EEVblog #522 - Rigol DS1000Z Oscilloscope Quick Look

[marmad]

I don't need any example to know a zig-zag rendering of min-max values is a wrong method.  But, given my strong reaction in the first place, I think it would be obvious I've come from an environment where it was done differently.

As evident from the published documentation describing the technique, the minimum and maximum values are separated in time. Displaying anything else when magnifying the discrete samples would be wrong - period. And as far as I know, every DSO maker (that implements it) does it the same way.

[sync]

As evident from the published documentation describing the technique, the minimum and maximum values are separated in time. Displaying anything else when magnifying the discrete samples would be wrong - period.

I think it`s just the cheapest way to implement it. An alternative way could be to save minimum and maximum for a sample interval and display a bar.

[evanh]

Here's a screenshot directly from my old scope.  I've intentionally made it super noisy with only 250 sample points to show off what the simplest correct Peak (labelled Envelope there) mode rendering should look like.

EDIT:  Those 250 points are, of course min-max pairs.  In Rigol parlance that would be 500 zig-zag'd sample points, but only from a rendering pov.  Rigol's acquisition hardware is still converting the same interval spacing per pair as my old scope.  Assuming, that is, Rigol uses what the Tek document says.  There is a chance it could be incorporating a time shift with each stored sample, which changes the meaning from straight min-max to something a little more complicated.  Not unlike the idea of interlacing for TV.

[marmad]

I think it`s just the cheapest way to implement it. An alternative way could be to save minimum and maximum for a sample interval and display a bar.

It has nothing to do with being cheap (it's just software) - it has to do with correct implementation of the described acquisition technique.

Here's a screenshot directly from my old scope.  I've intentionally made it super noisy with only 250 sample points to show off what the simplest correct Peak (labelled Envelope there) mode rendering should look like.

That's NOT correct Peak Detect mode - it's Envelope mode, which COMBINES many individual Peak Detect waveform acquisitions into a single displayed waveform.

As mentioned a few times already, Peak Detect mode uses the highest and lowest of all the samples contained in TWO CONSECUTIVE SAMPLE PERIODS. When magnified to the sample level, a triangular shaped waveform is PRECISELY what you should see (if it's implemented correctly).

[evanh]

It has nothing to do with being cheap (it's just software) - it has to do with correct implementation of the described acquisition technique.

Cheap in this context means quick and dirty, not much thought put into it.

That's NOT correct Peak Detect mode - it's Envelope mode, which COMBINES many individual Peak Detect waveform acquisitions into a single displayed waveform.

Com'on, how can you say that?  They're the same thing.

As mentioned a few times already, Peak Detect mode uses the highest and lowest of all the samples contained in TWO CONSECUTIVE SAMPLE PERIODS. When magnified to the sample level, a triangular shaped waveform is PRECISELY what you should see (if it's implemented correctly).

The actual number of sampled acquisitions is an entirely different figure.  In my example image there is 25 samples stored per division, that's 200 usec per interval.  The actual acquisition hardware runs at 200 Ms/s so there is 40,000 acquisitions per min-max pair.  The Rigol, in single channel, is 1 Gs/s so that would be 200,000 acquisitions per min-max pair.  Both scopes store the same meaning in the same amount of space.

The example that Tek gives is not about what happens at the feeble point of one acquisition sample to one stored sample but is just a simplified drawing with minimal detail to carry the idea.

Btw:  My old scope deals with that feeble area gracefully also.  As you wind the timebase in to faster speeds it automatically drops out of peak acquisition mode and goes back to normal mode.  And when winding the timebase back out to slower speeds it will go back to my setting of Envelope.  Which is why it displays the current mode up the top of the display.

[sync]

I just had a wrong measurement with my Rigol 1102E because of that zig-zag rendering
Thanks to this thread I instantly realized what's going on. I inspected a oscillation in a part of a wave form. I stopped acquisition and zoomed into the oscillation. This is what I got. The damned thing measured the frequency of that peak detect zig-zag! FAIL!



Here the real oscillation. Measured without zooming.



That zig-zag rendering sucks. It's wrong! Sorry, it should be bars.

[marmad]

That's NOT correct Peak Detect mode - it's Envelope mode, which COMBINES many individual Peak Detect waveform acquisitions into a single displayed waveform.

Com'on, how can you say that?  They're the same thing.

