Sub: Rigol's DHO800 Oscilloscope (Gibbs Effect & Aliasing Misunderstanding)

[Martin72]

That is why I kepp repeating that DHO800 samples properly and DHO900 does not unles in very limited usage scenario.
DHO800 with 100 MHz BW should be largely immune to this. Unless you manually reduce sample rate or memory length at long timebases.

Like this.

[Fungus]

that's sin(x)/x for you.

yes i provided the link in my post. sin(x)/x is a highly theoritical and mathematical. iirc i read long ago that you cannot reconstruct back the sampled real signal back on monitor the way its appeared on electrical circuit if you dont do infinite terms reconstruction

Using only the central part of the sinc function doesn't help but the main problem is lost harmonics due to limited bandwidth.

Those harmonics are what keeps the line flat before/after the pulse.

[tautech]

That is why I kepp repeating that DHO800 samples properly and DHO900 does not unles in very limited usage scenario.
DHO800 with 100 MHz BW should be largely immune to this. Unless you manually reduce sample rate or memory length at long timebases.

Like this.

Stock or hacked ?

[iMo]

There are 4 things not to be mixed here:

1. the "aliasing" creating the fake artifacts/spurious (caused by hw&sw because of "BW vs. sampling rate")
2. the "sinc" function used for "creating the nice pictures" via the "sinc interpolation" as seen on the LCD screen (each discrete point sampled is multiplied by sinc function and all that results are added, thus you get "additional points in between", you draw them all on the screen thus you see a solid line instead of just several sampled points, like 5 points for example)
3. the process  of a "reconstruction/filtering of a square wave" with help of FFT, like sqrwave->fft->filter->ifft->sqrwave
4. the "ringing" of the hw circuitry (caused by steep filters, unstable circuits, impedance mismatch)
Edit: typos..

[2N3055]

that's sin(x)/x for you.

yes i provided the link in my post. sin(x)/x is a highly theoritical and mathematical. iirc i read long ago that you cannot reconstruct back the sampled real signal back on monitor the way its appeared on electrical circuit if you dont do infinite terms reconstruction

Using only the central part of the sinc function doesn't help but the main problem is lost harmonics due to limited bandwidth.

Those harmonics are what keeps the line flat before/after the pulse.

Correct! And keep the top flat too..

And somebody mentioned infinite series to properly reconstruct. That is not needed as soon as you reach resolution of screen, noise floor or step response specifications of the scope. 

[2N3055]

That is why I kepp repeating that DHO800 samples properly and DHO900 does not unles in very limited usage scenario.
DHO800 with 100 MHz BW should be largely immune to this. Unless you manually reduce sample rate or memory length at long timebases.

Like this.

THANK YOU!!

[Mechatrommer]

That is why I kepp repeating that DHO800 samples properly and DHO900 does not unles in very limited usage scenario.
DHO800 with 100 MHz BW should be largely immune to this. Unless you manually reduce sample rate or memory length at long timebases.

Like this.

so both siglent and rigol can show false signal at lower sampling rate i was interested that much earlier in this thread, the very false (sinc) signal was used to argue that rigol displayed signal is incorrect, ie hiding "accurate" signal representation, it turned out, that "accurate" representation is just artifact from sinc interpolation this was my intention, to clear things up and lets argue fairly apple to apple... and not messing up with other aspects as iMo mentioned just above. another factor already answered by Kleinstein about differing input impedance and respond inside the dso front end filter... https://www.eevblog.com/forum/testgear/rigols-new-dho800-oscilloscope-unbox-teardown/msg5076118/#msg5076118 ps: i need to find where that sinc interpolation setting in rigol and turn it off, its no use anyway at trying to represent "accurate" signal by interpolation.

[Martin72]

Stock or hacked ?

Right out of the box (70Mhz).

[Fungus]

it turned out, that "accurate" representation is just artifact from sinc interpolation

The sinc function is only a very minor player here.

i need to find where that sinc interpolation setting in rigol and turn it off, its no use anyway at trying to represent "accurate" signal by interpolation.

sinc isn't the problem, lack of bandwidth is.

