REVIEW - Rigol DS2072 - First Impressions of the DS2000 series from Rigol

[marmad]

Was wondering if anyone could confirm this bug aswell:

Like most DSOs, ALL measurements from the left menu are software measurements on display memory (NOT sample memory) - they can only be considered accurate if you can see the wanted measurement parameter clearly defined on the displayed waveform.

OTOH, the counter from the Measure menu is a hardware frequency counter - and will be accurate, regardless of the displayed portion of the waveform.

[WVL_KsZeN]

ofcourse, but there's more than 100 oscillations on the display.. no need for >1% error.

And if it's only depending on the display, how does changing the mem depth make such a big difference? the displayed signal is exactly the same.

@ auto mem depth


@ 56mpts mem depth

[Bugware]

Sure that is a scaling problem. But I think all data in the display memory must be scaled?! So maybe there is no change in the update rate, so my thought...

There is the normal scaling done between sample memory and display memory. The scaling for statistics is clearly done after this (since it's doubling lines), as part of the overlay (adding measurements, screen icons, etc).

I see. Sorry I have misunderstood the bug #17 in the list. So then this is only a confirmation of that.   BTW. the offset is not only with AC Coupling but also for LF and HF Reject Coupling...   

Yes, I knew, but the bug clearly affects filtered triggers - so fixing it for one will likely fix it for all - but I added *filtered* to the bug description to make it clearer.

Thank you for your explanations. I have now understood the backgrounds. Again learned something! This helps newbies like me very much. 

[WVL_KsZeN]

yes.. other measurement also seem to be affected. I think it's weird that the mem depth is having an influence, how could that be? As you can see by my screenshots, they're almost 99.9999% identical, so the scope is not (only) measuring 'on the screen'.

This is weird? a post by Marmad seems to have disappeared?

[marmad]

ofcourse, but there's more than 100 oscillations on the display.. no need for >1% error.

And if it's only depending on the display, how does changing the mem depth make such a big difference? the displayed signal is exactly the same.

The memory depth changes the amount that the DSO decimates to the display memory - either 13.8MB decimated to 700 bytes - or 56MB decimated to 700 bytes - in your images. You can imagine there will be changes in the frequency decimated to because of rounding errors.

Personally, I would NEVER use frequency from the measure menu (because of accuracy) - just the counter.

But it seems we have this discussion about the flaky measurements on DSO every few months. The truth of the matter is this:

Unless you have an expensive DSO which measures in sample memory, DSOs suck at measurements. All oscilloscopes, traditionally, have been crude measurement devices (on old analog scopes we counted graticule lines): they are, first and foremost,  VISUALIZATION devices. Now, DSO manufacturers have figured out a cheap, simple, and fast way to tack on measurements by using display memory - but honestly, if measurement accuracy is important - use an instrument designed for accurate measurement - or - capture/save the sample memory and measure using external software.

[marmad]

Just found another brand new feature added to the latest firmware (which, honestly, should have been there from the start):

CURSORS in X-Y Mode:

[marmad]

Yesterday, I reported that I had (finally) found a situation in which the Anti-Aliasing ACTUALLY WORKED - but I deleted that post once I discovered the DOTS mode bug (which, as mentioned below, is tied to it).

But I re-investigated the issue just now... and it's funny. I guess switching on Anti-Aliasing is just causing the DOTS mode to display the waveform as it SHOULD - given the memory size and sample rate - but it simulates a "correct" anti-aliasing technique. 

Look at the following series of images made in DOTS mode:
1) MemDepth: 1.4M (no DOTS bug)/ Anti-Aliasing: OFF = obvious aliasing.
2) MemDepth: 1.4M (no DOTS bug)/ Anti-Aliasing: ON = absolutely no change to the aliasing.
3) MemDepth: 56M (DOTS bug)/ Anti-Aliasing: OFF = obvious aliasing.
4) MemDepth: 56M (DOTS bug)/ Anti-Aliasing: ON = aliasing is fixed!
5) MemDepth: AUTO (DOTS bug)/ Anti-Aliasing: OFF = obvious aliasing.
6) MemDepth: AUTO (DOTS bug)/ Anti-Aliasing: ON = aliasing is fixed!

[NikWing]

yeah, not many have cursors in XY mode, I guess that's another reason for people to buy Rigol XD
I hope mine comes tomorrow or friday ...
it's on the way, oh yay


edit: marmad, might you explain the background on the anti-aliasing problem? it would help me understand the problem

[marmad]

edit: marmad, might you explain the background on the anti-aliasing problem? it would help me understand the problem

Search back through this thread, my friend. There is PLENTY of discussion about it 

[Fagear]

Now I'm in!
I went to local distributor in Moscow in bought the DS2072A device.

My device details:




I'm ready to do some tests and anything else except opening the device. So no JTAG dump (and I even had no JTAG programmer).

I did some weird "parallel decoding" with internal calibration signal (attachment 4).

I have adjusted compensation on both probes (they are RP3300A, 10:1 only, 300MHz) and performed an self-calibration with nothing on BNCs.

