Minimum Oscilloscope Bandwidth?

[0culus]

Someone blew up one channel of our very expensive Keysight scope at work last year. Never found out who did it; Keysight repaired it under warranty.

Someone blew up MY OWN MSOX6004A at work without letting me know (but I watched surveillance and know who he is). He then graduated and fled.
Keysight replaced its acquisition board under warranty, so I didn't pursue further.

Grrrr, people can be so dumb. On the bright side, ours also got fixed for no charge AND proper use of the thing is going to be included in the mandatory training that will be required, no exceptions, for using the lab once they finish remodeling. 

If you use common sense and the right tools you won't blow it up.

And a fixed 10x (or 100x) probe - none of that 1x rubbish.

Yeah, and an HV probe too. I have a couple of Tektronix P6015s...sadly out of Freon but 1000x divider if you respect the frequency derating curve, and still quite useful with the derated spec when not filled with a dialectric fluid.

[bd139]

Ah that reminds me of university. We had two channel analogue scopes and had to play “which channel still works” every lab    Then they got the 54600’s in and we were not allowed to use them

[nctnico]

If you use common sense and the right tools you won't blow it up.

And a fixed 10x (or 100x) probe - none of that 1x rubbish.

1x is necessary when looking at small signals. And I wonder how you can blow up a 1M Ohm input. Most are rated for 300V to 400V  so they should be able to withstand mains.

[Fungus]

1x is necessary when looking at small signals.

Sure, but I'd rather not be using something with a tiny switch on the side when I'm probing hundreds of volts.

And I wonder how you can blow up a 1M Ohm input. Most are rated for 300V to 400V  so they should be able to withstand mains.

Mains has transients, valve amplifiers have higher voltage then mains.

But sure, go ahead. Risk your entire 'scope for the sake of a $10 probe. 

[RoGeorge]

And I wonder how you can blow up a 1M Ohm input. Most are rated for 300V to 400V  so they should be able to withstand mains.

In a school lab?
I wouldn't be surprised if some students will have the initiative to visualize the waveform of a spark from a piezo lighter.

[Wallace Gasiewicz]

On blowing up scopes:
There are many old analog scopes around with one channel burned out. This is usually the dual FET on the input. A lot of old scopes have a dual six lead FET on input on each channel. The really old solid state ones have two separate FETs.
Easy to fix. I have fixed a few. I have not worked on any of the newer scopes which are surface mount.
I knew a guy who loved to buy these damaged scopes and fix them. He thought it was easy money.

If you use a scope in DC mode or 50 ohm DC mode it is relatively easy to blow the input. FET amplifier and the 50 ohm circuit and the protection, if there is any, can easily be fried. Even in AC mode you can fry the input.

Wally KC9INK

[bd139]

Indeed.

The most fragile bit is usually the 100X attenuators in a Tek 4xx as well. Seen many of them toasted!

[vinlove]

I have seen some people using 10 khz max bandwidth tube scopes quite happily doing all sort of things, especially working on audio, amp, radio and curve tracing etc etc.

[bd139]

Yep. Similar to signal tracing. Can use one with a demodulator probe.

[tggzzz]

Nope. Sin(x)/x reconstruction only needs "more then two" points (less than three still works, eg. five points every two cycles is still plenty).

The explanation: There's only one possible sine curve that fits through those points, it can be extrapolated from them.

What happens if its not a sine wave ?

As I'm sure you are aware, any signal can be decomposed into a number of sine waves.

The statement applies to the sine wave with the highest frequency; all other sine waves will contain more points.

[Fungus]

Nope. Sin(x)/x reconstruction only needs "more then two" points (less than three still works, eg. five points every two cycles is still plenty).

The explanation: There's only one possible sine curve that fits through those points, it can be extrapolated from them.

What happens if its not a sine wave ?

Everything is made of sine waves.

So long as there's no sine wave above the Nyquist frequency in your signal then there's only one possible signal that goes through all your sample points.

(and that's the whole of sampling theory in a nutshell  )

[bd139]

Any periodic function that is, not everything

[Fungus]

Any periodic function that is, not everything

No, everything.

They'll come and go if the function is non-periodic but they're still sine waves.

(and that's how audio compressors work)

[gf]

Nope. Sin(x)/x reconstruction only needs "more then two" points [...]

