Contact an Expert

Thread: Why are Agilent scopes not able to measure in the sample memory?


Permlink Replies: 19 - Pages: 2 [ 1 2 | Next ] - Last Post: Apr 20, 2014 8:09 PM Last Post By: John Miles Threads: [ Previous | Next ]
0xdeadbeef

Posts: 8
Registered: 06/01/12
Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 1, 2012 12:28 PM
Click to report abuse...   Click to reply to this thread Reply
Don't get me wrong, I just ordered a DSOX3024 for our department (honestly as there was not enough money for a better option) and it is indeed a very neat little scope for the price that I like much more than the more expensive MSO7000A someone bought some years ago. Indeed a lot of things that I hated about the 7000A were improved. Yet it still shows a few quirks of the older models:

1) It measures in the display memory. So automatic measurements are usually inaccurate until you zoom in long enough. There is no indication whatsoever for the user to know when a measurement is inaccurate or not. You just have to try.
I really wonder who designed this and why. With a scope that as up to 4GS/s sample rate, you would expect that you can easily measure in the nanosecond range or (slightly) below. Instead automatic measurements are never more accurate than 10ns and depending on the display resolution, you can easily end up in the microsecond range even though the sampling rate is > 1GS/s.

2) There is no gating implemented. Ok, you can zoom in and do the (auto) measurement in the zoom window, but this is not gating as you can't move the left and right side of the gating area independently. Now really, every other bigger brand supports proper gating - why doesn't Agilent?

3) There is no dedicated key for switching between auto and normal trigger mode. Agreed, at least the DSOX3000 let's you define the quick action key to switch between these trigger modes, but why the heck doesn't Agilent finally put a dedicated "Auto/Normal" key next to the Run/Stop and Single buttons?

I really wonder if Agilent scope developers simply don't care about these issues or if there is some kind of ideology behind it. For #1 I could imagine that this is a limitation based on the signal path needed to create the high waveform update rate. But not everybody thinks that you need a million waveforms per second and sacrificing measurement accuracy for pseudo-analog behavior is a pretty bad idea for a lot of applications. Also be aware that e.g. LeCroy salesmen are well aware of these limitations and use them to sell their scopes.
Willco788

Posts: 110
Registered: 05/28/11
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 2, 2012 9:42 PM   in response to: 0xdeadbeef in response to: 0xdeadbeef
Click to report abuse...   Click to reply to this thread Reply
I'm not sure what you are trying to measure but I think your problem has to do with triggering. Auto mode normally requires adjustment of the Holdoff control. Another thing that can affect the measurement is the trigger position. Sometimes I forget to center the trigger by pushing the Horizontal Position knob and the trigger position is off the screen.
0xdeadbeef

Posts: 8
Registered: 06/01/12
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 4, 2012 4:29 PM   in response to: Willco788 in response to: Willco788
Click to report abuse...   Click to reply to this thread Reply
Nope, this has nothing to do with triggering. It's just the simple fact that Agilent scopes measure in the display memory and thus the measured values change when you zoom in and out the (already) sampled data. At least they are so fair to describe this in the manual. So they are aware that this behavior is wrong. They just don't seem to intend to fix it.

I really wonder why someone would come up with this approach in the 1st place. As they put a hint in the manual, they know for sure that this invalidates the measured results to some degree. So there must be a reason for not fixing this and I'm desperately trying to understand why someone would do this on purpose. Besides I find it fascinating that so few people seem to notice this flaw. Everybody I know who used e.g. a DSO7000A, recognized sooner or later that its measurements can't be trusted. And now I see that with the DSOX3000 (despite some other improvement) this is just exactly the same. Then again, in forums like this, people are either not aware of this or even deny it even though this is instantly reproducible and even described in the manual.

This is about as unexplainable to me as the refusal of Agilent developers to finally implement a proper gating mechanism for measurements now that even lower end scopes from nearly all the competitors have it implemented. I mean they already have already implemented 95% of the stuff needed for this. They'd just need to let you move two cursors and measure between the cursors instead of the whole (zoom) window. How hard can this be? They must be able to understand that this approach is superior,so why do they insist that measuring in zoom mode was sufficient?