Some measurement of noise with open BNCs and different settings: "Full (70 MHz) BW, 1 Meg", "20 MHz BW, 1 Meg", "Full (70 MHz) BW, 50 ohm", "20 MHz BW, 50 ohm". Results are in attachments.

[marmad]

Now I'm in!
I went to local distributor in Moscow in bought the DS2072A device.

Welcome! There is newer FW than the version installed on your DSO. You can download it from here.

[NikWing]

alright, I'll look again
though I don't remember reading anything that explains the anti-aliasing itself or how I would recognize this as a "problem" looking at the curves lol
1st thing I would think of is a trigger problem (if the curve isn't steady)

[marmad]

alright, I'll look again
though I don't remember reading anything that explains the anti-aliasing itself or how I would recognize this as a "problem" looking at the curves lol

Here is a link to a previous post, which has the original patent papers for Agilent's (orignally Hewlett Packard's) technique for using stochastic (random) sampling to eliminate aliasing. This followed a long argument discussion  between zibadun and myself, in which he claimed that the technique I had been describing was a gimmick and possibly not even used by Agilent. If you follow the discussion backwards from there, you can see a full explanation of why Rigol's AntiAliasing doesn't work - and what they could do to fix it - assuming they got around Agilent's patent 

[marmad]

alright, I'll look again
though I don't remember reading anything that explains the anti-aliasing itself or how I would recognize this as a "problem" looking at the curves lol

Here is one more link - using images to clearly show the problems of aliases when a 100kHz sine wave is undersampled at 100-200kSa/s - and how the Agilent (but NOT the Rigol) deals with it perfectly.

[NikWing]

thanks! I'll check them out
(though I currently/still wouldn't see it if I would have this problem in front of me lol)

[Fagear]

Some additional photos of unboxing...

It was in two boxes: one inside of another.

BNC-2-crocodile clip on 2nd photo is not from this box, I just bought it separately with my DS2072A.

[Fagear]

And some close-ups...

Going to sleep now, 1:50 (am) in Moscow... Later I'll update FW to the latest one.

[neslekkim]

So the probes that comes with it are 300mhz?, I thought they was supposed to be 350?, and they are fixed to 10x? not switchable?

[Hydrawerk]

I wonder why my DSOX2002A bought in 4/2013 was not packed in a plastic bag. 

[Fagear]

So the probes that comes with it are 300mhz?, I thought they was supposed to be 350?, and they are fixed to 10x? not switchable?

Manual for the probes says "DC~300MHz". They are not switchable, 10:1 only. Compensation trimmer is on box with BNC. And there is plastic srewdriver in the bag.

[neslekkim]

Hm, datasheet for the 2000A series states:

2 Passive Probes (350 MHz)
And the manual, page 1-5 says so also, at least the one I downloaded.. strange..
Mine will arrive late January so I'll see what's packing by then..

[marmad]

Hm, datasheet for the 2000A series states:

2 Passive Probes (350 MHz)
And the manual, page 1-5 says so also, at least the one I downloaded.. strange..
Mine will arrive late January so I'll see what's packing by then..

There have been previous reports in this thread that sometimes the DSO comes with the RP-3300A probes - instead of the RP-3300 probes, which most of us got.

[neslekkim]

hm, strange.. are those useless when upgrading to 300mhz then?

[Fagear]

I have updated FW to latest (00.02.01.00.03) and recalibrated my DS2072A. Everything looks fine. Trial minutes are still intact.

Here are some changes in info:



Note some differences in "Installed options".

[WVL_KsZeN]

I also got two RP3300A with my scope (DS2072 with hw1.0, but only 4 weeks old), while two RP3300 were advertised. Now I really wanted some 1x probes aswell, so I complained a bit (a lot) and finally got 2 extra RP3300 sent home for free. I didnt have to send my RP3300A back. Where I bought my scope, they're asking 45 euro's for a single RP3300 probe! Don't think they made any money from me, I got a 10% discount on my scope (765 euro) and two extra RP3300 (90 euro), so I only paid +- 675 euro for my scope.

Now, since I've got both, I can also compare them! :-)

The RP3300A have a much better feel than the RP3300. Build is better and they're also better isolated. With the RP3300 I get some glitches in the signals when I hold them in my hands the wrong way (finger close to the 1x/10x switch, probably picking up 50Hz in that case). The RP3300A are also a little bit smaller and thinner than the RP3300. The RP3300 have this yellow cap, to 'shield' the ground when probing on a PCB, but it falls off way too easily (for both probes, do others have this aswell?). I don't have a high frequency generator, so can't compare the bandwidth. Strangely, on some pages on the Rigol homepage, the A are rated at 350MHz, while on others they are rated at 300MHz.

If I'd have to chose between the RP3300 and RP3300A, then I'd go for the A! I'm not going to use the RP3300 in 10x mode, the RP3300A feel much better quality and is better isolated, so the RP3300 will be only used for low bandwidth 1x probing.

Btw, I doubt that either the RP3300 or the RP3300A are very good at 300MHz, but maybe someone already tested the bandwidth.