Question about analog oscilloscopes vs digital oscilloscopes

[Fungus]

As for DSOs being easier to interpret, easier to make measurements, and easier for a beginner to understand, have a look at the screen dump in the link below. I would hope and expect that even a beginner might question the measurement.

Hopefully even the greenest of newbies would realize there's no wires connected to channel 2's input BNC, that maybe the number can be ignored.

Yeah, it's not a great feature but picking up AC mains (or whatever) is to be expected whenever you have a floating input to a very sensitive op-amp.

The 'bug' isn't in the number on screen, the 'bug' is that it doesn't simply display zero whenever ground coupling is enabled (would that be cheating?)

But what about a less obviously "suboptimal" measurement that might be lurking?

If there were any other "suboptimal" measurements? I'm sure you'd be posting them here every five minutes.

[tggzzz]

As for DSOs being easier to interpret, easier to make measurements, and easier for a beginner to understand, have a look at the screen dump in the link below. I would hope and expect that even a beginner might question the measurement.

Yeah, it's not a great feature but picking up AC mains (or whatever) is to be expected whenever you have a floating input to a very sensitive op-amp.

The 'bug' isn't in the number on screen, the 'bug' is that it doesn't simply display zero whenever ground coupling is enabled (would that be cheating?)

In the context of that screendump, I cannot see the relevance of either statement.

But what about a less obviously "suboptimal" measurement that might be lurking?

If there were any other "suboptimal" measurements? I'm sure you'd be posting them here every five minutes.

Measurements of DC and sinusoidal waveforms are trivial on either an analogue or digitising scope. Where a DSO's measurement capabilities ought to be a significant advantage are in more complex waveforms, e.g. non-sinusiodal and/or with various forms of noise.

With such waveforms I don't know how I would notice a 10% error, let alone a beginner.

[Fungus]

In the context of that screendump, I cannot see the relevance of either statement.

Yet you posted the screendump as evidence of something? 

[joeqsmith]

     " have a look at the screen dump in the link below. "  Read his post.

In the context of that screendump, I cannot see the relevance of either statement.

Yet you posted the screendump as evidence of something? 

[Martin.M]

If there were any other "suboptimal" measurements? I'm sure you'd be posting them here every five minutes.

of coarse  also engineers want to smile sometimes.

[tggzzz]

   " have a look at the screen dump in the link below. "  Read his post.

Quite. I don't think he had had his first dose of caffeine

[rsjsouza]

Being the owner-user of an hp 8505A vector voltmeter and numerous other vintage test gear that required understanding and ability from the user, not just knowing how to access each and every feature on the display or writing code to drive each and every measurement.

Unfortunately understanding is not highly prized.

As for DSOs being easier to interpret, easier to make measurements, and easier for a beginner to understand, have a look at the screen dump in the link below. I would hope and expect that even a beginner might question the measurement. But what about a less obviously "suboptimal" measurement that might be lurking?

A bit more subtle error that emphasizes the importance of correctly interpreting the waveform on screen before blindly trusting measurements is in the snip of one of my videos below:
https://youtu.be/mw9H1NinDeQ?t=23m14s

I suspect that chances are higher that a less experienced user will suspect the signal source is wrong simply because "the number" tells it so.

[tggzzz]

Being the owner-user of an hp 8505A vector voltmeter and numerous other vintage test gear that required understanding and ability from the user, not just knowing how to access each and every feature on the display or writing code to drive each and every measurement.

Unfortunately understanding is not highly prized.

As for DSOs being easier to interpret, easier to make measurements, and easier for a beginner to understand, have a look at the screen dump in the link below. I would hope and expect that even a beginner might question the measurement. But what about a less obviously "suboptimal" measurement that might be lurking?

A bit more subtle error that emphasizes the importance of correctly interpreting the waveform on screen before blindly trusting measurements is in the snip of one of my videos below:
https://youtu.be/mw9H1NinDeQ?t=23m14s

I suspect that chances are higher that a less experienced user will suspect the signal source is wrong simply because "the number" tells it so.

Wow. That's ... um, surprising.

Any idea what causes it? (My Spanish isn't up to understanding anything you say!)