So why not remove the last few flaws from a pretty good product and make it outstanding?

Edited by: 0xdeadbeef on Jun 4, 2012 4:29 PM
algoss


Posts: 594
Registered: 11/03/06
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 5, 2012 9:56 AM   in response to: 0xdeadbeef in response to: 0xdeadbeef
Click to report abuse...   Click to reply to this thread Reply
This "feature" is a definite design decision made by the Agilent SW/HW designers. It is this way on every current Agilent scope. They heard many complaints from customers about the "gating" use model and decided to use the display as the "gate". On the Infiniium scopes, where a Zoom window is available, the Zoom window can be used to gate measurements.

This is related to another design difference between Agilent scopes and those made by others. Most others have no concept of 'Off-Screen' memory. If you want to capture a deep trace on other scopes, you have to zoom way out, which causes the display update rate to go way down. On most Agilent scopes, the displayed data is decoupled from the captured data, allowing very deep captures, with fast update rates. Once the scope is stopped, the user can change the time/division setting and look at different portions of the captured data.

Al
0xdeadbeef

Posts: 8
Registered: 06/01/12
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 5, 2012 11:59 AM   in response to: algoss in response to: algoss
Click to report abuse...   Click to reply to this thread Reply
Well, admittedly, other scopes, especially windows based, slow down quite a lot regarding refresh rate when you increase the record length as they transfer and render the whole buffer. Then again, you need to transfer all the data if you want to do sensible statistic measurements.
So scopes that capture more than they display in non-zoom-mode trade in the possibility of sensible statistics for measurements on the whole buffer for a higher refresh rate.
In an ideal world, this would be a user option to let the user decide which tradeoff he wants to make.

Honestly, I prefer to see the whole capture buffer on screen at the price of a slower refresh rate with deep memory setups. As deep memory is primarily meant for single shot, I don't really care if you get only two or three updates per second in this mode. If you want higher rates to observe some realtime behavior, you can simply reduce the memory size without changing anything else.
Well, not on an Agilent of course, but other scopes let you set either the memory depth or the sample rate and offer the auto mode only as third option.

And regarding the zoom mode as gating window: this is not nearly the same. The zoom workaround works just for simple cases, but the idea of the real gating cursors is that you can set them independently and in a good implementation also in deep zoom modes. When using the zoom window as gating window you'll easily get in situations where the resolution of the zoom is not fine enough to eliminate that one edge from your measurement. Besides, setting the two gating cursors around the area of interest is just the natural way to do it. Shifting and Zooming until you find a zoom window that contains only the area of interest is a PITA in comparison.

Now really, e.g. the DSOX3000 line is such a slick design in several points that it kinda hurts that some designers at Agilent cripple the scope with some real dumb implementations of basic functionality. It's even more disturbing that this whole concept is also present in the very highend stuff.
Willco788

Posts: 110
Registered: 05/28/11
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 7, 2012 4:37 PM   in response to: 0xdeadbeef in response to: 0xdeadbeef
Click to report abuse...   Click to reply to this thread Reply
1) It measures in the display memory. So automatic measurements are usually inaccurate until you zoom in long enough. There is no indication whatsoever for the user to know when a measurement is inaccurate or not. You just have to try.
I really wonder who designed this and why. With a scope that as up to 4GS/s sample rate, you would expect that you can easily measure in the nanosecond range or (slightly) below. Instead automatic measurements are never more accurate than 10ns and depending on the display resolution, you can easily end up in the microsecond range even though the sampling rate is > 1GS/s.

I'm confused! I don't understand why " automatic measurements are usually inaccurate". What is an automatic measurement? When I capture a "one shot" triggered event, reguardless of the time base or magnification setting, the measure appears to be accurate on my scope. Is there something I'm missing? Maybe if you post a picture or two, I can understand your problem.
algoss


Posts: 594
Registered: 11/03/06
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 8, 2012 7:38 AM   in response to: Willco788 in response to: Willco788
Click to report abuse...   Click to reply to this thread Reply
Thanks, Willco788. I guess I should have asked the same question early on.

