Siglent SDS2000X Plus Bandwidth & Aliasing Application Note

[2N3055]

Nice measurements...

Well, may show the HW setup as may you simple measure a 50E chain using a given cable...

As looking for overshoots in real gear (as LVHC logic) things fails soon by the used probe.. 

This means using a 300Mhz probe witch has a to high cap as ~20pF will draw more overshoots as using a FET/1GHz as ~1pF.

Not only the BW is important, the phase shifts will cause or show the overshoots. So phase is currently neglected.

Attached a Spectrum of a 10MHz clock up to 1.5GHz using as SSA juts to see that we get or have a large fourier content.
 

50Ohm source impedance and 20pF would form additional RC filter with -3dB point at cca. 160MHz. Good 10x probes at 300MHz and up would be more in a 12-14pF range. Good 500 MHz ones maybe even 10-11 pF.

I again fail to see your point. What are you trying to say: that manufacturers should stop making any scopes with less than 2 GHz bandwidth and that passive probes shouldn't be made and used anymore and we should be allowed only to use active probes..

Or are you complaining about phase response of the scope (i.e. overshoots) when using a 200 MHz scope to look at signal that has 1.5GHz content?

Or you are complaining that BW measurement of scope at 50 Ohm with coax directly from signal generator is not representative of using a scope with passive probe?
It is representative of 50 Ohm signal path that has to be measured also. And that is how scopes are specified and measured. So we can compare with other manufacturers on equal footing.

There are several topics where Performa did detailed characterization of how this scope behaves with several passive probes, done to industry standard too..

[hpw]

@2n3055: My point is, as it shows using the given probes (using shortest connections) gives false results (overshoots) for fast clocks.

So using a 1GHz FET probe using shortest connection trusted me. Also the phase behavior about is an opened task. IMHO.

50E basic measurement is one task, using probes the second task... and so on.

As the Topic on the given PDF: "Siglent SDS2000X Plus Bandwidth Discussion"... is IMHO a starting point.

Home workers may have a limited/reduced requirements, so to deal on real digital world required BW and Phase behavior to consider.

Even Siglent with the new DSO 6K model and 1..2 GHz FET probes does not provide the required information how the probes performs.

While currently I am missing a clear advice about required DSO BW & probes to measure more or less exact overshoots.

[2N3055]

@2n3055: My point is, as it shows using the given probes (using shortest connections) gives false results (overshoots) for fast clocks.

So using a 1GHz FET probe using shortest connection trusted me. Also the phase behavior about is an opened task. IMHO.

50E basic measurement is one task, using probes the second task... and so on.

As the Topic on the given PDF: "Siglent SDS2000X Plus Bandwidth Discussion"... is IMHO a starting point.

Home workers may have a limited/reduced requirements, so to deal on real digital world required BW and Phase behavior to consider.

Even Siglent with the new DSO 6K model and 1..2 GHz FET probes does not provide the required information how the probes performs.

While currently I am missing a clear advice about required DSO BW & probes to measure more or less exact overshoots.

Ok thank you for clarifications..

It seems that you have missed previous few topics (they were some time ago) where it was discussed exactly that.

If you want to look at signals that have 200 ps edges you need 4 GHz scope and corresponding active probe.  That is only way you will get what you want: a pristine pulse response. Period 
Any time signal has edge that is faster than the risetime of the scope one of two things will happen:

1.) if scope has strong AA filters and aggressively cuts  of (lowpass filters) any frequency  that is above Nyquist by 40 dB or more (for a 8 bit scope) and it will show perfectly shaped squarewave with no overshoots but completely wrong, without any high frequency content and edges that will be as fast as a scope and not in any relation to what signal edges are. It will be complete lie, but it will look pleasing because it will look perfect. Which is complete bullshit because original signal is not perfect and 1/3 of energy content from that signal is missing.

2.) You can have response like most modern digital scopes have. That one will try to show as much of frequency content they can. If you feed that one with a signal that has slower edge than scope has, it will show it nicely without overshoots. It will show the same as scope from 1. point. But if you feed it signal that has very fast edge (like here where we feed signal with 40ps risetime to scopes with nominal risetimes from 500-800ps, 10-15 times faster), scope will show overshoots. You say that is bad. I say it is not. It's great. Scope is having problems and is not hiding the fact.

Those peaks are information that my edge is too fast for my scope and that I know scope is lying to me. Scope in point 1. will show clean edges without overshoot all the time, when I have nice signal and when I have too fast signal with many thing hidden inside. I wont know a difference ..

So with my digital scope, if I see that a signal has an edge that is 400 something ps (edge of what my 1GHz scopes can do) and I see overshoots, I know signal is too fast for my scope. If  I see the signal has an edge that is 400 something ps and there are no overshoots, I know it is right there at the edge. And what I see is true. And if I see 500-600 ps edge and overshoots I know it is DUT, not a scope.

