Siglent SDS 1202X-E sampling anomaly?

[MacMan72]

Can anyone tell me why, when I zoom in on a trace in the 1202, it has this “digital” appearance? I have the Aquisition set to Normal and the mem depth set to 1.4M. If I set the Aquisition type to Average, I get what I would expect to see when looking at the ripple in the test calibration signal. Is this expected behavior or is there a setting/filter I’m missing here?

[tautech]

4 things scream at me why ?
Small mem depth.
Trigger off the display
20mV/div
Photo not the much smaller screenshoot ?

Press H Pos to zero the trigger position and get it back onto the display.
Check probe and input attenuation settings match and set appropriate vertical settings
Mem depth won't matter much for this but leave it at max 14Mpts and lt the scope manage memory depth.

In Utility, last page IIRC set Horizontal Ref Pos to Position, then the marker stays on screen regardless of timebase settings.

Use Print to USB to capture screenshots and there is no reason to need Stop the scope.

[Martin72]

Another question is, what do you want to see from a 1khz signal at 10ns/div...

[DavidAlfa]

You sampled the signal at 500mV/div, then zoomed in all the way to 20mV/div, what did you expect?

It's a 8bit ADC so this is completely normal, you have only 0-255, 256 different values.
500mv*8div = 4000mV  -> 4000/256 = 15.6mV vertical resolution.
So of course it looks like that.

When you later sampled the signal at 20mV/div, the story changes, now it makes:
20mv*8div = 160mV  -> 160/256 = 0.62mV vertical resolution.

But as you can see when zooming in, at 500mV the resolution is actually 20mV, so it's storing a bit more than 8 divisions:
20mv*256 = 5120mV  -> 5120/500 = 10.24 vertical divisions.

[bdunham7]

Can anyone tell me why, when I zoom in on a trace in the 1202, it has this “digital” appearance?

You're zooming vertically with the scope stopped?  So at 500mV/div, it has 8 bits of resolution, which is 256 levels from top to bottom, or 32 levels per division (at most).  When you zoom in to 20mV/div--25X--on this already captured trace, those levels get expanded so that there are only about 10 levels from top to bottom, or a bit more than 1 per division at most.  In practice, the capture window may extend off the top and bottom of the screen and it appears in this case that the result is that the trace is displayed at about 1 level per division. 

[MacMan72]

I just got the scope and was basically playing around with it to get familiar with the interface. I had also just gotten a bk precision pr150sa probe and was calibrating it and was zoomed in a bit to get it just right and it looked a bit “digital” to me so I zoomed in further to see what was going on. I was having trouble getting it to trigger correctly so I ran a single trigger and zoomed in. I can do the same on my old gw gds-1074b and it zooms in perfectly well and shows the results in the attached image (I’m typing on my phone, hence the photo instead of usb screen capture). It made me wonder why the difference between the two scopes so I figured I’d ask. Is this an aquisition artifact or a display artifact? Can it be corrected/adjusted? Like I said, getting to know the oscilloscope. I’d assumed at 2x the bandwidth and sampling rate, that the siglent would be “better” than the instek in this regard. I still cannot zoom in on the horizontal portion of the trace and get a good trigger so I guess I have more to learn.

[MacMan72]

Also, why does averaging show the expected “analog” ripple but normal shows the “digital” plot. I’m trying to understand the underlying math here.

[Martin72]

so I guess I have more to learn.

It would already be a progress if you would take regular screenshots instead of photographing the scope screen.

[bdunham7]

Your issues are all triggering.  The Siglent is not retriggering after you zoom because the trigger level is out of range.  You're comparing the scopes somehow but you aren't using the exact same settings.  Pick a set of settings and apply to both.  For this signal, I'd suggest 20mV/div, 40mV trigger (rising edge, which you have), 200ns/div horizontal and perhaps 500ns horizontal offset (delay).

