Comparing Agilent InfiniiVision 2000 and 3000 X-Series Oscilloscopes

[yodhe]

The Wavesurfer 3000 does have an AWG option.

http://teledynelecroy.com/options/productseries.aspx?mseries=472&groupid=144

[Someone]

Ahh. Choices, choices. I didn't feel like the memory depth would have been a big issue (especially when using channel one and three on the four channel version) but I can see how it could potentially become a problem in the future. I do recall Dave mentioning in one of his reviews that he quite likes the idea of having memory managed automatically, and coming from a 2000A it was easy for me to agree with him (whether that is justified or not is a separate matter altogether, haha).

Memory depth controls can be important on scopes where the acquisition rate, measurements, or user interface slow down with longer memory depths, Tek and Rigol in particular but most scopes do this to some extent. The Agilent/Keysight X series didn't need the option to reduce the acquisition depth as they offer so much speed even at full memory depth, its only reduced below maximum when using the segmented capture mode.

[Faith]

The Wavesurfer 3000 does have an AWG option.

http://teledynelecroy.com/options/productseries.aspx?mseries=472&groupid=144

Err no idea how I missed that... think I got that bit mixed up with Rohde & Schwarz at some point >,<"...

Its best to make your own value choices rather than relying on others such as Mr W, he likes to spin all aspects of Lecroy products as advantages even when they are undesirable features. Memory depth controls can be important on scopes where the acquisition rate, measurements, or user interface slow down with longer memory depths, Tek and Rigol in particular but most scopes do this to some extent. The Agilent/Keysight X series didn't need the option to reduce the acquisition depth as they offer so much speed even at full memory depth, its only reduced below maximum when using the segmented capture mode.

Yep, ultimately I will make a purchasing decision based on my own experiences and comfort levels.

While I did some homework here and there I did eventually end up between the MSOX3024T & MSOX4024A; thus my original question, which has been completely derailed by now

Of course 99% of my judgement is based from my familiarity with the 2000A, but I never really found its 1M memory depth to be too big of an issue, and hence why I didn't think much of it when I saw that the 3000T/4000A are 4M.

[nctnico]

Memory depth controls can be important on scopes where the acquisition rate, measurements, or user interface slow down with longer memory depths, Tek and Rigol in particular but most scopes do this to some extent. The Agilent/Keysight X series didn't need the option to reduce the acquisition depth as they offer so much speed even at full memory depth, its only reduced below maximum when using the segmented capture mode.

Speed is relative. With my DSO7000 I have to do some measurements several times because it is not very clear what samplerate is used and how much memory is actually available. All in all I'd be finished quicker using an oscilloscope which has more memory, is clear about the memory length & samplerate even though the UI might be slower.

[Wuerstchenhund]

PicoScope 6000 series excels at every points you have listed.

No, they don't. You get stuff like more in sample memory, and (depending on your computer's monitor) a larger screen. What you don't get are advanced glitch finders like WaveScan, or even so basic things in this class like an active probe interface. And you get an UI that requires having mouse and keyboard on your bench.

We use PicoScopes at work (mostly for automated test systems where they are great), but even their most ardent fans wouldn't suggest them as a replacement for a proper standalone scope.

The main reason for considering Keysight is because I am already familiar with the 2000A. And you know how it is, some of us don't like to change what we are used to or familiar with

I know, and that is a perfectly valid argument.

On the other side, it's a lot of money and I guess most people would want to get the most bang for their bucks, and at the end of the day it helps to know all available options even if you then go for original choice.

I did take a brief look at Teledyne LeCroy and more specifically their WaveSurfer 3000 but if I recall correctly it does not have a built-in Arbitary Function Generator. This is one feature I would certainly like as it would mean that there is less equipment occupying my limited bench space.

It does have an AWG, as stated in the list:
http://cdn.teledynelecroy.com/files/pdf/wavesurfer-3000-datasheet.pdf

Ahh. Choices, choices. I didn't feel like the memory depth would have been a big issue (especially when using channel one and three on the four channel version) but I can see how it could potentially become a problem in the future. I do recall Dave mentioning in one of his reviews that he quite likes the idea of having memory managed automatically, and coming from a 2000A it was easy for me to agree with him (whether that is justified or not is a separate matter altogether, haha).