PS: There won't be a way to turn it off. It would be a disaster to just draw lines between the dots.

[wasedadoc]

LC circuit cannot produce "non-causal" effect, can it?

Yes it cetainly can.  Ask anyone* who is competent to design high quality low pass filters for video signals.  Passive circuits with capacitors and inductors.

( * I am not such a designer but I worked with those who were.)

[TimFox]

"Non causal" usually means that the output can chang before (in time) the input does:  not realizable in the time domain.
https://en.wikipedia.org/wiki/Causal_filter
A causal filter's output depends only on past and present inputs, not future values of the input.
Digital filters operating on a data set can be non-causal.

[Mechatrommer]

sinc isn't the problem, lack of bandwidth is.

so then my 230MHz "DHO924" is in better hand than your 100MHz "DHO814"

PS: There won't be a way to turn it off. It would be a disaster to just draw lines between the dots.

thats too bad. DS1000Z can be switched off. this is useful to visualize and verify (qualitative) what we are seeing is true or possibly a false interpolation. what kind of disaster if we turn it off? at worst we see aliasing. or apparent low BW signal.. interpolation is meant to attempt reconstruct accurate representation, but its proven its not. you are on the side that think the "non-causal"/precursor/gibbs ear ringing is for real do you? i'm on the side with R&S and Lecroy are you saying directly download raw sampled data and plot point to point in PC is a disaster? i've had fair amount of time doing that, apart from sharp edges/triangulated plot instead of nice curvy interpolated plot, there is nothing to be afraid of, brain can certainly make better judgement

[EEVblog]

Two things are important here, in my view:

(a) It's not the case that the ringing is actually present on the Bodnar pulser's output and the DHO800 somehow "glosses it over" while Siglent shows the truth. Tautech had claimed that early in the discussion a few weeks ago, but dropped that claim later.

I showed in my review video two R&S scope responses, MXO4 and RTB2000 and both of them matched the DHO800. The Siglent was indeed the outlier.
The MXO4 is a 1.5GHz bandwith scope, so it should show everything on the Leo Bodnar generator warts and all, yet there is nothing there.
The Siglent clearly has the problem.
I can do more scopes if needed, but if I'm going to trust any scope in my lab it's the MXO44.

[Mechatrommer]

The Siglent was indeed the outlier.

its clear to me now so far, the problem is imperfect sinc reconstruction. on both siglent and rigol if condition is met. so the argument if this scope is more accurate representation than the others is moot.

[Mechatrommer]

Try it with a single channel and 1.25GSa/sec. It should look the same as my 200Mhz DHO804 if you do that. 

can we upgrade to DHO824 meh? i read its impossible, there is no BT7TO20 option, or did i get outdated? btw i saw your leo pulse when hacking to DHO924, its the same with me now (extra overshoot) but last time i read you downgraded to 100MHz DHO814 right?

Nope, you've set up a pathological case there and you know it. What you're seeing there is aliasing.

aliasing on fast risetime square 10MHz pulse? i suspect you are confused right now. or am i misinterpreting?

[tautech]

Two things are important here, in my view:

(a) It's not the case that the ringing is actually present on the Bodnar pulser's output and the DHO800 somehow "glosses it over" while Siglent shows the truth. Tautech had claimed that early in the discussion a few weeks ago, but dropped that claim later.

I showed in my review video two R&S scope responses, MXO4 and RTB2000 and both of them matched the DHO800. The Siglent was indeed the outlier.
The MXO4 is a 1.5GHz bandwith scope, so it should show everything on the Leo Bodnar generator warts and all, yet there is nothing there.
The Siglent clearly has the problem.
I can do more scopes if needed, but if I'm going to trust any scope in my lab it's the MXO44.

Mr Bodnar sees otherwise as does my 10 GSa/s 2 GHz SDS6204A.

From Leo's website a capture from a fast Tek clearly displaying the Gibbs ears other scopes choose to suppress.



