Is Bandwidth/memory depth a waste of money in oscilloscopes?

[Fungus]

If you are comparing different DSOs, then the judgment of easeness and intuitiveness is quite valid. If you are coming from a CRO, all modern DSO interfaces will seem unintuitive.

It's not that they're "unintuitive",  it's that they'd look like this if they had a button for every function and a potentiometer for every adjustment:


The features you use most are usually on hand or one (maybe two) buttons away.

Most people now don't "come from a CRO" anyway so they don't ever see this. The questions they ask are more about all the things it can do.

Case in point: it takes a single push of a button to recall a waveform in a tek 2232: MEMORY#, and it takes two to save it: SAVE + MEMORY#. In a modern DSO you've got to go quite deep into submenus to achieve the same, it's so cumbersome that often you simply end up not using waveforms.

But a $400 DSO can save multiple waveforms it to a USB stick so you can recall them later (and even email them to somebody on the other side of the world so they can load them into their 'scope and see them over there).

Plus you show a waveform on screen and still have multiple channels left over to overlay on top of it.

Functionality-wise, your old Tek is very limited, that's why it can have a simple interface.

[Pete F]

The bandwidth of an oscilloscope has a very specific definition, and it is completely false to believe that an oscilloscope of a specific bandwidth can only display sine wave signals to that bandwidth. The definition is that the displayed waveform will be 3 dB lower than the actual signal amplitude. The oscilloscope can (and probably will) continue to display signals way above the quoted bandwidth.

The rise time of an oscilloscope is also normally completely independent of the bandwidth, which is why the rise time is also quoted in the oscilloscope specifications.

[Fungus]

The rise time of an oscilloscope is also normally completely independent of the bandwidth, which is why the rise time is also quoted in the oscilloscope specifications.

No, Rise time will be strongly related to the bandwidth.

(it's approximately the reciprocal of it)

[Pete F]

The rise time of an oscilloscope is also normally completely independent of the bandwidth, which is why the rise time is also quoted in the oscilloscope specifications.

No, Rise time will be strongly related to the bandwidth.

(it's approximately the reciprocal of it)

That is incorrect

[TheUnnamedNewbie]

The rise time of an oscilloscope is also normally completely independent of the bandwidth, which is why the rise time is also quoted in the oscilloscope specifications.

No, Rise time will be strongly related to the bandwidth.

(it's approximately the reciprocal of it)

That is incorrect

Unless you can demonstrate why I doubt many people will believe you.

[Pete F]

This is a Tektronix paper that explains bandwidth and rise time in oscilloscopes, however there is no doubt many other sources. They are NOT related, and two oscilloscopes with the same bandwidth may have different rise times, and visa versa.

https://www.ece.ubc.ca/~robertor/Links_files/Files/TEK-Understanding-Scope-BW-tr-Fidelity.pdf

[Fungus]

This is a Tektronix paper that explains bandwidth and rise time in oscilloscopes, however there is no doubt many other sources. They are NOT related, and two oscilloscopes with the same bandwidth may have different rise times, and visa versa.

https://www.ece.ubc.ca/~robertor/Links_files/Files/TEK-Understanding-Scope-BW-tr-Fidelity.pdf

That paper gives you the formula for converting rise time to bandwidth:


Yes, there's always complications in real life do to extra capacitances in the wires, etc., but there's no way you can claim they're "completely independent" when there's a simple, widely used formula to approximate the relationship.

(and at sub-300MHz frequencies that formula is probably quite accurate)

The paper mostly talks about "Bandwidth for < 3% error", which is another thing entirely.

[rsjsouza]

If you are comparing different DSOs, then the judgment of easeness and intuitiveness is quite valid. If you are coming from a CRO, all modern DSO interfaces will seem unintuitive.

It's not that they're "unintuitive",  it's that they'd look like this if they had a button for every function and a potentiometer for every adjustment:

But they aren't built like this, so they are unintuitive to the person that is used with a 1:1 button assignment interface.

Intuitiveness is completely different than complexity.

Case in point: it takes a single push of a button to recall a waveform in a tek 2232: MEMORY#, and it takes two to save it: SAVE + MEMORY#. In a modern DSO you've got to go quite deep into submenus to achieve the same, it's so cumbersome that often you simply end up not using waveforms.

But a $400 DSO can save multiple waveforms it to a USB stick so you can recall them later (and even email them to somebody on the other side of the world so they can load them into their 'scope and see them over there).

Yes. Despite you could always take a photograph of the screen, stating that you end up not using it is a bit too much. The usefulness is quite high.

