Understanding Entry Level Scopes? (Limitations & Uses)

[nctnico]

Better use RMS, peak-peak measurement + statistics for determining the noise floor.

[David Hess]

@rf-loop: You had details on the noise floor of the SDS1000X and SDS2000X series. There was a big difference between both series and the SDS1000X was better in some areas than the SDS2000X series, but then again there was an issue with the noise on the offset level.

Where did you get these details? From Siglent or by your own measurements?
Can you summarize these details once more with more details on how you obtained that information.

I have only seen details on the noise floor for Rigol scopes, but never details about the noise on the offset level and wonder how Rigol compares there against Siglent.

Maybe that information can also be obtained with hands-on measurements?

It would be good to have the comparison in place between Rigol and Siglent

I read that discussion in detail and did not find any practical information on noise performance.

The peak-to-peak and RMS measurement capability (and standard deviation - thanks nctnico) of these oscilloscopes should be able to directly measure their input noise accurately but for reasons unclear in that discussion thread, this was not reported.  Relying on the index graded display is not the way to do it even on an analog oscilloscope unless tangential measurement is used which I am pretty sure cannot be done on a DSO because of limitations on how the display is rendered.

The processing involved to produce a display from high record acquisition rates may alter noise measurements although if done correctly it should not.

[David Hess]

Even Siglent SDS2000X have 4 channel + Ext trigger (this Ext trig input is back side)

But, here need also tell that Ext trig performance is not like main channels trig.

In many new scopes trigger system is very different what previously.
There is two main working principles. Analog side pathway trigger system where trig circuit take signal from analog front end just before ADC and then some kind of comparator system produce trigger.

More modern way is full digital side trigger engine (like example in SDS2000X or SDS1000X and same in other brands) what give lot of more performance for many kind of complex triggering.
This can do only with fully digitized channel. Time ago it was very difficult and mostly only in expensive scopes. 

In this kind of  scope class if there is Ext trigger, it is afaik still  "old school" analog trigger with limited accuracy and limited features. Also acquisitions fine positioning (interpolation between sample points and adjust to trigger point) is far below good full digital side triggering system (yes it can also be good but this kind of high quality complex circuits are not typical in cheap oscilloscopes)

The old school analog triggers are usually limited by noise in the source rather than internal jitter and digital triggers have the same limitation.  The digital triggers produce nice displays showing zero trigger jitter (LeCroy used to make a big deal about this in their advertising) but it is an illusion; the jitter shows up before and after the trigger point.  Even when not limited by noise in the source, analog triggers still product less jitter than can be observed unless you are using a real sampling oscilloscope where you get down into the 10s of picoseconds.

The reason digital triggers displaced analog triggers is simply because once you have enough digital integration, they are cheaper and of course you also get the benefits of complex triggers but again, that is because of increasing digital integration.  Moore's Law does not apply to analog triggers.

Does the external trigger on the Siglent SDS2000X use interpolation or is its timing resolution limited by the sampling clock?  I would worry about the same problem when using MSO inputs for triggering; the SDS2000X datasheet says 500 MSamples/sec for the digital inputs implying 2 nanoseconds of jitter if they are used for triggering.  For the external trigger input if it is tied to the sampling clock, then I would assume 1 nanosecond of jitter.

Note: It took me 15 minutes to get this reply posted after writing it.  I hate this forum software.

[David Hess]

As long as we are posting photographs which reveal nothing about noise performance, here are a couple which illustrate a potential problem.

The 2440 and 2232 were used to observe the same set of signals from the startup of a Tektronix 7000 series switching power supply.  Both oscilloscopes are using peak detection which raises apparent noise.  The 2232 photograph (the horizontal position is deliberately shifted so as not to obscure the cursor readout) shows much less noise especially on the lower trace.  Guess what is going on here.

[nctnico]

I guess the 2440 samples at a much lower frequency but modern DSOs use the maximum samplerate which is at least a couple of times the bandwidth to do peak detect.

edit: make that the 2232 which samples slower.

[Jwalling]

Any chance it could be crosstalk between channels? (WAG)

[julian1]

As a beginner, I'm very happy with my Rigol 1054Z.