Back in 2012 when the DSOX came out many low-end scopes required the user to manually select between short and long memory (i.e. you had two choices). The MegaZoom manages the memory automatically, i.e. the user has no choice how the memory is used, and the scope keeps the used amount of memory at minimum required to fill the display to increase the waveform rate.

The WaveSurfer lets you select the amount of sample memory manually, but you don't have to.

Its best to make your own value choices rather than relying on others such as Mr W, he likes to spin all aspects of Lecroy products as advantages even when they are undesirable features.

Memory depth controls can be important on scopes where the acquisition rate, measurements, or user interface slow down with longer memory depths, Tek and Rigol in particular but most scopes do this to some extent. The Agilent/Keysight X series didn't need the option to reduce the acquisition depth as they offer so much speed even at full memory depth, its only reduced below maximum when using the segmented capture mode.

  There's so much wrong with that it's not even funny.

First, you completely ignore that the time it takes to fill a certain amount of sample memory at a certain sample size is fixed:

                  capture time [seconds] = sample memory size [pts] / sample rate [Sa/second]

This basic law hold true even for a Keysight DSOX InfiniiVision scope. So on your (I believe) DSOX3kA, at 4GSa/s it takes 1ms to fill 4Mpts of sample memory. Even if we had a perfect trigger circuit immediately re-arming, the time it takes to fill the 4Mpts alone means that even with that perfect trigger the update rate would be limited to 1000 per second. There's no way around that, it's a physical reality.
 
What MegaZoom does is using windowing to reach its high update rates. Windowing is used during normal operation where the scope only uses enough memory to fill the period shown on screen. That means even if it shows you it's using 2Mpts it's actually just using a few kpts. Only when you stop the acquisition system or enable zoom then the last acquisition uses all of the displayed sample memory. It's pretty much a trick to circumvent the fact that using all the displayed memory at the displayed sample rate would make it physically impossible to attain high sample rates. This is also the reason the InfiniiVision doesn't allow manual controls.

[Someone]

Memory depth controls can be important on scopes where the acquisition rate, measurements, or user interface slow down with longer memory depths, Tek and Rigol in particular but most scopes do this to some extent. The Agilent/Keysight X series didn't need the option to reduce the acquisition depth as they offer so much speed even at full memory depth, its only reduced below maximum when using the segmented capture mode.

  There's so much wrong with that it's not even funny.

First, you completely ignore that the time it takes to fill a certain amount of sample memory at a certain sample size is fixed:

                  capture time [seconds] = sample memory size [pts] / sample rate [Sa/second]

This basic law hold true even for a Keysight DSOX InfiniiVision scope. So on your (I believe) DSOX3kA, at 4GSa/s it takes 1ms to fill 4Mpts of sample memory. Even if we had a perfect trigger circuit immediately re-arming, the time it takes to fill the 4Mpts alone means that even with that perfect trigger the update rate would be limited to 1000 per second. There's no way around that, it's a physical reality.
 
What MegaZoom does is using windowing to reach its high update rates. Windowing is used during normal operation where the scope only uses enough memory to fill the period shown on screen. That means even if it shows you it's using 2Mpts it's actually just using a few kpts. Only when you stop the acquisition system or enable zoom then the last acquisition uses all of the displayed sample memory. It's pretty much a trick to circumvent the fact that using all the displayed memory at the displayed sample rate would make it physically impossible to attain high sample rates. This is also the reason the InfiniiVision doesn't allow manual controls.

If you want to frame it in terms of time period captured on the screen then the update rate does has a ceiling of:
1/capture time
But most scopes will degrade their realtime waveform capture rate below the theoretical maximum when you increase the memory depth, which was the key differentiator of the "always on" memory and not having a memory depth control on the Agilent/Keysight X series scopes since they processed the waveforms to the display in the ASIC and wouldn't run substantially faster if the memory depth was reduced. I even plotted up the collected data from this forum to make it easy to compare:
https://www.eevblog.com/forum/testgear/rigol-mso4000-and-ds4000-tests-bugs-firmware-questions-etc/msg973064/#msg973064
All the data points at whatever the sample rate is gets to the screen, even though on the Agilent/Keysight design it is rendered at a lower resolution and you can only zoom in on the last capture and not the accumulated graduated display. As an alternative method didn't the Wavejets let you zoom and pan the graduated capture along with their segmented captures?

But to say that manual control of memory depth is a better thing to have when its just used to hide other downsides to a product is once again disingenuous. Ideally you would want to operate your scope at maximum memory depth all the time! And Agilent/Keysight delivered this.

