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

[pascal_sweden]

Did you finish the side by side review?

[NCG]

10 bit and input noise defined it above all for me.

[Fungus]

What happens to the infamous Rigol noise if you turn on (eg.) 4x averaging?

With 350Mhz front end and 8GS/sec sample rate you have samples to spare for things like this.

I'd imagine it drops right down, maybe even giving something approximating 9-bits resolution.

[tautech]

10 bits plus an additional ERES 3 bits on an already low noise front end leaves most others in its wake.

[TurboTom]

10 bit and input noise defined it above all for me.

Even though the SDS2000X Plus appears to provide a lower noise floor than some of its competitors, it's an eight bit oscilloscope by hardware. Siglent uses two MXT2001 dual 1GSPS 8bit ADCs in its frontend that can be cascaded to provide 2GSPS when only two channels of the scope are in use.

Every marketing talk of 10bit or even more refers to things that happen in software (and should be specified as such). This doesn't mean that this option isn't useful. But for comparison purposes, one should always keep that in mind. There are only a few scopes in the entry or mid level range that actually contain ADCs that sample more accurately than eight bits -- a few Owons (utilizing the HMCAD1520 that can be internally reconfigured to provide 8/12/14 bit sampling depth, yet at reduced sampling rate - good choice for audio / sensor jobs) or the RTB2000 (Rohde&Schwarz) with its proprietary 10 bit full-speed ADC (up to 2.5GSPS).

[mawyatt]

10 bit and input noise defined it above all for me.

Even though the SDS2000X Plus appears to provide a lower noise floor than some of its competitors, it's an eight bit oscilloscope by hardware. Siglent uses two MXT2001 dual 1GSPS 8bit ADCs in its frontend that can be cascaded to provide 2GSPS when only two channels of the scope are in use.

Every marketing talk of 10bit or even more refers to things that happen in software (and should be specified as such). This doesn't mean that this option isn't useful. But for comparison purposes, one should always keep that in mind. There are only a few scopes in the entry or mid level range that actually contain ADCs that sample more accurately than eight bits -- a few Owons (utilizing the HMCAD1520 that can be internally reconfigured to provide 8/12/14 bit sampling depth, yet at reduced sampling rate - good choice for audio / sensor jobs) or the RTB2000 (Rohde&Schwarz) with its proprietary 10 bit full-speed ADC (up to 2.5GSPS).

Agree that if software enhancement is utilized the "expand" the apparent ADC resolution it should be identified as such. The ADC & front end is the core of the DSO and should be evaluated on what this core circuitry can perform. With this in mind the real discriminator is how well this core ADC & front end can perform stand alone without software enhancements.

Invoking ENOB or Effective Resolution Bits (ERB) performance of the core gets to the performance that is fundamentally hardware limited. Howard Long did an actual comparison of DSOs based upon his measurements that was reported here which created lots of interesting discussions.

https://www.eevblog.com/forum/testgear/benefits-of-going-with-all-siglent-setup/

Later I did some measurements of the DSO I had and reported here,

https://www.eevblog.com/forum/testgear/benefits-of-going-with-all-siglent-setup/50/

I can't speak for the Rigol or other DSOs but the SDS2102X Plus I have has impressed me, and it's ENOB, ERB and input referred noise are quite respectable IMO.

Edit: Didn't know the Owon DSO were true 12 bits and used this Analog Devices ADC, just checked and the XDS3102A and 3202A apparently utilize this ADC. Need to check on how these Owon DSOs perform.

Best,

[Fungus]

Every marketing talk of 10bit or even more refers to things that happen in software (and should be specified as such).

Yep, but tautech's a siglent salesman who has to pretend that all Siglents are devices from heaven.

Technically speaking you can turn on 4x averaging on the Rigol and still have the same sample rate as the Siglent but with a lot lower noise. Hence my question above.

(and also my wondering whether all the noise about the Rigol's noise is really productive or not, ie. does it really make much difference in practice?)