[Pete F]

This is a Tektronix paper that explains bandwidth and rise time in oscilloscopes, however there is no doubt many other sources. They are NOT related, and two oscilloscopes with the same bandwidth may have different rise times, and visa versa.

https://www.ece.ubc.ca/~robertor/Links_files/Files/TEK-Understanding-Scope-BW-tr-Fidelity.pdf

That paper gives you the formula for converting rise time to bandwidth:


Yes, there's always complications in real life do to extra capacitances in the wires, etc., but there's no way you can claim they're "completely independent" when there's a simple, widely used formula to approximate the relationship.

(and at sub-300MHz frequencies that formula is probably quite accurate)

If you feel you know better than the test equipment manufacturers you can knock yourself out. The "paper" you posted even stated that!

It used to be widely believed that putting butter on burns was the correct treatment too, however times change. The rise time of an oscilloscope will depend on the roll-off of higher frequencies beyond the nominal bandwidth. One again I will say that it is NOT simply related to an inverse relationship as you've claimed, from what I've seen, that fact is widely published by various highly respected manufacturers in the field. What you choose to believe is of course entirely up to you, but my money is on them knowing a bit about the topic

[GeorgeOfTheJungle]

But a $400 DSO can save multiple waveforms it to a USB stick so you can recall them later (and even email them to somebody on the other side of the world so they can load them into their 'scope and see them over there).

Yes. Despite you could always take a photograph of the screen, stating that you end up not using it is a bit too much. The usefulness is quite high.

Lol, a pendrive? A photograph of the screen? What for? It seems to me neither of you two have ever used waveforms... for what they are.

[Fungus]

a $400 DSO can save multiple waveforms it to a USB stick so you can recall them later (and even email them to somebody on the other side of the world so they can load them into their 'scope and see them over there).

Yes. Despite you could always take a photograph of the screen, stating that you end up not using it is a bit too much. The usefulness is quite high.

Yes, and you can easily zoom in using a magnifying glass. Analog 'scopes rule!

[Pete F]

This is HP's take on rise time/bandwidth.

http://literature.cdn.keysight.com/litweb/pdf/5988-8008EN.pdf

For the average person using an oscilloscope like I just purchased, the difference between the theoretical (ie the traditional) and actual rise time is likely a moot point, however most decent 'scopes I've seen will have their rise time quoted in the specs, and this is the reason they do so.

[nctnico]

The rise time of an oscilloscope is also normally completely independent of the bandwidth, which is why the rise time is also quoted in the oscilloscope specifications.

No, Rise time will be strongly related to the bandwidth.

(it's approximately the reciprocal of it)

That is incorrect

Unless you can demonstrate why I doubt many people will believe you.

It is very simple: it depends on the input filter. If that is Guassian then the 'regular' formula applies. If the filter is more like a brick-wall filter then the risetime will be different. Look at the Keysight MSOX3104 for example.

So: NO, the risetime and bandwidth are not related.

[Fungus]

If you feel you know better than the test equipment manufacturers you can knock yourself out. The "paper" you posted even stated that!

OK, let's keep reading:



Ah, so it can be as high as 0.45?

That's a whole 28% higher than 0.35 when you start get into the GHz range.

Yes, I agree. The reciprocal relationship is completely broken. Not.

[Fungus]

It is very simple: it depends on the input filter. If that is Guassian then the 'regular' formula applies. If the filter is more like a brick-wall filter then the risetime will be different. Look at the Keysight MSOX3104 for example.

So: NO, the risetime and bandwidth are not related.

Technically true, but not really relevant. Most oscilloscopes are Gaussian if we're honest.

[Specmaster]

The rise time of an oscilloscope is also normally completely independent of the bandwidth, which is why the rise time is also quoted in the oscilloscope specifications.

No, Rise time will be strongly related to the bandwidth.

(it's approximately the reciprocal of it)

That is incorrect

Unless you can demonstrate why I doubt many people will believe you.

It is very simple: it depends on the input filter. If that is Guassian then the 'regular' formula applies. If the filter is more like a brick-wall filter then the risetime will be different. Look at the Keysight MSOX3104 for example.

So: NO, the risetime and bandwidth are not related.

They are related in a way and I'm not an expert on scopes, but I do know that manufacturers are not going to spend a lot of money developing a really fast rise risetime on say a 50 Mhz bandwidth scope, they will provide a risetime suitable for frequencies upto and just maybe over the bandwidth of the said scope. Therefore it makes perfect sense that as the bandwidth of the scope increases, the makers have to provide a risetime capable of handling at least the frequency stated for the bandwidth of that scope otherwise its going to be incapable of displaying the waveform correctly above a certain frequency.