"Zoomed in" and "Zoomed Out" measurements may be different, but that doesn't mean that either of them is wrong.

Please post examples, preferably something using WaveGen, so that they can be repeated by others.

Al
Guest
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 9, 2012 8:37 AM   in response to: 0xdeadbeef in response to: 0xdeadbeef
Click to report abuse...   Click to reply to this thread Reply
Yes, please post some examples.

I am now referring to the X-3024A scope.
It would be nice if the status info on the right of the screen indicated which window is the source of a particular measurement, i.e., Main or Zoom.
Unless I'm missing something, the user has to enter the measurements menu, settings option, to see this. Or you have to infer the source based on the values that you see.

For example, in Auto mode measure AC RMS Cycle and Vpp. As you narrow the Zoom window it becomes clear that the scope is measuring Vpp in the Zoom window and AC RMS in the Main window. (where the cursors are helps also) In auto mode the scope seems to make a good choice based upon the measurement type and the amount of waveform captured in the Zoom window. But it would be nice to know the measurement source at a glance from the screen status on the right.

This is a minor point, in my opinion, but it would enhance usability.

Ed
0xdeadbeef

Posts: 8
Registered: 06/01/12
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 9, 2012 3:39 PM   in response to: Guest in response to: Guest
Click to report abuse...   Click to reply to this thread Reply
Simple test: connect a frequeny generator with a square wave at 9999Hz. Use a samplingrate of ~1GS/s.
Measure the period. At higher zoom levels, the correct value of 100.01µs is displayed. At lower zoom levels, 100.00µs is displayed with the unchanged sample buffer.
I forgot the at which time resolution this happens exactly, but the behavior is obvious and easy to reproduce. This also happens with e.g. an MSO7000A and the manual of the DSOX3000 mentions this behavior and explains that this is due to the measurement in the display buffer (instead of the sample buffer).
Willco788

Posts: 110
Registered: 05/28/11
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 11, 2012 8:53 AM   in response to: 0xdeadbeef in response to: 0xdeadbeef
Click to report abuse...   Click to reply to this thread Reply
0xdeadbeef wrote:
Simple test: connect a frequeny generator with a square wave at 9999Hz. Use a samplingrate of ~1GS/s.
Measure the period. At higher zoom levels, the correct value of 100.01µs is displayed. At lower zoom levels, 100.00µs is displayed with the unchanged sample buffer.
I forgot the at which time resolution this happens exactly, but the behavior is obvious and easy to reproduce. This also happens with e.g. an MSO7000A and the manual of the DSOX3000 mentions this behavior and explains that this is due to the measurement in the display buffer (instead of the sample buffer).

In other words, When you use MegaZoom, the accuracy is .1% off. What's wrong with that? Why is that even an issue? As we always say in engineering "Close enough for the Government!".
algoss


Posts: 594
Registered: 11/03/06
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 11, 2012 9:27 AM   in response to: Willco788 in response to: Willco788
Click to report abuse...   Click to reply to this thread Reply
OK, I just did some measurements, and the answer is "Zoom Mode".

If you use Zoom Mode, and you zoom in so a small number of cycles (maybe 10) are in the Zoom window, then the resolution of 'time' measurements, such as Period, is limited by the sample rate. Sampling at 100MHz or above, the resolution is 10 nsec, or .01 usec. Sampling at 40MHz it's 25nsec, at 10MzH it's 100nsec (0.1 usec).

Because this is different than how other manufacturers may choose to do it, some people might not like it. There are many other things to like about Agilent scopes (yes, this is pure marketing hype). I don't, personally, see this as a big reason to switch back to another scope. All of the functionality and accuracy are (is?) there, just in a different implementation.