What happens to the infamous Rigol noise if you turn on (eg.) 4x averaging?

I'd imagine it drops right down, maybe even giving something approximating 9-bits resolution.

[tautech]

Every marketing talk of 10bit or even more refers to things that happen in software (and should be specified as such).

Yep, but tautech's a siglent salesman who has to pretend that all Siglents are devices from heaven.

Says he that abandoned Rigol and bought a Micsig. 

[mawyatt]

Every marketing talk of 10bit or even more refers to things that happen in software (and should be specified as such).

Technically speaking you can turn on 4x averaging on the Rigol and still have the same sample rate as the Siglent but with a lot lower noise. Hence my question above.

(and also my wondering whether all the noise about the Rigol's noise is really productive or not, ie. does it really make much difference in practice?)

What happens to the infamous Rigol noise if you turn on (eg.) 4x averaging?

I'd imagine it drops right down, maybe even giving something approximating 9-bits resolution.

I seriously doubt that you can achieve anywhere near 9 bits Effective Resolution with the Rigol ADC & Front End as Howard had shown it struggles to achieve an Effective Resolution of 5 bits, the ENOB will be even worse!

Another potential guage is the accuracy of the ADC and front end, which I have not seen much posted on various DSOs. Might be nice if someone could do actual accuracy measurements and post the results as Howard has done here. Table shown below.

https://www.eevblog.com/forum/testgear/benefits-of-going-with-all-siglent-setup/


Best,

[TurboTom]

That's what i meant when I posted in some contribution that this 5 bits of ENOB for the DS5k will now stand cast in iron. There was an error in the interpretation of the RMS figures, making the whole measurements for scopes with some millivolts offset mute. See here.

[Fungus]

Effective Resolution of 5 bits...

Here we go again... 

That chart has been debunked several times now.

[Fungus]

Says he that abandoned Rigol and bought a Micsig. 

What does that have to do with anything? 

Was I not allowed to buy something new? I must have missed that memo.

[mawyatt]

That's what i meant when I posted in some contribution that this 5 bits of ENOB for the DS5k will now stand cast in iron. There was an error in the interpretation of the RMS figures, making the whole measurements for scopes with some millivolts offset mute. See here.

Yes that is a concern reading the lower level rms vs. SD values, I had forgot about the DC offset issue which BTW doesn't seem to be a problem with the DSO I have, the rms and SD read similar on all vertical scales. However, shouldn't the higher level rms and SD be almost identical on the DS5k or is the offset just that bad everywhere on the vertical scales?

Best, 

[mawyatt]

Effective Resolution of 5 bits...

Here we go again... 

That chart has been debunked several times now.

Yes directly from the Rigol fanboy or girl that thinks averaging fixes everything

Best,

[Fungus]

Yes directly from the Rigol fanboy or girl that thinks averaging fixes everything

a) I don't own anything made by Rigol

b) I was asking what happens to the noise figures, not telling anybody anything. Averaging is a perfectly valid technique for reducing noise.

[mawyatt]

Yes directly from the Rigol fanboy or girl that thinks averaging fixes everything

a) I don't own anything made by Rigol

b) I was asking what happens to the noise figures, not telling anybody anything. Averaging is a perfectly valid technique for reducing noise.

a) Didn't you own some Rigol equipment in the past, maybe a DSO?

b) Quote from a comment above.

"Technically speaking you can turn on 4x averaging on the Rigol and still have the same sample rate as the Siglent but with a lot lower noise. Hence my question above."

Technically speaking seems like a statement to me!

and more on this.

"and also my wondering whether all the noise about the Rigol's noise is really productive or not, ie. does it really make much difference in practice?"

