Rigol DS1000Z series buglist continued (latest: 00.04.04.04.03, 2019-05-30)

[Njk]

Greetings!

DS1054Z is a new instrument to me. Owned Tek TDS3054B previously. In comparison, Rigol seems noisy but it also has more memory, more SW candies so it has its pros and conrtras, as usual. The scope is of its latest HW generation (I presume), board version 0.1.4.
I've enabled all the options, including 500 uV/div (which works fine, BTW, taking its limitations into account) and installed the firmware of the day, v.00.04.05.SP2.

Now I'm evaluating the FFT capabilities. It's not essential feature for a scopes but anyway.

In DS1000Z, the FFT function can operate in two modes, the Trace mode and the "Memory*" mode. The Trace mode looks like a finished product, At least, it works in a consistent manner. But it's actually a legacy mode, with poor resolution in the frequency domain. The user's manual for MSO5000 series provides better coverage for the matter.

FFT resolution = (FFT sampling rate) / (points number)

In the Trace mode, the FFT function fetches a points from the display memory, so the FFT sampling rate is actually equal to the beam sweep data rate, 100/(T scale) in points/s, and the points number is equal to 1200 (12 horizontal grid divisions, 100 two-pixel point images per division). From that, max. theoretical FFT resolution bandwidth for each T scale can be calculated. For instance, on the smallest T scale:

5 ns/div -> 20 GPts/s -> 16.666 MHz, while the finest FFT F scale setting is 20 MHz/div.

On the largest T scale (not considering the slow sweep scales):

100 ms/div -> 1 kPts/s -> 0.83 Hz, while the finest possible F scale is 1 Hz/div.

So in the Trace mode, on each T scale, the FFT resolution is almost equal to the smallest F scale. Not good, but it would be pointless to try to increase the resolution adding some AI rather than more data, because that would likely result in even more poor spectrum plot. Instead, Rigol finally decided to implement fetching a points from the acquisition memory in their entry level scopes, adding the "Memory*" mode in DS1000Z. In that mode, FFT sampling rate is equal to the instrument's current sampling rate, and the max. points number is equal to 16K, which is about 14 times greater, allowing for a much more accurate FFT.

Specifically, there are two features in DS1000Z, that sources their input data in the acquisition memory, the Frequency Counter, and the FFT function. The Counter works fine, but my experience with FFT in the Memory mode is mixed. I've a feeling that there is a large room for improvement in the SW.

For instance, the book reads "You can adjust the center frequency and horizontal scale at the same time using the horizontal scale knob". It's natural for a user to expect that the displayed readings for the frequency and sampling rate FFT scales will change accordingly. In the Trace mode, that's in fact so. In the Memory mode, when the instrument's sampling rate has changed by the T scale knob rotation, the readings does not change. More precisely, sometimes they changes, sometimes not and sometimes they changes in wrong way. It's not consistent, and the only reliable way to update the readings is to turn the FFT function off and on again, losing all the cursors, configured for the Math window.

But that's the most innocent defect, though it's annoying. On rare occasions, the spectrum trace does not appear at all when the function has cycled on/off/on. Typically, it does, and in that case I've noticed similar inconsistency trying to control the FFT settings. With the same combination of sample rate, mem depth and T scale, sometimes it's possible to configure FFT for desired center frequency and scale, sometimes not, because of too high granularity or because the banner "Parameter Limited!" pop-ups.

I'm failed to figure out a pattern so far. Perhaps it depends on how lucky the scope was at the boot time or how long the user played with the buttons and knobs. Once I witnessed the "Memory leak in FFT" pop-up message. Something is definitely wrong there. Poor software, but it's good enough to make a problems difficult to report because of the sometimes keyword.

I'm hoping someone here who has more experience with that instrument can provide more insights so we can work out a meaningful communiqu to Rigol's attention.

---
WBR

[Njk]

Ok, found a way to navigate around the defects and use the feature.

The trick is in proper sequence. When you'd initially switched to the Memory mode, it's necessary to set the Center frequency at zero, and then set the F scale as desired. Then you can set desired center frequency. Otherwise, it could not be possible because of high granularity of allowed center frequency values. Pretty weird. Looks like missing initialization.

A post without images looks boring now. This is the FFT plot for 1 V 1.1 kHz sine wave signal (img0.png):
" alt="" class="bbc_img" />

Notice the discrepancy in measured amplitude values between the FFT and the RMS item. It's a pure sine wave from a pure analog high quality signal generator, there are no other harmonics, why the 3 dBv difference?, Actually, it's also weather-dependent, could be 0 dB on another day. Perhaps a calibration issue, that's Rigol, never mind.