[nctnico]

I did some tests with a 2.1GHz burst (8us burst, 50ms silence) fed into a 1GHz 2Gs/s scope in ET mode. In the first one I forgot to switch the probe factor back to 1:1 because the signal is fed in from a 50 Ohm source.

[Fungus]

   " have a look at the screen dump in the link below. "  Read his post.

Quite. I don't think he had had his first dose of caffeine

The first clue is in the tiny symbol shown on channel 2. I know it's small so I've highlighted it here:


That seems very relevant to my second statement: "the 'bug' is that it doesn't simply display zero whenever ground coupling is enabled"

Understanding the rest of it will mean you have to read the entire thread where this issue was discussed. Bummer, right?

[rsjsouza]

Being the owner-user of an hp 8505A vector voltmeter and numerous other vintage test gear that required understanding and ability from the user, not just knowing how to access each and every feature on the display or writing code to drive each and every measurement.

Unfortunately understanding is not highly prized.

As for DSOs being easier to interpret, easier to make measurements, and easier for a beginner to understand, have a look at the screen dump in the link below. I would hope and expect that even a beginner might question the measurement. But what about a less obviously "suboptimal" measurement that might be lurking?

A bit more subtle error that emphasizes the importance of correctly interpreting the waveform on screen before blindly trusting measurements is in the snip of one of my videos below:
https://youtu.be/mw9H1NinDeQ?t=23m14s

I suspect that chances are higher that a less experienced user will suspect the signal source is wrong simply because "the number" tells it so.

Wow. That's ... um, surprising.

Any idea what causes it? (My Spanish isn't up to understanding anything you say!)

My best idea is that the amplitude was too low for the frequency counter to trigger properly, but I didn't think about that the day I shot the video nor I remember now.

BTW, you were close... the language is Portuguese and I can babble as much as Dave.

[tggzzz]

BTW, you were close... the language is Portuguese and I can babble as much as Dave.

Damn! I looked at "souza", considered Portuguese, and rejected it because of the US flag!

[vk6zgo]

Next time, before you tell others they don't know what they're talking about, I recommend you get a at least a basic Yes, a 1GHz burst requires a higher sample rate than 2.5GSa/s (or 5GSa/s, for that matter), as due to its non-continuous properties it will contain frequency components that are much higher than f₀. But it also requires an analog bandwidth of a lot more than 1GHz, so the 7904 is utterly useless for such signals. You might see something, but it won't have much to do with how the input signal really looks like.

[

A burst does not necessarily contain "much higher frequency components" than a CW signal of the same frequency.
It is all dependent upon the envelope shape of the burst.
For instance,the 4.433MHz colour burst in PAL TV contains very little harmonic energy--line & field rate sidebands,yes,harmonics no.

OK,that is a specially "tailored" envelope,but naturally occurring intermittent bursts of oscillation tend to build up over time,rather than just suddenly appear full amplitude,with the result that the envelope rise is also fairly gradual.

Add to this,the frequency limitations of the circuit the problem appears in,& those "much higher frequency" components will be well attenuated before you go anywhere near it with an Oscilloscope.

[G0HZU]

GK :  Why do you think that reputable manufacturers like Lecroy make DSOs with real-time sampling rates ten times the -3dB (analogue) bandwidth if it is "nonsense" even to sample at five times the BW?

Wuerstchenhund:   Simple, really:

- As nctnico said, most scope series contain higher bandwidth models, and a 1GHz scope requires a faster sampling rate than the 350MHz model. You'll also find that for some scope series, the lower BW models come with a much lower sample rate (i.e. the LeCroy WaveRunner2 LT, WR6k(A), WR6zi, WP7Zi(-A))

- Higher than required sample rates allow the use of oversampling techniques to increase vertical resolution to up to 16bit (i.e. R&S RTE, RTO Series)

- While low-end scopes usually use standard components for ADCs, mid-range and high-end scopes usually use ADC hybrids that are inhouse developments of the scope manufacturers, which makes them pretty expensive. To keep costs down, they are often re-used across scope ranges, with as few as possible variants

- A high sample rate looks impressive on the spec sheet

Wuerstchenhund: Asking for a 5GSa/s sample rate for 1GHz BW is nonsense.

