Old analogue scopes beat new digital scopes, for bit-bashed digital outputs

[tggzzz]

Having come from a 10 MHz scope prior to my Tek 485, I had to come up with some tricks to figure out how to display things that were beyond the bandwidth.

One approach that always works is to slow down the clock.  Get the system to work and then see what happens when the clock is restored. Does it still work?

When peripherals and system operation allow it, and the source code is available allow, yes those are good standard techniques.

Another approach is to code simple loops over the offending peripheral.  Even with the 485, I can't see an SPI transaction that occurs once per second with a clock of 10 Mhz.  It's done and gone before the trace is even visible.  Loops help.  So does a DSO with single shot mode.

Yes indeed; the killer use-case for digitising scopes. Analogue storage scopes always were a pain in the rear end

XMOS seems cool, I might find an application for it some day.  It's a beginner level device in the same way an FPGA is a beginner level device.  It depends on where you are beginning from.  There are some folks who majored in hardware design in grad school...

The XMOS devices are much much easier to "get into" than FPGAs. If you want to do two things at once in an MCU, then thr XMOS devices are also easier than learning an RTOS.

The OP has been around here a long time (over 5,000 posts) so it seems a little disingenuous to not know that the DS1054Z is easily, and properly, unlocked to 100 MHz.  It's not a hack, it's not hardware changes, it is simply selecting a different analog filter that is known to exist.  A feature designed into the hardware so the same device could be sold with two different bandwidths.  He knows this but chooses to ignore it.

I am aware of it, but it is non-standard and I don't have such a hacked scope. I cannot test it to see its effectiveness or otherwise.

And, yes, I expect the Arduino to be the uC of choice for beginners for a very long time.  There's a ton of projects on the Internet and a lot of support groups.

Yes, the ecosystem is of critical importance - that's why the 80x86s are still dominant.

[tggzzz]

Now, there's something that hasn't yet been mentioned but which is quite important.  And that is that it's probably not an accident that entry level microcontrollers are relatively slow.

<snip> 

For these reasons, I don't expect beginner territory to change all that radically even if microcontrollers can be faster.

We already have uCs that are very much faster than the Arduino.  The Blackfin is about 600 MHz and it isn't even in the running for FAST.  But you're right, they aren't for beginners.  The learning curve is high and steep.

Yes, and interesting.

I was very pleasantly surprised at how easy it is to create things in the XMOS world. That is partly because the theoretical concepts (CSP) are so simple and obvious - and usable in practice.

From downloading the development environment to the first version of my software-only frequency counter (2*6MHz input frequencies) with the results being displayed on the host PC's screen was 1-2 days effort. I wouldn't have been able to do that any faster on the Arduino. Add another couple of days learning about the hardware and debugging mechanisms, experimenting and tweaking, and it is up to 2*15MHz.

What the Arduino has going for it (as the Basic Stamp had before the Arduino came along) is reasonable performance and a tremendous support group.  Programming the device is made easy by the library code and there are thousands of example projects on the Internet (no, I didn't count them!).  OTOH, the Parallax Basic Stamp is probably the best documented uC ever sold.  Parallax has wonderful documentation.

Experienced users kind of chuckle at Arduino users but that Arduino is a heck of a lot faster than the IBM 1130 I programmed back in '70.  And it was state of the art for entry level minicomputers.

Unless that is a major shift in thinking, I see the Arduino as the beginner's CPU of choice for a very long time.

Agreed.

When I returned to embedded programming via Arduinos, I was simultaneously satisfied and depressed at how little had changed in 20-30 years. The XMOS devices are definitely a change for the better, both hardware and software.

[tggzzz]

I DON'T agree that the analogue scope is so useful for signal integrity (as regards rare phenomenon). Because there can be occasional and super rare events such as glitches/transients/metastability etc which can flip bits (to incorrect states), when they occur. But may be very difficult to detect with analogue (non-storage tube and/or digital) oscilloscopes.

Yes, but digitising scopes don't make that any easier. The problem is finding a suitable trigger/detector, and then waiting!

With a digital scope (with sufficiently powerful enough capabilities), it can be triggered (one-shot), when the MCU (or FPGA or whatever), detects bit errors, bit collisions, etc.
These may be extremely rare (e.g. metastability), and take many hours to occur (trigger).
Nevertheless, these issues need to be sorted, in order for the hardware to work properly, in most cases.

Just so.

The top MCU (that the OP seems to be using), can go to a claimed 4000 MIP's. I don't know its clock frequency, but that COULD be 250 pico seconds (4 GHz). So I can imagine the OP, with a signal mostly at 0, but at 1 for just 250 pico seconds, with him trying to display it on a 50 MHz oscilloscope. Then creating another thread, complaining about why he can't see his 250 pico second signal, on his borrowed 50 MHz oscilloscope.

I don't know how the XMOS marketing department came up with the 4000MIPS! They've mentioned "dual issue", but I don't think the StartKit processor cores have that.