It's interesting the amplitude value calculated by the FFT can be more accurate, probably because it uses data from the acquisition memory (img1c.jpg):
" alt="" class="bbc_img" />

The Frequency counter also seems working fine for that frequency (img2c.jpg):
" alt="" class="bbc_img" />

Finally, let's increase the amplitude above the channel's clipping level to see the harmonics (img3.png):
" alt="" class="bbc_img" />



Edit: actually, the interdependency of Center frequency and F scale settings is documented, but in the user's manual for DS2000A series: "Note: Setting the horizontal position can changing the center frequency of the FFT operation results indirectly."
---
WBR

[Njk]

Another puzzle. What prevented Rigol from (optionally) dismissing the input signal trace for full screen FFT plot? Displaying both of them can result in total mess on the screen.

Found a workaround. Switch the display in the Dots mode and reduce the beam intensity down to zero. Fortunately, the intensity control does not affect a math traces. For better effect, play with the acquisition mode setting (seemingly it does not affect FFT when it's at the Memory mode). The result: a nice full-screen view.

" alt="" class="bbc_img" />

My apologizes if that trick is long time obvious.

---
WBR

[tautech]

 
Don’t these have a Exclusive FFT mode that doesn’t show the source waveform ?

[RoGeorge]

That's a nice trick  , I don't think DS1054 has the option to display FFT only.

In fact, the FFT functionality in Rigol DS1054Z is unusable in practice.  It works, but it's too cumbersome to operate the settings, too low resolution, and too slow to make any use of it, other than maybe a didactic demonstration.

The most useful trick with FFT on DS1054Z is to avoid the FFT entirely. 

[Njk]

Don’t these have a Exclusive FFT mode that doesn’t show the source waveform ?

No and that's the problem

---
WBR

[Njk]

In fact, the FFT functionality in Rigol DS1054Z is unusable in practice.  It works, but it's too cumbersome to operate the settings, too low resolution, and too slow to make any use of it, other than maybe a didactic demonstration.

It could be better but I'm trying to be positive. Some niche areas, e.g. education..

---
WBR

[Fungus]

In fact, the FFT functionality in Rigol DS1054Z is unusable in practice.  It works, but it's too cumbersome to operate the settings, too low resolution, and too slow to make any use of it, other than maybe a didactic demonstration.

It does work, but it takes a LOT of fiddling with the parameters.

Then if you change the timebase for a different signal you pretty much have to start over. 

The DS1054Z isn't the 'scope for you if FFT is something you use a lot. This has never been a secret.

[Njk]

It does work, but it takes a LOT of fiddling with the parameters.

I was of the same opinion, initially. But now I've to say that actually it's an efficient and elegant UI design, for given environmental constraints. I can't propose anything better (maybe the now popular touch screen technology provides more opportunities, but its advantages are offset by a scratched screen surface, dirty with fingerprints. Not everybody likes that). It's something like a scientific calculator, which uses stack-oriented input notation. Clumsy and inconvenient, but only to those without the idea. To all others, it's very convenient and super efficient.

What is really missing is the documentation. Every new user have to learn how it work. BTW the similar situation is in open source. Looks like the OSS guys are playing their game preventing the things from easy commercialization by competitors. But the Chinese are not dumb, they're thinking in they own way. Perhaps it's as difficult for them to explain something to a westerner, as it was for me to locate an affordable restaurant with casual European kitchen in Taipei. So omission is they favorite game as well.

Concerning the FFT function, the problem is how to configure the two numerical values, the frequency scale (Fscale) and the central frequency (Fc). The latter ranges from zero to the FFT sampling frequency divided by two. It's quite a bandwidth and if you want to achieve high precision and to avoid inconvenience and excessive mechanical wear, two tuning methods are requited, coarse and fine. But there are two controls in all, one is for Fscale and the other is for Fc. Instead of adding one more control, Rigol had decided to overload the existing controls with tuning method selection.

For each FFT sampling frequency, six Fscale values are available. Each Fscale value is associated with unique Fc increment/decrement (step) value. Consider 1 GSa/s sampling frequency as an example:

Fscale = 1 MHz   ->  Fc step is 20 kHz;
Fscale = 2.5 MHz -> Fc step is 50 kHz;
Fscale = 5 MHz    -> Fc step is 100 kHz;
Fscale = 10 MHz  -> Fc step is 200 kHz;
Fscale = 25 MHz -> Fc step is 500 kHz;
Fscale = 50 MHz -> Fc step is 1 MHz.

The Fc value can be set in the range of 500 MHz with 20 kHz precision. The procedure is as follows:

1. Select the highest Fscale value. Coincidentally, pressing the multi-function knob for Fscale does just that.
2. Tune close to desired Fc. That's coarse tuning.
3. Select the finest Fscale.
4. Tune exactly to desired Fc. That's fine tuning.
5. Finally, set Fscale at desired value.

