I will quote this (below) here for continue this here about Siglent SDS2000X HD model wfm speed and segment speed many things. (so that this do not make OT in other thread) Normal mode wfm/s speed is not so important but also it have some things good to know including also normal history mode (what really is not suitable for segmented acquistions without enough knowledge about its limits). Its burst-type acquisition operation is good to understand, then the user can perhaps avoid some traps.
But Sequence mode is very different.
There is also good to know absolutely trusted limits
for know that wanted triggering event does not drop out from acquisition. The average speed is often almost insignificant due to the nature of the needs for which fast sequence mode is used.
One simplified example about trap. Lets think there is 100ns pulses in continuous stream 1ms -2ms period (100ns pulses freq 0.5-1kHz) some people may think it is easy peace of bisquit and think he capture every these pulses using normal mode and after example one minute stop and look every pulse and its time position from history. Then he is watching time stamps for looking delta time between pulses. And his test goes soon to garbage when he fid there is something weird. Now one tell him there is missing pulses, no one turn on red light there is missing pulses. And there can be, due to oscilloscope (if DUT do not have missing pulses, how he know which one is quilty). Every time pulse occurs in this time when acquisition keep pause this pulse is missing from history - of course.
Sequence mode is different and better tool for this. But with it, we need know its limits with different t/div scales and channels in use. Only then we can be sure that these events we want capture are really captured and not dropped out. (even more when Memory management is not in Auto mode but user set "fixed" sample rate or "fixed" memory.)
It is too heavy work to do table for all setup combinations so user need somehow test it for his dedicated purpose. If can not find ready data for it. Things can also change some amount by new FW versions.
In this said example case it do not drop out pulses until maximum amount of user defined segments are captured.
ETA: Testing this is very time consumpting. Specially if want find even somehow accurate guaranteed maximum limit and so that it is reliable.
I use method where amount of segments is maximum and then pulse generator generate pulse bursts where one burst include exactly same amount of pulses as is segment count. This is repeated enough long time for find if even in rare case single pulse dropped. It is possible it capture 100 or 1000 sequence perfectly but then in one sequence there drop is missing one pulse. This can easy do so that generator run in burst mode and automatically repeat this burst.
Other oscilloscope follow tested oscilloscope trigger out. Main window so that whole sequence in inside screen. Zoom window just display last pulses and display persistence infinite. If pulse speed is too high it can detect from 2'ns oscilloscope screen.
Other possible method is 2 channel counter. Counter count burst pulses and oscilloscope trig out pulses. If after burst, trig out is less than pulses in burst game is clear. Speed is too high. This is not full guaranteed speed of sequence. Drop burst pulse frequency until count match every time... when go to near maximum it may need run hundreds of sequence or even more until we can write it is guaranteed speed what is small amount less than last clear long run test.. In normal mode, the structural and programmed fact of such an oscilloscope is that the wfm/s rate is an average with some peak value, it operates in burst mode. Trigger...save...trig..save...etc and then break this burst and do screen update (captures stacked to display) and after then repeat new burst cycle. Now the situation can be that inside the burst a speed can be example so that it can capture new wfm every 10us. But then it keep pause, what can be example 2ms. During this time, the oscilloscope will not start a new acquisition if trigger condition is met. This potential new wfm just went down the drain.
In the fast Sequence recording, on the other hand, there is no pause during the Sequence. Nor does the oscilloscope do anything extra to enable the fastest possible re-triggering and acquisition. Fastest trigger interval is 2us in this mode. But it is true only when 1 channel is in use and t/div is 50ns/div. Segment length is in this case 1 kpts. These kind of segments can be in one Sequence max 80000.
-----------------
Here the table with the wfms/s.
As mentioned before this can´t be the normal mode on the rigol.
Ultra and "Normal" are looking very similar, the blindtime between the packages are greater in the ultra mode.
Second part of my summary will following in the evening..
Martin
In Sequence mode there is important to measure guaranteed maximum speed. This what matters.
In Sequence mode we need trust what is minimum trigger interval when not any single trigger event do not drop out! (reason is, imho, obvious)
With 50ns/div SDS200XHD is most fast.
Its guaranteed maximum speed is (my data) ~511,5 ksegment/s
How to test it. You need burst signal. (I have measured it using 80000 pulse burst and segments in oscilloscope sequence 80000. And other t/div / memory values using maximal amount of segments in sequence.
Set scope for 50ns/div, 1k memory length. No matter if dots or vectors and if interpolation is Sinc or x.
Set sequence for 80000 segments
Then without signal start oscilloscope Sequence (now it is waiting trigger)
Start burst. As long as it capture all 80000 signal signal speed is below sequence acquisition guaranteed max speed.
Set now busrt (example pulses, freq 568180kHz as is value in your table)
Somewhere before 80000 it stops acquisition (and waiting more pulses) because some pulses are dropped out , not captured.
Now drop this burst frequency until it can reliable capture every single pulse. Repeat it over ten times until you are sure it never drops any single pulse (segment).
I can say guaranteed maximum speed is 510 ksegment/s (perhaps tiny bit over but there need be some marginal if promise something)
Then about normal mode wfm/s speed. In your table it looks like there is peak value what exist inside acquisition cycle burst.
When we talk wfm/s speed we need handle it using average speed (this we can use example for propability calculations for glitch hunting etc)
SDS2000XHD maximum average wfm/s speed is 98kwfm/s (display mode dots, 50ns/div, single channel in use and trigger just basic edge)
Peak value inside acquisition burst is just bit over 108kwfm/s as is in your table. This is not wfm/s speed it is peak value.
I have measured average speed using HP53131A (with very accurate reference).
Here below is part from my old tests (it is still "preliminary" before full table ready and revised)
Then yesterday and today I did some check about these my values and I did not find errors... (naturally values are bit rounded (down)) :
ETA: Later now after carefully look my collected data... there are some inexplicable exceptions what need further tests and with some different test method. Specially errors are in guaranteed max speed of sequence what is not displayed in this image.
During normal wfm/s test display looks this below.
Next image below. SDS2000X HD Trig out is connected to SDS2000X Plus. Here can see how it looks like.
Same trig out out signal what go to SDS2kX+ input is splitted and go to HP53131A input and there is used 3s gate time for get value what can see in my table.
Using these kind of fast universal counters need use enough long gate time because signal is not continuous wfm. Also when test other wfm speed with other settings counter Auto trigger leveling may lead to very weird results >> manual level set.
I will not comment on Rigol's results. But the things said earlier about Siglent also raised some small doubts about the Rigol numbers.
@Martin72
What is your SDS2000X HD HW version?
So lets continue here. (also I will some day finish or least expands and completes the table I presented above, which has been in the preliminary state for a long time)