However the XMOS devices can up to 32 cores each running at ~100MHz. Given that they have effectively embedded CSP in hardware (think of CSP as removing the need for an RTOS), all cores run independently and simultaneously except when they synchronise to exchange information over a "channel". The result is you can dedicate a separate core to controlling each peripheral; when there's nothing to do the core simply halts - and then restarts with latencies in the nanoseconds. FFI see XMOS literature.

The I/O runs at up to 250Mb/s.

There are some very good reasons why, analogue oscilloscopes (brand new), have partly or fully disappeared from the market place.

Yes indeed, but that doesn't mean that decent working analogue scopes should be ignored. There is a gap, constantly changing, where they are a very viable option.

[kcbrown]

yup except that when you expect a 50-100MHz 1GSps scope to show good fidelity on 1-2ns rise/fall time, then its just ridiculous.

It would indeed be ridiculous.

But for reasons best known to yourself, you have chosen to ignore paragraph 4 in the first post "The first point...". Hence your point is a strawman argument.

Yes, but if you're talking about a modern microprocessor (hey, those were your words, not mine  ), then you're talking about a GHz-level device.  Even your 350MHz analog scope isn't going to be capable of dealing with one of those.

That's incoherent technobabble; clearly something has been lost in the translation.

I think we're all in violent agreement on your main point: choose the right tool for the job.

What we're not in agreement about is whether or not your example shows that an analog scope with higher bandwidth is a better choice than a digital scope with lower bandwidth for a beginner to use with digital signals.  Yes, the analog scope can theoretically show the problems with the signal integrity of your example, but in practice it will be difficult (very difficult, I'd wager) for a beginner to use an analog scope for that purpose unless the glitch itself is very repeatable (i.e., shows itself many, many times per second).   Put another way, for digital signals, the analog scope is probably not the right tool for that job, even if it might have enough bandwidth.

Absent those very specific conditions that make the analog scope workable at all, the real takeaway here is that if you are working with signals where the pulse rate is in the tens of MHz (more precisely, where the bandwidth requirements of the signals are high enough that the bandwidth or sampling rate required to see glitches would exceed what the scope is reasonably capable of capturing), then you need a scope with greater bandwidth and sampling rate than the entry-level scopes that are typically recommended.  And guess what?  It also means that you have to start worrying about probing techniques and other things that most beginners haven't even considered.

But if you're really working with such signals, chances are high you are not a beginner!  Or, if you are, you're probably (way) over your head, because if your signals are that fast, you'll probably have to go to extra lengths to avoid signal integrity issues in the first place, lengths that you wouldn't even have to think about with hardware typically used by beginners, and you'll have to go to extra lengths to ensure that, firstly, your probing methods aren't introducing glitches into the signal and, secondly, that they are appropriate for the signals you're trying to evaluate.

What this means, of course, is that one of the recommendations that must be given to beginners is that they work with relatively slow hardware, precisely because doing that would allow them to concentrate on things other than how to avoid the problems introduced by the transmission line characteristics of signal carriers -- and would allow them to use breadboards and other convenient prototyping mechanisms, to boot.

[MK14]

I DON'T agree that the analogue scope is so useful for signal integrity (as regards rare phenomenon). Because there can be occasional and super rare events such as glitches/transients/metastability etc which can flip bits (to incorrect states), when they occur. But may be very difficult to detect with analogue (non-storage tube and/or digital) oscilloscopes.

Yes, but digitising scopes don't make that any easier. The problem is finding a suitable trigger/detector, and then waiting!

They make it easier, because they can capture a single (erroneous/suspicious) event. As long as you can set up a suitable trigger. E.g. Failed CRC32 comms error, sets a port pin, which is used to trigger the digital oscilloscope.
Hence you can potentially see the glitch/transient/metastabilty etc, which caused the bit error(s) in the high speed communications stream.

Although there did use to be analogue storage scopes, which could do a similar function. Assuming their limited write speed was acceptable and the resultant image fade, would last long enough (e.g. it would probably reduce in focus/intensity/readability, after about 10 minutes or so, maybe less, maybe longer, in my experience) to see the error.

The top MCU (that the OP seems to be using), can go to a claimed 4000 MIP's. I don't know its clock frequency, but that COULD be 250 pico seconds (4 GHz). So I can imagine the OP, with a signal mostly at 0, but at 1 for just 250 pico seconds, with him trying to display it on a 50 MHz oscilloscope. Then creating another thread, complaining about why he can't see his 250 pico second signal, on his borrowed 50 MHz oscilloscope.

I don't know how the XMOS marketing department came up with the 4000MIPS! They've mentioned "dual issue", but I don't think the StartKit processor cores have that.

However the XMOS devices can up to 32 cores each running at ~100MHz. Given that they have effectively embedded CSP in hardware (think of CSP as removing the need for an RTOS), all cores run independently and simultaneously except when they synchronise to exchange information over a "channel". The result is you can dedicate a separate core to controlling each peripheral; when there's nothing to do the core simply halts - and then restarts with latencies in the nanoseconds. FFI see XMOS literature.