An exact sinc reconstruction of the original signal needs in fact an infinite number of points, since a sin(x)/x kernel has an unlimited extent. I.e. each sample influences the whole (infinite) timeline of the reconstructed signal. Infinite is not practical, though, so in practice it is rather approximated by a truncated sinc kernel with a finite extent (with say a dozen of samples) - but strictly, it is no longer a "true" sinc interpolation then.

[Fungus]

Nope. Sin(x)/x reconstruction only needs "more then two" points [...]

An exact sinc reconstruction of the original signal needs in fact an infinite number of points

Let me clarify that: "More than two points per cycle of the highest frequency sine wave in the signal"

[David Hess]

An exact sinc reconstruction of the original signal needs in fact an infinite number of points, since a sin(x)/x kernel has an unlimited extent. I.e. each sample influences the whole (infinite) timeline of the reconstructed signal. Infinite is not practical, though, so in practice it is rather approximated by a truncated sinc kernel with a finite extent (with say a dozen of samples) - but strictly, it is no longer a "true" sinc interpolation then.

If the displayed reconstruction goes through the original displayed sample points, then it is close enough.

[gf]

An exact sinc reconstruction of the original signal needs in fact an infinite number of points

Let me clarify that: "More than two points per cycle of the highest frequency sine wave in the signal"

That's correct.
No, not nore than two points per cycle of the highest frequency in the signal, but an
With infinite I meant the total number of samples, i.e. an infinite number of samples preceeding the "time of interest" and an infinite number of samples succeeding the "time of interest" where we want to obtain the reconstructed value.

If the displayed reconstruction goes through the original displayed sample points, then it is close enough.

In practice, truncation is fine, of course.

[Fungus]

Let me clarify that: "More than two points per cycle of the highest frequency sine wave in the signal"

No, not nore than two points per cycle of the highest frequency in the signal, but an infinite number of samples preceeding the "time of interest" and an infinite number of samples succeeding the "time of interest" where we want to obtain the reconstructed value.

And a minimum of two samples per cycle...

[bd139]

Any periodic function that is, not everything

No, everything.

They'll come and go if the function is non-periodic but they're still sine waves.

(and that's how audio compressors work)

Nope. It is within a fixed interval which is a stinking massive cheat. It’s periodic outside that interval.

You can’t approximate a single pulse in infinite time with Fourier series. The domain of the transform is infinite if you don’t use a fixed interval and the sine function is periodic.

Even if you could it wouldn’t be very practical unless you were Q or something.

[Fungus]

You can’t approximate a single pulse in infinite time with Fourier series.

A pulse is a discontinuity in the signal.

[james_s]

Someone blew up one channel of our very expensive Keysight scope at work last year. Never found out who did it; Keysight repaired it under warranty.

Someone blew up MY OWN MSOX6004A at work without letting me know (but I watched surveillance and know who he is). He then graduated and fled.
Keysight replaced its acquisition board under warranty, so I didn't pursue further.

That's super annoying, I think I would have gone to the police about something like that. I mean it's one thing to damage someone's personal property but it's quite another thing to flee the scene without saying anything. Not to mention it's really disrespectful to the person who owns the gear.

As far as blowing up scopes, I've never done it myself, but I've certainly seen plenty of damaged scopes. The problem is that people *aren't* sensible about what they're doing and in some cases you can have much higher voltages than you realize. Even though I have a proper Tek HV differential probe I don't let my scope near my HV experiments, it's not worth the risk of possibly damaging expensive gear.

[gf]

Let me clarify that: "More than two points per cycle of the highest frequency sine wave in the signal"

No, not nore than two points per cycle of the highest frequency in the signal, but an infinite number of samples preceeding the "time of interest" and an infinite number of samples succeeding the "time of interest" where we want to obtain the reconstructed value.

And a minimum of two samples per cycle...

Sorry, the phrasing of my last post does not say what I actually wanted to say. I have corrected my post.

The Nyquist limit at fs/2 applies of course, and the condition is "<" and not "<=", i.e. fmax < fs/2 (or fs > 2*fmax). So your statement "More than two points per cycle of the highest frequency sine wave in the signal" was indeed correct. I.e. a sine wave of say 49.999999999 MHz sampled at 100MSPS can be theoretically reconstructed exactly (but you you need an infinite number number of samples). When doing an approximate reconstruction via truncated sinc interpolation from a finite number of neighbor samples, fmax must not be that close to fs/2, but more headroom is required.