Final Questions Before I Get My First Oscilloscope

[ebastler]

You have to see the screenshots in context. They have been made on a 3Ghz 20GSa/s high end scope, and are merely an illustration of the relevance of FFT sampling points (all other parameters are static). If you look at the lower right-hand corner you'll see the frequency resolution (the same as RBW on a SA) which for that sample rate is 19.53MHz at 1kpts and 1.19kHz at 16Mpts. Obviously, because the DS1054z samples much slower it's 16k sample points will capture a longer interval, which affects the frequency resolution.
[...]
Acoustics and vibration analysis are pretty low frequency and for that the 16k are adequate (and you'd see little improvement from a larger FFT sample size), but you'll have difficulty appropriately resolving a MHz signal or probably even a 100kHz signal.

Hmm, I don't think that's how it works. As long as you do not go into a frequency range where the scope's sampling frequency limits its ability to faithfully capture the waveform, everything should look the same regardless of frequency. (All other things being equal, of course -- especially the same number of waveform cycles being sampled.)

So I did get out the function generator after all... The first screenshot shows my earlier measurement of the 1 kHz signal, and the second the same experiment with a 1 MHz signal. The main difference is that my signal generator is not that great, so the even harmonics are present in the 1 MHz sample. But the FFT resolution looks fine to me.

[Fungus]

So how about we focus on the OP's question instead?

You mean drop all talk of FFTs? That's a good idea.

As to the four channels, yes its great but it's still a bottom-of-the-barrel scope with a sample rate that drops to 250MSa/s with all four channels enabled, which isn't enough for lots of tasks where four channels would be beneficial.

I'm not saying he should buy a DS1054Z, just to consider very carefully if 2 channels is enough. Number of channels is something that's difficult to work around.

Anyway....

Going back to the original post: He was choosing between the Rigol DS2072A and the Rigol DS2072A-S (this was before the "FFT" brigade leaped into the thread).

Opinion was that with $1200 to spend it's better to get a separate signal generator. I think that makes sense.

Next there was an argument made for getting R&S instead of Rigol. Let's see:

The baseline R&S is about $1200 so it's within budget, yes. It's definitely better, nobody disagrees with that. But ... $1200 is still $400 more than the Rigol and it quickly costs an awful lot more when you start adding options (it could easily double the price) whereas Rigol has all the options for free.

So for $1200 you can have:
a) A 300MHz Rigol DS2072A with all options unlocked + a real SigGen, or
b) Baseline 100MHz R&S HMO1202.

Anybody care to argue the case for option (b)?

[Muxr]

So for $1200 you can have:
a) A 300MHz Rigol DS2072A with all options unlocked + a real SigGen, or
b) Baseline 100MHz R&S HMO1202.

Anybody care to argue the case for option (b)?

Yes, read this thread: https://www.eevblog.com/forum/testgear/final-questions-before-i-get-my-first-oscilloscope/
But I can recap it for you. Price per Mhz of bandwidth is the only thing going for Rigol. HMO is a superior scope in every other aspect. And it's arguably the best scope in the up to $1200 price segment.

[markone]

I would argue that there are two major use cases for FFT, and that the Rigol scopes are not suited for the one, but very useful for the other:

• Modulated carrier, where the modulation bandwidth is typically small vs. the carrier frequency, and you want to see a narrow spectral window around the carrier. This is where large FFT windows are needed, and the Rigols (and other scopes with just a few kPoints FFT) will simply not resolve enough spectral detail around teh carrier.
• Unmodulated signal, where you want to observe harmonic content -- e.g. in acoustics, vibration analysis, to check spectral purity of an unmodulated carrier etc.. A few kPoints are perfectly fine for me to assess harmonic content in these applications, so my DS1054Z is certainly useful to me.

Acoustics and vibration analysis require much higher dynamic range and much higher FFT's speed plus programmable HW anti alias filtering, the suited amplitude resolution is 24bit and for vibration sensors the IEPE interface is quite a must.