Peak Detect is a single waveform; Envelope is a combination of numerous waveforms, like Averaging - so no, they're not the same thing. That's why some DSOs allow you to select Peak Detect mode OR Envelope mode..

[Electro Fan]

It would be great to hear from more DS1000Z users, but what's the thinking so far among Rigol (and other) scope users?  Is the DS1000Z a worthy competitor/alternative to the DS2000? 

Give up some performance, screen space, and nav/review functionality and take the 2 extra channels?

Or stick with the legendary DS2000?

Anyone out there happen to have the opportunity to own or play with both the 1000Z and the 2000 series?

Dave - how about an in-depth "not a review" video of the DS1000Z? 

[evanh]

Give up some performance, screen space, and nav/review functionality and take the 2 extra channels?

Or stick with the legendary DS2000?

There's no doubt the DS2000's have the sweet spot for price/performance/quality on two channels, and if there was a four channel version I'm sure everyone would be praising it.  But I presume the extra channels would have a notable dollar figure attached.  It would seem Rigol chose to lower some specs, putting it in the DS1000 bracket, rather than raise the price.

I can't blame them, apart from some firmware bugs to be ironed out, I'm happy with the compromise and truly impressed, even if I've poked some fun and raised questions, with the capabilities of the DS1000Z.  Four channels and deep memory is a prerequisite in my books.  That combo is more important to me than 100+ MHz analogue bandwidth.

You never know, there still might be a DS2000Z out there in the future ...

[creyc]

If Rigol offered a cheaper 2-channel version on the DS1000Z platform (with the large high-res screen) I think that would see a lot of hobbyist interest.  In fact, add signal generation and a "DS1072Z-S" for around $700 would be the scope I'd get!

Although I suppose this might step on the toes of the already affordable 1000E lineup, but I don't see why the larger screen has to be associated with 4-channels??  I think their product lineup needs a bit of a restructuring to be honest.  With 12 models under $1000 and some strange gaps and feature overlap, choosing an affordable Rigol scope isn't an simple task!

[marmad]

If Rigol offered a cheaper 2-channel version on the DS1000Z platform (with the large high-res screen) I think that would see a lot of hobbyist interest.  In fact, add signal generation and a "DS1072Z-S" for around $700 would be the scope I'd get!

You would pay $700 for a two-channel 'DS1072Z-S', but you wouldn't pay $118 more for the same thing with two extra analog channels? That's some strange logic - those two extra channels are a massive bargain at that added price.

[Electro Fan]

If Rigol offered a cheaper 2-channel version on the DS1000Z platform (with the large high-res screen) I think that would see a lot of hobbyist interest.  In fact, add signal generation and a "DS1072Z-S" for around $700 would be the scope I'd get!

You would pay $700 for a two-channel 'DS1072Z-S', but you wouldn't pay $118 more for the same thing with two extra analog channels? That's some strange logic - those two extra channels are a massive bargain at that added price.

Marmad, thanks for all you do to keep us informed and thinking, and Happy Friday too.

Just checking, setting aside the function generator, would you trade a 2072 even up for a 1104Z?  Or trade a 2072 for a 1104Z plus $100? 

EF

[marmad]

Just checking, setting aside the function generator, would you trade a 2072 even up for a 1104Z?  Or trade a 2072 for a 1104Z plus $100? 

I don't know - I think I'd want to try it first before I made a decision. I'm quite happy with the DS2000 - and even though it would be nice to have an extra 2 analog channels (for me, it's not that often that I would need them - maybe 2% of the time - even less if they made the EXTERNAL trigger function as input to all trigger types), I'd hate giving up the benefits of the DS2000, especially the screen size: I want bigger screens on everything 

[Electro Fan]

Just checking, setting aside the function generator, would you trade a 2072 even up for a 1104Z?  Or trade a 2072 for a 1104Z plus $100? 

I don't know - I think I'd want to try it first before I made a decision. I'm quite happy with the DS2000 - and even though it would be nice to have an extra 2 analog channels (for me, it's not that often that I would need them - maybe 2% of the time - even less if they made the EXTERNAL trigger function as input to all trigger types), I'd hate giving up the benefits of the DS2000, especially the screen size: I want bigger screens on everything 

What's the reason they don't make the External trigger function as an input to all/other trigger types?  Just cost/price? Or something else?

To get a bigger screen do you ever use an external monitor?  Or does that not really address the issue (how the scope displays overlapping menus, etc.)?