[Wuerstchenhund]

But most scopes will degrade their realtime waveform capture rate below the theoretical maximum when you increase the memory depth, which was the key differentiator of the "always on" memory and not having a memory depth control on the Agilent/Keysight X series scopes since they processed the waveforms to the display in the ASIC and wouldn't run substantially faster if the memory depth was reduced.

Most scopes will decrease their update rate because they are actually using the memory they claim to use for every single acquisition.

The DSOX (like it's InfiniiVision predecessors) only manage to reach faster because they are *not* using the full memory most of the time. They have to, as it is physically impossible to reach high waveform numbers while using all of that paltry 4MB memory, as shown in my previous post.

I even plotted up the collected data from this forum to make it easy to compare:
https://www.eevblog.com/forum/testgear/rigol-mso4000-and-ds4000-tests-bugs-firmware-questions-etc/msg973064/#msg973064

All the data points at whatever the sample rate is gets to the screen, even though on the Agilent/Keysight design it is rendered at a lower resolution and you can only zoom in on the last capture and not the accumulated graduated display.

Which is one limitation of the MegaZoom platform.

As an alternative method didn't the Wavejets let you zoom and pan the graduated capture along with their segmented captures?

I believe so, but then I haven't had much contact with them, and since they are essentially Iwatsu rebadges a few things work differently on these scopes than on 'real' LeCroy scopes.

But to say that manual control of memory depth is a better thing to have when its just used to hide other downsides to a product is once again disingenuous. Ideally you would want to operate your scope at maximum memory depth all the time! And Agilent/Keysight delivered this.

No, they did not! They delivered a system which creates the impression to the (uninformed) user that it always uses maximum sample memory when in fact it is cheating its way out. This also has consequences, i.e. for recording pre-trigger events.

I've no doubt that the way MegaZoom works makes it easier for beginners. But spinning the absence of manual controls as an advantage is preposterous. It's like saying a scope that only has an auto-set button is better than one that has manual knobs and buttons because the auto-setup system always produces a stable image (well, it doesn't, but let's ignore that for a moment). The point of the various controls of a test instrument is that the user can setup the instrument to fit the measurement situation, and that works better with more control than with less. I'm not against automatics in scopes (unlike many of my peers), but only as long as they don't get in the way of giving the user control of the instrument. They do become a hindrance when automatics are the only option, even more so when the 'price' for that automatism is a comparably tiny sample memory.

Of course Keysight knows this very well, after all pretty much all of its high-end scopes (aside from the DSOX6004, but that's pretty much just an expensive variant of the DSOX3k which is an entry-level scope) allow manual control of memory and sample rate. The DSOX doesn't because it was pretty much designed for the highest possible update rate at the price of everything else, which was pretty much Agilent's marketing spin back in 2012. But there's a reason why other scopes (even keysight's own scopes) didn't follow that path, and more than compensate their lower waveform rate with advanced triggering and analysis tools.

The more I have to carry my MSOX3104A across lab, home and office, the more I wish I have bought a PS6000. PS have some quite advanced triggers that can automatically analyze something, but I have never really looked into it. Talking about active probe IF, it is really a PITA. However, the active probe IF on my MSOX is also next to useless -- the probes quickly sum up more than the price of the scope. It is common to see a probe selling for $1000, and the tips $4000, which is ridiculous. Thanks to the custom coax interface, even if I have a prove and a scope, without its proprietary tips, it is next to useless.

Well, that's normal as stuff that is labelled Agilent/Keysight is pretty sought after, as is Tek. Simply because when thinking about scopes Agilent/Keysight and Tek are for many people the only names that come to mind. But just because some seller wants $4k for something doesn't mean it's selling. Of course, if its compatible with the DSOX it will fetch extra money, thanks to Agilent/Keysight making some probes artificially incompatible with the InfiniiVision scopes and thereby limiting the choice for InfiniiVision scope users.

On the other side, I've bought several active LeCroy probes (AP033, AP034, HPF Series) for a couple of hundred dollars, as have other forum users. Thanks to most people focussing on Agilent/Keysight and Tek there are a lot more deals for LeCroy probes (although it does require patience). It also helps that LeCroy's active probe interface hasn't changed since the '90s, and because of the unified software stack there are pretty much no compatibility issues with older probes on newer scopes.

But yes, if you have to schlepp around the scope regularly then the PicoScope might be the better option, even more so when you're already using a PC on the bench.