A high sample rate is good for various reasons and one of them is to prevent alias terms from corrupting the displayed waveform. A typical 500MHz scope will have a graceful frequency rolloff with flat group delay so the amount of alias rejection will be limited if the sampling rate is only 2 or 3GHz. Therefore, if I had to choose a DSO to replace a 500MHz analogue scope and I wanted the same frequency (and flat group delay) response then I'd want a high sampling rate to minimise the influence of alias terms when looking at complex waveforms that have spectral content up in the GHz region.

So I think 5GHz sampling would be a good/safe choice. Go down to 2.5GHz sampling and I think that the alias effects will begin to degrade the display of the wideband waveform compared to what you would see on the analogue scope. I guess it depends on how accurate you want the display to be? Some people will be happy to live with the degraded display at the lower sample rate. They might not care if it isn't quite right and the degradation is negligible to them?

The other option would be to choose a scope with a steeper filter rolloff (eg max flat response) as this will be less prone to alias issues. So you could use a lower sample rate and have less problems with alias terms. But this won't have the same frequency response characteristics as the analogue scope. It's up to the user to decide if this is a good/bad thing for the range of things they want to use the scope for.

[Rupunzell]

Heading down a road previously traveled...
https://www.eevblog.com/forum/testgear/recommendation-request-new-rigol-ds1054z-or-used-tektronix-2467bhd/msg798701/#msg798701

https://www.eevblog.com/forum/testgear/my-first-oscilloscope/msg645726/#msg645726


Little if anything has changed.

Bernice

[rsjsouza]

Another thing that usually does not change... 

Blimey, a bloke asks about buying an old Ford Cortina, which will do the job ok providing he can wield an oil can and a set of spanners and we get pages of discussion about the various merits of a Ford RS200 vs a Toyota GT86

[nctnico]

A high sample rate is good for various reasons and one of them is to prevent alias terms from corrupting the displayed waveform. A typical 500MHz scope will have a graceful frequency rolloff with flat group delay so the amount of alias rejection will be limited if the sampling rate is only 2 or 3GHz.

The older digitizing scopes like the Tektronix TDS500/700 series DSOs (and probably others) typically have such a Gaussian roll-of but these use equivalent time sampling to overcome the limits of their sampling rates (500Ms/s to 2Gs/s). For looking at the waveforms you are suggesting they work just as well as an analog scope WITH all the benefits of a DSO.

[GK]

A high sample rate is good for various reasons and one of them is to prevent alias terms from corrupting the displayed waveform. A typical 500MHz scope will have a graceful frequency rolloff with flat group delay so the amount of alias rejection will be limited if the sampling rate is only 2 or 3GHz. Therefore, if I had to choose a DSO to replace a 500MHz analogue scope and I wanted the same frequency (and flat group delay) response then I'd want a high sampling rate to minimise the influence of alias terms when looking at complex waveforms that have spectral content up in the GHz region.

So I think 5GHz sampling would be a good/safe choice. Go down to 2.5GHz sampling and I think that the alias effects will begin to degrade the display of the wideband waveform compared to what you would see on the analogue scope. I guess it depends on how accurate you want the display to be? Some people will be happy to live with the degraded display at the lower sample rate. They might not care if it isn't quite right and the degradation is negligible to them?

The other option would be to choose a scope with a steeper filter rolloff (eg max flat response) as this will be less prone to alias issues. So you could use a lower sample rate and have less problems with alias terms. But this won't have the same frequency response characteristics as the analogue scope. It's up to the user to decide if this is a good/bad thing for the range of things they want to use the scope for.

