DS1054Z: Weird Screen Clipping Behavior

[swgertsch]

New to oscilloscopes, so I have been reading through the User Manual and testing my understanding of what I read. Came across the option to "invert" the signal. (Not sure when I would want to invert the signal, but I wanted to see it in action.)

I was curious about how well it worked so I connected 2 probes to the calibration source and inverted one, then added them together using the Math A+B function. In theory the 2 signals (CH1 & CH2) should cancel each other out with the resultant being a flat line; I wanted to see this happen.

Worked perfectly as shown in the DZ1Z_expected.jpg. But this is where things got weird...

I changed the vertical scale of CH1 from 2V to 5V. There was a noticeable relay click sound indicating a major circuit path change taking place internally in the DS1054Z. The noise level of the A+B resultant increased significantly as shown in DS1Z_5V.jpg. Not sure why this happened.

Now here is the really weird part...

Changed the vertical scale of CH1 to 200mV which caused the top of CH1 to extend beyond the display extents. CH1 obviously got clipped from the display, but the part of the signal that was clipped got removed from the Math A+B resultant and shown in DZ1_200mV.jpg.

Being a newbie, I do not think this is the correct behavior. If I had to guess, I'll bet the software is not discriminating between clipping for internal signal value limits/bounds and display bounds; there are equating the two - display clipping affects the internal value of the signal and therefore is influencing all the internal math.


     

[onlooker]

The "violet (math)" is in 2V scale and the "yellow" is in 5V scale.

That is, "math_thickness" ~= 2.5 x "yellow_thickness" + 1 x "blue_thickness".

Your "math_thickness" is actually doing better than this or, in other words, nothing is really wrong.

Edit: for independent random noises, one can use root mean square to superposition the 2 signals. If the 2 signals are opposite in phase, the plus above will be minus.

[Tim F]

The noise increased because your signal to noise ratio became worse when one of the signals was sampled at 5V/div as noise is now relatively larger compared to the signal at the analog to digital converter. If you changed both channels to 1V/div the traces should become yet cleaner.

Not quite sure what you mean about the clipping scenario as the pic is missing. If you mean when you clip the ADC the Math trace becomes messed up then this is normal behaviour. The 1054Z doesn't know when it is clipping, it is the users responsibility to ensure that the vertical scale is set appropriately.

[Mechatrommer]

DZ1_200mV.jpg.

do we have to check the missing item for you?

[swgertsch]

I messed up attaching the DZ1_200mv.jpg.  Here it is...

[swgertsch]

The noise increased because your signal to noise ratio became worse when one of the signals was sampled at 5V/div as noise is now relatively larger compared to the signal at the analog to digital converter. If you changed both channels to 1V/div the traces should become yet cleaner.

Not quite sure what you mean about the clipping scenario as the pic is missing. If you mean when you clip the ADC the Math trace becomes messed up then this is normal behaviour. The 1054Z doesn't know when it is clipping, it is the users responsibility to ensure that the vertical scale is set appropriately.

I do not understand how the display V/div value has any affect on the recording precision of a signal. If the sampling rate of the signal remains constant the quality should not be affected. Yes, changing the display scale V/div can make a waveform visually appear more noisier but it is the same noise value. The Math A+B trace scale was not changed, yet the noise level of the resultant increased when display scale of CH1 was increased from 2V to 5V.

See the correct DS1_200mV.jpg. (I screwed up the initial upload.)

[w2aew]

The bulk of the noise is in the scope channel, not on your signal. And since vertical scale changes are mostly done by changing an attenuator in the signal path, the resulting noise is usually a fairly constant percentage of the screen (with exceptions and variations, mainly at the extremes). Thus, the noise voltage is higher at higher V/div settings.

[miguelvp]

I messed up attaching the DZ1_200mv.jpg.  Here it is...

The math is done on the captured values in the digital domain, so if one is clipped it will reflect on the math operation. So the math channel doesn't operate with the analog signal on a DSO. What you are seeing is normal.

[fubar.gr]

I do not understand how the display V/div value has any affect on the recording precision of a signal. If the sampling rate of the signal remains constant the quality should not be affected. Yes, changing the display scale V/div can make a waveform visually appear more noisier but it is the same noise value. The Math A+B trace scale was not changed, yet the noise level of the resultant increased when display scale of CH1 was increased from 2V to 5V.

See the correct DS1_200mV.jpg. (I screwed up the initial upload.)

Sampling rate is about signal fidelity in the horizontal axis. But you also have to take into account the vertical resolution.

A typical vertical resolution in entry level DSOs is 8 bits. This effectively means that there are 2^8 = 256 discrete points from the bottom to the top of the screen. If the scope has 8 divisions from top to bottom, that's 32 points per division

Now imagine that you have a 1Vpp signal and the DSO set to 1 Volt per division. The scope is using 32 points to display that 1 volt. Now, if you set the scope to 500mV per division, it will be sampling the same 1 Volt with 64 points, hence greater resolution.

So when displaying waveforms, it is best practice to set the scope such as the waveform fills as much of the screen possible.

[swgertsch]

It has become obvious to me that I my understanding of how a DSO functions is based on presumptions that are flawed. In order for me to properly use this tool (get accurate measurements and understand the precision limits of a DS1054Z) I need to step back and get a more fundemental understanding.

So consider this post closed. I will start another post to have some basic DSO operation questions answered.

[alsetalokin4017]

Aww... aren't you even going to try looking at the Math trace at 500 ns/div horizontally?

[fubar.gr]

It has become obvious to me that I my understanding of how a DSO functions is based on presumptions that are flawed. In order for me to properly use this tool (get accurate measurements and understand the precision limits of a DS1054Z) I need to step back and get a more fundemental understanding.

So consider this post closed. I will start another post to have some basic DSO operation questions answered.

Don't get frustrated!

Oscilloscopes are nothing like eg. a multimeter, which you can learn how to use in 5 minutes. Oscilloscopes are high precision, highly complicated instruments and it takes years to learn all of their intricacies and how to use them properly.

So keep experimenting with your new scope. It's the only way to learn.