[Howardlong]

So I have set up permanently MSOX3054A, MSOX7104B, MDO3054, and 54832D on the bench. Since I put the MSOX7104B on a slide out swivel mount above the desk, that gets by far the most use, but it used to be the MSOX3054A.

Depite not being able to freely manually set the acquisition sample rate and/or memory depth, by far most of the time, I find the way the MSOX series work are most conducive to my workflow.

With the MDO3054 and 54832D I can manually set those parameters, but I actually find that by far most of the time, having to attend to that step gets in the way of what I'm trying to do, and I do find myself needing to alter the memory depth quite a lot particularly on the Tek where it's a compromise between sluggish performance with long memory or faster response with short memory. To some extent that can also apply to the 54832D, but there are also other problems with the Tek: most notably the UI as a whole which as well as being unresponsive is a bit of a potpourri, the crazy paving of UIs, it feels like everything's a fight.

Although on occasion being able to adjust the memory settings might seem a reasonable thing to want to do, in practice I just find that I use a slightly different workflow depending on the scope. If the proof is in the pudding, almost all the time I find myself using that of the MSOX series, but others may have different needs. I certainly don't find it a limitation -most- of the time.

[Wuerstchenhund]

With the MDO3054 and 54832D I can manually set those parameters, but I actually find that by far most of the time, having to attend to that step gets in the way of what I'm trying to do, and I do find myself needing to alter the memory depth quite a lot particularly on the Tek where it's a compromise between sluggish performance with long memory or faster response with short memory. To some extent that can also apply to the 54832D, but there are also other problems with the Tek: most notably the UI as a whole which as well as being unresponsive is a bit of a potpourri, the crazy paving of UIs, it feels like everything's a fight.

As a fresh owner of a Tek MDO3054 I can certainly understand your plight, but that is pretty much due to the MDO being a very poor scope with a slow platform and a horrendous UI designed by someone who probably hates humans.

As to UI, the 58432D isn't really a lot better with its mouse-driven UI. My DSO8064 uses the same, and although it has a higher resultion display with touch, I find its menu system cumbersome, especially for basic things like changing sample memory. I leave it most of the time on auto as well but, despite all criticism, at least there's the option to change the auto settings manually. 

Just to compare, on my LeCroy it's just a tab on the Acq box in the lower right corner to get to the memory settings.

Although on occasion being able to adjust the memory settings might seem a reasonable thing to want to do, in practice I just find that I use a slightly different workflow depending on the scope. If the proof is in the pudding, almost all the time I find myself using that of the MSOX series, but others may have different needs. I certainly don't find it a limitation -most- of the time.

Completely understandable, but this is probably more down to the fact that neither the Infiniium 54832D nor the Tek MSO3054 are good every-day scopes than to the InfiniiVisions being particular great. But whatever works for you.

[Someone]

Really, have a look at the LeCroy WaveSurfer 3000. It's a lower mid-range scope which comes with a few advantages over the DSOX3000T, for example (copied from an older posting of mine):

• ...
• The WaveSurfer allows automatic and manual sample memory/sample rate management while the DSOX3kT is automatic only
• ...

Since you seem to have wandered way off and reframed the discussion several times to distract from this, can you provide examples of when it is desirable to use less than the maximum memory possible? I can't think of any situation where I would want to use less memory depth on an Agilent/Keysight X series scope, there is no downside to having the maximum memory enabled at all times. There may be downsides on other scopes but there are none here, the absence of a memory depth control is not a negative point and rather points to deficiencies of other products which require it.

[nctnico]

Really, have a look at the LeCroy WaveSurfer 3000. It's a lower mid-range scope which comes with a few advantages over the DSOX3000T, for example (copied from an older posting of mine):

• ...
• The WaveSurfer allows automatic and manual sample memory/sample rate management while the DSOX3kT is automatic only
• ...

Since you seem to have wandered way off and reframed the discussion several times to distract from this, can you provide examples of when it is desirable to use less than the maximum memory possible? I can't think of any situation where I would want to use less memory depth on an Agilent/Keysight X series scope, there is no downside to having the maximum memory enabled at all times. There may be downsides on other scopes but there are none here, the absence of a memory depth control is not a negative point and rather points to deficiencies of other products which require it.

If you forget that on average the actual record length in the Keysight scopes is 10 to 50 times shorter than in other scopes then you are right.