But I'm even happier I arranged my budget to purchase a sig-gen with it.

The combination allows me to think about and prototype circuits as devices consisting of inputs and outputs. The siggen is used to create the test input, while the oscilloscope provides visibility on the output.

[David Hess]

I guess the 2440 samples at a much lower frequency but modern DSOs use the maximum samplerate which is at least a couple of times the bandwidth to do peak detect.

I would consider these "modern" DSOs since they do the same thing; the maximum sample rate is used for peak detection.  The 2440 is sampling at 500 MSamples/second (per channel) and the 2232 is sampling at 50 MSamples/second (per channel) because both are in peak detect acquisition mode.  That zig-zag looking aspect of the 2232 display where the 2440 display is filled in has to do with limitations of the display controller in the 2232 and nothing to do with the signals.

No aliasing is present in either photograph and there is no crosstalk; it is very difficult to generate aliasing on these oscilloscopes when peak detection is used.  The critical part to look at is the upper trace before the 3rd horizontal division and the entire lower trace on the 2232.  Compare both to 1st horizontal division of the traces shown by the 2440 which is before the power supply is even turned on.  How can the 2232 traces be essentially noise free especially when peak detection is used?  They cannot be; something else is going on.

Update: Nothing is going on which has to do with sample rate.  No external noise sources were present; connections were coaxial to minimize noise pickup both from external sources and the power supply itself.  For all practical purposes, input noise was zero before the power supply was turned on.

[vk6zgo]

Which could be equally said for modern mid-range and high-end scopes.

But it doesn't make current bottom-of-the-barrel scopes any more "mid-range". They still are what they are, the cheapest (reasonable) scopes money can buy.

Indeed,but they are better in so many ways than the things Tektronix & HP tried to foist on us in the dawn of the DSO era.

Yes, but again, this is due to normal technical progress and cost savings through high integration and cheap labor, not because the modern variant of a low-cost scope is inherently the better product (it is only 'better' because of the technical progress, after all HP & Tek had to work with the technology that was available back then). I'd even go as far as to say that inherently, these old Teks and HPs were the better products, because they offered better reliability and maturity as well as support. But then, hardly any of the early DSOs were classed (or priced) as entry-level scopes.

When most Oscilloscopes were purchased by organisations rather than individuals,the  "Entry level" would have been that for small Radio/TV service shops,where something similar in performance to the BWD, or the Telequipment S31would have been (just) adequate.
The next level up would be that of Unis & Tech Schools.

Up from that would have been large Communications & Broadcasting organisations,which went from Tek 545s & the like,up through Tek 7000 series,to the last generation of Tek & HP Analogs. (mid level?)

In many cases,it wasn't the wider bandwidth which was important,so much as better triggering,Delayed timebases & so on.

In the late 1980s/early'90s,most people in this type of work were quite familiar with digital based equipment (for the youngsters reading--yes,there was digital equipment before microprocessors).

Perhaps for that reason,we expected more from the early DSOs we were offered,without considering that the design of such instruments was a lot more difficult than,say that of a digital video device which only had to operate 0-5MHz.

Alas,we were disappointed---the early DSOs were incapable of performing standard video tests,due to the large reduction in sample rate at PAL field rate frequency.
They were,for our purposes,useless pretty boxes,so we had no way to measure their reliability or support.

Modern DSOs could,in the main perform such tests,which although analog TV is gone,have their equivalents in other fields.

[rf-loop]

Even Siglent SDS2000X have 4 channel + Ext trigger (this Ext trig input is back side)

But, here need also tell that Ext trig performance is not like main channels trig.

In many new scopes trigger system is very different what previously.
There is two main working principles. Analog side pathway trigger system where trig circuit take signal from analog front end just before ADC and then some kind of comparator system produce trigger.

More modern way is full digital side trigger engine (like example in SDS2000X or SDS1000X and same in other brands) what give lot of more performance for many kind of complex triggering.
This can do only with fully digitized channel. Time ago it was very difficult and mostly only in expensive scopes. 