Noise generally isn't productive and most designers attempt to reduce it unless your striving for a noise generator

Putting all this fun aside, think we should all ask Dave or Howard to do a proper side by side unbiased evaluation since they seem to be the only folks with direct access to a number of these mid-level DSOs. Since Howard had indicated he was somewhat of a Rigol Fanboy I placed a little more faith in his evaluation than I normally do, but as TurboTom pointed out the rms vs. SD with offset bias issue. In all honesty I would have probably made that same mistake before realizing the DSO rms reading includes the DC offset since most RMS DVMs read AC waveform rms and don't include the DC term. What's funny is my Fluke and Keysight handheld DVMs say True RMS Reading, which isn't True!! They both read AC rms, since a True rms should be waveform agnostic and include all sources AC + DC

Anyway, we wouldn't be discussing any of this if we didn't have such high performance, quality instruments available at moderate prices from newer resources such as Rigol, Siglent, Micsig & others. Fun times for those younger engineers and scientist that now can afford these instruments at home. I know when I was younger the usual Tektronix and HP equipment was way off limits price-wise. You had to wait for a much older bargain instrument on the used market that likely needed repair.

Now to study up on the Owen DSO that was mentioned which includes the 8/12/14 bit Analog Devices HMCAD1520 multi-mode ADC, Dave's review wasn't encouraging tho

Best,

[Fungus]

"Technically speaking you can turn on 4x averaging on the Rigol and still have the same sample rate as the Siglent but with a lot lower noise. Hence my question above."

Technically speaking seems like a statement to me!

I don't think there can be any argument over whether averaging reduces noise or not. Or whether 8 divided by 4 equals 2. 

The question is: How is the noise afterwards?

[Fungus]

as TurboTom pointed out the rms vs. SD with offset bias issue. In all honesty I would have probably made that same mistake before realizing the DSO rms reading includes the DC offset since most RMS DVMs read AC waveform rms and don't include the DC term. What's funny is my Fluke and Keysight handheld DVMs say True RMS Reading, which isn't True!! They both read AC rms, since a True rms should be waveform agnostic and include all sources AC + DC

Yep, it's a very easy thing to overlook and very difficult to 100% cancel out.

Yet more proof that you should never trust any number on the screen of your DSO.

[mawyatt]

"Technically speaking you can turn on 4x averaging on the Rigol and still have the same sample rate as the Siglent but with a lot lower noise. Hence my question above."

Technically speaking seems like a statement to me!

I don't think there can be any argument over whether averaging reduces noise or not. Or whether 8 divided by 4 equals 2. 

The question is: How is the noise afterwards?

Averaging isn't guaranteed to reduce all "noise" as expected, it depends on how the "noise" correlates between the averaging samples. If the "noise" is from a switching type source (SMPS), then averaging may not work as well since the switching noise will likely have a high correlation between samples. Pure random white noise does follow the usual thinking about noise averaging since it has zero correction between samples.

The DSO input referred noise could have components from it's SMPS, digital clocking, display drive, the Mains, and so on, these may not average out as expected since they are not truly random and likely depend on the sweep rate, triggering and scale factors. A well designed front end shouldn't have serious noise issues at any scale factor that reduce the display quality and create uncertainty in the observed waveform evaluation.

Waveform averaging should be utilized to improve the measured waveform noise properties, and shouldn't be a requirement for the measuring instrument to subdue its own input noise. During a discussion on higher performance ADCs & DACs a Keysight Fellow told me that the instrument needs to be at least 10X better than the measurement, which places a reasonable limit on the DSO input noise performance.

The DSO I have has front end noise of ~77uvpp with a SD of ~17uv @ 1ns/div and 163uvpp &  23uv @ 1us/div respectively on the 500uv/Div. Using the 10X factor mentioned indicates that the DSO input noise should not be greater than 8*Scale Factor/10 or 8*500uv/10 or 400uvpp.