[Someone]

Really, have a look at the LeCroy WaveSurfer 3000. It's a lower mid-range scope which comes with a few advantages over the DSOX3000T, for example (copied from an older posting of mine):

• ...
• The WaveSurfer allows automatic and manual sample memory/sample rate management while the DSOX3kT is automatic only
• ...

Since you seem to have wandered way off and reframed the discussion several times to distract from this, can you provide examples of when it is desirable to use less than the maximum memory possible? I can't think of any situation where I would want to use less memory depth on an Agilent/Keysight X series scope, there is no downside to having the maximum memory enabled at all times. There may be downsides on other scopes but there are none here, the absence of a memory depth control is not a negative point and rather points to deficiencies of other products which require it.

If you forget that on average the actual record length in the Keysight scopes is 10 to 50 times shorter than in other scopes then you are right.

Long memory can be important and/or useful, but that doesn't explain why a scope needs to have a memory depth control. You're welcome to have a crack at the question too rather than derailing the thread further.

[nctnico]

Setting the memory depth can be handy for speeding up things like math traces, measurements, etc. Still it would be a great help if Keysight showed the memory length and samplerate so (especially when using segmented recording) you have an idea on whether the amount memory/samplerate is going to be adequate or not.

[nfmax]

So: memory depth...

If I'm looking at a repetitive signal, one which repeats often enough to fool the eye, then so long as I have enough memory to fill the screen (and handle the pre trigger delay I've set), having more memory doesn't improve anything much. If I want to zoom in or pan, I just adjust the sweep rate control & trigger delay, and the next trigger event fills the memory with my chosen settings. In fact, having too large a memory slows the waveform update rate down (since each trigger must fill all of memory) which brings its own problems, namely the risk of missing transient signal which I'm not triggering on. Operating in this mode, I can set infinite persistence on, and wait for the odd event to happen. If I can see it on screen, I can set a trigger condition for it.

Things are different in single shot mode. Now, I'm triggering on the rare event, and when it happens, I'd like to capture as much of it as possible. Memory depth is now really important, and waveform update rate generally doesn't matter much at all. Memory depth sets the amount I can zoom in, post acquisition, and also the time range I can pan over. Memory depth is king.

So if I'm working in single shot mode, I should set the memory depth to maximum, but if I'm working in repetitive trigger mode, I should set it to minimum (there really doesn't seem to be much case for setting it to any other value).

As I understand it, the MegaZoom 4 system in all the current InfiniVison scopes does this for me, automatically. In single shot mode, I get to use all the memory available - the 1Mpt (2000 series) or 4Mpt (3000/4000 series) per channel is filled by the trigger & I can pan & zoom around as I want (though of course, more memory would be nice, as it decidedly on the small side). In repetitive acquisition (Run mode), the memory is split in two and used in swinging buffer fashion. One half is used for acquisition while the other half is being rendered to screen. The rendering is done by the ASIC and is very fast. In the 2000 series, there is a single ASIC and it takes no more than 50 microseconds to update the screen and free the buffer for the next acquisition. In the 3000 series there is a pair (or maybe a quartet?) of ASICs and the screen is updated inside one microsecond. So there is little benefit to be gained by reducing the memory size below half the maximum value.

Hence neither of these scopes has, or needs, an explicit memory size setting - only as part of the run/single shot configuration.

I've used both 2000 and 3000 InfiniVision scopes, and this seems to be the way they work. If Daniel is following this thread maybe he could confirm/clarify this. What happens in segmented mode I'm not so sure. I've only used the 2000 in segmented mode and it looks as if you only get to use half the total memory to store all your segments, because effectively you are operating in run mode so that the dead time between segment triggers is minimised. This is a pity, because the screen is not being updated, so it looks as if you ought to be able to use it all. Maybe the 3000 series work this way? (We don't have the segmented memory license for the 3000)

[nctnico]

You forget one thing: you don't want to hit the 'single' button for each acquisition. I regulary use the normal trigger mode to capture a signal and only zoom in on the interesting acquisitions. This leaves me with half the memory depth and sometimes this means the signal is not as detailed as I would like.

[nfmax]

I agree, if you just hit 'stop' when you see something interesting, that's what you'll get. Assuming you hit 'stop' fast enough!

Using the run controls as intended, you can get twice the memory. it's a bit like relying on the 'auto set' button instead of setting your channels, trigger and sweep manually (auto setup isn't particularly effective on these scopes, BTW. Oddly enough, the DSO1000  family does it better - presumably it's Rigol software in those instruments!)

