Rigol DS4000 next gen

[nctnico]

What happens if you zoom in? It makes sense the scope only shows/uses relevant digits because the user can't see beyond the pixels on the screen.

[Gunb]

I think that most Agilent scopes use the display memory for measurements. http://www.home.agilent.com/owc_discussions/thread.jspa?threadID=33426&tstart=120
Not a big problem IMHO.

Yes, most do.

The opener of the thread you've linked is the one who asked me about the HMO and the Rigol a few month ago exactly about this issue. He told me that Agilent's DSO3000X would measure in sample memory instead of using the display buffer. He was complaining because he had a pulse of about 1s and a second one with 1.001s and that most scopes cannot measure the difference with cursors without changing the timebase first and adjusting the X1 and X2 exactly to the edges. He wished to keep the complete pulse on screen without changing the timebase and set the cursors to the edges to see the 1.001s directly. If measuring in sample memory that would be possible, at the moment it's not because a timebase of 200ms/DIV enables to measure in the range of 2 fractions only, the 1ms would be the 3rd fraction and that is not displayed.

I did not read the complete thread you've linked but it seems that he was also wrong concerning the DSO3000X, hmmm.

Concludingly it shows, that my assumption above seems to be not right, so no advantage for Agilent at this point, and no disadvantage for Rigol.

Not a big problem IMHO.

No, it's not a big problem, because you can use the trigger for that purpose. Relating to the pulse example above of 1.001s width I trigger to the first edge where the trigger is set to 0s. Then I shift it to the left until the trigger marker is set to the 2nd edge. The trigger value shows very exactly the width of the pulse.
I mean, 0xdeadbeef is not wrong, it would be comfortable to measure the width with high accuracy with the cursors directly. But if most scopes use the display buffer it can't be done without either adjusting the cursors exactly to the edges subsequently or using the trigger as I've described above.
But I agree, not really a big problem 

And there are counters which can do the job also....

[Gunb]

What happens if you zoom in? It makes sense the scope only shows/uses relevant digits because the user can't see beyond the pixels on the screen.

Yes, exactly, that's what it does.

[Gunb]

Just FYI:

Last year, when I had the discussion with 0xdeadbeef about this issue I noticed that the Rigol DS4012 had a bug concerning this. Let's take the example of the 1.001s pulse:

1. timebase at 200ms/DIV -> 1 pulse on screen
2. zoom in to the 1st edge and adjust cursor X1 
3. zoom out and shift left until the 2nd edge is centered
4. zoom in again to the 2nd edge and adjust cursor X2 very accurate to it.
5. deltaX shows the difference very accurate

The bug of the Rigol DS4000 was that it lost the position of the first adjusted cursor when adjusting the 2nd one. Reported that issue to the German support, they came back and confirmed. 2 months ago I've received a mail with an update - now this issue is fixed and the Rigol behaves as the Agilent!

I'm very happy to see support working!

[Gunb]

A real argument that speaks for a few recent Agilents is measuring in the sample memory instead of the display memory.

Which Agilents are you referring to?

I was thinking of the DSO6034A which I've besides other scopes in my office. When I was writing the comment yesterday I was at home, checking this just a moment ago reveals that there are many digits of the fractional part but only the first two are used. Uaaah, so no advantage for Agilent anymore regarding this, seems to use the display buffer for cursor measurement, too.

Just a moment ago I've compared both scopes with the same pulse of 1s and I've noticed what lead me to the statement yesterday - maybe also interesting to know:

-> there is a difference between the older 54622D and the newer DSO6034A which seems to be the reason for 0xdeadbeed's argumentation:
when adjusting the cursors to the edges in the way as described in my last comment the Agilent 54622D reduces the fractional part to 2 digits when zooming out, i.e. to 200ms/DIV. The DSO6034A keeps the accurate cursor value that was adjusted when zooming into the edges.

For the pulse example it means that the 1.001s is still displayed on screen of the DSO6034A even if timebase is set to 200ms/DIV, the 54622D shows only 1.00ms because it looses the 3rd fraction at a timebase of 200ms/DIV. Checking this procedure with 1µs difference behaves the same, the DSO6034A keeps the value on screen.

Not the same as measuring in sample memory, but better than cutting the digits off.

Kind rgds
Gunb

[AndyC_772]

Its not difficult to find a scope with similar specs but the aim is to find something significantly better. In my opinion a significant improvement would be a display which has way more pixels in height (the height is a limiting factor when displaying multiple traces with some amount of detail). A faster user interface and more bandwidth would also be nice.