The I/O runs at up to 250Mb/s.

These XMOS devices, seem very interesting. Some of what you have described (e.g. accurately switching port pins, on a timed basis), can be done with modern timer hardware, on many MCU's. These timers have their limits (e.g. one or two transitions, then they need new values from the MCU), so being able to do it in real time, on a dedicated core (if you can affordably have 8/16 or 32 cores, hence some can be dedicated for the hardware I/O).

Interrupts on conventional MCUs, take time. So the very high speed nature of these XMOS devices, potentially allows them to perform well, in some problem areas. E.g. very high performance PWM applications, with complicated capabilities.

There are some very good reasons why, analogue oscilloscopes (brand new), have partly or fully disappeared from the market place.

Yes indeed, but that doesn't mean that decent working analogue scopes should be ignored. There is a gap, constantly changing, where they are a very viable option.

A good point.

Analogy:
Linear bench power supplies (although still potentially available brand new), are being increasingly replaced by modern switched mode power supplies. So if you want a linear bench power supply (because of its potential for low noise/accurate outputs, simplistic/easier repairability and are use to them, from older, past experience). Your best bet, might be to buy a nice used one on ebay (or elsewhere).

So I agree, a decent used analogue scope, is useful for some people.

[snoopy]

The bandwidth of this Rigol scope is only 50 MHz and your analog scope is 350-400MHz ? That in itself a huge discrepancy !!

Of course it is! The point is that a £85 old scope has a much more faithful measurement (in significant respects) than a new £275 scope. While that's never the only consideration, it is significant.

Good luck to you if you find your analog scope more useful for measuring protocols and counting bits. You'd be the exception to the rule.

I don't know what you mean by "measuring protocols and counting bits". That scope (and many others like the Tek 2xxx ranges) measures time/frequency/voltage.

When I want to interpret the analogue voltages as a character or number, no, I certainly don't use a scope for anything more than a trivial quick look. If I want to look at digital information then I use digital domain tools: logic analyser or printf statements.

In this case I would be very surprised if a protocol decode correctly interpreted the signal in the second post. And, quite frankly, it shouldn't be able to interpret it correctly since it appears to be an invalid digital signal.

Most beginners aren't going to be decoding serial or SPI buses running at 50MHz. Most of the time they will be running at 1MHz so even that old Rigol will be more than adequate for the job.

Also whilst you might pickup up one of those old analog scopes cheap if it doesn't work it can be a nightmare to fix and a huge repair bill as the old custom chips are no longer available and have to be scrounged from other scopes. It's really false economy to say that an old analog scope is good value just because it is cheaper than a new one.

[David Hess]

yup except that when you expect a 50-100MHz 1GSps scope to show good fidelity on 1-2ns rise/fall time, then its just ridiculous. not to mention invalid setups to prove a point where in one measurement, ringing (on the falling edge) is much more prominent in the said DSO. here look again... with the right mind we should expect some illusive aliasing (wave superimpose) on the signal packet... signal dis-integrity due to questionable measurement setup.

I would expect it to display a waveform consistent with a 50 MHz bandwidth limit.  The 1 GS/s sample rate is plenty high.

What aliasing?  And at a 1 GS/s sample rate with a 7ns transition time?  The 2465 shows a flat waveform after the last edge and a 50 MHz oscilloscope should have shown the same thing with a 7 nanosecond fall time on the trailing edge.  If there was not a problem with the DSO or experimental setup, then I would suspect a problem with the interpolation algorithm, digital frequency/phase compensation, or "bandwidth enhancement".

Right, I can understand your point.  The only thing I am suggesting is that there may be reasonably priced vintage DSOs that would also offer some pretty big advantages over an analog or low end DSO.

I can think of several vintage DSOs like that and I have several of them.  I do not usually recommend them because they are less reliable and more difficult to maintain than analog oscilloscopes of the same age simply because of their extra complexity.

Do they have any advantages over modern budget DSOs?  They can be very inexpensive.  Some have full documentation.  They usually (always?) make measurements on their processing record unlike several modern instruments and many Rigol models.  They often have "retina" level display resolution making up somewhat for lack of index grading.

I actually find it difficult to believe that the 50 MHz DS1052E (or any oscilloscope) produced that result unless it was broken; to me it looks more like a probe problem.  tggzzz, your original post says that you used a 1.5GHz Z0 (low-z?) probe but could you describe the probe in more detail? How was it connected to the DS1052E? Was it through the needed feedthrough termination? It should not matter but was the same feedthrough termination used on the 2465 or was its internal termination used?  Could the DS1052E have been damaged before the test?

A sensible doubt, sensible observations, and sensible questions.