[Pete F]

Yep, you keep digging that hole "Fungus" 

Meanwhile the rest of us will move on I think.

In regard the intuitiveness of the Siglent 1202, I would have to say I didn't find it especially intuitive in many areas. However that shouldn't necessarily be seen to be a criticism of Siglent, and it could well be that they are no better or worse than other similar offerings from other manufacturers. In my case I have spent many years, well decades unfortunately, using old-school analogue 'scopes so this is certainly quite a change for me. There's no denying in my mind however that the power of instrument is quite mind blowing in my opinion (especially considering the prices!), and would be completely unthinkable if I was to put it up against CROs from the same era as say my 2215. As mentioned above, I'm sure I will get used to the commonly used functions and accessing them quickly, but they will never be as quick as pretty much any analogue 'scope I've used in the past. For a hobbyist it's probably no big deal, but if I were still doing it professionally I think I would find it a bit frustrating to be honest; time is money, and when fault-finding you're changing functions a lot.

[Fungus]

Yep, you keep digging that hole "Fungus" 

What hole?

In practice brick wall filters cannot be achieved in hardware, they have to be done in software.

This means that if you want a 'scope with a a non-gaussian filter you have to sample at a much higher rate than Nyquist then digitally post-process it.

ie. you have to massively underrate the bandwidth which is printed on the front - inside the oscilloscope it has to actually have the bandwidth needed to match the sample rate.

In other words, it's a bit of a con game played in "high-end" oscilloscopes. That Tek paper which was posted earlier is really just a sales brochure in disguise ("You want a 'scope with only 3% error in rise time instead of 30%? We sell those!").

The reality is that a 'scope with 3% error is just a scope with 30% error (ie. gaussian filter) that's been derated, has a software filter applied internally, and a different label stuck on the front.

If you want to believe those scopes are breaking math, well... that's up to you.

[rf-loop]

Yep, you keep digging that hole "Fungus" 

What hole?

In practice brick wall filters cannot be achieved in hardware, they have to be done in software.

Where from you get this trumpth. Even I have designed and done some analog "brick wall" filters and I have done these time when Intel 4004 was young child and Mot 6800 just born. Do you think this era it need DSP. 

[Pete F]

Yep, you keep digging that hole "Fungus" 

What hole?

In practice brick wall filters cannot be achieved in hardware, they have to be done in software.

This means that if you want a 'scope with a a non-gaussian filter you have to sample at a much higher rate than Nyquist then digitally post-process it.

ie. you have to massively underrate the bandwidth which is printed on the front - inside the oscilloscope it has to actually have the bandwidth needed to match the sample rate.

In other words, it's a bit of a con game played in "high-end" oscilloscopes. That Tek paper which was posted earlier is really just a sales brochure in disguise ("You want a 'scope with only 3% error in rise time instead of 30%? We sell those!").

The reality is that a 'scope with 3% error is just a scope with 30% error (ie. gaussian filter) that's been derated, has a software filter applied internally, and a different label stuck on the front.

If you want to believe those scopes are breaking math, well... that's up to you.

I tell you what, why don't you write to Tektronix/Keysight and tell them how they have it all wrong, and you know, for a fact, that it's all a big conspiracy to sell more high end oscilloscopes. Let is know how you get on 

Incidentally I didn't "massively underrate" the bandwidth that is printed on the front, in fact I specifically stated "The oscilloscope can (and probably will) continue to display signals way above the quoted bandwidth."

Like I said, the difference between the actual, calculated, and archaic theoretical rise time is probably a moot point for most people, and I doubt many people here would have the actual capacity to accurately measure the rise time of a fast oscilloscope in any case. However the point is that the rise time may be of significance for some people in their applications and they may specifically seek that when looking for their test instrument. It is incorrect to assume all oscilloscopes of the same bandwidth will have the same rise time. Tektronix say that. Agilent say that. Keysight say that. In fact the only person who seems to not be saying that is you 

In practice the solution if the rise time is too slow is simply to get a higher bandwidth oscilloscope. However one shouldn't assume a specific rise time based merely on quoted bandwidth, as that is not always a correct assumption.

BTW my background was largely in RF, and I'm offering this emoji to the assertion that filtering needs to be done in software 

[David Hess]

At higher bandwidth, it becomes more difficult to maintain a Gaussian response so even with analog oscilloscopes, some higher bandwidth instruments deviated from the 0.35 rule.  In some cases, the same instrument was available tuned for either Gaussian response sacrificing bandwidth or higher bandwidth sacrificing transient response.

