It is the difference of the clock rate of the DSO, and ofcourse related to the sample rate,
and the input frequency. So you can also measure the real clock frequency of the DSO
My DSO is about 3 Khz off at 2 Ghz, so that gives an error of less then 0,25 %

If you can, slowly change the input frequency just below till just above the multiple of the internal clock,
you can also see some beat frequency wave in the non zoom window.

And in zoom mode there can be only one trigger point, that why when 1 trigger point you dont see that.
Then the DSO can compensates for the phase differences.

It is the difference of the clock rate of the DSO, and ofcourse related to the sample rate,
and the input frequency.

Clock rate of the DSO? Frequency at which the signal is digitally processed or something like that. 
I edit to correct it, sorry.

All these test equipment has a master clock, on my DG4102 i use a Rubidium master clock,
and also for my HP equipment. And on most equipment is an external connection for a master clock.

It is a pitty that the DS2000 has no input for a master clock, that makes i more difficult
to use it in combination with other equipment. ( phase differences... )

It is the difference of the clock rate of the DSO, and ofcourse related to the sample rate,
and the input frequency.

@Wim: I don't see any difference in the variation based on the input frequency.

I just ran a test @ 100us/div [Zoom 500ns/div] testing different frequencies from 10kHz to 1MHz. The size of the jitter/offset is constant - exactly what I mentioned before - of plus or minus one sample period, i.e. in this case ±100ns (100MSa/s).

@Marmad, thats is corect, only the speed of the phase difference changes.
Because the DSO has only one reference.

So if you vary the inout freq. the speed of the phase diferences will change,
you can see that also, change the DSO screen you see a alias.

It is the difference of the clock rate of the DSO, and ofcourse related to the sample rate,
and the input frequency.

I would say that "the frequency of drifting depends on relation between input frequency and sampling clock", where the total jitter (in both sample clock and input signal as well) is responsible for the maximum drift amplitude. The minimum drift amplitude (well, that's the maximal resolution) is defined by the sampling rate.

Sorry, man, I don't think you're correct.

i can only recommend the R&S article to undertsand the difference between analog and digital trigger.

http://www.ecnmag.com/articles/2011/07/oscilloscope-performance-digital-triggering
http://www.rohde-schwarz.de/file/Benefits_of_RTO_digital_trigger_system_2.pdf

Compare your delayed trigger picture to the typical analog trigger jitter picture in the pdf, they look very similar.
I know that Rigol said "DS2000 have digital trigger", the question is how they implemented it.

Maybe someone can check this: the LMH6518 aux outputs, they routed somewhere. I think they routed to FPGA (the one over the ADC), i think to lvds inputs and their complementary inputs (of the FPGA) are routed to the DAC/MUX. That would be for analog or mixed trigger (levels analog, patterns digital). For pure digital (the implementation described by R&S) there should be no single signal path, however lmh6518 aux is for sure routed somewhere (but that didnt means anything, it can be used as well for hardware frequency counter).

So i can only repeat : we .. talking about ... the jitter added by trigger circuit, signal paths, FPGA design and potential firmware implementation. We don't know if Siglent have delayed trigger and how they measured, we don't know if they using exact the same design as Rigol, we know the FPGA are similar - but that only one less error source, we don't know exact used pins (which is as well good jitter source), we don't know hw they design looks like, we have no idea how rigol implemented the delay trigger (simply imagine the FPGA jitter x amount of delay elements and you will destroy any adavntage of digital triggering).

So i would say, check the aux routing, send Rigol nice email then with link to R&S pdf and ask them again about their trigger implementation.

For me this looks like analog level/digital pattern trigger and delayed trigger implementation with no errors (the smallest jitter amplitude is equal to sample rate, so that's ok - but as well proof of analog triggering). But maybe i'm wrong, maybe Rigol simply screewed up delay line and maybe they do have pure digital trigger and simply forgot to advertise that in DS2000 datasheet 

i can only recommend the R&S article to undertsand the difference between analog and digital trigger.

I understand well the difference between analog and digital trigger, have read the R&S article before, and have even posted the link to it two or three times on this forum already 

Compare your delayed trigger picture to the typical analog trigger jitter picture in the pdf, they look very similar.

But horizontal jitter is horizontal jitter - it all looks the same. The question here is whether the TRIGGER (or it's path) itself is the source of that jitter - or something further down the line of post-processing. If it's not the trigger - than analog or digital makes no difference (in the R&S block diagrams, analog and digital are the same past the "Trigger System" module). As I mentioned before, the Siglent has a clear offset in the single GIF that Herman posted - exactly like the offset (except bigger) as in my image of the Rigol attached below - no jitter in the image: just offset. So it seems as if they are "suffering" from a similar problem - we just don't know if the Siglent actually "jitters" at certain sample rates (or the extent of it's offset) because Herman didn't post any other info about it.

What I don't understand is this: on an analog scope, a delayed sweep involves ACTUALLY delaying the horizontal deflection amp for the B sweep - thus the reason for it's name (tradition). This involves a delay time comparator, gate, etc, - all of which could introduce jitter/error into the process.

On a DSO, correct me if I'm wrong, there is no actual delay of anything - the entire process is just done in firmware. The DSO just does it's normal capture of the main time base samples - and then extracts a subset of those samples for the "zoomed window". There is no extra comparator, gate, added circuitry, etc. - just software.

If this is true, what difference would the trigger type matter?

---

But horizontal jitter is horizontal jitter - it all looks the same.

yeah, unfortunately.