I do not see at all  any 8 bit scope in this field, with sound you can do much better with any medium quality 192KSa/s 24 bit sound card while for any vibration related serious job, with real (maybe calibrated) accelerometers, you have to look to something like the NI USB-4431, but it's going to cost some Kilo Euro, the sensor cost is also quite high.

Otherwise for vibrations you can always look for mems accelerometers with integrated ADC and serial inteface, like the AD (http://www.analog.com/en/products/mems/mems-accelerometers.html), they cost a fraction, of course the performance is also scaled down, especially for noise floor.[/list]

[markone]

And it's arguably the best scope in the up to $1200 price segment.

Mmmh, i like a lot that scope, but this statement seems to be a bit pretentious.

[ebastler]

Acoustics and vibration analysis require much higher dynamic range and much higher FFT's speed plus programmable HW anti alias filtering, the suited amplitude resolution is 24bit and for vibration sensors the IEPE interface is quite a must.

I do not see at all  any 8 bit scope in this field, with sound you can do much better with any medium quality 192KSa/s 24 bit sound card while for any vibration related serious job, with real (maybe calibrated) accelerometers, you have to look to something like the NI USB-4431, but it's going to cost some Kilo Euro, the sensor cost is also quite high.

You've got a point there. The spectral resolution of simple scopes' FFT is adequate for these applications (which was my main point), but a decent sound interface or dedicated A/D converter will give a much better balance between dynamic range and sampling rate.

[zapta]

You mean drop all talk of FFTs? That's a good idea.

FYI,  somebody created recently a thread dedicated to FFT.

[nctnico]

R&S wasted a whole lot of money slapping a PSU together, putting it in a shielded box and get it certified where GWInstek just orders the PSUs without the headache of certifying it or (due to the PWM method) needing a lot of shielding. It is all about seeing the big picture while doing clever engineering. BTW I don't think I have ever seen a PSU designed by Tektronix themselves in any of their DSOs or logic analysers. It is not their core business so they leave it to someone else who is good at it!

Considering the complexity of what R&S produces from decades, say that they wasted a whole lot of money to design a single 12v PSU in metal shield box sounds hilarious at very best.

I'm starting to think that you never even come close to any of their products, i did, i wrote some control application for RF test benches with R&S equipment (mainly RF network analyzer &  signal generator) and i have a lot respect for them, sure not DIY toys.

You think the same team which designs the high end R&S RF gear is working on low end oscilloscopes? I'm 100% sure they are not. If the HMO1202 was build to the R&S high end specs it would have a full aluminium inner chassis, the board mounted firmly onto that chassis and the input BNCs bolted to the chassis as well. But it is not... It is a Hameg (which never was high-end; more like Europe's Rigol) with an R&S sticker on it.

[Mechatrommer]

You mean drop all talk of FFTs? That's a good idea.

do you mean? drop all the BS gwInstek FFT talk? That's a good idea. 

[Muxr]

And it's arguably the best scope in the up to $1200 price segment.

Mmmh, i like a lot that scope, but this statement seems to be a bit pretentious.

I said arguably, maybe you don't understand the meaning of the word, but it can mean probably.

[nctnico]

Perhaps we should make a list with scopes which cost $1200 and then compare spec-by-spec and bug-by-bug.

[Wuerstchenhund]

You have to see the screenshots in context. They have been made on a 3Ghz 20GSa/s high end scope, and are merely an illustration of the relevance of FFT sampling points (all other parameters are static). If you look at the lower right-hand corner you'll see the frequency resolution (the same as RBW on a SA) which for that sample rate is 19.53MHz at 1kpts and 1.19kHz at 16Mpts. Obviously, because the DS1054z samples much slower it's 16k sample points will capture a longer interval, which affects the frequency resolution.
[...]
Acoustics and vibration analysis are pretty low frequency and for that the 16k are adequate (and you'd see little improvement from a larger FFT sample size), but you'll have difficulty appropriately resolving a MHz signal or probably even a 100kHz signal.