If I routinely want to look at 100 ps edges I need to buy 4GHz+ scope...


So much about that.

As for probing, that has been discussed even more times. 
That is something that we all agree is a most important and most misunderstood topic in using scopes.
Most of the times quality of measurements is directly related to probing issues. And realities of it.
Many users don't understand that when you connect probe it becomes part of the circuit and changes it. Even active probes, just to a bit smaller extent (if you chose right probe for the job). As frequencies go up, it gets much, much worse.
There is no perfect probe or a scope. It is all about education and knowing your instrument. Both weak and strong points. And working around those.

[hpw]

@2n3055: My point is, as it shows using the given probes (using shortest connections) gives false results (overshoots) for fast clocks.

So using a 1GHz FET probe using shortest connection trusted me. Also the phase behavior about is an opened task. IMHO.

50E basic measurement is one task, using probes the second task... and so on.

As the Topic on the given PDF: "Siglent SDS2000X Plus Bandwidth Discussion"... is IMHO a starting point.

Home workers may have a limited/reduced requirements, so to deal on real digital world required BW and Phase behavior to consider.

Even Siglent with the new DSO 6K model and 1..2 GHz FET probes does not provide the required information how the probes performs.

While currently I am missing a clear advice about required DSO BW & probes to measure more or less exact overshoots.

Ok thank you for clarifications..

It seems that you have missed previous few topics (they were some time ago) where it was discussed exactly that.

If you want to look at signals that have 200 ps edges you need 4 GHz scope and corresponding active probe.  That is only way you will get what you want: a pristine pulse response. Period 
Any time signal has edge that is faster than the risetime of the scope one of two things will happen:

1.) if scope has strong AA filters and aggressively cuts  of (lowpass filters) any frequency  that is above Nyquist by 40 dB or more (for a 8 bit scope) and it will show perfectly shaped squarewave with no overshoots but completely wrong, without any high frequency content and edges that will be as fast as a scope and not in any relation to what signal edges are. It will be complete lie, but it will look pleasing because it will look perfect. Which is complete bullshit because original signal is not perfect and 1/3 of energy content from that signal is missing.

2.) You can have response like most modern digital scopes have. That one will try to show as much of frequency content they can. If you feed that one with a signal that has slower edge than scope has, it will show it nicely without overshoots. It will show the same as scope from 1. point. But if you feed it signal that has very fast edge (like here where we feed signal with 40ps risetime to scopes with nominal risetimes from 500-800ps, 10-15 times faster), scope will show overshoots. You say that is bad. I say it is not. It's great. Scope is having problems and is not hiding the fact.

Those peaks are information that my edge is too fast for my scope and that I know scope is lying to me. Scope in point 1. will show clean edges without overshoot all the time, when I have nice signal and when I have too fast signal with many thing hidden inside. I wont know a difference ..

So with my digital scope, if I see that a signal has an edge that is 400 something ps (edge of what my 1GHz scopes can do) and I see overshoots, I know signal is too fast for my scope. If  I see the signal has an edge that is 400 something ps and there are no overshoots, I know it is right there at the edge. And what I see is true. And if I see 500-600 ps edge and overshoots I know it is DUT, not a scope.

If I routinely want to look at 100 ps edges I need to buy 4GHz+ scope...


So much about that.

As for probing, that has been discussed even more times. 
That is something that we all agree is a most important and most misunderstood topic in using scopes.
Most of the times quality of measurements is directly related to probing issues. And realities of it.
Many users don't understand that when you connect probe it becomes part of the circuit and changes it. Even active probes, just to a bit smaller extent (if you chose right probe for the job). As frequencies go up, it gets much, much worse.
There is no perfect probe or a scope. It is all about education and knowing your instrument. Both weak and strong points. And working around those.

Thank you for repeating the old good stuff  may your given content should be added to the PDF??!!

While was about an 20G/s Keysight DSO but the support question made the end   

[2N3055]

Thank you for repeating the old good stuff  may your given content should be added to the PDF??!!

While was about an 20G/s Keysight DSO but the support question made the end

Thanks!
Yes unfortunately Keysight seems to have gone the way of IBM. Instead of greatest T&M company in the world, they are now cloud/synergy/solution partner for emerging synergistic 5G solution providing electronics synergy solution T&M workflow optimizing profit solutions something....
T&M instruments are not a focus anymore, they are made just to enable consulting and support sales.

Basically, you used to buy a scope from them ( a primary product) and if you wanted some support you got some for free and extended support you had to pay for (additional revenue).