I think you'll struggle - and that's without taking too much notice of the cost. Plugging in an external monitor to get a bigger, brighter, clearer display was about the best £20 I've ever spent on lab kit.

I've just upgraded my TDS754D to an Agilent MSO-X3054A, more for the sake of ongoing, long-term reliability than anything else. I can't afford to be without a scope, and the Agilent came up at a very good price from the agilentused Ebay store.

The Agilent does have some 'nice to have' new features compared to the Tek. For example:

- the ability to display traces in colour at the same time as the intensity graded, high speed display
- saving PNG files to USB instead of using RLE and floppy discs
- logic analyser and waveform generator
- serial and parallel bus decoding
- a very welcome increase in available lab bench space
- quieter operation
- doesn't make the lab uncomfortably hot when it's been on for a while

If that comes across as a pretty disappointing sounding selection for 15 years' worth of scope development, then I'd entirely agree - though I think that's a testament to just how good the TDS7xx range was rather than any failing in the Agilent.

It's worth mentioning how much more usable the long memory in the Agilent is than the Tek. I gave up trying to fight with the Tek's awkward, clumsy UI in respect of setting the memory depth, zooming in and out, and using the segmented memory. In the Agilent, the memory depth is always selected optimally, the sample rate is kept as high as possible given the time base, and the horizontal scale & position controls always work exactly as I expect them to. The search & navigate features are really fast and intuitive too, even though there are (inexplicably) many more event types that can be used for triggering than can be searched for. I'd have expected the opposite, if anything.

The Agilent's UI is more responsive overall, as you'd expect, though in some ways it's actually less quick to use. Maybe the difference is just familiarity, but for me, pressing a couple of buttons in sequence is quicker than pressing a button and then turning a knob to select from a sub-menu. I think it'll take a while to get used to the trigger controls too; the relatively small number of soft keys means the Agilent has one (hard) button to select the trigger source, type and slope, and another to select mode (auto/manual), coupling, noise rejection and so on.

I do miss equivalent time sampling, which the Agilent doesn't have. I've found a few really surprising firmware bugs in it too, though nothing I can reproduce reliably enough to describe and report.

I couldn't say I'm actually disappointed with the 3000X, because it's clearly a very capable instrument and will no doubt serve me very well - but if your TDS744A is working well then I'd stick with it. Unless I can persuade you to upgrade to a recently retired TDS754D?   

[jpb]

I do miss equivalent time sampling, which the Agilent doesn't have.

I suppose this is one drawback of having the system designed for max waveforms/sec. To do ETS they would need to bypass the normal samples-to-screen route and also I guess that the digital trigger is designed for speed rather than high time resolution.

But visually presumably in dots mode the points are filled in by the multiple curves being shown at once - you just can't save the data as a single curve?

[Gunb]

Even the Rigol DS4000 and DS6000 series have an 800x480 display. In the past decade the major improvement of oscilloscopes has been the overlap with a logic analyser. For that you need as much pixels in height as possible (*). Why the displays didn't scale up is beyond me.

*My logic analyser has an 800x600 display which is reasonable.

At this point I like the Hameg HMO which offers at least a compromise. You have 20 virtual DIVs vertically, which can be switched on/off. If set to ON the knob besides the display enables the user to scroll up/down and place analog as well as digital channels with sufficient space.

[marmad]

....All in all it packs a neat bag of tricks despite its getting close to 20 years old. DSOs where already a mature product back then.

The specs of your DSO are definitely nice - although I think you're bumping up your waveform update rate by about 120%   "Tektronix TDS744A Performance Characteristics: Display Update Rate   180000 wv/sec". But yes, there's no chance you'll get something significantly better (or even as good) for what you paid through eBay - I was just suggesting that the $15k figure was much too high for something similar today.

In the past decade the major improvement of oscilloscopes has been the overlap with a logic analyser. For that you need as much pixels in height as possible (*). Why the displays didn't scale up is beyond me.

Well, there is always a battle between dead time and pushing pixels to the display so it's understandable to some degree.

The amount of memory in a scope is of limited use. At some point the user interface gets too clumsy too handle it. OTOH it is handy for decoding buses.