[nctnico]

I agree, if you just hit 'stop' when you see something interesting, that's what you'll get. Assuming you hit 'stop' fast enough!

If there is only one event to trigger on (generated by pushing a button manually), then there is plenty of time to hit the 'stop' button.

[nfmax]

I agree, if you just hit 'stop' when you see something interesting, that's what you'll get. Assuming you hit 'stop' fast enough!

If there is only one event to trigger on (generated by pushing a button manually), then there is plenty of time to hit the 'stop' button.

...and plenty of time to push the 'Single' button, too. Think of it as the 'maximum memory' button if that helps.

[nfmax]

I agree, if you just hit 'stop' when you see something interesting, that's what you'll get. Assuming you hit 'stop' fast enough!

Trigger inhibit delay and segmented memory are there to help.

Agreed - these two facilities together are very powerful!

[nctnico]

I agree, if you just hit 'stop' when you see something interesting, that's what you'll get. Assuming you hit 'stop' fast enough!

If there is only one event to trigger on (generated by pushing a button manually), then there is plenty of time to hit the 'stop' button.

...and plenty of time to push the 'Single' button, too. Think of it as the 'maximum memory' button if that helps.

So instead of one button I need to push 2 buttons? Excellent ergonomics! I rather user a different scope with more memory.

[kcbrown]

Memory depth controls can be important on scopes where the acquisition rate, measurements, or user interface slow down with longer memory depths, Tek and Rigol in particular but most scopes do this to some extent. The Agilent/Keysight X series didn't need the option to reduce the acquisition depth as they offer so much speed even at full memory depth, its only reduced below maximum when using the segmented capture mode.

  There's so much wrong with that it's not even funny.

First, you completely ignore that the time it takes to fill a certain amount of sample memory at a certain sample size is fixed:

                  capture time [seconds] = sample memory size [pts] / sample rate [Sa/second]

This basic law hold true even for a Keysight DSOX InfiniiVision scope. So on your (I believe) DSOX3kA, at 4GSa/s it takes 1ms to fill 4Mpts of sample memory. Even if we had a perfect trigger circuit immediately re-arming, the time it takes to fill the 4Mpts alone means that even with that perfect trigger the update rate would be limited to 1000 per second. There's no way around that, it's a physical reality.

Oh yes there is.  Or, there is unless I'm missing something vital.

If the trigger mechanism is based on the actual samples (as is the case with modern DSOs if I'm not mistaken), then the scope can simply treat the memory as a circular buffer and mark the memory locations at which the trigger conditions were satisfied.  The display mechanism can then display the appropriate region of the memory by examining the list of triggered locations and selecting whichever one is most suitable (ideally, it would be a region that is far removed from the region that the sampling mechanism is currently writing to so as to avoid collisions).   With that scheme, hitting the "stop" button would stop the sampling mechanism at something like t + 1/2 * memory-fill-time, where t is the time of the last trigger event, and would cause the display engine to display the waveform at time t.  To get away with this, you'd need to either use dual-ported RAM or RAM that is at least twice as fast as the fastest sampling rate (which implies a memory controller that is also suitably fast).

If you do that, then the trigger rearm mechanism would be completely independent of the buffer size.  It would obviously have to be able to keep up with the current sampling rate in order to ensure that it doesn't miss any events, which means being able to keep up with the maximum sampling rate.  If it can't keep up with the sampling rate, then there would be some chance of missing a triggering event.  Even then, once the sampling is stopped, the triggering mechanism can walk backwards through the samples to find previous occurrences of triggering events that it missed.

What MegaZoom does is using windowing to reach its high update rates. Windowing is used during normal operation where the scope only uses enough memory to fill the period shown on screen. That means even if it shows you it's using 2Mpts it's actually just using a few kpts. Only when you stop the acquisition system or enable zoom then the last acquisition uses all of the displayed sample memory. It's pretty much a trick to circumvent the fact that using all the displayed memory at the displayed sample rate would make it physically impossible to attain high sample rates. This is also the reason the InfiniiVision doesn't allow manual controls.