Hmm, I don't think that's how it works. As long as you do not go into a frequency range where the scope's sampling frequency limits its ability to faithfully capture the waveform, everything should look the same regardless of frequency. (All other things being equal, of course -- especially the same number of waveform cycles being sampled.)

Aside from the analog bandwith (which limits which frequency components actually make it into the scope) of course, you're right that the sample frequency limits the ability to capture all relevant aspects of a waveform. However, the sample rate also determines how many sample points are used per time interval, and thus for a given sample memory size determines the length of the time slice that is captured. For FFT, the time slice that is captures impacts the frequency resolution (i.e. the RBW).

To demonstrate the impact the sample rate has on FFT I've set up another quick test. This time I used a 1Mhz sine wave from a Keysight 33522B generator as a source. I used similar settings (i.e. 500kHz/div) as you did on your DS1054z, plus I fixed the number of FFT points to 16k as it is now the new limit in the DS1000z Series. I couldn't replicate everything 100% (i.e. the Rigol's graticule has 12 divisions while my LeCroy has just 10) but it should demonstrate the impact (the upper trace shows the FFT, the lower is zoom).

First, let's sample with 25MSa/s (which is close to the 20MSa/s that is shown for FFT on your scope):




Looks OK, doesn't it? As you can see in the lower right-hand corner, the resolution (RBW) is 1.53kHz and the available bandwidth is 12.5MHz. Now let's zoom into the FFT spectrum, say 25x (20kHz/div):




You can clearly see that the detail level is limited which is caused by the limited frequency resolution (1.53kHz), however you can still somewhat make out the peak, so it's not completely useless.

Now let's repeat that with a sample rate of 1GHz (everything else being the same, incl the 16k FFT) and zoom in:




We see that, because of the higher sample rate, the FFT bandwidth has increased to 500Mhz. However, the frequency resolution has decreased from 1.53kHz to 61.04kHz. And this shows in the FFT spectrum, and even more so in the zoom trace (25x as before):




Because of the low frequency resolution the same zoom trace is now pretty useless, it's just a large bump.

So I did get out the function generator after all... The first screenshot shows my earlier measurement of the 1 kHz signal, and the second the same experiment with a 1 MHz signal. The main difference is that my signal generator is not that great, so the even harmonics are present in the 1 MHz sample. But the FFT resolution looks fine to me.

It does, but simply because you have a very narrow span (6MHz) and a very low sample rate which severely limits the usable bandwidth (which is 10Mhz). That is fine for looking a narrow peak from a sine wave like the one used in our examples but not enough for looking at a reasonable part of the spectrum at a reasonable resolution (which is what you need to do for pretty much any more complex waveform). (FFT) bandwidth requires higher sample rates, and higher resolution at a given sample rate require more FFT points. Simples 

[markone]

Acoustics and vibration analysis require much higher dynamic range and much higher FFT's speed plus programmable HW anti alias filtering, the suited amplitude resolution is 24bit and for vibration sensors the IEPE interface is quite a must.

I do not see at all  any 8 bit scope in this field, with sound you can do much better with any medium quality 192KSa/s 24 bit sound card while for any vibration related serious job, with real (maybe calibrated) accelerometers, you have to look to something like the NI USB-4431, but it's going to cost some Kilo Euro, the sensor cost is also quite high.

You've got a point there. The spectral resolution of simple scopes' FFT is adequate for these applications (which was my main point), but a decent sound interface or dedicated A/D converter will give a much better balance between dynamic range and sampling rate.

Give a try to Spectrum Lab and your sound card, it's free and powerfull, from the german Ham DL4YHF :

http://www.qsl.net/dl4yhf/spectra1.html

[markone]

And it's arguably the best scope in the up to $1200 price segment.

Mmmh, i like a lot that scope, but this statement seems to be a bit pretentious.