Probe: HP10020A.
50ohm termination: Greenpar BNC inline/feedthrough which ought to be good to at least a GHz. I haven't seen such effects when I've used them in the past, but that is a weak statement.
Rigol damage: I've borrowed the Rigol, but I believe it hasn't been used much. I've no reason to believe damage, and both channels are the same.
2465: internal termination.

Like you I think feedthrough-vs-internal termination "should not matter", but that is undoubtedly an unnecessary extra variable. I'll check it at the earliest opportunity (not tonight).

Unfortunately I haven't got any active probes to use as a comparison, and I certainly wouldn't trust my high impedance passive probes with a 1ns edge rate.

A low 50MHz bandwidth will conceal any differences between the terminations unless one of them is very broken so comparing them on my 2465 will be more than sufficient.  Still, it is something I would check and I often break out my PG506 pulse generator (super clean roughly 500ps rise/fall time) to verify that whatever probe and oscilloscope combination I am using is working correctly.

Active probes include more things to go wrong so your x10 HP10020A is probably ideal if its 500 ohm input resistance is acceptable.

Although there did use to be analogue storage scopes, which could do a similar function. Assuming their limited write speed was acceptable and the resultant image fade, would last long enough (e.g. it would probably reduce in focus/intensity/readability, after about 10 minutes or so, maybe less, maybe longer, in my experience) to see the error.

I have a Tektronix 7834 and its write speed is plenty fast to support the needed bandwidth to find glitches and metastability which is exactly what I use it for.  It was much less expensive than than a modern DSO of equivalent bandwidth and acquisition rate.

There are some very good reasons why, analogue oscilloscopes (brand new), have partly or fully disappeared from the market place.

What is odd is that they have not; they are still produced for the lowest end of the market.  Above that they cannot compete simply because of economics even if you ignore the difference in features.  It is *cheaper* to produce a DSO because of the specialized non-scalable technology required for high performance analog oscilloscopes while DSOs can take advantage of increasing semiconductor integration.

[MK14]

Although there did use to be analogue storage scopes, which could do a similar function. Assuming their limited write speed was acceptable and the resultant image fade, would last long enough (e.g. it would probably reduce in focus/intensity/readability, after about 10 minutes or so, maybe less, maybe longer, in my experience) to see the error.

I have a Tektronix 7834 and its write speed is plenty fast to support the needed bandwidth to find glitches and metastability which is exactly what I use it for.  It was much less expensive than than a modern DSO of equivalent bandwidth and acquisition rate.

You are perfectly CORRECT. (I.e. I should have clarified my post above, by saying the historical time periods, when the write speeds were not that good, so needed alternatives, if you needed to see faster signals).

What it is (I think), is that at some points in time in history (when these oscilloscopes and similar, were the current state of the art), e.g. 1968 (take the exact time period with a big pinch of salt). The analogue storage scopes (e.g. Tektronix), had somewhat limited write speeds (according to their own documents, if I remember correctly).

So because of the limited write speeds, for some applications (probably things like very early TTL 74 logic, and a bit later, very fast ECL logic).
Tektronix used alternative techniques (I think it was photomultiplier crt tube technology), to allow engineers to see much higher speed signals.
E.g. The Tektronix 485 (if I remember correctly).
It was not a storage scope as such, but had ways of getting round it, (persistence of screen phosphor, or similar/something), so that one off events, COULD be seen, even if briefly.
I can't remember if it was one-shot or needed repetition to work well.

Later (in history), it seems, they produced storage oscilloscopes, which could work at much higher write speeds.

Anyway, sorry, if any of the above is a bit wrong. It is because I am trying to remember stuff, and maybe getting it wrong.

There are some very good reasons why, analogue oscilloscopes (brand new), have partly or fully disappeared from the market place.

What is odd is that they have not; they are still produced for the lowest end of the market.  Above that they cannot compete simply because of economics even if you ignore the difference in features.  It is *cheaper* to produce a DSO because of the specialized non-scalable technology required for high performance analog oscilloscopes while DSOs can take advantage of increasing semiconductor integration.

Now you mention it, I remember seeing/hearing about those. They seem to be just for the education (and maybe hobbyist) market. But these are very simple/basic oscilloscopes, which are not really what the OP has in mind, for measuring very high speed pulses etc.

[1audio]

Tek had pretty good storage scopes even in the late 1970's. The faster ones had a mode where the screen "shrunk" to be able to write faster one shot events. And then there was always the camera which used with the special phosphor could capture very fast one off events. Uncle Sam used variations on this for measurements of nuclear tests. In the 1970's and 1980's when I was bringing up new computers and chips the camera and the scope hood were essential. A 4 trace scope and knowing what to trigger it from was all you had but you could make it work with lots of squinting into the dark. And computers like CDC 7600 Cray's and IBM mainframes were pretty fast.