Modern DSOs with lower sampling rate to -3dB bandwidth ratios (say only 4:1) have tailored, non Gaussian responses which flatten the response prior to the -3dB point, but provide a steeper roll off and greater attenuation afterwards. This helps prevent/minimize aliasing issues, even when using sinx(x) interploation. This modified response is discussed in that article joeqsmith linked to; where the traditional (Gaussian) relationship BW=0.35/RT no longer holds. I suggest a 500MHzBW-5GSa/s speced DSO as an equivalent "real time" performer to my 500MHzBW 7905 as this is the kind of DSO that is most likely to have an analogue frequency response beyond -3dB best matching that of the old analogue. Additionally, I did not cite 1GHz as the highest frequency visible on the CRT of a 7904. It just is what I have actually seen and it also happened to be a convenient figure to quote because I was completely anticipating being otherwise duly informed that a 500MHz DSO only needs a 1GSa/s sample rate (due to what Nyquist allegedly says) to display anything that my beloved 7904 can. I'm sure that my 7904 is already rolling off at more than a -6dB rate by 1GHz, but I wouldn't be surprised if I could see significantly further than this if I had a test signal source of such frequencies handy. That visible ~1GHz parasitic op-amp oscillation I mentioned was from an old GHz-GBWP part #"CLCxxx" (Comlinear corporation?) CFA (in a DIP!) that I dead-bugged as a voltage follower with too small a feedback (f-compensation) resistor between the output and the inverting input pin, so it wasn't entirely happy. Come to think of it I might borrow one of our 3G radio telemetry/data modems from work with an SMA-to-BNC adaptor. Incidentally at a sampling-to-bandwidth ratio of 10:1 or more sinx(x) interpolation is mostly superfluous. An old boat-anchor DSO with that ratio need only have the processing power for passable-performance linear interpolation.

Incidentally#2: The distributed deflection plate CRT system of the 7904 is -3dB at ~1GHz on its own. I forget the model number but Tek made a passive vertical "amplifier" plug-in that was installed by disconnecting/bypassing the vertical amplifier in the mainframe, connecting the passive plug-in directly to the deflection plates. You needed a soldering iron to do the installation. The specified vertical bandwidth was then 1GHz min. but there was obviously not much flexibility in the sensitivity.

[G0HZU]

A high sample rate is good for various reasons and one of them is to prevent alias terms from corrupting the displayed waveform. A typical 500MHz scope will have a graceful frequency rolloff with flat group delay so the amount of alias rejection will be limited if the sampling rate is only 2 or 3GHz.

The older digitizing scopes like the Tektronix TDS500/700 series DSOs (and probably others) typically have such a Gaussian roll-of but these use equivalent time sampling to overcome the limits of their sampling rates (500Ms/s to 2Gs/s). For looking at the waveforms you are suggesting they work just as well as an analog scope WITH all the benefits of a DSO.

Yes, I have an old HP 54540C here (500MHz 2Gs/s) and it has a mode called 'Repetitive' and I assume that this is a form of ETS. I've not used this scope much but the repetitive mode seems surprisingly useful. I expected it to be a fussy waste of time but it copes quite well with quite a range of waveforms. Obviously, ETS can/will fall over on lots of waveform types but on an old scope like this it does help a lot sometimes.

[Martin.M]

I forget the model number but Tek made a passive vertical "amplifier" plug-in that was installed by disconnecting/bypassing the vertical amplifier in the mainframe, connecting the passive plug-in directly to the deflection plates.

this means a 7A21N  (N of coarse, a passive input can`t generate any readout)
7A21N is not more useful then a good restorated 519 

greetings
Martin

[G0HZU]

I suggest a 500MHzBW-5GSa/s speced DSO as an equivalent "real time" performer to my 500MHzBW 7905 as this is the kind of DSO that is most likely to have an analogue frequency response beyond -3dB best matching that of the old analogue.

Yes, I'd agree with that. However, I'm really not a scope guru or enthusiast so I only know the basics about any potential tradeoffs wrt sampling/filtering/adc/interpolation/etc when choosing the right DSO for a task.

Here's some basic info about my old HP54540C. This isn't a scope I'd recommend to most people for various reasons but it has 2Gs/s and a 500MHz BW when used on a single channel.

The frequency response does look to have a gradual rolloff and if I put it in repetitive mode and stick a sig gen into it the response falls by about 17dB by 1300MHz. It still displays and measures the frequency correctly in this 'repetitive' mode. But 1300MHz is obviously way past the Nyquist limit for 2Gs/s when in real time mode.

I really don't think you would be impressed by this DSO if you saw it in real time mode when trying to display a 500MHz sinewave. It looks bad in real time mode once you get up above 300MHz. I assume the interpolation struggles a bit here as there is a lot of wobble on the signal. However, I don't have another one to compare it to so this degree of wobbliness may be a slight fault.