I said arguably, maybe you don't understand the meaning of the word, but it can mean probably.

Yes, you are right, it's my fault, but i'm learning 

[Muxr]

R&S wasted a whole lot of money slapping a PSU together, putting it in a shielded box and get it certified where GWInstek just orders the PSUs without the headache of certifying it or (due to the PWM method) needing a lot of shielding. It is all about seeing the big picture while doing clever engineering. BTW I don't think I have ever seen a PSU designed by Tektronix themselves in any of their DSOs or logic analysers. It is not their core business so they leave it to someone else who is good at it!

Considering the complexity of what R&S produces from decades, say that they wasted a whole lot of money to design a single 12v PSU in metal shield box sounds hilarious at very best.

I'm starting to think that you never even come close to any of their products, i did, i wrote some control application for RF test benches with R&S equipment (mainly RF network analyzer &  signal generator) and i have a lot respect for them, sure not DIY toys.

You think the same team which designs the high end R&S RF gear is working on low end oscilloscopes? I'm 100% sure they are not. If the HMO1202 was build to the R&S high end specs it would have a full aluminium inner chassis, the board mounted firmly onto that chassis and the input BNCs bolted to the chassis as well. But it is not... It is a Hameg (which never was high-end; more like Europe's Rigol) with an R&S sticker on it.

Hey now, Hameg was a good brand. Old analog Hamegs were great. They weren't as advanced as Tektronix scopes but they were quality scopes nonetheless.

Also it's clear R&S worked on this scope. The firmware and the menus are very similar to other R&S scopes.

[ebastler]

Aside from the analog bandwith (which limits which frequency components actually make it into the scope) of course, you're right that the sample frequency limits the ability to capture all relevant aspects of a waveform. However, the sample rate also determines how many sample points are used per time interval, and thus for a given sample memory size determines the length of the time slice that is captured. For FFT, the time slice that is captures impacts the frequency resolution (i.e. the RBW).

I see -- so in your earlier examples, you had kept the very high sampling rate, and reducing the sample to a few kpoints meant capturing just a few cycles of the input signal! I had indeed overlooked that, although it is very clear from your screenshots, which do include the time domain signal. In that situation, it is clear that the resolution will be terrible. Spectral resolution of the FFT is proportional to the length of the time domain signal that gets processed, after all.

But that's not how one applies the limited kPoints available for FFT on a low-end scope: I can always chose the time base to get the best compromise between spectral resolution and bandwidth.

It does, but simply because you have a very narrow span (6MHz) and a very low sample rate which severely limits the usable bandwidth (which is 10Mhz). That is fine for looking a narrow peak from a sine wave like the one used in our examples but not enough for looking at a reasonable part of the spectrum at a reasonable resolution (which is what you need to do for pretty much any more complex waveform). (FFT) bandwidth requires higher sample rates, and higher resolution at a given sample rate require more FFT points. Simples 

The available span is actually 10 MHz in this example (just zoomed in a bit on the display to show the resolution limit more clearly), so we have an FFT with 10 MHz bandwidth and 2.5 kHz resolution. For an analysis of harmonic content that seems plenty, even for complex signals.  But as mentioned earlier, I agree that there are very relevant uses of FFT where a few kPoints will be too limiting.

[markone]

You think the same team which designs the high end R&S RF gear is working on low end oscilloscopes? I'm 100% sure they are not. If the HMO1202 was build to the R&S high end specs it would have a full aluminium inner chassis, the board mounted firmly onto that chassis and the input BNCs bolted to the chassis as well. But it is not... It is a Hameg (which never was high-end; more like Europe's Rigol) with an R&S sticker on it.

Yes, i knew it already, it's a revised Hameg DSO, IMHO in its past story Hameg had pretty decent equipments on its catalog, many schools used them for young students with proved reliability.

I already said that i share your doubts about the front BNC connector mounting method, i placed the question to Dave, let's wait for AU daytime to see what he has to say about it.

