Best 350MHz scope in a hackable world (Siglent SDS2104X Plus or Rigol MSO5072)

[kahuna0k]

Terminating both scopes with a 50\$\Omega\$ bnc terminator, limiting to 20Mhz at the 2mV/1us range, I get the following:

	Vpp	StdDev
MSO5074 Normal	1.21mV	187uV
SDS2104X+ 8b	194uV	30.1uV
MSO5074 HiRes	1.14mV	184uV
SDS2104X+ 10b	140uV	20.4uV

Bonus data: If I set the channel to GND coupling the Rigol actually gets 0.0 for both while the Siglent still gets 132uV and 17uV for Vpp and StdDev 

Edit: correct typo (thanks Martin72

[Martin72]

194mV by 2mV/Div. ?

[mawyatt]

Think this is in uv not mv!

If so, I get similar results for both 8 and 10 bit modes

8 bit      ~178uv PP and ~29.3uv SD
10 bit    ~132uv PP and ~19.5uv SD

Thanks for doing these tests

Best 

[Martin72]

Will do this tomorrow too, then we have three...

If I set the channel to GND coupling the Rigol actually gets 0.0 for both while the Siglent still gets 132uV and 17uV for Vpp and StdDev

Will check this too.

[Fungus]

Bonus data: If I set the channel to GND coupling the Rigol actually gets 0.0 for both

If so, it must be 'cheating' in software...

What do you see on screen? A perfectly flat line?

[Fungus]

	Vpp	StdDev
MSO5074 HiRes	1.14mV	184uV

How many samples were being averaged there? If the 20Mhz limiter is on then you can easily go to 32x (or even 64x).

[Martin72]

Will do this tomorrow too, then we have three..

Done, after warmup and self-cal, using the internal 50ohm and no averages.

[rf-loop]

Why peoples are using 2mV/div vertical scale when lowest full resolution is 1mV/div.

[TurboTom]

IIRC, Rigol's phoenix chipset initially had been released without a boxcar averaging (or highres) mode. After complaints by the early adopters, it had been added in firmware -- as it appears, as an afterthought -- and as we now see, there's apparently still something broken with it.

There had been a lot of speculation about the nature of the digital part of their sampling engine, if it's a genuine FPGA or rather some hard-wired gate array. In case of the latter, it may explain the poor performance of Rigol's phoenix chipset's high-res mode, and would be a real bummer and a bad error on behalf of them.

Since apparently, none of their scopes utilizing this chipset, works much better than the MSO5k in highres mode or has much less noise in the high input sensitivity ranges, it seems that it's an inherent problem to their proprietary chipset, which may turn out a real problem for Rigol. They probably have bound all of their entry and mid range scopes to this platform for years to come and if this platform isn't performing competitively, they simply won't sell as anticipated. They seem to be in a similar boat as Agilent when they introduced their MegaZoom chipset, limiting it to the relatively small internal sample memory -- with the difference, that when the MegsZoom chipset hit the market, its memory size was competitive and it performed very well otherwise, too.

I'ld be surprised that if there was an easy (firmware-based) fix for Rigol's recently released scopes' deficiencies, Rigol wouldn't correct the problems in no time. Right now, one must get the impression that Rigol cannot play in the same league as their main competitor, Siglent. For the laws of the market, that's bad news...

[Martin72]

Why peoples are using 2mV/div vertical scale when lowest full resolution is 1mV/div.

For comparision with the two prior user-measurings.*
But we all could repeat it with 1mV/Div. 

*) "Remember, only when all competitors will start at the the same time, a comparison of the times makes sense" (J.Malmsheimer)

[tv84]

Bonus data: If I set the channel to GND coupling the Rigol actually gets 0.0 for both

If so, it must be 'cheating' in software...

I hate this!

[Martin72]

@TurboTom

I got a taste of it as I did the question to the rigol support concerning the next higher range, the 7000s, will they got lowe noise.
And the answer was, they got the same frontend....
MSO5000, I can understand when it´s a bit noisier than others.
But the more expensive 7000s Range also....And what about the 8000 ?
Seems, rigol have made a bad mistake.