It's not a nice scope (or a great performer) overall although it has some powerful analysis modes and the keypad system is very powerful/useful with the blue shift key. The fan noise is particularly annoying as it emits a low pitched drone from the back that easily dominates all my other test gear. I don't think the fan is worn or faulty, I just think the size of the fan and the overall scope enclosure gives it this droning quality. I salvaged it at work as it was destined for the skip. Nobody else wanted it and I think I now know why 

[Jwalling]

Here's some basic info about my old HP54540C. This isn't a scope I'd recommend to most people for various reasons but it has 2Gs/s and a 500MHz BW when used on a single channel.

I'd be very surprised if you couldn't upgrade that scope to a 54542C and sample @ 2Gs/s on all four channels simultaneously. That scope uses the same attenuators and A/D FISOs that are on the 54810A through 54825A scopes. See a post I made to the HP/Agilent group on Yahoo a while back:
https://groups.yahoo.com/neo/groups/hp_agilent_equipment/conversations/messages/60910

You'll have to look around to find the strapping resistor -  look for a 50MHz oscillator for starters. If you'd like, I'll have a look at a 54542A ACQ board I have in my scrap pile and see if I can identify where it's located. Let me know...

Jay

[G0HZU]

Here's some basic info about my old HP54540C. This isn't a scope I'd recommend to most people for various reasons but it has 2Gs/s and a 500MHz BW when used on a single channel.

I'd be very surprised if you couldn't upgrade that scope to a 54542C and sample @ 2Gs/s on all four channels simultaneously. That scope uses the same attenuators and A/D FISOs that are on the 54810A through 54825A scopes. See a post I made to the HP/Agilent group on Yahoo a while back:
https://groups.yahoo.com/neo/groups/hp_agilent_equipment/conversations/messages/60910

You'll have to look around to find the strapping resistor -  look for a 50MHz oscillator for starters. If you'd like, I'll have a look at a 54542A ACQ board I have in my scrap pile and see if I can identify where it's located. Let me know...

Jay

Yes, I'd be interested in this, thanks

I did have a look through the firmware a while back to see if it could be done this way but I think it would be too time consuming to work out the code changes. The hardware does look to be very similar in the documentation.

I've not taken the covers off it so I don't know how accessible all this stuff is but I really should take it apart to check the PSU caps for leakage etc.
If the upgrade is possible then it would be nice to get the full performance on all 4 channels.

I'm not too worried about bricking it if things go wrong. It didn't cost me anything.

[Jwalling]

Yes, I'd be interested in this, thanks

I did have a look through the firmware a while back to see if it could be done this way but I think it would be too time consuming to work out the code changes. The hardware does look to be very similar in the documentation.

I've not taken the covers off it so I don't know how accessible all this stuff is but I really should take it apart to check the PSU caps for leakage etc.
If the upgrade is possible then it would be nice to get the full performance on all 4 channels.

I'm not too worried about bricking it if things go wrong. It didn't cost me anything.

OK, My money would be on R744. Note the bit of dicking about on the solder - it looks like there was a resistor there at one point. R745 also appears to have been removed - not sure about this one. There's only about five resistors with reference designations, the rest have nothing. Very telling, in IMHO.

Messing with these resistors will be more difficult than on the 548XX scopes where all you have to do is remove the cover. You'll have to pull the ACQ board, which is about a 30+ minute job depending on your expertise in this particular scope. IOW a PITA 

EDIT: (trimmed), One thing-  I'm not completely sure this is a 54542A and not a 54540A, but looking at your ACQ board will reveal a difference, hopefully.

Jay

[G0HZU]

Thanks for the info and the closeup image

I wonder if some of the labelled resistors may be select on test resistors to optimise something during production test? The value of R703 (and the shape of the PCB track it connects to looks a bit odd for something that is a config resistor.

I'm assuming that the idea is to change a clock multiplier to step up the sampling rate?

There's an image of the 54540C board on a current ebay auction. it says 542 in places on the board but it has a white 54540 sticker on the PCB edge.

http://www.ebay.com/itm/54540-66511-Main-PCB-for-HP-54540C-/261677319734?hash=item3ced2f0636:g:5-wAAOSwubRXIiV0

I'll have a close look at it and I'll have a look in the manual to see how easy it is to take it all apart.