About Rigol i will dare to say that among all known chinese TE producers they currently have the best building quality, take a look to tear down videos for DS2000, DS1054Z, DSA815TG, DSG815 to say some, and look to pcb layout, soldering quality, psu component topology, case and shielding mechanical precision/robustness, electrical harness care and so on.

We are talking about a brand that actually design in house what it sells, i have an old DS1022CD that still works on my table, at the time of the purchase all the rest of chinese products in that price segment was pure cr@p or bad Rigol copy.

So saying that Hameg was like the "Europe's Rigol" it's like to make a compliment.

Sadly i cannot say the same thing for Siglent, at least for what concern metal chassis, soldering quality and care of overall details.

I mean, all those soldering oops & balls, hammered tincan chassis, PSU transformer ferrite core loose clip and so on spotted on SDS1202X tear down, a unit (i think) supposed to be under Dave's magnifying glass till it's departing from china, well not a smart move.

It's a pity, because this model has an interesting performance / price ratio, if the FW is up to HW potential.

[nctnico]

The Hameg I used at my first employer needed some slapping before working. That particular model had problems with some of the internal connections. Back then I already had a Tektronix DSO at home which I very much preferred over to the Hameg.

[Helix70]

If that is what you think then you have missed the point. The point is that features on some oscilloscopes may be nothing more than checkbox items with very little practical use.

I haven't missed the point at all.

The point is that you need to assign a value to each of those features for your own personal use.

According to this thread: "FFT" has a value of 10/10 to everybody. That's idiotic.

I'm sorry to have to say this but the only thing that is idiotic is your statement . *No-one* in this thread has argued that the DS1000z is a bad scope or not worth buying because its FFT sucks and serial decode is poor, nor has anyone suggested that good FFT is the most important thing in a beginner's scope. Only god knows why the hell you constantly attempt to turn the discussion in a direction as if that was the case. I wonder why this is, I doubt it's a language isse so it looks like there's some kind of agenda here.

The simple point was that FFT on the DS1000z is pretty poor, and while the recent upgrade from 4k to 16k (which in Rigol's term is "High Resolution FFT Deep Memory FFT" ) is nice, it doesn't change the fact that it's still not very useful, something any prospective buyer should be aware of. Does it make the DS1054z a bad scope? Not al all. But it's one (of several) limitations which come from being a very low cost bottom-of-the-barrel scope from a Chinese B-brand. Is that really so hard to understand for you?

For me personally, 4 channels wins.

Right, and because it does so for you it seems you think this must be the same for everyone else, and you already made it clear that you seem to consider any discussion about any of the shortfalls that don't affect you personally as "silly". Talk about a person with blinkers on 

You seem have some kind of agenda here (like Mechatrommer), and that doesn't do you any favors to get across any reasonable argument or contribution you might have.

The OP wants a 5 year scope, I reckon he will have a need for 4 channels before any silly FFT. Hard to debug an SPI bus conversation with only 2 channels. Impossible to monitor the power supply, or the CS line at the same time without 4 channels. It is horses for courses. This argument about the FFT is laughable. No one is seriously using an entry level scope as a substitute for a real spectrum analyzer.

The idea that a beginner (probably the biggest audience of these scopes) will not ever want to use the scope's FFT and use a real spectrum analyzer instead is honestly utterly moronic (someone who can't afford more than a $400 scope has the funds for a real SA? Seriously?  ). The simple reality is that actually FFT is a very good aid for a beginner as it helps to understand basics i.e. that all non-sine waveforms are actually made up from a set of sine waves. It also helps to identify interference and other issues in your circuits. But for that FFT needs to implemented in a way that's actually useful. Plus having a tool to watch the frequency domain in your scope is often pretty handy, and not all frequency domain measurements are best done with a SA.

As to the four channels, yes its great but it's still a bottom-of-the-barrel scope with a sample rate that drops to 250MSa/s with all four channels enabled, which isn't enough for lots of tasks where four channels would be beneficial.