Best,

[mawyatt]

as TurboTom pointed out the rms vs. SD with offset bias issue. In all honesty I would have probably made that same mistake before realizing the DSO rms reading includes the DC offset since most RMS DVMs read AC waveform rms and don't include the DC term. What's funny is my Fluke and Keysight handheld DVMs say True RMS Reading, which isn't True!! They both read AC rms, since a True rms should be waveform agnostic and include all sources AC + DC

Yep, it's a very easy thing to overlook and very difficult to 100% cancel out.

Yet more proof that you should never trust any number on the screen of your DSO.

That was simply a user error using the rms rather than the sd value as pointed out by TurboTom. I suspect the Rigol reported both those values correctly, including the peak to peak but I don't know since I didn't make those measurements.

YOU may not trust the numbers on YOUR DSO screen, but I trust those on my DSO screen! Here's a few numbers from the DSO screen I do trust since a new Keysight 34465A was used, and results supported with a HP34401A and an Agilent 34401A.

KS34465A             SDS2102X Plus
9.99999VDC          9.966VDC
7.04192VDC          7.0092VDC
0.999964VDC        1.0003VDC
100.0005mVDC     98.987mVDC
5.00026VDC          4.9987VDC
2.49994VAC           2.4996VAC (250.000Hz)

https://www.eevblog.com/forum/testgear/benefits-of-going-with-all-siglent-setup/

And a few more "noise" & ERB measurements including FFT noise floor.

https://www.eevblog.com/forum/testgear/benefits-of-going-with-all-siglent-setup/50/


Best,

[Fungus]

"Technically speaking you can turn on 4x averaging on the Rigol and still have the same sample rate as the Siglent but with a lot lower noise. Hence my question above."

Technically speaking seems like a statement to me!

I don't think there can be any argument over whether averaging reduces noise or not. Or whether 8 divided by 4 equals 2. 

The question is: How is the noise afterwards?

Averaging isn't guaranteed to reduce all "noise" as expected, it depends on how the "noise" correlates between the averaging samples. If the "noise" is from a switching type source (SMPS), then averaging may not work as well since the switching noise will likely have a high correlation between samples. Pure random white noise does follow the usual thinking about noise averaging since it has zero correction between samples.

Now who's speaking "technically"?

All I'm after is somebody with an MSO5000 to actually do it and say what happens.

Because ... if the noise drops way down to manageable levels then this whole issue would seem moot to me.

I think Siglent's not being able to zoom out after capturing a signal would probably bother me more  - I need that much more often than I'd need to look at mV signals. YMMV.

Edit: If the input signal interacts badly with the sample averaging then there's also the option of applying a low pass filter on a math channel... use the tools provided!

[Noy]

If Dave can do a video, maybe rigol has than some spare software engineers left after the DS70000 stuff is "completed" for an initial release... and than they can improve/polish something in the 5k-8k series (also they should look / copy  Bode plot things from siglent ones. The pressure must only be high enough.. ;-)

[Martin72]

Hi,

One of my last measurings with the 5000 was ripple-noise from a 60Vdc linear power supply.
Specified with 1mVrms, you see nothing from it on the scope, because of it´s own noise, nevertheless what you try (high res, bandwithlimit, capacitor directly on the input, etc..).
But then I´ve used averaging, in this case 64 times and now I could see something..

Will repeat this with my siglent in the next days (holidays, yeah), expecting not to need to average the signal.

Martin

Edit:

I need that much more often

For what ?

[Fungus]

But then I´ve used averaging, in this case 64 times and now I could see something..

There you go...!

The noise isn't a fatal problem. I wouldn't pay $500 extra for the occasional times I need to look at that sort of signal, not when there's a simple workaround.

[2N3055]

But then I´ve used averaging, in this case 64 times and now I could see something..

There you go...!

The noise isn't a fatal problem. I wouldn't pay $500 extra for the occasional times I need to look at that sort of signal, not when there's a simple workaround.

That is a crutch. It will only show repetitive signal. It won't show spread spectrum signal or real noise, in fact it will suppress it...
So if you want to measure something like that it is useless without preamp that costs much more than a scope...