[evanh]

I just got my private key a product key today and went to Rigol's webpage for license key generating and found it ain't all that friendly towards Firefox.  What should have been a simple non-scripted HTML form, turned into a scripting mess that involved me having to click in the appropriate entry box before each and every character entered.  In the end, I felt lucky it actually generated the license key.

So, their website is even more buggy that their scopes!

[evanh]

I have used the date-stamping on my old scope.  Particularly when leaving it to auto-store rare trigger events.  I haven't had enough time with the Rigol to say yet, but the lack of such a feature in the DS1000Z might be a bit of a downer ...

Okay, after a phone call, the DS1000Z does save it's files with a date-stamp ticking up from power on.  The files are stored from the FAT32 epoc of 1 Jan 1980.  So, this is usable.  For the most part just knowing the diffs between recordings is enough for problem solving but, knowing roughly when the scope was turned on each time, the absolute data-stamp could also be administered.

[evanh]

Got the scope back for a bit ... Discovered a couple of humdingers with serial decodes:

 - 1)  The serial decoders only display results with a limited set of samples in the trace - from reading the settings it's meant to be 2x the display resolution (Which should allow up to 600 bits, ~60 characters, to be decoded on display.) but in reality, from the visual point of transition, it looks more like 1/4 of display resolution - allowing a max of something like seven 10-bit characters.  I'm not impressed at all!  Poor effort there Rigol.

 - 2)  The serial decoders auto-disabled when the timebase is set low enough to cause scanning display mode.  That's really bad.  I thought it was bad that the rolling display mode has no triggering but this also means all slow timebases exclude many of the scopes features!

And to top it off, along with many other similar behaving settings, the setting also doesn't re-enable when moving back to faster timebase.

The workaround, for this particular one, is to do the needed capture at slower timebase then stop acquisition.  Then you can zoom in to the trace using the timebase dial.  Once past the equivalent of 200ms/div timebase, even though timebase is not relevant when stopped like this, the decoders can be turned on again.


Okay, more rendering problems that can fool the unprepared.  And I'm certainly not sure I've got a complete handle on this yet ...

High Res acquisition mode had me a little perplexed right from the start.  It doesn't look as clean as I expected.  I could also see High Res mode aliasing an awful lot earlier than Normal acquisition mode.  This is a tell-tail indicator of what is wrong.  High Res mode should combine multiple consecutive samples instead of Averaging mode's combining of multiple consecutive sweeps.  The trick is to do this as a rolling window or with decimating feedback on a sample by sample basis, aka digital filters.  The trade-off is High Res inherently lowers the low-pass cutoff.

First up, the rendering at screen resolution seems to be done at something like four sample points per pixel, ie:  Not using all in-range sample points.  This is very cheap, it creates it's own aliasing and throws away a lot of the look of the higher precision.  Even Peak mode has this behaviour, albeit to a lesser extent.   Rendering of Normal mode is same as Peak.  Bad Rigol!

The acquisition of the trace may actually be done correctly but, just like the zig-zag peak-to-peak rendering method, Rigol messed it at the final stage of displaying the results.  See attachment - showing a full length trace as a meandering thin line while the zoomed in view shows it's actually a full height 50 Hz trace.


As a bonus, there's something wrong with Channel 4 selection.  There is no relay clicking sound when toggling Ch4 on/off.  As if it's always on.  Solved, I had channel 4 set for triggering.  Oops.

There is also no effect on the Sin(x)/x setting either, this setting is forced on when a single channel is on but defaults to off with multiples channels sampling.  That is, except for Ch4.  I can have Ch4 on with any other single channel and the Sin(x)/x setting stays forced on.  As if it doesn't exist.  I still don't understand Sin(x)/x but this one is now consistent at least.  It actually seems to need three channels on before allowing a setting.

[evanh]

The mystery deepens.  Attached (Unnecessary_Aliasing2.png) is another example of the same signal source and same timebase but a deeper capture memory.  In this case the apparent aliasing has massively changed.  It's not an ideal image but it's clearly a dense trace being represented.

So, it's not as simple as I first surmised because if the scope rendered by using a fixed number of samples per pixel then this image wouldn't have changed from the last one.

There's a possibility the scope is trying to perform the hi-res function at display time rather than during acquisition.  This would be bad too as that further reduces the low-pass cutoff, which could be what I'm seeing.  Due to the lowish sample rate of the previous trace, any further filtering of it would quickly cut into the 50 Hz oscillation.

