Your problems with them from the past are meaningless to this discussion, IMO - virtually every single company has made mistakes with some product at some point or another. The point is rather that you started this theory of yours in response to a MISTAKE that was posted in this thread by an owner (Fungus) about the way the DSO dealt with with sin(x)/x.
They are meaningful in questioning the credibility of Rigol and especially so since they have continuously made the same mistake for years and have not corrected it yet. Perhaps I am too strident at times.
I admit that I mistakenly thought Rigol was disabling sin(x)/x at low sample rates to conceal digitizer non-linearity and sampling jitter. Now I am suspicious that they are doing the same at higher sample rate to conceal the same problems when interleaving is used.
What other oscilloscopes disable sin(x)/x reconstruction at high sample rates where it is still useful? Oscilloscopes operating with equivalent time sampling do so because they can fill the entire waveform record with real sample points. These 1 Gsample/second real-time only DSOs certainly cannot do that.
It does NOT make sense for it to be disabled (or enabled) to avoid aliasing because sin(x)/x reconstruction neither causes nor increase aliasing. It merely makes it more apparent.
As has been mentioned before: for sin(x)/x interpolation to be accurate, you have to have an analog input signal that has no frequency content above the Nyquist frequency - which, when 3 or 4 channels are on, is 125MHz. The normal frequency response of the DS1000Z does not roll-off fast enough to minimize aliasing for sin(x)/x interpolation ...
I agree.
... - i.e. LINEAR interpolation should be used - or- to put it another way, there exists a good reason for being able to manually keep sin(x)/x turned OFF when 3 or 4 channels are on, if you need to. ...
The aliasing produced in the digitizer between an input signal which is completely below the Nyquist frequency and the sample clock occurs whether sin(x)/x reconstruction is used or not. If you were to collect the sample points over multiple acquisitions, they would show thickening of the waveform which is *not* caused by noise.
I have often wondered if DSOs which support some form of persistence and appear to show excessive noise actually suffer from this but that is a different discussion.
I also wonder if this is why they do not support equivalent time sampling with their digital triggers but that is also a different discussion.
... On the other hand, if you enable the 20MHz bandwidth limiter for each channel that's turned on, you can use sin(x)/x interpolation without problems on the higher sample rates
I agree but for a specific reason; with the bandwidth limit engaged or with a lower frequency signal, the mixing products between the signal and the sample clock produced by the non-linearity and sample jitter in the digitizer are *also* below the Nyquist frequency so no aliasing is produced.
I would love to link to a set of screen shots or videos showing if the aliasing problem exists or not in a Rigol DSO but I do not have one to test. I can show it on other (old) DSOs and in Agilent's application notes but that is not very helpful except to show that the problem exists in a general sense. Agilent pointed the problem out to distinguish themselves from Tektronix.
Again, where are these application notes? I want to see a document describing turning off sin(x)/x interpolation because of interleaving problems at the fastest real-time sampling rates. I can link to reams of literature about the problem of aliases in sin(x)/x interpolation, if you like.
When I first ran across this problem with my Tektronix 2440 which needed CCD calibration, Agilent had an application note describing the issue perfectly but I did not keep a copy and have not been able to find it again online. I did find a couple of others:
This LeCroy application note mentions the sin(x)/x reconstruction problem in connection with interleaving on page 15 with a Tektronix DSO. Agilent likes to pick on Tektronix about this as well:
http://cdn.teledynelecroy.com/files/whitepapers/wp_interpolation_102203.pdfThis Agilent application note discusses the interleaving problem and distortion from the digitizer itself as the source of aliasing when the input signal is completely within the Nyquist bandwidth. I think I linked this one earlier:
http://literature.cdn.keysight.com/litweb/pdf/5989-5732EN.pdfI even found a discussion thread on EEVblog with accompanying video which shows exactly the problem I described with a Rigol DS1102E. I do not agree with the analysis that I have read so far there but am still working through the discussion:
https://www.eevblog.com/forum/chat/rigol-ds1000e-series-possible-errorfail-in-sin%28x%29x-interpolation/At about 46 seconds into the video, it shows the results of sin(x)/x reconstruction on a sine wave that started out below the Nyquist frequency were higher frequency components created by mixing between the sample clock and the signal in the digitizer cause aliasing.
No company in their right mind is going to include as a reason for turning off sin(x)/x interpolation that it is to conceal aliasing made worse by interleaving done to increase the real-time sample rate.
Do you mean that no company will have published literature about this made-up theory of yours?
Well, they are certainly not going to advertise it in their oscilloscope specifications or documentation unless for the purpose of distinguishing themselves from their competitors and maybe not even then.