What's on the screen is just a subsampling of what has been captured.  That clearly has to be the case since the screen is resolution-limited, and thus is guaranteed to be able to display only a limited number of points (whatever the horizontal resolution of the screen is).  But clearly, the sampling interval has to be at most the amount of time the screen represents divided by the horizontal resolution if the screen's resolution is to be utilized to its fullest (if the time base is set small enough, of course, then there will be no choice but to display a smaller number of points than represented by the horizontal resolution, and some sort of interpolation would be needed to fill in the gaps).

Regardless, with the circular buffer model, I see no advantage to using a smaller amount of memory than what the scope actually makes available, since doing so only ensures that the buffer is cycled through more quickly by the sampling engine.  Since manufacturers obviously change the size of the buffer to achieve better performance, I must be missing something crucial here, but I don't have any idea what that is.

[nfmax]

Isn't there a problem with this approach that when you hit Stop, you won't know what memory depth you will get? If the first trigger has only just occurred, then there may only be just enough samples to see what is already on the screen, so while you can zoom, you can't pan outside the screen window. Only if there have been enough triggers to completely fill the sample memory will you get the full depth. This could be very confusing to the user. It seems better to maintain the one trigger = one memory buffer rule for consistency.

If you can search the memory for trigger events after stopping the acquisition, it's difficult to see what practical advantages this approach brings - beyond increasing the data sheet waveforms/sec number, for bragging rights, of course! 

[Keysight DanielBogdanoff]

Isn't there a problem with this approach that when you hit Stop, you won't know what memory depth you will get?

Ok, so the ASIC on the InfiniiVision X-Series uses what we affectionately call "ping pong" memory. With 4M of total space, it uses alternating halves (2M) during normal Run operation. When you hit "Stop" the scope immediately stops capturing and displays the latest 2M capture (or whatever memory applies to the screen time/div settings). At fast time/div settings there's still 2M of data, but the screen may show less info than that. The scope only processes the data that's on the screen, but does this even when it's not actively acquiring data. When you're stopped and then zoom out, if there's more data there it will re-plot the whole 2M signal capture.

If you hit "Single" you get the full 4M of memory, but it will take a new capture instead of keeping the current 2M one.

This might be too basic for you guys, but I did a brief video on "Run/Stop vs Single" that discusses the memory buffer on our YouTube channel:

[kcbrown]

Isn't there a problem with this approach that when you hit Stop, you won't know what memory depth you will get? If the first trigger has only just occurred, then there may only be just enough samples to see what is already on the screen, so while you can zoom, you can't pan outside the screen window. Only if there have been enough triggers to completely fill the sample memory will you get the full depth. This could be very confusing to the user. It seems better to maintain the one trigger = one memory buffer rule for consistency.

That won't be a problem with the circular buffer approach.  That's because with the circular buffer, once enough initial time has passed to fill the entire buffer with samples, the buffer is always full, which means you can always scroll backwards, at the very least, to see what has happened before, for roughly however much time it took to fill the buffer.  The real question is how much additional sampling the scope should do if it is set up to stop on a trigger and it encounters the triggering condition.  For that, it seems logical that it will depend on the amount of memory that has already been filled and the amount of time it would take to fill the remainder.  If the amount of memory that has been filled is small, then there's little point in actually stopping until the remainder is filled, as long as doing so doesn't take too long (that cutoff time is something that could be defined by the user).

If you can search the memory for trigger events after stopping the acquisition, it's difficult to see what practical advantages this approach brings - beyond increasing the data sheet waveforms/sec number, for bragging rights, of course! 

The main advantage is that you're guaranteed to have the maximum possible amount of history at your disposal, while also being able to run the triggering system at its maximum speed.  It makes the speed of the triggering system independent of the memory depth, too.  And it makes it possible to use higher sampling rates with longer timebases.

The approach is so blindingly obvious to me that I must be missing something crucial here, as I would have expected scope manufacturers to already be implementing it if it didn't have some sort of showstopper property that I'm missing.  But it does sound like Keysight does something similar.  Their solution does take care of the case where trigger events are rare, and could be integrated into the single circular buffer approach, by splitting the memory into two pieces in the event you don't see another trigger event after half the buffer fill time since the previous trigger event.

[nctnico]

Because you don't need so much samples to shows on the screen (1024 samples is enough) this memory can be included in the ASIC and can operate independantly of the main memory. Surely circular buffering can be used to fullfill prefill requirements.

@Daniel: the memory is only a quarter (or less) of 4Mpts with 4 channels enabled and in normal run mode on the MSOX4000 series. See pages 220 and 324 of the user manual.