EEVblog #1213 - The Oscilloscope Interpolation Trap

[EEVblog]

Oscilloscope Sin X/x Interpolation can be a trap for young players, find out why.

[vk6zgo]

It's a trap for old players, too!

The display at 2:40 or so, looks for all the world like that  from a "2T" "sine squared pulse" when testing an analog video system with good, but not excellent, characteristics.

The trap for anyone who had spent a lot of time on analog TV testing would be that they would assume that the original pulse was intended to be of that shape, "because we are used to it!" & think that the designer of the DUT had worked out some "cunning plan" to obtain correct timing/data retrieval from a signal of that shape.

The other comments about sample rate -v-"time/div" is what we kept running into with early DSOs, where the sample rate would "slow down to a crawl" at settings like 5ms/div & slower, due to the tiny sample memory they possessed.

[EEVblog]

It's a trap for old players, too!
The display at 2:40 or so, looks for all the world like that  from a "2T" "sine squared pulse" when testing an analog video system with good, but not excellent, characteristics.
The trap for anyone who had spent a lot of time on analog TV testing would be that they would assume that the original pulse was intended to be of that shape, "because we are used to it!" & think that the designer of the DUT had worked out some "cunning plan" to obtain correct timing/data retrieval from a signal of that shape.

Very true!

[TD-Linux]

Another excellent video about digital sampling by Monty of Xiph, this time in respect to audio (but it's the exact same problem):

https://xiph.org/video/vid2.shtml

youtube:

[EEVblog]

Another excellent video about digital sampling by Monty of Xiph, this time in respect to audio (but it's the exact same problem):
https://xiph.org/video/vid2.shtml
youtube:

That's a really great video

[bitwelder]

That's not a Sin(x)/x graph, that's an incoming YT strike!

[ggchab]

Very interesting video  Thank you

Some time ago, I discovered this book that starts from the very beginning. I like it.
"The Scientist and Engineer's Guide to Digital Signal Processing" by Steven W. Smith (http://www.dspguide.com/)

[Kleinstein]

In the interpolated signal there is a dead give away that the ringing is due to some kind of digital filtering / interpolation: there was ringing before the transition. Ringing from a probe or poor signal quality has ringing only after the transition.

[Fungus]

Another excellent video about digital sampling by Monty of Xiph, this time in respect to audio (but it's the exact same problem):
https://xiph.org/video/vid2.shtml
youtube:

That's a really great video

Should be required viewing.

(Bonus: It has lots of old test gear)

[tautech]

Who says memory depth and sample rate are just sales gimmicks ? 

[Fungus]

Who says memory depth and sample rate are just sales gimmicks ? 

Tektronix salesmen?

Rigol MSO5000 haters...?

[EEVblog]

In the interpolated signal there is a dead give away that the ringing is due to some kind of digital filtering / interpolation: there was ringing before the transition. Ringing from a probe or poor signal quality has ringing only after the transition.

Yes, i should have mentioned that.

[Fungus]

In the interpolated signal there is a dead give away that the ringing is due to some kind of digital filtering / interpolation: there was ringing before the transition.

Yes, i should have mentioned that.

Yep. 

Needs emphasis: Anything before the rising edge is from the oscilloscope's interpolation, always, not from the signal or the probes. It will probably have a counterpart at the top of the rising edge, too.

See: "Gibbs phenomenon" (horrible page full of math that doesn't explain much)

[0xdeadbeef]

Hm, is is just me or does that red sinc drawing look a bit "pubertal"? I mean I would totally expect things like that in a school toilet but I guess some people (who don't even bother to look into the video) could be offended by this on Youtube.

[Fungus]

Hm, is is just me or does that red sinc drawing look a bit "pubertal"? I mean I would totally expect things like that in a school toilet but I guess some people (who don't even bother to look into the video) could be offended by this on Youtube.

It's almost as if you haven't watched the video. 

[0xdeadbeef]

Admittedly I was only interested in the SDS5104X part since I didn't expect to learn anything new about sinc/linear interpolation or subsampling. So I skipped over the rest of the video.
I don't get the point though. I'm not offended that easily personally, I just wondered if it's such a good idea to use that kind of drawing on the intro screen since there are a lot of people who are easily offended out there.

[Brumby]

As Dave said, it's "the 10 year old" in him that saw that.  I didn't see it myself until after watching the video and hearing that comment from Dave.

Personally, if people want to see that, then I think they are looking for it and complaining would be their childish tantrum way of making themselves feel good.  Edit:  I was not referring to you, 0xdeadbeef!

I'd have been a little happier if it hadn't been pointed out - but, hey, Dave's channel ... Dave's call.  I can live with it.