Unnecessary_Aliasing3.png is for comparison of same setup to the first image except for acquisition mode changed to Normal.  You can see there the full trace looks sensible, and similar to the second image.

So, it's probably not aliasing at all.  But rather filtering the trace into oblivion.


Evan

[evanh]

With no parameters and no info, High Res mode is near worthless.  However, with a more honest treatment, this mode could be put to good use via a firmware update.  One feature of my old scope is a quickly adjustable digital input filter on each channel ... and, funnily, this has the side-effect of increasing vertical resolution.

How about it Rigol?  Move this broken "High Res" feature out from the acquisition menu and put it in each channel individually with individually adjustably parameters for it's cut-off frequencies ... turning it from a confusion into a real useful feature!


I've had another shot at the UART decoder and discovered the problem is only with 300 baud.  Which just happened to be what I tried first.  Doh!  Attached is example of problem and proof of valid data plus examples of good behaviour at 600 and 1200 baud.

Going by the results of 1200 baud the decoder looks to be able to top out at 30 characters on display at once.


Evan

[marmad]

Move this broken "High Res" feature...

It's not "broken". As mentioned in the other thread, the implementation appears to work exactly the same as Agilent's, Tektronix's, etc - but perhaps not the way YOU would prefer it worked.

[jwm]

I recieved my 1072Z and love it. The main reason I got it was the intensity graded display, I have been keeping an old tek analog scope around just because of the weird aliasing effects of my other DSO (atten 1062CAL). The 1074Z easily replaces both in less space.

Anyone want me to attempt something and post it? Unfortunately, I didn't realize the trial option countdown happens while the scope is turned on, not just when using them, so by leaving it on overnight by accident I killed my ability to test decoding/advanced trigger/etc functionality which is too bad, it was something I was considering getting but now can't test. Sort of lame that the countdown starts right away, I spent at least a day testing out the standard features against my other scopes and signal sources and playing around with 4ch possibilities before attempting the new territory of the trial options at which point they are gone.

[Electro Fan]

I recieved my 1072Z and love it.

Unfortunately, I didn't realize the trial option countdown happens while the scope is turned on, not just when using them, so by leaving it on overnight by accident I killed my ability to test decoding/advanced trigger/etc functionality which is too bad, it was something I was considering getting but now can't test. Sort of lame that the countdown starts right away, I spent at least a day testing out the standard features against my other scopes and signal sources and playing around with 4ch possibilities before attempting the new territory of the trial options at which point they are gone.

Yes indeed, Rigol is just frustrating users instead of helping to sell upgrades; they should let the user decide when to turn the trial options on rather than insisting on the options being installed continuously from the get-go.

[vk6zgo]

I recieved my 1072Z and love it. The main reason I got it was the intensity graded display, I have been keeping an old tek analog scope around just because of the weird aliasing effects of my other DSO (atten 1062CAL). The 1074Z easily replaces both in less space.

Anyone want me to attempt something and post it? Unfortunately, I didn't realize the trial option countdown happens while the scope is turned on, not just when using them, so by leaving it on overnight by accident I killed my ability to test decoding/advanced trigger/etc functionality which is too bad, it was something I was considering getting but now can't test. Sort of lame that the countdown starts right away, I spent at least a day testing out the standard features against my other scopes and signal sources and playing around with 4ch possibilities before attempting the new territory of the trial options at which point they are gone.

Try looking at an analog video waveform at field rate-----this was the killer for many earlier DSOs.
You probably have a DVD player lying around which still has analog outputs.

[evanh]

... I spent at least a day testing out the standard features against my other scopes and signal sources and playing around with 4ch possibilities before attempting the new territory of the trial options at which point they are gone.

https://www.eevblog.com/forum/testgear/sniffing-the-rigol%27s-internal-i2c-bus/msg326217/#msg326217

[sync]

My DS1074Z arrived today. Ordered yesterday! 

I did some wfrm/s measurements. Input was a 20MHz sine wave. One channel active. Measured the trigger out with a frequency counter. For some time base settings the wfrm/s wasn't stable. I used a rough average then.

I also observed that the wfrm/s are depended on the displayed waveform. If the scope has more to draw it slows down. My initial signal was 4div high (peak to peak). I done another run with 1div. It was faster. This behavior makes measuring/ comparing the wfrm/s difficult.