Fixing bad Oscilloscope Waveform Shape caused by poor Interpolation

[bb1]

Attached pdf contains oscilloscope screenshots showing bad waveforms spoiled by poor interpolation at not high enough sampling rate.
The pdf describes very simple method to exchange long waveform length to high time resolution hugely exceeding sampling rate.
Very much improved waveform plot, limited only by the scope analog circuitry bandwidth, but not the sample rate, is shown.
Examples use waveforms captured by the RIGOL DS2000A scope, but the described method can be applied to any scope.

[Multimetertools]

Poor interpolation can cause an oscilloscope to make waveforms that are not accurate or consistent. Interpolation adds extra points between existing points to provide better resolution and accuracy. When this process isn't done correctly, it can lead to incorrect measurements or distorted waveforms.

1. Understanding Interpolation:

It is essential to understand the concept of interpolation to successfully fix oscilloscope waveform shape caused by poor interpolation. Interpolation is adding extra points between existing points to provide more accurate resolution and measurement of signal behavior.



2. Identifying Poor Interpolation:

The first step in resolving oscilloscope waveform shape caused by poor interpolation is to identify the source of this problem. Visual inspection can determine if there are distortions within the waveform or if it appears jagged or uneven. Also, measurements like amplitude, frequency, and rise time can be taken at certain points on the waveform to see if bad interpolation is to blame.



3. Adjusting Interpolation Settings:

Once bad interpolation has been found to be the cause of the shape of the waveform, the next step is to change the interpolation settings on the oscilloscope. To get an accurate and consistent shape from different types of waveforms, you may need to use different interpolation settings. For example, a sine wave will require slightly different settings than a square wave or triangle wave. Experiment with various combinations of Interpolation type (Linear/Cubic) and Interpolator Length until an optimal waveform is achieved.



4. Troubleshooting Other Causes:

If changing the oscilloscope's interpolation settings doesn't make the waveform look better, it's important to think about what else could be wrong. Issues such as improper grounding, incorrect probe connection, or an interference source may be causing the distortion and should be thoroughly investigated and addressed.



5. Verifying Measurements:

Once the oscilloscope waveform has the shape you want, it's important to take accurate measurements to ensure it's right. At certain points on the waveform, you can check the amplitude, frequency, and rise time to ensure they are in acceptable ranges. If any of these numbers don't seem right, you may need to do more troubleshooting to fix the problem.



If you understand interpolation and follow these steps to fix bad oscilloscope waveform shape caused by bad interpolation, you can make accurate and consistent waveforms. This will give the most accurate readings and ensure the oscilloscope is used safely and well.

LINK REMOVED

[2N3055]

Attached pdf contains oscilloscope screenshots showing bad waveforms spoiled by poor interpolation at not high enough sampling rate.
The pdf describes very simple method to exchange long waveform length to high time resolution hugely exceeding sampling rate.
Very much improved waveform plot, limited only by the scope analog circuitry bandwidth, but not the sample rate, is shown.
Examples use waveforms captured by the RIGOL DS2000A scope, but the described method can be applied to any scope.

Or you just buy Siglent scope and put it in a dot mode....

Research term ETS (Equivalent time sampling) Sequential Sampling and Random interleaved sampling...

[nctnico]

Attached pdf contains oscilloscope screenshots showing bad waveforms spoiled by poor interpolation at not high enough sampling rate.
The pdf describes very simple method to exchange long waveform length to high time resolution hugely exceeding sampling rate.
Very much improved waveform plot, limited only by the scope analog circuitry bandwidth, but not the sample rate, is shown.
Examples use waveforms captured by the RIGOL DS2000A scope, but the described method can be applied to any scope.

It doesn't have to be applied to any scope because most oscilloscopes have sin x/x reconstruction (it is not interpolation!) implemented correctly. But it is not out of the ordinary to find improper sin x/x implementations in alpha version DSO firmware so I guess Rigol is not an exception.

@2N3055: again, dot mode is not ETS (equivalent time sampling). There is much more to ETS than just plotting some dots on screen. For starters ETS will take a known / predictable amount of samples per time interval. Or put differently: the effective samplerate is pre-determined which you won't get with dot mode.

[bb1]

Quotes from the Tektronix User manual:
"Your oscilloscope can interpolate between the samples it acquires. Like for
equivalent time sampling, it does so only when it cannot obtain all the real
samples it needs to fill up its waveform record."
"Sin(x)/x interpolation computes record points using a curve fit between the actual
values acquired."

[nctnico]

Quotes from the Tektronix User manual:
"Your oscilloscope can interpolate between the samples it acquires. Like for
equivalent time sampling, it does so only when it cannot obtain all the real
samples it needs to fill up its waveform record."
"Sin(x)/x interpolation computes record points using a curve fit between the actual
values acquired."