Now support, services and consulting is the primary product. Equipment is only there so they have an object as an anchor to sell you the services for it.
That will (and already has) mean that they will ignore all market segments (customers and products) that does not have potential for large volume of services and paid support.
That means all low and mid range equipment, small business etc etc..

[Performa01]

I've only very lightly touched the topic of probing in my document:

High bandwidth can also have its drawbacks. We need better probes alongside with better probing techniques,
better cables and connectors – which all sums up in additional cost and effort. This deserves more and more
attention at increasingly higher frequencies.

The article was not intended as a tutorial for high-speed probing, but to show the capabilities as well as the limits of a modern upper entry level instrument like the Siglent SDS2000X Plus series. We need to know what our tools can do for us in the first place, then we can look into a proper probing solution.

For me personally the 50 ohms probing is quite a natural thing. It is far better to design some 50 ohms test ports into the prototype in order to get easy access to the strategic important signals and limit the fiddling with probes to the less important signals that need not be accessed that often.

One goal was to establish some basic understanding what really dictates the required bandwidth – after all I've experienced "professionals”, also here in the forum, who seriously claimed that digital oscilloscopes are crap, because a 50 MHz square wave didn’t look square at all on a 100 MHz DSO…
(…and then there’s also the fact that this of all things would not be any different on their beloved old CRO)

Taking this one step further, the fundamental frequency is almost completely irrelevant if a close representation of the real signal is of any importance. This has been demonstrated by using a 10 MHz pulse, which requires at least 1 GHz bandwidth (a hundred times the fundamental frequency!), just because it has fast 500 ps edges. Of course, the spectrum extends to even higher frequencies, but everything below some -40 dBc becomes irrelevant – at least for an 8-bit DSO.

In the vast majority of applications, it should not matter whether a pulse overshoot is measured as 2 % or 1.5 % - or even 3 % for that matter, so the -40 dBc figure should give plenty of margin, even on a 12-bit DSO. We just don’t want to see a significant overshoot from the scope itself when fed with a signal spectrum within its bandwidth. Regarding my 500 ps tests, other than for 1 ns and above, I do not have a perfect signal available; all overshooting came from the signal source, as is confirmed by the reference measurement on the 2 GHz scope.

[mawyatt]

Something useful for those without a proper fast edge sub-nanosecond pulse generator is the inexpensive RF Signal Source based upon the ADF4351 chip. This is much better than the kludge avalanche transistor junction we were using before!! Someone else posted this on another thread which I can't locate, so maybe they'll spot this post and take credit. These are handy little sources than operate from USB, shown below being USB battery powered.

Here's the response of an "Enhanced" SDS2104X+, enhanced to 500MHz which also has the 500ps/div sweep extension. Note the screen capture is shown as SDS2504X Plus.

Best,

[blurpy]

Since there is some confusion about the model name, it's been discussed before: https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg3088181/#msg3088181

Performa01 likely has an old configuration file from before Siglent officially added 500MHz model name.

[trp806mo]

Now support, services and consulting is the primary product. Equipment is only there so they have an object as an anchor to sell you the services for it.
That will (and already has) mean that they will ignore all market segments (customers and products) that does not have potential for large volume of services and paid support.
That means all low and mid range equipment, small business etc etc..

+1 and more when you read the topic https://www.eevblog.com/forum/testgear/keysight-officially-lost-the-plot-dont-buy-if-youre-a-hobbyist/msg3865562/#msg3865562

Currently if I'm not able to spend at least 50K€ of"service/hardware" every year I have the same service as 10 years ago when I were a private person with a 1K€ old stuff.

[Johnny B Good]

There is no such thing as an SDS2504X Plus:

What does your scope identify itself as in the system menu? 

Unsurprisingly it says: "SDS2354X Plus". The model name in the screenshots also comes from the scope itself.

My SDS2104X+ with "features" say SDS2504X plus in system menu

Yes, AFAIK SDS2354X+ with the 500 MHz BW SDS2000XP-4BW05 option applied should display SDS2504X+ in the system menu.
Maybe other members need also confirm this.

 I upgraded my SDS2104X+ to the 500MHz BW option over a year ago (mid 2020), initially landing up with the serial number showing a series of xs in place of the original and I think the displayed model number showed as SDS2354X+ (I'd done the upgrade in steps) until  a couple of the forum members provided the required keycode to restore my 'scope's original serial number which also changed the displayed model number to correctly reflect the 500MHz BW upgrade (or not - see edit below).

 The xxxxx serial number was, iirc, a Siglent testing mode. If you've upgraded to the 500MHz BW but the model number isn't being displayed as SDS2504X+, I'd expect your 'scope would be displaying a series of xs in place of the original SN (or not - see below).