That's it.

---
WBR

[Njk]

That's funny, my FFT experiments resulted in the scope's death. The RSoD syndrome. Definitely I'm not the first one. There is an advisory note of Aug 1, 2022 at the Rigol NA web site with the procedure the user is advised to follow in that case before service application. Did not worked. The Rigol FAE guy was confident that the problem will be easy to fix. Then he realized that in my case they have to order a replacement main board from China.

In the thread about MSO5000 the users reported that the scope is quite hot when operating. It was not like that, but I never took the unit apart to check how hot the parts are. So it might be a coincidence

---
WBR

[Njk]

The scope is back after main board replacement.

State of the art version data, for reference.

Board version: 1.0.4

Extended information (menu menu force menu):

SW version: 00.04.05.02.03
Boot version: 0.0.1.9
FW version: 0.2.3.12
CPLD version: 1.1
Build date: Feb 15 2022 15:54:18 | 0 0:0/ 0 0:30

At first glance, it works not worse, not better. Failed to get release notes for this SW version.

[rob040]

According to me SW version 00.04.05.02.03 has not been published (yet).
Latest was/is: 00.04.05.02.02.

[Njk]

According to me SW version 00.04.05.02.03 has not been published (yet).

That's why I'd asked the local Rigol rep about the release notes. But he also has no information

[Adrian_Arg.]

According to me SW version 00.04.05.02.03 has not been published (yet).

That's why I'd asked the local Rigol rep about the release notes. But he also has no information

Hello, see if this problem was corrected, with the new virmware that you installed. there I give you the link

https://www.eevblog.com/forum/testgear/rigol-1054z-and-1074z-timing-problems-at-500ms-timescale-and-24m-memory-depth/

[Njk]

No miracle. With that version, I was able to reproduce it

[Ben23]

My apologies if this has already been mentioned, but I've skimmed several threads and couldn't find any mention of it, and it's not an easy one to search for.

I've just bought the DS1054Z, its software is up-to-date (00.04.05.SP2, or 00.04.05.02.03 in the expanded info view).

The bottom two vertical measures in the left sidebar last page are weird.
"AC.Vrms" does indeed turn on AC.Vrms, but the Help page brought up by that button is "Variance".

And then there's "3". What is that supposed to be? The Help for "3" says it's AC.Vrms, but what it actually turns on is Period, which isn't even a vertical measure! The icon suggests it should be some other sort of RMS measurement? Or should it be the missing "Variance"?

Also AC.Vrms and a bunch of the horizontal measures don't appear in "Measure All", but I guess they just have a loose definition of "All"...

[RoGeorge]

Mine is one version behind, 00.04.05.02.02, and it doesn't have any icon on the 6th place.  First 4 icons are the same, as in your pics, the 5th is called 'Variance', the last place is empty.

What does the help function says about the 6th icon called "3", if you press 'Help' then '3'?

[Ben23]

Mine is one version behind, 00.04.05.02.02, and it doesn't have any icon on the 6th place.  First 4 icons are the same, as in your pics, the 5th is called 'Variance', the last place is empty.

Interesting.

What does the help function says about the 6th icon called "3", if you press 'Help' then '3'?

It gives the help for AC.Vrms.

[Njk]

Yep. I confirm the "3" icon with the 02.03 version. Pressing on it results in "Period=" item appearance in the bottom. It's so innocent problem I was not sure it's specific to this version.

BTW, my progress in finding the difference is very limited. But from what I've found so far, I think that v.04.05.02.03 is a build dedicated to system stability improvement. With the previous versions, the best way to demonstrate frozen UI was to use FFT. With this version, it does not hang there. That does not mean any FFT defect is fixed. Instead of hanging, the FFT trace just disappears while the scope remains responsive. But after that, the whole math trace "channel" is lost, it's not possible to get a data from the scope (points data, screenshot, etc.), and who knows what else stopped working properly. The scope is not trusted any more and it needs to be power cycled anyway. I don't know what is better. Looks like they don't have access to the low-level FFT code. Perhaps it was received in the binary form with OS/SDK/BSP from Infineon or a third-party vendor. So they can't fix it but the attempt to control the damage is evident.

Overall, no news from them over long time. According to the press release, they recently went public. So new leadership, new vision, new priorities, a lot of empty positions in the local office. Add to this the international unrest, the covid scam, the ship shortage. Seems now they have something more interesting to do rather than just to make the things.

[RoGeorge]

My DS1054z is used rather rare, but I don't recall any hangs with the firmware it has, 00.04.05.02.02.  Used FFT a couple of times, no hangs.  Maybe downgrade back to 00.04.05.02.02 if you find 00.04.05.02.03 to be faultier.

For now, I wouldn't upgrade to 00.04.05.02.03, thanks for the warning.