I would disagree with this - more memory is always a good thing - if the firmware can take advantage of it. For example, looking through the User Manual of the TDS744A, I don't see any mention of any analysis, etc. which can be used on captured segments; in fact, FastFrame seems quite crude compared to segmenting on my Rigol DS2000. So perhaps while many features have been around for 20 years, the implementation of some of them has matured.

[marmad]

-> there is a difference between the older 54622D and the newer DSO6034A which seems to be the reason for 0xdeadbeed's argumentation:
when adjusting the cursors to the edges in the way as described in my last comment the Agilent 54622D reduces the fractional part to 2 digits when zooming out, i.e. to 200ms/DIV. The DSO6034A keeps the accurate cursor value that was adjusted when zooming into the edges.

For the pulse example it means that the 1.001s is still displayed on screen of the DSO6034A even if timebase is set to 200ms/DIV, the 54622D shows only 1.00ms because it looses the 3rd fraction at a timebase of 200ms/DIV. Checking this procedure with 1µs difference behaves the same, the DSO6034A keeps the value on screen.

I've never really understood the basis for these complaints.

Firstly, DSOs are not the best measurement devices - and they've never been designed to be. There are always better instruments available for more accurate measurements if needed.

Secondly, in terms of the number of digits displayed, DSO screen real estate is at a premium. If you need more precision just read the floating point number from the DSO and display it on another screen; IMO, it takes less time to do that than it does to come on forums and complain about the issue.   As an example, the image shows my software reading Vtop from the DSO and displaying it on a PC screen (with 5 least significant digits ) - while the DSO displays 2.50V on it's screen (limited to 2 least significant digits).

[Gunb]

I've never really understood the basis for these complaints.

Me neither.

Firstly, DSOs are not the best measurement devices - and they've never been designed to be. There are always better instruments available for more accurate measurements if needed.

Yes, and that was also my argumentation - could not convince him, he started this discussion at German mikrocontroller.net, moved it to this blog and - as Hydrawerk revealed this morning - even Agilent community suffered from it. The best was that he asked me about the details at mikrocontroller.net and sold these information here as his own - funny guy 

If I want to measure timings with accuracy of non-periodic signals I'll use a counter with an oven which can be triggered instead of a scope. If a scope should be used anyway it's not a big deal to follow the steps I've described above using the trigger.


 

[AndyC_772]

I think you're bumping up your waveform update rate by about 120%   "Tektronix TDS744A Performance Characteristics: Display Update Rate   180000 wv/sec"

That's about what I get on my TDS754D. I find it mildly interesting that this parameter didn't change significantly through several generations of TDS7xx scope.

FastFrame seems quite crude compared to segmenting on my Rigol DS2000. So perhaps while many features have been around for 20 years, the implementation of some of them has matured.

I think you're comparing two opposite extremes here. The early Tek implementation is about as clunky as I've seen their otherwise excellent UI ever get, to the extent that I completely gave up on it as a diagnostic tool. On the other hand, the waveform record/playback feature on the DS4000 (I presume the 2000 is the same) is a fine example of how to make segmented memory easy to use, and ironically, it's something which the mainstream Western manufacturers would do well to copy from the Chinese.

Who'd have thought it? 

[nctnico]

I think you're bumping up your waveform update rate by about 120%   "Tektronix TDS744A Performance Characteristics: Display Update Rate   180000 wv/sec"

That's about what I get on my TDS754D. I find it mildly interesting that this parameter didn't change significantly through several generations of TDS7xx scope.

That is because the TDS500, 600, 700 and 800 series are basically the same design with some minor updates over the years. Each series has different acquisition hardware though. About the wfms/s: the 400k figure is what I read from the InstaVu specs (an acquisition mode with almost zero dead-time due to back-to-back triggering).

[marmad]

About the wfms/s: the 400k figure is what I read from the InstaVu specs (an acquisition mode with almost zero dead-time due to back-to-back triggering).

I understand. There seems to be conflicting information floating around the net - with some people listing the TDS 744A spec as 180k - and others as 400k. But looking more closely, it seems the 400k rate is the best rate for the fastest DSO in the series - so it might apply to the TDS 784A.  Strangely, the Tektronix document: "TDS 684A, TDS 744A, & TDS 784A Digitizing Oscilloscope Performance Verification and Specifications" doesn't mention the waveform update rate a single time.

Anyway, my rule of thumb - when two conflicting numbers exist for the same specification - is to assume the worst one is the correct one 

[function]

I contacted Rigol North America and they anticipate September or later in the year for the MSO ds4000, no word on the ds2000 though.