Keysight and Tektronix are not making those high bandwidth 0.45 oscilloscopes without reason.  They are easier and cheaper to make than one with a 0.35 Gaussian response and bandwidth sells.  If you are using your oscilloscopes as a frequency domain instrument, this even makes some sense which is why it was sometimes done in the past with analog oscilloscopes.


Modern oscilloscope user interfaces are not poorer because of complexity when even basic functions have been buried in menu structures or other poorly thought out interface methods; they just have poor human factors engineering.  But complexity makes a good excuse.

The same problem comes up with anything modern which uses skeuomorphic interface elements.  What genius, other than Apple who can do no wrong, thought that was a good idea?

[David Hess]

I tell you what, why don't you write to Tektronix/Keysight and tell them how they have it all wrong, and you know, for a fact, that it's all a big conspiracy to sell more high end oscilloscopes. Let is know how you get on 

The only people who will respond are from marketing so why bother?

Like I said, the difference between the actual, calculated, and archaic theoretical rise time is probably a moot point for most people, and I doubt many people here would have the actual capacity to accurately measure the rise time of a fast oscilloscope in any case. However the point is that the rise time may be of significance for some people in their applications and they may specifically seek that when looking for their test instrument. It is incorrect to assume all oscilloscopes of the same bandwidth will have the same rise time. Tektronix say that. Agilent say that. Keysight say that. In fact the only person who seems to not be saying that is you 

The people who are going to spend $20,000+ on a high bandwidth DSO with a non-Gaussian response are not going to be reading this thread for advice.  That leaves hobbyists who do not know any better and buy toys and everybody else who will only be able to afford a crummy Gaussian response oscilloscope.

In practice the solution if the rise time is too slow is simply to get a higher bandwidth oscilloscope. However one shouldn't assume a specific rise time based merely on quoted bandwidth, as that is not always a correct assumption.

Wasn't the argument that rise time has nothing to do with bandwidth?  Have you changed sides?

[Pete F]

Excellent points David. One of the functions I change quite often is to look at the DC/AC component of a test point ie changing the input coupling. I found it a bit clumsy when changing that in each channel vs what I'm previously used to which is just flicking a switch.

By the same token I'm probably being a bit critical. Technology wise it's a little like winning the lottery, the complaining when the winnings are delivered in $10 bills instead of $100s I am still amazed at what that little machine is capable of doing, it's just that I'm glad I didn't throw the old Tek out when I had the chance!

No David I haven't changed sides, and I have never said it has "nothing to do with bandwidth". What I said was that the rise time was NOT strictly dependent on the bandwidth. I've said that quite a number of times and provided specific data from manufacturers to support that statement.

[rsjsouza]

But a $400 DSO can save multiple waveforms it to a USB stick so you can recall them later (and even email them to somebody on the other side of the world so they can load them into their 'scope and see them over there).

Yes. Despite you could always take a photograph of the screen, stating that you end up not using it is a bit too much. The usefulness is quite high.

Lol, a pendrive? A photograph of the screen? What for? It seems to me neither of you two have ever used waveforms... for what they are.

Perhaps the 2232 performs additional functions but, according to its manual, the waveform storage looks no different than a modern DSO saving the actual data (either in .csv or a proprietary format) to a pendrive and being able to perform advanced analysis on it, or keep it in memory to serve as a reference waveform for simple visual comparison or automated pass/fail tests. In addition to that, DSOs can also save bitmapped images of the screen (.jpg, .png, .bmp, etc.).

My surprise with your statement was that you mentioned that waveforms in a DSO are rarely used due to the added functionality that requires additional steps to properly configure it. At least in my experience that is not the case at all.

[Fungus]

The reality is that a 'scope with 3% error is just a scope with 30% error (ie. gaussian filter) that's been derated, has a software filter applied internally, and a different label stuck on the front.

Incidentally I didn't "massively underrate" the bandwidth that is printed on the front, in fact I specifically stated "The oscilloscope can (and probably will) continue to display signals way above the quoted bandwidth."

Seems like you're not sure which side you're one.

nb. Even a Rigol DS1054Z will display something up to about 300mHz. It'll be very very attenuated but it will be there.

BTW my background was largely in RF, and I'm offering this emoji to the assertion that filtering needs to be done in software 

That's not what I said at all.

Gaussian filtering obviously doesn't need software - we spend a lot of time and money trying to eliminate it in our hardware.

Brick-wall filtering? If you know how to make a nice brick-wall filter in hardware then I'm sure Tek., et., al. will be very interested.