And it is wrong terminology. Interpolation means adding points that estimate the signal. Sin x/x doesn't do that. Sin x/x is basically a filter function that helps visualise a signal by drawing a graph that intersects with all the sampled points. The word visualise is important here because sin x/x reconstruction only serves the purpose of presenting a signal that we, humans, can see. For a computer the acquired samples are good enough to represent the signal (given the sampling was done within the mathematical boundaries set by Nyquist et al).

The only problem is that implementing sin x/x correctly is not a given and most likely the cause of the poor signal representation on the DS2000.

[wasedadoc]

Quotes from the Tektronix User manual:
"Your oscilloscope can interpolate between the samples it acquires. Like for
equivalent time sampling, it does so only when it cannot obtain all the real
samples it needs to fill up its waveform record."
"Sin(x)/x interpolation computes record points using a curve fit between the actual
values acquired."

And it is wrong terminology. Interpolation means adding points that estimate the signal. Sin x/x doesn't do that. Sin x/x is basically a filter function that helps visualise a signal by drawing a graph that intersects with all the sampled points. The word visualise is important here because sin x/x reconstruction only serves the purpose of presenting a signal that we, humans, can see. For a computer the acquired samples are good enough to represent the signal (given the sampling was done within the mathematical boundaries set by Nyquist et al).

The only problem is that implementing sin x/x correctly is not a given and most likely the cause of the poor signal representation on the DS2000.

Sin(x)/x is the theoretically best interpolator to reconstruct intermediate values from a set of samples that have been acquired properly.  How does that not fulfil the two functions cited by nctnico? 

[2N3055]

Attached pdf contains oscilloscope screenshots showing bad waveforms spoiled by poor interpolation at not high enough sampling rate.
The pdf describes very simple method to exchange long waveform length to high time resolution hugely exceeding sampling rate.
Very much improved waveform plot, limited only by the scope analog circuitry bandwidth, but not the sample rate, is shown.
Examples use waveforms captured by the RIGOL DS2000A scope, but the described method can be applied to any scope.

It doesn't have to be applied to any scope because most oscilloscopes have sin x/x reconstruction (it is not interpolation!) implemented correctly. But it is not out of the ordinary to find improper sin x/x implementations in alpha version DSO firmware so I guess Rigol is not an exception.

@2N3055: again, dot mode is not ETS (equivalent time sampling). There is much more to ETS than just plotting some dots on screen. For starters ETS will take a known / predictable amount of samples per time interval. Or put differently: the effective samplerate is pre-determined which you won't get with dot mode.

And again, I never said it was. It is similar in effect to  RIS..

[bdunham7]

And it is wrong terminology. Interpolation means adding points that estimate the signal. Sin x/x doesn't do that.

I try to avoid semantic arguments, but....

It may not be what you typically think of as interpolation and it certainly isn't linear interpolation, but sinx/x reconstruction does add points--enough of them to create the appearance of a continuous curve--to display what the algoritm thinks is the signal.  The difference between an 'estimation' and your claim of perfection in an ideal sinx/x case is not a debate worth having, IMO, but...

The only problem is that implementing sin x/x correctly is not a given and most likely the cause of the poor signal representation on the DS2000.

I doubt they've done the math wrong, although anything is possible and I'll have to reread the OPs .pdf again  From what I've seen, failures of sinx/x to provide a clean, accurate signal representation are due to other issues like sampling and quantization noise, jitter, over-BW signal components and things like that.  And...

again, dot mode is not ETS (equivalent time sampling)

I don't think anyone is saying it is as good as 'proper' ETS and they don't call it that, but if you have a fast edge that looks jumpy and noisy because it is over-BW, switching to dot-mode will give you a stable display, and as long as the frequency of the signal isn't close to a multiple of the sampling rate, the display fills in quite nicely.  The 'ETS effect' is a random feature and is a bit hit-or-miss as to the density of the dots, but even if they are sparse you can nicely visualize their stability.

[nctnico]

The only problem is that implementing sin x/x correctly is not a given and most likely the cause of the poor signal representation on the DS2000.

I doubt they've done the math wrong,

They have. Just like GW Instek and R&S have done in early versions of their DSO firmware. But unlike Rigol, they fixed it (and rather quickly). One of the first things I test in a DSO is whether sin x/x works as it should. When sin x/x is implemented properly, a DSO should have no problem showing a sine wave with a frequency up to fs / 2.5.

[tggzzz]

Poor interpolation can cause an oscilloscope to make waveforms that are not accurate or consistent. Interpolation adds extra points between existing points to provide better resolution and accuracy. When this process isn't done correctly, it can lead to incorrect measurements or distorted waveforms.
1. Understanding Interpolation:
It is essential to understand the concept of interpolation to successfully fix oscilloscope waveform shape caused by poor interpolation. Interpolation is adding extra points between existing points to provide more accurate resolution and measurement of signal behavior.
......

That's a good example of a "statistical parrot" producing word salad bullshit.

Good decision to ban the bugger

Unfortunate!y I don't have confidence that I could easily detect it in a different field, e.g. organic chemistry