[EDIT]  I may be conflating the destroyed SN issue with a different issue to the one relating to the failure of the 500MHz BW update to correctly update the model number. This was seemingly a typographical error in a cfg file which Siglent corrected in a later version as reported by tv84 here:

https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg3088181/#msg3088181

[Performa01]

I had to replace the document one more time, see changelog below:

EDIT2 20220107: Document revised again:

• Misleading "Safe Signals" and "Conclusion" sections corrected - in half channel mode, the SDS2000X Plus can handle at least the 200 ps rise-time signals without artifacts.
  
• Quoted Conclusions in this posting adapted according to the revised document.
  
• Artifacts can be Gibbs phenomenon and/or aliasing, both caused by high frequency portions of the input signal that vastly exceed the Nyquist bandwidth of the DSO.
  
• Lower bandwidth models most likely do not have a digital filter.
  
• Version number added to the title page (V1.02).
  

[tv84]

Yes, AFAIK SDS2354X+ with the 500 MHz BW SDS2000XP-4BW05 option applied should display SDS2504X+ in the system menu.

Do these scopes display the 500 MHz option in the options menu?

As I've described in the scope's thread awhile ago:

It seems there was an initial batch of SDS2000X+ that contained an "incorrect" .cfg file (maybe on purpose, at the time...  ).

I find it no surprise that, as some other members here, Performa01 has got one of those. When this was discovered, some members corrected the thing by replacing the .cfg file with the newer one (or simply upgrading the FW with the recent .cfg file).

So, although Siglent doesn't show in its site the model SDS2504X+, the fact is Siglent from long ago started including that designation in the scope's config files (which is what most of the EEVBLOG users see in their system menu). Nitpicking: for me it would be  clearer/simpler if the document stated SDS2504X+ instead of SDS2354X+ with 500 Mhz option.

Attached are the 2 .cfg files (one from the old batch and the newer one) and their respective parsing so that you see their content's little difference.

BTW, nice document! 

[hpw]

@Performa01

Lets do an "HpW G/s Equation" to calculate the required DSO BW so handy for the sales team

DSO BW as : 20..50 / (5 x of fastest rise of a single slope 10..90%) [1/s]

 this will shows more IMHO or less the findings within the document...

as 500ps = 20 / (5 * 500ps) = 8 G/s

as 1ns    = 20 / (5 * 1ns) = 4 G/s

as 50ps = 20 / (5 * 50ps) = 80 G/s 

For the probes to use, may 1/2 of the DSO BW

About why 5 times:

1. the fastest slope

2. repeat 1 as second slope

3. & 4 as the top & bottom level time equal to 1 & 2

5. to compensate the other 0..10% and 90..100% time

so to get more or less 5 times the fastest slope as our signal frequency


So lets 

[Performa01]

Yes, scope manufacturers tend to recommend the use of an instrument that has only one tenth of the signal rise time.

Of course they do, because it will increase their income, as engineers will approach their bosses and say: "look, what Tek specifies here in this application note! So we need to buy a 100 GHz highest end DSO in order to properly look at our signals with 50 ps rise time".

This could easily backfire though, because bosses might reply: "ooops, do they even have such a beast? If not, we need to look elsewhere! Can you bring the LeCroy and Keysight catalogs?" 

Of course, it is the same thing everywhere. Buy a measurement solution that is ten times faster and/or ten times more accurate than what your DUT can deliver, and you should be able to trust all your measurements blindingly ... if you know what you're doing and how to avoid systematic errors due to probing or environmental effects, that is.

Yet there are situations, where we need to leave that comfort zone and make do with an instrument, that is only little faster and/or little more accurate than our DUT, or not even that – it is also an art to determine whether an instrument is up to the task. In any case we need to use a calculator (or even spreadsheet) in such cases.

We can calculate the rise time of e.g. the 500 ps pulse that is measured as 540 ps on the SDS6204 by taking the 230 ps rise time specification of the scope and do the math: the result is roundabout 490 ps. Close enough?

Even the result for the 200 ps signal, which is already faster than the scope itself, is still pretty close.

And with regard to the bandwidth, I guess no one can argue against the practice to check the spectrum and then decide that a scope covering all signals down to a certain level is sufficient for a certain application. Of course, if we want the maximum fidelity, we need to cover the dynamic range defined by the ENOB, which might be more than 45 dB on an excellent 8-bit system - and not that much more even in 12-bit high bandwidth systems.

[Performa01]

Only now I have found a supplemented version of my document, which for some reason didn't make it to the public back in January 2022. I have now updated the initial posting and the attachment.

[BillyO]

Nice document! 

There are  (quite) a few places where you refer to the SDS2504X-P as the SDS2304X-P, but otherwise I found it informative.


Thanks!