With any scope you need to know when its lying to you. With an analog scope the problems are pretty easy in recognize and don't need complex analysis to identify. A DSO has a number of tools to help see complex and difficult phenomena BUT that means many ways to "lie". Not an issue if you know what you are looking at. If you don't (which is the usual reason for getting out a scope) having several different ways to see or not see what you are looking for can lead to mistakes and problems.  A lot can be attributed to learning curve. Switching between DSO's makes for complex learning curves, especially when the scope is very complex and time spent looking through menus.

A real pet peeve is that today's SOTA scope has a short service life, low resale and are almost unrepairable. Watching a $120K investment become a boat anchor because a motherboard/OS upgrade can't be done is not good (unless you are the manufacturer). Meanwhile my 1990's era Tek 7854 (same class of product) can still do a lot of useful stuff.

[MK14]

then there was always the camera which used with the special phosphor could capture very fast one off events

That I think, is EXACTLY the part that I was trying to remember.

When the writing speed of analogue storage scopes was NOT fast enough. The special phosphor capture + camera, was the preferred method.

These days, digital storage scopes, easily capture, even extremely fast, one-off events. Putting such methods (analogue storage and/or using cameras) into the history books.

I think the earliest oscilloscopes, were just pens, (rotated with electric motors) paper and electro-magnetic movements. So that the oscilloscope wrote directly the trace, onto paper. Obviously only at extremely low frequencies, then (a very long time ago).

https://en.wikipedia.org/wiki/Oscilloscope_history

[1audio]

CRT scopes date back to the 1920's I believe. GR was there pretty early on. Tek pioneered the triggered sweep and the ability to actually measure values from a scope. Before that you used a scope for observing waveform shapes but not much else.

The challenge with really fast scopes (20 to 50 GHz) is storing the samples. At those sample rates is not a simple task.

[David Hess]

What it is (I think), is that at some points in time in history (when these oscilloscopes and similar, were the current state of the art), e.g. 1968 (take the exact time period with a big pinch of salt). The analogue storage scopes (e.g. Tektronix), had somewhat limited write speeds (according to their own documents, if I remember correctly).

So because of the limited write speeds, for some applications (probably things like very early TTL 74 logic, and a bit later, very fast ECL logic).
Tektronix used alternative techniques (I think it was photomultiplier crt tube technology), to allow engineers to see much higher speed signals.
E.g. The Tektronix 485 (if I remember correctly).
It was not a storage scope as such, but had ways of getting round it, (persistence of screen phosphor, or similar/something), so that one off events, COULD be seen, even if briefly.
I can't remember if it was one-shot or needed repetition to work well.

Later (in history), it seems, they produced storage oscilloscopes, which could work at much higher write speeds.

Anyway, sorry, if any of the above is a bit wrong. It is because I am trying to remember stuff, and maybe getting it wrong.

You got it mostly right but I have detailed notes. 

First Tektronix used the brute force method of increasing the CRT acceleration voltage which produces a brighter and sharper display with a faster writing rate.  People who have not used an oscilloscope like a 485 or 7904 are probably not even aware of this aspect of their performance which applies to both high frequency *and* low repetition rate signals.

Next was analog storage CRTs with transfer meshes although I rarely use the transfer mode on my 7834; variable persistence is almost always fast enough for me.

Then the 7104 with its microchannel plate image intensifier came out which can produce a single shot mark one eyeball visible trace at a writing rate faster than the speed of light.

1969   7704 150MHz 24kV
1972   485 350MHz 21kV Portable
1973   7704A 200/250MHz 24kV
1973   7904 500MHz 24kV
1977   7834 400MHz Reduced Scan Transfer Storage
1979   7104 1GHz MCP
1987   2467 350MHz MCP

[MK14]

CRT scopes date back to the 1920's I believe. GR was there pretty early on. Tek pioneered the triggered sweep and the ability to actually measure values from a scope. Before that you used a scope for observing waveform shapes but not much else.

The challenge with really fast scopes (20 to 50 GHz) is storing the samples. At those sample rates is not a simple task.

The history, gets complicated.

You mentioned the 1920's. Which is potentially perfectly correct, I'm not necessarily disagreeing, here.

But I think oscilloscopes (or similar), go back way older than that. But it then becomes more and more debatable, as to what exactly constitutes an "oscilloscope".

E.g. if you look at Wireless World from 1911, on the bottom of page 8. It shows an oscilloscope screen, along with its complicated image. I DON'T know specifically how they did it, it may have been some kind of camera, and valve/tube-early-partial-crt phosphor thingy or something.

Feel free to take a look (sorry, 1911 was the earliest wireless world I could easily find on the internet).
http://69.195.158.203/Archive-Wireless-World/MARCONIGRAPH/Marconigraph-1911-04.pdf

[ebastler]

E.g. if you look at Wireless World from 1911, on the bottom of page 8. It shows an oscilloscope screen, along with its complicated image. I DON'T know specifically how they did it, it may have been some kind of camera, and valve/tube-early-partial-crt phosphor thingy or something.