In this kind of  scope class if there is Ext trigger, it is afaik still  "old school" analog trigger with limited accuracy and limited features. Also acquisitions fine positioning (interpolation between sample points and adjust to trigger point) is far below good full digital side triggering system (yes it can also be good but this kind of high quality complex circuits are not typical in cheap oscilloscopes)

The reason digital triggers displaced analog triggers is simply because once you have enough digital integration, they are cheaper and of course you also get the benefits of complex triggers but again, that is because of increasing digital integration.  Moore's Law does not apply to analog triggers.

Do we now talk anything about same things.

Please read what is conventional (aka analog) trigger circuit principle in digital oscilloscopes and what is modern digital trigger.
It can read example here.(R&S RTO Oscilloscope's Digital Trigger)

Reason is not to be cheap. Reason is Improve performance over old ancient conventional and cheap analog trigger in digital oscilloscopes. This was time ago possible only in expensive oscilloscopes. Today (as you told we get more integration and lower cost for functions) we can have this also in lower end of oscilloscopes.
Still due to much more easy old  simple conventional analog trigger (in this R&S paper chapter 1. Conventional trigger)  is used in bottom level digital scopes. Good examples are Rigol DS1052E and example Siglent SDS1102CML etc.

Siglent SDS1000X/X+  SDS2000X use this principle what is in this R&S paper, Chapter 2: Digital Trigger


First advantage in digital trigger is that there is one signal pathway from input to ADC and acquisition data for display is exactly same what trigger get.  No different delay and frequency bandwidth. In conventional (analog) system there is two signal pathway after analog front end amplifier. One to ADC for sampling and one to trigger circuits. Trigger circuit may see very different signal what is signal after ADC. More explained in this paper.

About digital channels and triggering. What can interpolate if  have only and exactly two states 0 and 1.
There is not method to do oversampling and interpolate if there is not more information than sample clock and comparator result 0 or 1.  Not even in theory until there is some more information available. In digitized analog channel there is.

Sidenote: Also nearly end of this paper, R&S  explain also one more  thing what Siglent have in these X series models. (3.7 History View Function)

[pascal_sweden]

@rf-loop: Can you revisit this topic?

@rf-loop: You had details on the noise floor of the SDS1000X and SDS2000X series. There was a big difference between both series and the SDS1000X was better in some areas than the SDS2000X series, but then again there was an issue with the noise on the offset level.

Where did you get these details? From Siglent or by your own measurements?
Can you summarize these details once more with more details on how you obtained that information.

I have only seen details on the noise floor for Rigol scopes, but never details about the noise on the offset level and wonder how Rigol compares there against Siglent.

Maybe that information can also be obtained with hands-on measurements?

It would be good to have the comparison in place between Rigol and Siglent

I read that discussion in detail and did not find any practical information on noise performance.

The peak-to-peak and RMS measurement capability (and standard deviation - thanks nctnico) of these oscilloscopes should be able to directly measure their input noise accurately but for reasons unclear in that discussion thread, this was not reported.  Relying on the index graded display is not the way to do it even on an analog oscilloscope unless tangential measurement is used which I am pretty sure cannot be done on a DSO because of limitations on how the display is rendered.

The processing involved to produce a display from high record acquisition rates may alter noise measurements although if done correctly it should not.

[Wuerstchenhund]

When most Oscilloscopes were purchased by organisations rather than individuals,the  "Entry level" would have been that for small Radio/TV service shops,where something similar in performance to the BWD, or the Telequipment S31would have been (just) adequate.
The next level up would be that of Unis & Tech Schools.

Back then many of the scopes used by TV repair shops (in Germany usually some Hameg) could be found with many serious hobbyists as well.

Then there was a crop of even cheaper scopes coming from Eastern Europe (i.e. GDR, USSR) which were resold by Western electronics vendors under their own name for (back then) relatively low prices.

In the late 1980s/early'90s,most people in this type of work were quite familiar with digital based equipment (for the youngsters reading--yes,there was digital equipment before microprocessors).

Perhaps for that reason,we expected more from the early DSOs we were offered,without considering that the design of such instruments was a lot more difficult than,say that of a digital video device which only had to operate 0-5MHz.