At the end of the day, it should be remembered that this is not about what we use at home, it's about helping someone else to choose a right scope within his budget that satisfies his requirements, and part of helping is that we talk honest and open about the benefits *and shortcomings*. So how about we focus on the OP's question instead?

BTW, the OP has $1200 for a scope which means he isn't forced to go for the $400 bottom-of-the-barrel Rigol, so why should he have to live with its shortcomings when he can get something better for his money?

"Utterly Moronic" is the assumption that a beginner would get more benefit from a slightly better FFT than an extra two channels and $700 in his pocket. Seriously. And when used as a 2 channel scope, you get the same sampling rate as the others.

4 channels FTW. Everyone should have access to a 4 channel scope. so many jobs can not be handled without one, full stop. After you have your 4 channel scope, add any fancy FFT wielding 2 channel scope you like.

[Muxr]

"Utterly Moronic" is the assumption that a beginner would get more benefit from a slightly better FFT than an extra two channels and $700 in his pocket. Seriously. And when used as a 2 channel scope, you get the same sampling rate as the others.

4 channels FTW. Everyone should have access to a 4 channel scope. so many jobs can not be handled without one, full stop. After you have your 4 channel scope, add any fancy FFT wielding 2 channel scope you like.

No, the two channel scopes being compared all have twice the sampling rate. 2Gs/s vs DS1054z's 1Gs/s.

Also 4 channels @250 Ms/s vs 2 channels at 1 Gs/s is debatable. It entirely depends on your common use case.

[nctnico]

IMHO 4 channels scopes should be standard nowadays. If you are designing circuits then you'll never go back to a 2 channel scope once you have used a 4 channel scope. I don't think I have owned a 2 channel scope in the past 10 years.

[Muxr]

IMHO 4 channels scopes should be standard nowadays. If you are designing circuits then you'll never go back to a 2 channel scope once you have used a 4 channel scope. I don't think I have owned a 2 channel scope in the past 10 years.

I have a 4ch scope and a 2ch scope. The number of times I use 4ch is very rare, half the time I am too lazy to connect all 4 probes and I just change the probe points with the 2 probes. I am not going to say 4ch isn't a great feature, it is. But people often overstate it. I would personally rather have a faster sampling 2 channel scope than a 4 channel scope.

edit: One probe is my usual case heh, this is from earlier today, checking on a prototype board I put together to evaluate a chip. In the picture I am making sure that the RTC chip being powered by a small super cap is still oscillating after being powered off for a long time.

[georges80]

4 channels (or more than 2) is very useful when you have other probes.

In my case I have a tek current probe going to channel 4 on my dso. Channel 3 has a differential probe.

Channel 1/2 are 'typical' probes.

A common use scenario in my case would be channel 1 going to the signal I want to observe, channel 2 going to a trigger signal and channel 4 displaying the current probe output.

Not often I need 4 channels, but quite often I'm using 3.

Never use the FFT functionality, of no use in any of the design areas I work in.

And yeah, I've never owned a 2 channel scope.

cheers,
george.

[Muxr]

4 channels (or more than 2) is very useful when you have other probes.

In my case I have a tek current probe going to channel 4 on my dso. Channel 3 has a differential probe.

Channel 1/2 are 'typical' probes.

A common use scenario in my case would be channel 1 going to the signal I want to observe, channel 2 going to a trigger signal and channel 4 displaying the current probe output.

Not often I need 4 channels, but quite often I'm using 3.

Never use the FFT functionality, of no use in any of the design areas I work in.

And yeah, I've never owned a 2 channel scope.

cheers,
george.

Many of the 2ch scopes have an external trigger you can also use in your scenario. You can often adjust the trigger settings like you can on any channel (trigger level, coupling.. etc). So it can replace the normal channel you're using to trigger with.

[zapta]

teamSMITHusa@, did you get an answer to your question? Did it help you to make a decision? What is it?