[rf-loop]

Why peoples are using 2mV/div vertical scale when lowest full resolution is 1mV/div.

For comparision with the two prior user-measurings.*
But we all could repeat it with 1mV/Div. 

*) "Remember, only when all competitors will start at the the same time, a comparison of the times makes sense" (J.Malmsheimer)

here

1mV
  8bit
  10bit

2mV
  8bit
  10bit

[Noy]

Would be interesting to see what the DS70000 is doing there. Same chipset..

[rf-loop]

Why peoples are using 2mV/div vertical scale when lowest full resolution is 1mV/div.

For comparision with the two prior user-measurings.*
But we all could repeat it with 1mV/Div. 

*) "Remember, only when all competitors will start at the the same time, a comparison of the times makes sense" (J.Malmsheimer)

Perhaps because Rigol do not have  real 1mV. It is only display scale magnified from some other ... Rigol data sheet is bit weird. They tell 2 and 1mV is magnified from 4mV/div... (I bit suspect this data sheet. https://www.batronix.com/files/Rigol/Oszilloskope/MSO5000/MSO5000_DataSheet_EN-V2.0.pdf)

Vertical and there note 3.  Oh and note 3 do not include anything about 500uV. 

[Martin72]

500µV was later added by firmware - why ever...
In his first video about the rigol, Dave said you won´t use this scope for low level measurements.

[rf-loop]

500µV was later added by firmware - why ever...
In his first video about the rigol, Dave said you won´t use this scope for low level measurements.

Advertising manager need this number. This is only reason.

But if I believe this data sheet also 1 and 2 mV are just made by FW. What is real truth here.
Siglent SDS2000XPlus 500uV/div is sure just magnified from 1mV so it have  half resolution.
Also if measure noise it give just roughly same what 1mV/div. naturally because it is only pixelzoom.

[Fungus]

IIRC, Rigol's phoenix chipset initially had been released without a boxcar averaging (or highres) mode. After complaints by the early adopters, it had been added in firmware -- as it appears, as an afterthought -- and as we now see, there's apparently still something broken with it.

There had been a lot of speculation about the nature of the digital part of their sampling engine, if it's a genuine FPGA or rather some hard-wired gate array. In case of the latter, it may explain the poor performance of Rigol's phoenix chipset's high-res mode, and would be a real bummer and a bad error on behalf of them.

If that's true then it's an unbelievable mistake. Even an idiot like me can imagine adding a programmable FIR filter to a chipset for this sort of application.

[mawyatt]

Designing a chip-set such as Rigol's Phoenix isn't trivial and likely Rigol contracted this design out. Even tho Phoenix isn't anywhere near SOTA, even when introduced, still requires unique design skills and unlikely Rigol has those skills captured internally.

As many have found out contracting complex chip designs to 3rd parties isn't always straight forward & successful, and may require many costly respins to "get it right", or even "get it close". Depending on the CMOS or SiGe BiCMOS process utilized this quickly becomes an expensive recurring endeavor.

My guess this happened to Rigol and after a couple expensive respins, they decided that Phoenix was "good enough" but not what they had originally anticipated. Hopefully they've learned from this and can develop future custom chip-sets that are more robust in performance than Phoenix.

Being a retired IC designer I personally admire Rigol for taking this risk, and hopefully they will be successful in the next generation chip-set, placing a lot of pressure on Siglent and others to follow suit. Integration is the way to higher performance at lower recurring cost, but has much higher non-recurring cost & risk, so difficult to sell management considering the limited market and ROI. In the near future we may have new mid-level and priced DSOs with real 500uv/Div scale factors, and real 12-14 ADCs with 11~12 bit ENOB @ 10~20GSPS, if certain managements can be sold

Best,

[kahuna0k]

For completeness:
Terminating both scopes with a 50\$\Omega\$ bnc terminator, limiting to 20Mhz at the 1mV/1us range, I get the following:

	Vpp	StdDev
MSO5074 Normal	1.34mV	177uV
SDS2104X+ 8b	162uV	24.2uV
MSO5074 HiRes	1.24mV	174uV
SDS2104X+ 10b	125uV	18.1uV

In terms of noise there is no discussion, but IMHO, noise is not everything, not even the more important thing depending on what you are working with. There are things that I like in the Rigol, like for example: memory management (the zooming out thing), 1GSa/s vs 500MS/s in each digital channel, being able to search for patterns in the decoded channels, 4 math channels instead of 2, 2 channel AWG.

