafficlight.

[rf-loop]

No filtering etc, 350 Mhz Hameg just standard setup. 1mV/div 10 MHz sinewave from a HP8640 generator

Can you show withoiut signal, peak acquire mode, maximum memory, 1ms/div
1Mohm and 50ohm inputs, 1mV, 5mV and 10mV/div (all possible filtering off)
So we get some "reference" before someone show Rigol data.

Why ? That is a rather useless test, If I have much RFI here then it will show more noise as in a surroundig where no other gear is powered on. A better test is like my picture, a very small signal from a good generator in a 50 Ohm environment. That is what you want to know, how does it performe while measuring small signals. If there was a lot of noise in my trace it could only be the scope or the generator, nothing else ( I use good mil specs cables)

With open inputs I can be everything. Like comparing the readings of multimeters while connected to nothing.

And no this is no trick, look at my website, there are many screenshots from real meaurements made with this scope.

What you want is something like comparing cars idling while waiting for a trafficlight.

Perhaps it is too difficult to enclose inputs example with connected or not connected terminator.
Of course tests are good to do with good generator. I use in my tests HP8644B and or HP8642B.
But on thois forum there is many many peoples who do not have equipments.
This test with no signal is much more comparable as tests with what ever unknown signals.

Perhaps also Agilent do something wrong? They do not test with signals here.

No filtering etc, 350 Mhz Hameg just standard setup. 1mV/div 10 MHz sinewave from a HP8640 generator

Can you show withoiut signal, peak acquire mode, maximum memory, 1ms/div
1Mohm and 50ohm inputs, 1mV, 5mV and 10mV/div (all possible filtering off)
So we get some "reference" before someone show Rigol data.

Why ? That is a rather useless test, If I have much RFI here then it will show more noise as in a surroundig where no other gear is powered on. A better test is like my picture, a very small signal from a good generator in a 50 Ohm environment. That is what you want to know, how does it performe while measuring small signals. If there was a lot of noise in my trace it could only be the scope or the generator, nothing else ( I use good mil specs cables)

With open inputs I can be everything. Like comparing the readings of multimeters while connected to nothing.

And no this is no trick, look at my website, there are many screenshots from real meaurements made with this scope.

What you want is something like comparing cars idling while waiting for a trafficlight.

Perhaps it is too difficult to enclose inputs example with connected or not connected terminator.
Of course tests are good to do with good generator. I use in my tests HP8644B and or HP8642B.
But on thois forum there is many many peoples who do not have equipments.
This test with no signal is much more comparable as tests with what ever unknown signals.

Perhaps also Agilent do something wrong? They do not test with signals here.

No filtering etc, 350 Mhz Hameg just standard setup. 1mV/div 10 MHz sinewave from a HP8640 generator

Can you show withoiut signal, peak acquire mode, maximum memory, 1ms/div
1Mohm and 50ohm inputs, 1mV, 5mV and 10mV/div (all possible filtering off)
So we get some "reference" before someone show Rigol data.

Why ? That is a rather useless test, If I have much RFI here then it will show more noise as in a surroundig where no other gear is powered on. A better test is like my picture, a very small signal from a good generator in a 50 Ohm environment. That is what you want to know, how does it performe while measuring small signals. If there was a lot of noise in my trace it could only be the scope or the generator, nothing else ( I use good mil specs cables)

With open inputs I can be everything. Like comparing the readings of multimeters while connected to nothing.

And no this is no trick, look at my website, there are many screenshots from real meaurements made with this scope.

What you want is something like comparing cars idling while waiting for a trafficlight.

Perhaps it is too difficult to enclose inputs example with connected or not connected terminator.
If you afraid fat trace come from envinroment.

I recommend this test becouse most of peoples example here in forum do not have enough good signal sources. I have not problem with signal for tests, normally I use HP8644B and or HP8642B.

This test without signal is simple and very easy get image about noise what is comparable.
So for comparison reasons I ask if you can do it with hameg for some kind of reference.

Also I wonder why we can not see tests made with Rigol DS2000 what is claimed as very low noise front end.

Also I can not keep thios document as bullshit. Why they use this method without signal.
http://cp.literature.agilent.com/litweb/pdf/5989-3020EN.pdf

[PA4TIM]

From that document: Each scope was terminated into 50 Ohm

So no open input