Feel free to take a look (sorry, 1911 was the earliest wireless world I could easily find on the internet).
http://69.195.158.203/Archive-Wireless-World/MARCONIGRAPH/Marconigraph-1911-04.pdf

That picture probably shows a recording from a mirror galvanometer which projects a light beam onto photograpic film, not from a CRT.  Such devices were used around that time -- essentially improved, faster chart recorders.

[MK14]

E.g. if you look at Wireless World from 1911, on the bottom of page 8. It shows an oscilloscope screen, along with its complicated image. I DON'T know specifically how they did it, it may have been some kind of camera, and valve/tube-early-partial-crt phosphor thingy or something.

Feel free to take a look (sorry, 1911 was the earliest wireless world I could easily find on the internet).
http://69.195.158.203/Archive-Wireless-World/MARCONIGRAPH/Marconigraph-1911-04.pdf

That picture probably shows a recording from a mirror galvanometer which projects a light beam onto photograpic film, not from a CRT.  Such devices were used around that time -- essentially improved, faster chart recorders.

Some sources, seem to say that the CRT oscilloscope, comes from 1897, when its crt was also invented.

E.g.

Karl Ferdinand Braun
Karl Ferdinand Braun was the German physicist who shared the Nobel Prize for Physics in 1909 with Guglielmo Marconi for the development of wireless telegraphy. Braun is also known as the developer of the CRT oscilloscope. He demonstrated the first oscilloscope in 1897, after work on high frequency alternating currents.

Source1:
http://theinventors.org/library/inventors/blcathoderaytube.htm

I have seen other sources about it was well, on similar lines.

E.g.
Source2:
https://en.wikipedia.org/wiki/Karl_Ferdinand_Braun

In 1897 he built the first cathode-ray tube (CRT) and cathode ray tube oscilloscope.[2]

[MK14]

What it is (I think), is that at some points in time in history (when these oscilloscopes and similar, were the current state of the art), e.g. 1968 (take the exact time period with a big pinch of salt). The analogue storage scopes (e.g. Tektronix), had somewhat limited write speeds (according to their own documents, if I remember correctly).

So because of the limited write speeds, for some applications (probably things like very early TTL 74 logic, and a bit later, very fast ECL logic).
Tektronix used alternative techniques (I think it was photomultiplier crt tube technology), to allow engineers to see much higher speed signals.
E.g. The Tektronix 485 (if I remember correctly).
It was not a storage scope as such, but had ways of getting round it, (persistence of screen phosphor, or similar/something), so that one off events, COULD be seen, even if briefly.
I can't remember if it was one-shot or needed repetition to work well.

Later (in history), it seems, they produced storage oscilloscopes, which could work at much higher write speeds.

Anyway, sorry, if any of the above is a bit wrong. It is because I am trying to remember stuff, and maybe getting it wrong.

You got it mostly right but I have detailed notes. 

First Tektronix used the brute force method of increasing the CRT acceleration voltage which produces a brighter and sharper display with a faster writing rate.  People who have not used an oscilloscope like a 485 or 7904 are probably not even aware of this aspect of their performance which applies to both high frequency *and* low repetition rate signals.

Next was analog storage CRTs with transfer meshes although I rarely use the transfer mode on my 7834; variable persistence is almost always fast enough for me.

Then the 7104 with its microchannel plate image intensifier came out which can produce a single shot mark one eyeball visible trace at a writing rate faster than the speed of light.

1969   7704 150MHz 24kV
1972   485 350MHz 21kV Portable
1973   7704A 200/250MHz 24kV
1973   7904 500MHz 24kV
1977   7834 400MHz Reduced Scan Transfer Storage
1979   7104 1GHz MCP
1987   2467 350MHz MCP

That is quite amazing, the "faster than the speed of light", link! 

It is quite fascinating, reading the history, of how Tektronix, developed their various oscilloscopes, over the years.
Their equipment, had a real quality to it, and usually exceptional qualities (for the time period).

They were the Rolls Royce or BWM of the oscilloscope world, for a very long time.

I also very much admire HP's (who later changed their name) early history as well. But HP oscilloscopes, (in at least my mind), were not up to Tektronix oscilloscopes, in various respects.
But much of HP's other stuff was brilliant, for the time, just not so much Oscilloscopes.

A bit like Fluke vs Brymen, but maybe that is a bit unfair.

I'd probably love to have and play around with one of Tektronix's early valve/tube oscilloscopes, such as the 547. But their horrific size, weight and other impracticalities, FORTUNATELY stop me from getting one. Keeping it working, may also be difficult as well.

[Fungus]

yup except that when you expect a 50-100MHz 1GSps scope to show good fidelity on 1-2ns rise/fall time, then its just ridiculous. not to mention invalid setups to prove a point where in one measurement, ringing (on the falling edge) is much more prominent in the said DSO. here look again... with the right mind we should expect some illusive aliasing (wave superimpose) on the signal packet... signal dis-integrity due to questionable measurement setup.