Also although the UI is more polished and more responsive in the Siglent, I hate when I cannot use the physical controls to do something and I need to use the touch screen. And a minor complain, can we get the generic rotating control to scroll the menu when it does not fit in the screen? Scrolling the right menu by dragging it is slow and always ends in me selecting the wrong thing.

You can only say that one scope is better than the other if it is better in everything. The Siglent is better if you care about noise, but IMHO the Rigol is better if you care about digital signals. So go and look at the tech specs of both, check what you will use and what you will need, and if both have it, go for the Siglent, as the UI is better and the firmware updates more frequent.

[Martin72]

being able to search for patterns in the decoded channels

Just a matter of the software (like "Me, I´m fat and you´re dumb - I could get slim and what about you? " )...

memory management

On their list (with low priority)

4 math channels instead of 2, 2 channel AWG

Shoot, that couldn´t be equalized....Although I if I had a wish, I would like to have a noiseless frontend instead having a AWG build in.

go for the Siglent, as the UI is better and the firmware updates more frequent.

There are a little bit more pros like ext. trigger input, 50Ohm integrated, 1/10 sense (for me not so important), bigger and brighter screen and so on.
But you´re right, I got the rigol for over a year and could say, there´s nothing equal to its price of appx 1000 bucks.

Scrolling the right menu by dragging it is slow and always ends in me selecting the wrong thing.

Here, me too.... Must learn to use a mouse...

[kahuna0k]

	Vpp	StdDev
MSO5074 HiRes	1.14mV	184uV

How many samples were being averaged there? If the 20Mhz limiter is on then you can easily go to 32x (or even 64x).

Sorry, I don't understand the question. I was using the HiRes capture mode in the Acquire menu of the Rigol and the 10 bit mode in the Siglent.

In the Rigol, if I use average acquisition mode over 64 samples at 1mV/1uS I get 568uV Vpp and 57.2uV StdDev, and these numbers don't get better if I increase the number of samples.

I would expect those to be 0.0 for noise if it would be sampling over consecutive waveforms (as in fact it happens if I use the average math function in the Siglent) so it seems that you are right and it is averaging over consecutive samples of the same waveform. That would be a more fair comparison to the 10bit mode of the Siglent. Although, even with that, the noise is still not as good as the 10bit mode of the Siglent, but clearly better than the HiRes and Normal acquisition modes. So now the question is, wtf is the HiRes mode on the Rigol?

In Rigol states that the HiRes is actually sampling over the same capture waveform and average is capturing over different triggering events, but how can StdDev of that average over different triggering events do not average to 0 (even at 4096 sampling events)? I'm starting to think that not even Rigol understands how this new ASIC works 

[Martin72]

Like the 500µV, the hi-res mode was later on implemented.
As it came up, I couldn´t see any enhancements, but it was active...I´ll search my former post.

[Martin72]

Hi,

Found the post and...Well...

With my knowledge of today I must say, it was nearly almost BS what I´ve posted 1.5yrs ago...

But for the records, here it is:

https://www.eevblog.com/forum/blog/new-rigol-scope/msg2233980/#msg2233980

 

[kahuna0k]

Watching the video and looking at some tests I've done myself it seems that what the HiRes mode does is increase the dynamic range. It seems that I can get much finer detail in a 5V peak-to-peak signal in the Rigol than in the Siglent. I need to do more tests, as that somehow contradicts what Martin said about measuring power supplies ripple, where the Siglent performed much better.

Maybe the floor noise of the Rigol is bad for whatever the reason, but the dynamic range in HiRes is actually better (probably not 12bits but better than those famous 5-6 ENOB). I need to learn a lot to be able to properly discuss this, but I think that there is more to be uncovered here