Better add the falling edge screenshot for the sake of completeness.

[Fungus]

last last time i read you downgraded to 100MHz DHO814 right?

Yep.

Measured bandwidth is 200Mhz with the BW7T10 option. I'm not 100% sure what's going on inside the 'scope but I like it... 

I don't need 250Mhz so this keeps it a little bit further away from Nyquist. I also enjoy the lack of junk on screen and the perfect calibration.

aliasing on fast risetime square 10MHz pulse? i suspect you are confused right now. or am i misinterpreting?

With 250Mhz+ analog bandwidth and 312.5MSa/sec. sample rate? Yes, you'll get bad aliasing on a pulse.


There's a simple way to know if you're seeing aliasing: Turn some channels off and see if the displayed waveform changes.

(assuming you're not STOPped...)

[wasedadoc]

"Non causal" usually means that the output can chang before (in time) the input does:  not realizable in the time domain.
https://en.wikipedia.org/wiki/Causal_filter
A causal filter's output depends only on past and present inputs, not future values of the input.
Digital filters operating on a data set can be non-causal.

It is necessary to compare the relative time of filter input and output.  Agreed the filter output cannot begin to ring before the time of the input transition.  But looking at the output of the Ls and Cs filter there can be preshoot or ring before the main transition.  Perhaps that does not meet the strict definition of "non-causal" but has the appearance of non-causal.  The screenshots under discussion are such.  Only output waveform is shown.

[boggis the cat]

I showed in my review video two R&S scope responses, MXO4 and RTB2000 and both of them matched the DHO800. The Siglent was indeed the outlier.
The MXO4 is a 1.5GHz bandwith scope, so it should show everything on the Leo Bodnar generator warts and all, yet there is nothing there.
The Siglent clearly has the problem.
I can do more scopes if needed, but if I'm going to trust any scope in my lab it's the MXO44.

Have you tried exchanging the probes between scopes?

The probes are part of the measurement system.

[tautech]

I showed in my review video two R&S scope responses, MXO4 and RTB2000 and both of them matched the DHO800. The Siglent was indeed the outlier.
The MXO4 is a 1.5GHz bandwith scope, so it should show everything on the Leo Bodnar generator warts and all, yet there is nothing there.
The Siglent clearly has the problem.
I can do more scopes if needed, but if I'm going to trust any scope in my lab it's the MXO44.

Have you tried exchanging the probes between scopes?

The probes are part of the measurement system.

Bodnar pulser connects directly to the BNC channel input. 

[Fungus]

From Leo's website a capture from a fast Tek clearly displaying the Gibbs ears other scopes choose to suppress.

Weird that it doesn't have the "pre" ears like the Siglents do...

It actually looks a lot closer to the little Rigol DHO800's response than the Siglent's:

[tautech]

Yes well at 100ps/div like Mech showed earlier much more becomes visible than any slow scope displays.

[Fungus]

Yes well at 100ps/div like Mech showed earlier much more becomes visible than any slow scope displays.

Similar to a fancy Lecroy as well? That's cool. 

Edit: I'm not picking sides and saying which philosophy is "best" but I think it's cool that my little Rigol behaves like the big expensive scopes.

[tautech]

Yes well at 100ps/div like Mech showed earlier much more becomes visible than any slow scope displays.

Similar to a fancy Lecroy as well? That's cool. 

Edit: I'm not picking sides and saying which philosophy is "best" but I think it's cool that my little Rigol behaves like the big expensive scopes.

At 5ns/div what do you see 400ps before the rising edge ? Zip, zero, nada !

[asmi]

Mr Bodnar sees otherwise as does my 10 GSa/s 2 GHz SDS6204A.
From Leo's website a capture from a fast Tek clearly displaying the Gibbs ears other scopes choose to suppress.

You would do really well to get some reading comprehension. The discussion is about ringing BEFORE pulse, not AFTER it, and your screens shows none of it. Which is the way it's supposed to be.