Bandwidth issues aside; the first question I'd ask a 'beginner' who showed me that screenshot would be, "Are you using the little spring on the probe?"

[Mechatrommer]

yup except that when you expect a 50-100MHz 1GSps scope to show good fidelity on 1-2ns rise/fall time, then its just ridiculous. not to mention invalid setups to prove a point where in one measurement, ringing (on the falling edge) is much more prominent in the said DSO. here look again... with the right mind we should expect some illusive aliasing (wave superimpose) on the signal packet... signal dis-integrity due to questionable measurement setup.

I would expect it to display a waveform consistent with a 50 MHz bandwidth limit.  The 1 GS/s sample rate is plenty high.

1GSps will sample every 1ns, i'm suspecting risetime is quicker than 7ns (that you saw in the Tektronix 2465, both scopes in the OP is somewhat "BW limited or low passed" during transition). with sub ns risetime added with BW limited acting in the scene, you'd expect only few funny sampled points available for transition representation. granted with poor sinc(x) implementation of the rigol dso, pretty much it can be anything. keep in mind shannon theory, if signal contains freq element higher than 2X sample rate, no digitizer in this world will be able to contruct the original signal correctly.

What aliasing?  And at a 1 GS/s sample rate with a 7ns transition time?  The 2465 shows a flat waveform after the last edge and a 50 MHz oscilloscope should have shown the same thing with a 7 nanosecond fall time on the trailing edge.  If there was not a problem with the DSO or experimental setup, then I would suspect a problem with the interpolation algorithm, digital frequency/phase compensation, or "bandwidth enhancement".

look again closely in the OP pictures (do i need to photoshop to highlight it?), in 2465, there is no ringing after falltime of the 7th peak, otoh in rigol you can see low freq ringing after it (on the low side) that should not be there (about same rate as the data rate), interpolation algorithm cant just simply fabricate that, because those are part of sampled points. with that ringing, we can hypothize that the same ringing occured after the falltime or long period peak (0xFF data?) and got superimposed with the subsequent data (0b10101010101010)... as said this is only hypotesis, the truth is who knows what? to ensure whats really going on, the OP need to do either (1) try to terminate the ringing by changing/improving probing method (the OP managed to probe correctly using 2465 but failed on 1052 which is a missing part mystery) or (2) use higher than 1GSps DSO or higher than 0.45/risetime BW, which is somewhere 450MHz scope, thats bare minimum, for the purpose of greater clarity of signal integrity, he should use 1GHz scope or 10GSps DSO or greater. imho.

[David Hess]

I also very much admire HP's (who later changed their name) early history as well. But HP oscilloscopes, (in at least my mind), were not up to Tektronix oscilloscopes, in various respects.
But much of HP's other stuff was brilliant, for the time, just not so much Oscilloscopes.

The HP oscilloscopes I have studied strike me as having better circuit design.

I'd probably love to have and play around with one of Tektronix's early valve/tube oscilloscopes, such as the 547. But their horrific size, weight and other impracticalities, FORTUNATELY stop me from getting one. Keeping it working, may also be difficult as well.

I have a 547 and they are another one of those unique models being the last generation before scan expansion meshes became ubiquitous.  (1) So they have an unusually sharp trace without needed a high acceleration voltage and they also have no ghosting from a scan expansion mesh allowing a higher contrast display than later CRTs.

There is a story that the Tektronix Grass Valley Group which produced television products returned their 7704s in favor of their obsolete 547s because of this.  They could not discern the VITS (vertical interval test signal) on the fuzzier 7704 CRT without using the delayed sweep.

Bandwidth issues aside; the first question I'd ask a 'beginner' who showed me that screenshot would be, "Are you using the little spring on the probe?"

I considered that but I doubt tggzzz would make that mistake and the rising edge just before the horizontal center shows no sign of this.

(1) A scan expansion mesh allows for a shorter CRT and lower voltage CRT amplifier which is important for solid state output stages.

[Fungus]

Bandwidth issues aside; the first question I'd ask a 'beginner' who showed me that screenshot would be, "Are you using the little spring on the probe?"

I considered that but I doubt tggzzz would make that mistake and the rising edge just before the horizontal center shows no sign of this.

There's clearly something wrong with the probing, the way the wiggly bit gets taller as it progresses is also suspicious.



The simple solution is:
a) Output five 1's instead of four to see your peaks.
b) Don't try to characterize 150MHz signals using a 50MHz 'scope.

The 1001 advantages of a DSO over an analog boat anchor don't justify this thread.

(even if it weren't a troll, which it is - his refusal to output five 1's demonstrates that)

[Mechatrommer]

I considered that but I doubt tggzzz would make that mistake and the rising edge just before the horizontal center shows no sign of this.

probably the ringing of the high side got cancelled by the longing ringing of the low side before that? or impedance difference/mismatch between low and high output? who knows? making hard conclusion from limited range measurement is not convincing...

[Fungus]

impedance difference/mismatch between low and high output?