As I mentioned before, the Siglent has a clear offset in the single GIF that Herman posted - exactly like the offset (except bigger) as my image of the Rigol attached below - no jitter in the image: just offset.

offset is another thing, even on pattern or pure digital trigger one have still to adjust the hardware/firmware/software to display the triggerpoint in the middle of the screen. Some manufacturer doing this only based on hardcoded tables, others doing this based on selfcalibration. For me looks like Siglent simply adjusted wrong, easy to fix thing (remember the DS2000 500uV/DIV bug). When using zoom, there is maximal possible resolution, so depends on the implementation (calculated or hardcoded) there might be difference (and we should not forget as well rounding errors ... they very common)

we just don't know if the Siglent actually "jitters" at certain sample rates (or the extent of it's offset) because Herman didn't post any other info about it.

oh well, soon we will see a review(s), so what.

What I don't understand is this: on an analog scope, a delayed sweep involves ACTUALLY delaying the horizontal deflection amp for the B sweep - thus the reason for it's name (tradition). This involves a delay time comparator, gate, etc, - all of which could introduce jitter/error into the process.

On a DSO, correct me if I'm wrong, there is no actual delay of anything - the entire process is just done in firmware. The DSO just does it's normal capture of the main time base samples - and then extracts a subset of those samples for the "zoomed window". There is no extra comparator, gate, added circuitry, etc. - just software.

On a DSO there is always comparator or gate, no matter of you do this in hardware of firmware/software, e.g. lvds inputs, timer, delay line, counter - at the end you comparing values. And this in clocked circuit, so there will be jitter added as the FPGA and/or µC running own clock sources.

Even in pure digital way (like R&S RTO done in ASIC) where the trigger is calculated based on sampled data and with one clock source but multiple elements where their transition time variation, hysteresis and thermal noise are responsible for jitter as well (sure, at the end only fs jitter, but it is there).

what difference would the trigger type matter?

On a DSO, due the fact that the trigger is set/calculated in one (clocked or analog) path and the data is coming in second path there will be jitter. So it does matter trigger technology type (pure digital or semi-digital/analog). On DSO the trigger type (edge vs. delayed va. pattern) matters as well, as every path can be potential delay/jitter source.

Now we have two discussions,

1 phase jitter, because samples are taken at a certain frequency rate,
which gives a beat frequency in combination with the measured frequency.
Wave calculations, 1 freq + 1 freq= something between 0 and 4

2 trigger jitter, digital or analog comparator.
digital gives more jitter for the same reasons, depends on resolution of sampler.

Lot of jitter anyway, but dont mix there is already so much mixed up in the DSO

When using zoom, there is maximal possible resolution, so depends on the implementation (calculated or hardcoded) there might be difference (and we should not forget as well rounding errors ... they very common)

...which is what the jitter/offset seems like to me - calculation or rounding errors. There is no jitter in Normal sweep @ the same time base and sample rate - but there is when the DSO is triggering in exactly the same way - but extracting a subset of the captured samples for a "magnified" display - with an error of precisely ±1 sample (combined with Wim's 'phase jitter').

For me looks like Siglent simply adjusted wrong, easy to fix thing (remember the DS2000 500uV/DIV bug).

Complete speculation on your part - if Herman had posted GIFs of Delayed Sweep @ 500MSa/s, the Rigol would have also just seemed "adjusted wrong" - as in my posted image above. The truth of the matter is that the tiny amount of data released by him does not support any kind of hypothesis of why the Siglent is exhibiting the offset error - or any conclusions about analog versus digital trigger.

Don't get me wrong: the Siglent looks like a great upcoming DSO - but I think it's impossible to judge what error it might/might not exhibit over the range of sample rates in Delayed Sweep mode.

Can somebody explain why trace of channel 1 looks very different if channel 2 is on or off? Look at the pictures!

---

Maybe because it's only 1Gsa/s when capturing 2 channels?

Philipp

Maybe, I don't know. The over shoot with Rigot is also quite big. It should be under 3 %. Maybe the adaption with 50 Ohm feed through terminator is not very good.

Here is for comparison picture trom TEK TDS3032. Over shoot is 2.1 %.

---

There is always something like cross talk between channels.
To the book Rigol claims cross talk between ch1 and 2 better then 40 dB

Well on my DSO, at 200 Mhz, crosstalk is about 70dB,

See picture, ch2 is closed with only 50 ohm, nothing else.

---

Thanks Wim, it sounds reasonable!

In Pic-4 the average rise time is 1.187 ns and the average over shoot is 10.77 %. There I have used Rigols 1 MOhm 10x probe with 50 Ohm feed through terminator connected between the probe tip and generators output connector.

.... The over shoot with Rigot is also quite big. It should be under 3 %. Maybe the adaption with 50 Ohm feed through terminator is not very good.

---

Probe calibration?

Probe calibration?

His probe have HF compensation?

In mine, just when the relay switches. The difference is subtle, but the two internal compensation circuits are not exactly adjusted:

---

Can somebody explain why trace of channel 1 looks very different if channel 2 is on or off? Look at the pictures!

@EV
 Hi, Did you check the sample points?
 here are the samples @2ns with Chan 1 only and Chan 1& 2
also see the interpolation shown in pix3 of an 'Ideal' step, note the Sinx/x Overshoot

---

A touch of humor: 

---

@Teneyes
Hi, No I did not. I check tomorrow.

@EV
 Hi, Did you check the sample points?
 here are the samples @2ns with Chan 1 only and Chan 1& 2
also see the interpolation shown in pix3 of an 'Ideal' step, note the Sinx/x Overshoot

A touch of humor: 

That was good humor 

Here are the sample points.

@EV
 Hi, Did you check the sample points?
 

---