Alas,we were disappointed---the early DSOs were incapable of performing standard video tests,due to the large reduction in sample rate at PAL field rate frequency.
They were,for our purposes,useless pretty boxes,so we had no way to measure their reliability or support.

Understandable considering your field (TV/video). Back then for us they opened up doors that no analog scope could, simply because we finally had proper means to digitize waveforms at increasingly higher sample rates for further processing. And the '80s did had some really decent DSOs for us, i.e. the Philips PM3320.

[rf-loop]

@rf-loop: Can you revisit this topic?

@rf-loop: You had details on the noise floor of the SDS1000X and SDS2000X series. There was a big difference between both series and the SDS1000X was better in some areas than the SDS2000X series, but then again there was an issue with the noise on the offset level.

Where did you get these details? From Siglent or by your own measurements?
Can you summarize these details once more with more details on how you obtained that information.

I have only seen details on the noise floor for Rigol scopes, but never details about the noise on the offset level and wonder how Rigol compares there against Siglent.

Maybe that information can also be obtained with hands-on measurements?

It would be good to have the comparison in place between Rigol and Siglent

I read that discussion in detail and did not find any practical information on noise performance.

The peak-to-peak and RMS measurement capability (and standard deviation - thanks nctnico) of these oscilloscopes should be able to directly measure their input noise accurately but for reasons unclear in that discussion thread, this was not reported.  Relying on the index graded display is not the way to do it even on an analog oscilloscope unless tangential measurement is used which I am pretty sure cannot be done on a DSO because of limitations on how the display is rendered.

The processing involved to produce a display from high record acquisition rates may alter noise measurements although if done correctly it should not.

I can not see any dedicated reason for this.

SDS1000X and SDS2000X works normally as designed and manufactured and related to noise I can not see any issue what need talk about. Both they are quite good in this class of scopes if we talk about noise. Also I do not understand what you mean that SDS1000X and SDS2000X have "big differencies" and then that in some things SDS1000X is better in some areas. What?
Only "better" in SDS1000X is that it have real 1mV and 500uV/div sensitivity when SDS2000X have 2mV/div sensitivity (+1mV/div derived digitally from 2mV/div - just as also some Agilent do) 

I can not see any problem issues there. Not in noise and not in offset setting. They works just normally as designed without issues in noise and vertical offset etc.

Totally other question is if we start talking about what is nice to have and how things can develop better. You know that also good can develop better, also exellent can develop better, just as also crap can develop better. Even state of art level things can develop better.

If want develop example SDS1000X analog front end better it can of course do. And Then we can talk about what are things what need firs develop better if start developing it. 1/f noise can reduce (what ever is source of this) and Offset setting resolution and accuracy can develop better. Specially when we talk about 0.5mV/div but also 1 and 2mV/div.  After these next can develop lower wide band noise. Also 50ohm impedance can develop better so that it is more accurately  50 ohm impedance over whole frequency bandwidth and bit over.  Then next frequency response can develop better, so that it is as flat as possible over whole BW but then soon after nominal freq  like "brick wall"  sharply down.

And so on and so on.

And please do not now convert these all what I say that these are some issues/problems in these scopes! 

[tautech]

If want develop example SDS1000X analog front end better it can of course do. And Then we can talk about what are things what need firs develop better if start developing it. 1/f noise can reduce (what ever is source of this) and Offset setting resolution and accuracy can develop better. Specially when we talk about 0.5mV/div but also 1 and 2mV/div.  After these next can develop lower wide band noise.

I started compiling some screenshots today to show this, I won't be posting them in this thread only the SDS1kX thread and just one unit (SDS1kX+) I tested showed zero trace offset after an Autocal but was out somewhat before the Autocal.
I'll pull at least 1 more new SDS1kX unit out for the same tests, doing my pre-sales checks at the same time. 

A SDS2304X will get the same treatment and those screenshots will will get posted in yet another thread.

So Pascal will get his much desired screenshots, he'll find them easy as they'll be in threads that he's already posted in. Got a bit on so they'll take 24-48 hrs. 