Al
0xdeadbeef

Posts: 8
Registered: 06/01/12
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 11, 2012 10:08 AM   in response to: algoss in response to: algoss
Click to report abuse...   Click to reply to this thread Reply
What bothers me is that Agilent scopes vary their measurement accuracy with the zoom factor.
I tried it again today with 400MS/s and 9999Hz. Although the input signal is accurate to the picosecond range and the sample rate suggests a measurement accuracy of 2.5ns, the measurements vary between 100.00µs (no zoom), 100.1µs, 100.02µs and finally 100.01µs (one full period) depending on which zoom level you choose to watch the very same sample data. If there was a lot of jitter on the input data (which however isn't), you could argue that the number of averaging cycles differs, but even then the measurement with the most cycles should have the best result. It is however the other way around: the more you zoom out, the worse gets the measurement result on the same data.
algoss


Posts: 594
Registered: 11/03/06
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 11, 2012 11:02 AM   in response to: 0xdeadbeef in response to: 0xdeadbeef
Click to report abuse...   Click to reply to this thread Reply
One quick point... What we're talking about is resolution, not accuracy. Treatises have been written about the differences between those two concepts. The accuracy of a single-shot measurement will be related to the resolution, but other factors affect the accuracy as well, and maybe to a larger extent.

OK, so we're back close to the original question.

Oscilloscopes are, or maybe originally were, intended for display of analog signals. Making measurements was secondary. Before digital scopes, resolution was limited by the viewer, and generally the resolution of a measurement was about 3% of full scale (vertical or horizontal), since that is what the average viewer can discern. If you wanted to make a high resolution measurement of something like period, you used a different tool, maybe a high-speed counter. You used the scope to make sure the signal was there, but didn't rely on it for the measurement.

With the advent of digital scopes, and all of the measurements that have been added, many people have an expectation that they will be able to get reliable, repeatable measurements of many different types with just the oscilloscope. That is generally true, but...

For many different scopes, the resolution of a measurement is/was limited by the sample rate, and/or the interpolation. For performance reasons, the designers of the DSOX3000 series of scopes made the decision to make all measurements in display memory. If you look in the datasheet, the equation for time accuracy of measurements includes a term based on screen width (0.16%). In normal mode this would be the main trace, and in Zoom mode, it would be the width of the Zoom window (in time). Thus, as you zoom out, you lose resolution, whether you are in normal mode or Zoom mode.

So, if you want the best resolution for this measurement, you should use a separate piece of equipment, such as a counter. For other measurements, other pieces of equipment might be better. If you are using a scope, then you need to understand the limitations, and set it up with the correct settings to make the best possible measurement.

No matter what measurement you want to make or what tool you use to make it, you need to understand the possible sources of error, work to minimize them and adjust your expectations accordingly.

Al
0xdeadbeef

Posts: 8
Registered: 06/01/12
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 11, 2012 12:07 PM   in response to: algoss in response to: algoss
Click to report abuse...   Click to reply to this thread Reply
Indeed it was always my impression that Agilent is targeted too much backwards in the sense of trying to recreate analog features as good as possible instead of designing the best possible DSO.
And about that counter idea: not every signal is periodic. Again this is an analog idea in a way. DSOs were developed to overcome the analog limitations and not to simulate them. Perfect measurement possibilities in single shot mode should still be the main feature of every DSO before trying to simulate analog behavior. But for this you need measurement in the sample memory and a sensible gating implementation. Agilent refuses to implement either - and not only on the DSOX3000, but on all DSOs I ever tested. And this something that I just don't understand.
Willco788

Posts: 110
Registered: 05/28/11
Re: Why are Agilent scopes not able to measure in the sample memory?
Posted: Jun 11, 2012 1:01 PM   in response to: algoss in response to: algoss
Click to report abuse...   Click to reply to this thread Reply
algoss wrote:

Because this is different than how other manufacturers may choose to do it, some people might not like it. There are many other things to like about Agilent scopes (yes, this is pure marketing hype). I don't, personally, see this as a big reason to switch back to another scope. All of the functionality and accuracy are (is?) there, just in a different implementation.

Al


I'm reading complaints about the implementation and the fact that the measurements are off by a few microseconds when you zoom in but is that a real problem when your trouble shooting the setup and hold time on chip? Give some real world situations where the instrument is so far off that it's impossible to make an accurate measurement. If your dealing with bus speeds in the GHz then buy a better scope!

Point your RSS reader here for a feed of the latest messages in all forums