[noisyee]

This is an expected display behave.
The waveform display area is 700 pixel wide.
At 2us/div, the actual mem depth is 28k pts. They will be scaled down to fit in the 700 pixels display.
At 50ns/div, mem depth is 700 pts. One sample point will fit one display pixel.
Below 50ns/div, mem depth is less than 700 pts. They must be interpolated to full fill the 700 pixels display if Display Type is set to Vectors.
If Stop to zoom below 50ns/div, there is no incoming data to be interpolated. Scope will use memory data (already "Averaged" data in this case) to interpolate. That's where the “digital” appearance come.
Try to set Display Type to Dots or do not zoom below 50ns/div if you don't want to see that.

[MacMan72]

I figured it out… I think. First, I couldn’t zoom in as close as I thought I needed to because the waveform is 3v pk-pk and I ran into the vertical zoom limits at about 100mV/div. That’s what prompted me to stop scanning and zoom further in since it will let me zoom further on a saved waveform. I had initially tried to zoom in on the top of the waveform to look at overshoot but when I tried to zoom horizontally, the waveform sloped away (since the rise time of the calibration signal sucks  (first image) (properly captured) (you’re welcome). So I switched to the bottom of the waveform to see what that looked like. That’s when the weird digital effect became apparent. When I look at a similar waveform generated by my AWFG, it has a nice fast rise time and typical overshoot with settling ripple like I was expecting (Second and third images). So, looking at the lower portion of the waveform, I was zooming all the way in and, here’s the point finally, so close, I’m looking at individual datapoints and it seems as though the Instek is doing a polynomial fit between the points to maintain continuity while the Siglent is just playing connect the dots. At least it does that for Normal mode. Averaging looks to be employing a curve fit method similar to the Instek.

[tautech]

I’m looking at individual datapoints and it seems as though the Instek is doing a polynomial fit between the points to maintain continuity while the Siglent is just playing connect the dots. At least it does that for Normal mode.

Not quite.
Siglent scopes have 2 display modes, interpolated (joined dots) or Dot mode.

In Dot mode when Stopped you can clearly see the individual datapoint dots especially when zoomed in.
In Run and Dot mode, the trigger rate overlays the acquisitions so to look like an interpolated trace, however it is not.

Some even keep their DSO's in Dot mode but might switch to interpolated in a capture to see the trace better.
Keep exploring. 

[MacMan72]

That's what I'm saying... the Siglent is doing linear interpolation in Normal Aquisition mode and a polynomial fit (with more data points, I'm assuming from multiple sweeps determined by the Average # set by the user) in Average Aquisition mode. If you look at the images, the first is normal with dots, the second is normal with vectors (straight linear interpolation between datapoints). The third is dots with averaging, the fourth is averaging with vectors (polynomial fit).

[tautech]

If this worries you and rather than looking at these quite sensitive DSO's, instead investigate presence of any local RFI.

I have had customers call and question noise on the traces and working against their disbelief had them find a horrible noisy wallwart across the room behind a curtain was the real issue.

Probing technique, ditching the probe reference (GND) lead in favor of the spring in the probe accessories does markedly reduce the antenna effects of the reference lead.

[Martin72]

Probing technique, ditching the probe reference (GND) lead in favor of the spring in the probe accessories does markedly reduce the antenna effects of the reference lead.

See also this picture with no gnd connected, gnd connect via the normal lead/clip and gnd connected via the spring.

[Performa01]

Screen height is just 200 samples, the rest is overhead. As a consequence, there are 25 samples per division. At 500 mV/div we get 500/25 = 20 mV per LSB. Unsurprisingly, we get one LSB per division at 20 mV/div. Neither does this look pretty nor does it make much sense for any measurement.

Vector display mode has two sub-modes:

• linear interpolation
• sin(x)/x reconstruction

Make sure you have sin(x)/x selected instead of linear.

Average acquisition mode combines several acquisitions into one single data set. With 4x averaging we gain two additional bits of resolution. In higher class scopes we have a limit of 16 bits at the moment. If the same is true for the SDS1202X-E, then averaging up to 256x can be used in order to get a 16 bit acquisition. This will take some time though, because this acquisition mode is not HW-accelerated in the SDS1000X-E series.