[tggzzz]

My personal preference is to use a sampling scope (digitising or analogue) set to show me the samples only.

My brain can "join the dots" and interpret that information easily and, more importantly, not infer information that isn't actually there.

I've come across very irritating digitising scopes allow you to select "dots only" mode, but whenever you change any control it reverts to "hide dots" interpolated mode. Grrr.

[thinkfat]

Much truth is in the sentence: "before you measure, be sure about what you expect to see".

[tggzzz]

Much truth is in the sentence: "before you measure, be sure about what you expect to see".

Yes.

And in "trust, but verify".

And in "Think" - one of only two mottos that are worth a damn, the other being "Be Prepared".

[David Hess]

In the interpolated signal there is a dead give away that the ringing is due to some kind of digital filtering / interpolation: there was ringing before the transition.

Yes, i should have mentioned that.

Yep. 

Needs emphasis: Anything before the rising edge is from the oscilloscope's interpolation, always, not from the signal or the probes. It will probably have a counterpart at the top of the rising edge, too.

That is incorrect; preshoot can occur in the analog domain as well.  For instance lumped-parameter transmission lines create preshoot when excited beyond their cutoff frequency and this is visible on some analog oscilloscopes like the 300 MHz Tektronix 2465 and following models which pushed the capability of their delay lines.

It is easy enough however to distinguish this from the Gibbs phenomenon which is symmetrical around the transition.


Back when DSOs were new, Tektronix had an idea for detecting the problem Dave showed.

The processed trigger signal can be used to estimate the frequency content of the input signal.  If it approaches or is greater than the current Nyquist frequency, (1) then it is very likely that aliasing has occurred and a warning was displayed to the user.  This idea completely failed because it made it seem like Tektronix DSOs suffered from aliasing more than their competitor's DSOs, which they did not, so it was quietly dropped.

I seem to recall using some DSOs which displayed a warning when interpolation was used to fill in the display.  That seems like an easy solution but no doubt it suffers from the same problem.  Again, marketing triumphs over engineering.

(1) Or if any trigger period is less than twice the sample period.

[schmitt trigger]

In the interpolated signal there is a dead give away that the ringing is due to some kind of digital filtering / interpolation: there was ringing before the transition. Ringing from a probe or poor signal quality has ringing only after the transition.

Thanks Kleinstein, for the valuable tip.

Would that be caused by an artifact similar to an IIR filter?
EDIT : I watched the video. It explained it exactly.

[EEVblog]

I'm not offended that easily personally, I just wondered if it's such a good idea to use that kind of drawing on the intro screen since there are a lot of people who are easily offended out there.

So what if people are offended?
Whoopty-do.

[snoopy]

I'm not offended that easily personally, I just wondered if it's such a good idea to use that kind of drawing on the intro screen since there are a lot of people who are easily offended out there.

So what if people are offended?
Whoopty-do.

He doesn't understand Aussie humour LOL

[snoopy]

In the interpolated signal there is a dead give away that the ringing is due to some kind of digital filtering / interpolation: there was ringing before the transition.

Yes, i should have mentioned that.

Yep. 

Needs emphasis: Anything before the rising edge is from the oscilloscope's interpolation, always, not from the signal or the probes. It will probably have a counterpart at the top of the rising edge, too.

That is incorrect; preshoot can occur in the analog domain as well.  For instance lumped-parameter transmission lines create preshoot when excited beyond their cutoff frequency and this is visible on some analog oscilloscopes like the 300 MHz Tektronix 2465 and following models which pushed the capability of their delay lines.

It is easy enough however to distinguish this from the Gibbs phenomenon which is symmetrical around the transition.


Back when DSOs were new, Tektronix had an idea for detecting the problem Dave showed.

The processed trigger signal can be used to estimate the frequency content of the input signal.  If it approaches or is greater than the current Nyquist frequency, (1) then it is very likely that aliasing has occurred and a warning was displayed to the user.  This idea completely failed because it made it seem like Tektronix DSOs suffered from aliasing more than their competitor's DSOs, which they did not, so it was quietly dropped.

I seem to recall using some DSOs which displayed a warning when interpolation was used to fill in the display.  That seems like an easy solution but no doubt it suffers from the same problem.  Again, marketing triumphs over engineering.

(1) Or if any trigger period is less than twice the sample period.

Equivalent time sampling scopes don't have the same trouble with aliasing provided the signal you are observing is repetitive. Unfortunately Tek has dropped this feature on its newer scopes in favour of interpolation with all the issues of aliasing. It's a pity because they did it extremely well on the scopes that offered it.