This. The chip isn't driving low/high equally.

[nctnico]

I considered that but I doubt tggzzz would make that mistake and the rising edge just before the horizontal center shows no sign of this.

probably the ringing of the high side got cancelled by the longing ringing of the low side before that? or impedance difference/mismatch between low and high output? who knows? making hard conclusion from limited range measurement is not convincing...

The way the signal looks (slow amplitude rise and ringing when the signal goes away)  definitely points to something wanting to oscillate so there might be a problem in the probing solution used.

[David Hess]

1GSps will sample every 1ns, i'm suspecting risetime is quicker than 7ns (that you saw in the Tektronix 2465, both scopes in the OP is somewhat "BW limited or low passed" during transition). with sub ns risetime added with BW limited acting in the scene, you'd expect only few funny sampled points available for transition representation. granted with poor sinc(x) implementation of the rigol dso, pretty much it can be anything. keep in mind shannon theory, if signal contains freq element higher than 2X sample rate, no digitizer in this world will be able to contruct the original signal correctly.

If the analog bandwidth of the Rigil DSO is 50 MHz, then at 1 GS/s I expect about 7 sample point on every edge because the transition time cannot be faster than 7 nanoseconds.  And whatever reconstruction filter is being used should place every sample point on the reconstructed trace although we have seen before that Rigol is bad about this and maybe that is related to what we see here.

look again closely in the OP pictures (do i need to photoshop to highlight it?), in 2465, there is no ringing after falltime of the 7th peak, otoh in rigol you can see low freq ringing after it (on the low side) that should not be there (about same rate as the data rate), interpolation algorithm cant just simply fabricate that, because those are part of sampled points. with that ringing, we can hypothize that the same ringing occured after the falltime or long period peak (0xFF data?) and got superimposed with the subsequent data (0b10101010101010)... as said this is only hypotesis, the truth is who knows what? to ensure whats really going on, the OP need to do either (1) try to terminate the ringing by changing/improving probing method (the OP managed to probe correctly using 2465 but failed on 1052 which is a missing part mystery) or (2) use higher than 1GSps DSO or higher than 0.45/risetime BW, which is somewhere 450MHz scope, thats bare minimum, for the purpose of greater clarity of signal integrity, he should use 1GHz scope or 10GSps DSO or greater. imho.

I was thinking that the burst rate synchronized with distortion produced by the reconstruction filter.  That would explain why the initial part of the burst is lower amplitude and low amplitude ringing was produced after the burst but for this to happen, the reconstruction filter would need to be 40 sample points long.  Could the reconstruction filter do this while not distorting the longer pulses?  Does these low bandwidth Rigol DSOs use a bandwidth enhancement filter?

Reconstruction filters which include bandwidth enhancement are common in high bandwidth DSOs but I have never seen one on a 300 MHz or slower DSO.  It would be fun to make an amplitude versus frequency plot of this DSO and see if there is excessive ripple and if it has a single pole gaussian response producing a 0.35/risetime bandwidth.

I considered that but I doubt tggzzz would make that mistake and the rising edge just before the horizontal center shows no sign of this.

probably the ringing of the high side got cancelled by the longing ringing of the low side before that? or impedance difference/mismatch between low and high output? who knows? making hard conclusion from limited range measurement is not convincing...

Using an identical probe setup with both oscilloscopes should rule this out which is why I asked for details.  If I ran across this behavior, I would check my probe setup and leave everything exactly the same (including the feedthrough termination) while swapping oscilloscopes.  The Rigol should be showing a 50 MHz bandwidth limited version of what the 2465 shows.  Enabling the 100 MHz bandwidth limit on the 2465 would be informative.

[Mechatrommer]

Reconstruction filters which include bandwidth enhancement are common in high bandwidth DSOs but I have never seen one on a 300 MHz or slower DSO.

possibly this. freq respond plot of the china DSO is something to be desired. once venturing this subject, is not beginner territory anymore. but we hobbiest loved to use crippled/cheap tools to do stuffs at some level of "confidence"

It would be fun to make an amplitude versus frequency plot of this DSO and see if there is excessive ripple and if it has a single pole gaussian response producing a 0.35/risetime bandwidth.

as said, this is something to be desired. i've been wanting to do this since my DS1052E, but to do this we need super duper FG and DSO to verify the measurement. iirc there are some discussing here on fast pulse respond of the Rigol DSO, but that limited to that only, what we need to do is a complete freq sweep from say 10MHz to 200MHz, or preferably up to 1GHz range, but thats not cheap. for example this guy (below) did it with higher end gears (he is a eevblog member? not sure who he is), but sadly its still limited and of questionable validity, ie his measurement setup could be different from us, we say sometime need to measure on a short 50ohm terminated BNC, etc. his test is using standard bundled probe... with the proper respond plot, i believe we can make better measurement through signal recompensation. tedious job yes, but thats an effort if we want to make out something usefull, from something cheap...