[andyturk]

It's not more expensive than other scopes with the included features, rather cheaper. I've bought it 1.5 years ago for hobby purposes.

Slightly off-topic...

I just unwrapped a brand new DS4014 about an hour ago. What's the most recent firmware for the beast? It came with firmware 00.01.00 and hardware 1.2

[marmad]

I just unwrapped a brand new DS4014 about an hour ago. What's the most recent firmware for the beast? It came with firmware 00.01.00 and hardware 1.2

I don't know what the latest FW version is for the DS4000, but I'm guessing that what you posted is not the full number. Perhaps the same technique that's used on the DS2000 to see the actual version number works on the DS4000 as well (go to the bottom of the post).

[andyturk]

Right you are, marmad!

I've got:

Software version: 00.01.00.00.07
Hardware version: 0.1.2.3
FPGA version:
SPU: 03.00.06
WPU: 00.07.04
CCU: 01.40.05
MCU: 1.3

[marmad]

Right you are, marmad!

I've got:

Software version: 00.01.00.00.07
Hardware version: 0.1.2.3
FPGA version:
SPU: 03.00.06
WPU: 00.07.04
CCU: 01.40.05
MCU: 1.3

Great, Andy! Now you just need to find another DS4000 owner to see if it's the latest.

[marmad]

@andyturk: Any chance you can test the Anti-Aliasing (under Acquire menu) function on your DS4014? I've been testing it on my DS2000, and I can't get it to stop ANY aliasing. Wondering if it works on the 4000s.

[andyturk]

@andyturk: Any chance you can test the Anti-Aliasing (under Acquire menu) function on your DS4014? I've been testing it on my DS2000, and I can't get it to stop ANY aliasing. Wondering if it works on the 4000s.

Here's what I found:

With a 100kHz sine at 50ms/div, no aliasing is visible (regardless of whether A-A is turned on or not). The same is true with 10MHz and 10ms/div. What I did notice is that when A-A is turned on, the "phosphor" of the trace gets brighter. Signals came from a DG1022 and memory depth was auto at all times.

The only thing I could find that even looks like aliasing is a 10MHz signal viewed at 1s (one second) per div, with A-A turned off:


Same signal with A-A turned on:


10MHz, 500mS/div with A-A turned on crashes the scope!

[marmad]

With a 100kHz sine at 50ms/div, no aliasing is visible (regardless of whether A-A is turned on or not). The same is true with 10MHz and 10ms/div. What I did notice is that when A-A is turned on, the "phosphor" of the trace gets brighter. Signals came from a DG1022 and memory depth was auto at all times.

Sorry to trouble you again Andy    but it looks to me like you've got 70MPts as the sample length (due to AUTO setting) - that means you've got a 5M sample rate and wouldn't get any aliasing on a 100kHz wave (since it has to do with sample rate < input frequency). You need to set your sample length to the lowest possible setting (7 or 14k) - or raise your signal frequency much higher (20MHz or so) in order to get an accurate idea if it's working.

[andyturk]

Duh. I knew that (*cough*).

10MHz signal, 100us/div, 14K pts memory depth (as low as it goes), and it aliases like a mofo. The A-A button seems to have no effect on what's displayed. Whenever the sample rate is lower than the signal frequency, it gets confused.

I'll just leave A-A off, since best case, it doesn't work. Worst case, it crashes the scope's UI.

[marmad]

10MHz signal, 100us/div, 14K pts memory depth (as low as it goes), and it aliases like a mofo. The A-A button seems to have no effect on what's displayed. Whenever the sample rate is lower than the signal frequency, it gets confused.

Yes, that seems to track with reports from DS2000 users. I've passed the bug (or unimplemented feature?) along to Rigol (via my dealer) and we'll see what they say.

[Yaksaredabomb]

There's probably next to no chance this is right, but might it not have to do with the picture itself - not the waveform displayed?  I mean - how do you make something out of nothing?  If you don't have the sample rate necessary then you simply don't have the samples and you *will* get aliasing.  Maybe there's some math-magic you can do so I guess I wouldn't rule it out, but still.


I seem to recall the idea or speculation that AA is used on the picture - ie, like in gaming - to make diagonal lines appear smoother.  It's not just for gaming anymore - I believe Microsoft Office 2010 (and maybe 2007?) charts use anti-aliasing, so some people's approval and others' disgust.  This could be part of what makes the Rigol displays in the newer/higher-end models look so good.