[nctnico]

Better post them here for comparison. There is no use otherwise because the other threads will become littered with comparisons and making a non-comprehensible mess.

[tautech]

Better post them here for comparison. There is no use otherwise because the other threads will become littered with comparisons and making a non-comprehensible mess.

I won't. Period.

If anything they'll be OT and more worthy of a separate thread titled DSO Baseline noise and Offset.
Threads like this turn into pissing competitions, any images I post will be statements of fact and best left in dedicated threads.
If any reader can't then open another tab on their browser and compare competing brand screenshots side by side what hope have they got attempting to use a scope? 

Members need always consider how they word or title any post they make so content can easily be found with forum searches.

[nctnico]

Creating a different topic with noise & offset crossed my mind as well and would be benificial as a reference. But perhaps it is an option to use the pictures posted in various other threads to compose a comparison in a new topic.

I'm not quite sure the offset is interesting though. Normally a self- calibration after a warm up period should set this to zero and unless the scope is used at a much different temperature I expect the offset to remain zero for a while (a couple of months to a year).

[David Hess]

On the 2232 what is going on is that in peak detect mode, processing is used to remove residual noise apparently in a way similar to how a noise gate works.  The reason this is relevant to modern DSOs is that many include the same type of thing making self measured noise comparisons difficult even without the issued involved in interpreting the display. 

Tektronix documented this feature on the 2232 but it was easy to miss in the manual unless you were already aware of it.  It can be easily disabled in the user interface with two button presses, which is how I first became aware that it was in the manual somewhere, in which case it produces results which accurately match the 2440 and most other DSOs, or at least the ones which do not process the hell out of the data.  I first noticed it later after the photographs were taken; had I noticed immediately, I would have taken a third set for comparison purposes.  It is not a bad feature, works incredibly well, and is occasionally useful especially for documentation purposes, but I am glad Tektronix included the ability to disable it because it can be misleading.

So if you want to measure DSO noise, use the built in peak-to-peak, RMS, and standard deviation measurement capability; if they do not work on noise, then they are broken.  Then make sure they return consistent results with different operating modes, number of channels, and vertical and horizontal settings.  If you are ambitious, check them against an external noise standard.  If you have to, break out a dual channel analog oscilloscope (look up tangential noise measurement) or RMS sampling voltmeter to make a good RMS noise measurement of your noise standard for comparison purposes.

Also see if the input coupling mode affects noise with and without an external short since so many DSOs now misleadingly do not have actual ground coupling even though it is listed in the coupling menu.  I suspect this "feature" in DSOs is deliberate or if for economic reasons, then deliberately not documented, to make their noise performance look better.

[nctnico]

It is the first time I have heard about a noise gate in a DSO. If the 2232 is using a CCD for sampling it would make some sort of sense though because CCDs are very noisy. I doubt they use noise gates on modern DSOs which use a fast ADC. The amount of processing required would be substantial.

[David Hess]

Please read what is conventional (aka analog) trigger circuit principle in digital oscilloscopes and what is modern digital trigger.
It can read example here.(R&S RTO Oscilloscope's Digital Trigger)

I have read that paper before and it is easy to take out of context.  At 500 MHz and below, jitter from a properly operating analog trigger is insignificant.

The performance considerations discussed in that paper apply to a 4 GHz 10 GSample/second DSO.  A typical analog trigger even 30+ years ago supported 2 to 20 GSamples/second equivalent time sampling which requires trigger jitter in the 500 to 50 picosecond range and it is not difficult to achieve a jitter of 10 picoseconds.  This level of jitter is not significant at the bandwidths these triggers are designed to support.

Reason is not to be cheap. Reason is Improve performance over old ancient conventional and cheap analog trigger in digital oscilloscopes. This was time ago possible only in expensive oscilloscopes. Today (as you told we get more integration and lower cost for functions) we can have this also in lower end of oscilloscopes.
Still due to much more easy old  simple conventional analog trigger (in this R&S paper chapter 1. Conventional trigger) is used in bottom level digital scopes.  Good examples are Rigol DS1052E and example Siglent SDS1102CML etc.

