Sorry for quoting such a long & interesting post
Thanks to you Perfoma01 for your many "tech" post written in such a way my 10-yr-old nephew could understand (test ahead
)
Ah well, I saw this coming...
To really appreciate what’s going on, I’d like to recapitulate how a modern fully digital trigger system works. It is a completely separate processing chain, fed with the sample data stream of the trigger channel. This trigger system doesn’t care about the other channels and whether they are in use or not. Consequently, on the SDS5000X this data stream is always 2.5GSa/s with a 400ps sample interval. This makes it all the more remarkable when accuracy and stability down to single digit picoseconds can be obtained. Furthermore, the channel interleaving as well as the various decimation algorithms (normal, peak detect, average, Eres) are applied to the data channel exclusively at a later stage and do not affect the trigger path in any way.
Of course the objection that averaging also irons out some of the trigger jitter is valid, so I decided to setup one final test, using a highly stable sine wave instead of a pulse. This way we get rid of all the uncertainties with regard to the signal source and quality – and we get faster transitions too (at high frequencies).
Here is the same measurement as for the first test, but in normal acquisition mode, without averaging. Since I’ve already put away the original setup, skew is not exactly the same as in the first test.
1GHz sine wave, fed into channels 2 and 4 in parallel through 12.4GHz wideband power splitter. Some minor skew between channels because of subtle signal path length differences.
• Dots display mode
• Infinite persistence recording for >10 minutes to catch any jitter and noise effects.
(Attachment Link)
SDS5104X Sine 1GHz Norm Skew Pers10m NR
At 200ps/div, the visible jitter is still pretty much insignificant.
Let’s have a look at the automatic measurements again. We now have an error of 600fs for the trigger channel 4 and just 7ps pk-pk jitter for the measurement data. The standard deviation (RMS jitter) is <1.23ps.
The non-triggered channel 2 has a significantly higher pk-pk jitter of 15ps now, RMS jitter is <2.12ps.
As can be seen, the channel to channel skew is finally accurate now at some 15ps. The peak to peak variation of 18ps with a standard deviation of 2.4ps meets the expectation.
A very pleasing result in my book.
Yes, the trigger interpolator does a rigorous job and reduces noise and jitter to zero at the trigger point, but we can take the peak to peak jitter of channel 2 as a measure for the maximum trace width at 45° slope, and that’s only 15ps. If I interpret nctnico's screenshot correctly, then there the trace width is uniformly more than 100ps – but that might just be because of a poor quality signal source.
For comparison, here’s the 150MHz pulse from the SDG6052X again, no averaging, infinite persistence for more than 5 minutes.
(Attachment Link)
SDS5104X Pulse 150MHz Norm Skew Pers5m NR
Automatic measurements indicate the following:
-21ps Trigger time error, 18.5ps pk-pk and 2.6ps RMS jitter for the trigger channel 4.
-2.84ps Trigger time, 37.4ps pk-pk and 4.6ps RMS jitter for the independent channel 2.
Channel skew 18.2ps (almost spot on), 39ps pk-pk and 5.25ps RMS jitter between channels.
Even though the jitter values are much higher now, trace width at 45° slope is <38ps, which is more than twice as much than with a high quality sine source, but still a far cry from 100ps.
Setup you mentionned > a SDS5104X, and a nice wave gen plugged in.
Meas. from the scope :
we end up to an "answer" from the SDS5104X vs the input signal => "std dev" is <5ps, given persistance during x minutes etc...
Question :
how => "std dev <5ps" is relevant, given that we have a 1GHz BW scope ?
my basic answer would be => not relevant.
Be cool Performa01, this is not a complaint & any blablabla (I'm sure you're a meastro vs myself
), that's such a tech question to know if we are fooled by the scope figures
I'm posting this becauseI got a SDS2504X+ (although it has a sticker "2104X+" on it...
)
So, let's say I've a 500MHz, 2GS/s, MSO&DSO, and happy with it.
But I want more "fine details", time-domain only (MSO feat. I don't care), thus => jitter/eye diag. & so on analysis, on "high speed" signals (above 1GHz is minimum).
so... I went thru all post/topics about SDS6000A, specs of the SDS7000A series, plus any A-brand scopes...
=> after brainstorming like hell
=> I end up anyway to the Picoscope 9400 series !
Am I fool because there are no topic/post about these scope ?
Or,
are they a dead crazy BIG perf/price alternative to any benchtop scope on the market ?
(given that we deal with "repetitive" signal, thus "digital/clock/etc")
Conclusion :
when you have a SDS2504X+,
=> you end up to 9400 series from Pico (dea lower cost vs A-brands)
I missed smth ? or what...
Rgds