That is right; the 11403A shown is not a sampling oscilloscope. Those vertical amplifier plug-ins are not sampling heads; they are an improved version of the 7000 series vertical amplifier plug-ins.
Except for the 11A81 3GHz "presampler" vertical plugin.
I was not aware of the details of this plug-in although I knew it existed. It is pretty cool but the specifications given in the 1993 catalog on page 79 are not consistent with a sampling plug-in.
What happened is that the mainframe's interface connector cannot handle a 3 GHz bandwidth (1) so Tektronix implemented the attenuators, vertical amplifier, and a sample and hold in the plug-in. It is not a sampling input in the sense of a sampling oscilloscope and this can be seen in the specifications which show a 10ns overdrive recovery time and a 25 volt peak maximum input voltage.
The older 7854 with the 7T11A sampling timebase works in the same way with the mainframe's digitizer slaved to the sampling plug-in so Tektronix was aware of this possibility.
(Does anyone know the presampling rate?)
I could not find it in the sparse to non-existant available documentation for the CSA404, 11403A, and 11A81. It should be only limited by the sampling rate of the plug-in itself because the bandwidth between it and the mainframe is 1 GHz leaving plenty of time for settling between samples. That would make it 20 MS/s maximum just like any other vertical plug-in.
(1) It was a pretty big deal when Tektronix extended the 7000 series interface to 1 GHz in the 7104 and there were trade articles written at the time about how they did it.
I also has 11801 with SD24 sampling heads. This is 20 GHz bandwidth, but 20 M sps.
The 11K and SD specifications all say 200 kS/s maximum. I think the chief limit in these single sampling gate designs is the operating life of the avalanche pulse generator.
If the marketing departments of other companies want to confuse the issue, that is up to them.
It's not confusing the issue, and it's not even driven by marketing.
This is what Keysight has to say:
Oscilloscopes fall into two groups, real-time oscilloscopes and sampling oscilloscopes (also called equivalent-time oscilloscopes) and it is important to understand the difference between the two types. Real-time oscilloscopes digitize a signal in real-time. Imagine a repetitive AC signal - the real-time oscilloscope acts like a camera, taking a series of frames of the signal during each cycle. The amount of frames the real-time oscilloscope captures depends upon the bandwidth, memory depth, and other attributes that we will soon discuss. A sampling oscilloscope, on the other hand, takes only one shot of the signal per cycle. By repeating this one shot, but at slightly different time frames, the sampling oscilloscope can reconstruct the signal with a high degree of accuracy.
It is confusing and here is why. Sampling oscilloscopes in the sense that I am using the term have three features which are not found in other oscilloscopes:
1. They have zero overload recovery time. A sampling bridge does not care about what is going on during the time that the sampling gate is not open. A practical application of this is measuring settling time to high accuracy which is somewhere between difficult to impossible with a normal oscilloscope which otherwise has sufficient bandwidth.
2. They use a sampling efficiency much lower than 100% to exceed the RC bandwidth limitation of a sample or track and hold which uses a sampling gate time greater than its RC time constant to allow full settling. This results in a huge increase in bandwidth.
3. 2 above produces a non-linear sin(x)/x frequency response do to integration during the sampling gate time. The effective sampling gate width can be measured by looking for the null in the response.
Both 1 and 2 above lead to sampling oscilloscopes not using linear elements like amplifiers or delay lines before the sampler. Either would significantly lower the achievable bandwidth and the former would destroy the overload recovery capability.
Funny you mention PicoScope, as it seems they, too, disagree with your definition of Sampling Scope:
https://www.picotech.com/download/manuals/picoscope-9000-series-questions-and-answers.pdf:
A sampling oscilloscope is a special type of oscilloscope that uses a technique called sequential time sampling.
...
And in the very first sentence, Picoscope gets it wrong. There is nothing about a sampling oscilloscope which requires sequential sampling. Indeed, random sampling in a sampling oscilloscope allows viewing the triggering waveform without the use of a delay line or pretrigger. The only reason they were not more common is that most applications could live with these limitations.
As I said, I understand where you're coming from but there's no point to fixate of a some 30 years old definition by a single manufacturer while constantly referring to ancient kit that has long been obsolete when the rest of the world has moved on.
The point is that your definition confuses two different classes of oscilloscope which have distinct properties. If there is a better and less ambiguous term for sampling oscilloscopes in the older sense, then let us know.
I have posted few 11800/CSA803A frontend photos elsewhere but they are so beautiful they need to be here too. They don't care how they are called.
Leo
2 channel 20GHz sampling head
Note that the coaxial line on the top carries the sampling strobe from the mainframe to the sampling gates and the two coaxial lines on the bottom carry the divided sampling strobes back into the mainframe for timing measurement. This is required to achieve high horizontal timing accuracy. None of the coaxial lines carry the sampled signal which is low bandwidth after the sampling gates.
The real advantage of these scopes is in their low price compared to a real-time scope with the same BW.
I like them because they are immune to overload and have a predictable frequency response allowing them to be used to calibrate the transient response of other instruments. It is even possible to calibrate the transient response of a sampling oscilloscope without an external reference although it takes 3 sampling oscilloscopes to do this. Maybe I should have included this in my list of features above.