The advantage of digital triggers include:

1. With increasing integration, they are cheaper than analog triggering.
2. They can operate in real time whereas analog triggers have relatively long holdoff times.
3. They support exotic trigger modes.

But for cheap oscilloscopes, the overriding reason is that with increasing integration, digital triggers are effectively free while analog triggers require dedicated hardware.  If analog triggers were cheaper, than the cheapest (real) DSOs would be using them.

First advantage in digital trigger is that there is one signal pathway from input to ADC and acquisition data for display is exactly same what trigger get.  No different delay and frequency bandwidth. In conventional (analog) system there is two signal pathway after analog front end amplifier. One to ADC for sampling and one to trigger circuits. Trigger circuit may see very different signal what is signal after ADC. More explained in this paper.

I have never even seen this come up as a problem although based on reports I have heard, it might have been an issue with certain manufacturers like HP.  You might be surprised at how dissimilar the vertical and trigger signal paths can be before it affects anything.

It gets weird when you consider analog trigger designs which have *much* higher or lower bandwidth than the vertical signal path which is common with sampling oscilloscopes but some analog oscilloscopes have this as well.  The only time I have seen it make a difference is when the trigger source has an unusually high but narrow bandwidth which a low bandwidth trigger cannot even see.  A couple of times, I have used a 500 MHz 7B92A timebase in my 100 MHz 7603 mainframe just so I could externally trigger on a narrow digital strobe which the normal 100 MHz trigger could not reliably act on.

About digital channels and triggering. What can interpolate if  have only and exactly two states 0 and 1.
There is not method to do oversampling and interpolate if there is not more information than sample clock and comparator result 0 or 1.  Not even in theory until there is some more information available. In digitized analog channel there is.

Most 2-channel DSOs apparently digitize their external trigger input avoiding this problem and I would consider them completely broken if they did not.  So the only difference between their vertical inputs and external trigger input may be lack of signal conditioning and acquisition memory storage for the external trigger input.  This implies that the difference between a modern 2-channel DSO with external trigger input and a 4-channel DSO is less than the manufacturers charge for.

In the past there were logic analyzers which digitized their digital inputs so could in theory implement precise trigger timing.  The Tektronix TLS216 "Logic Scope" at 2 GSamples/second has a specified timing resolution of +/- 100ps so I assume it did this.  Apparently logic analyzers like this are invaluable in some applications.

http://www.tek.com/datasheet/tls216
https://www.eevblog.com/forum/testgear/tektronix-tls-216-16-ch-logic-scope-any-experiances-with-that-tool/

[nctnico]

In the past there were logic analyzers which digitized their digital inputs so could in theory implement precise trigger timing.  The Tektronix TLS216 "Logic Scope" at 2 GSamples/second has a specified timing resolution of +/- 100ps so I assume it did this.  Apparently logic analyzers like this are invaluable in some applications.

AFAIK most current high end logic analysers from Tektronix and HPAK work this way. The one from Tektronix I have samples the inputs at 8GHz which is then processed into sampled data. This allows to define precise setup and hold timing for a synchronous clock.

Still in a DSO you'll need to do some trigger interpolation to overlay subsequent acquisitions precisely otherwise the trace will jitter on the screen. 800x480 is a commonly used screen resolution. When using 600 horizontal pixels for the graticule at 10ns/div and 1Gs/s each pixel will account for 10ns/600/10 divisions=166ps. A 1ns jitter in the trigger will move the signal 6 pixels from left to right. You'll see that for sure.

[David Hess]

Still in a DSO you'll need to do some trigger interpolation to overlay subsequent acquisitions precisely otherwise the trace will jitter on the screen. 800x480 is a commonly used screen resolution. When using 600 horizontal pixels for the graticule at 10ns/div and 1Gs/s each pixel will account for 10ns/600/10 divisions=166ps. A 1ns jitter in the trigger will move the signal 6 pixels from left to right. You'll see that for sure.

That is something I have wondered about on DSOs which use a digital trigger but never had an opportunity to test.  What happens with a bandwidth limited edge?  Aliasing should move the trigger point around.