Choosing between entry-level 12-bit DSOs

[gf]

It's not (only) a bandwidth issue. It's the failure to interpolate the sample data points correctly.

EDIT: Rephrased the question.

@shapirus, assume that you probe an almost ideal square wave and a scope would show you either figure7 or figure8.
Which one would you prefer subjectively? [ I'll explain the difference later in order to get an unbiased vote. ]

[tggzzz]

Regarding the "beginner/non-beginner" argument: any beginner stops being such very quickly, but the tool tailored for beginners (assuming it is) will remain such forever.

I wouldn't call myself a beginner but I'm happy with my Rigol.

I'm simply never going to connect four 200MHz signals to it.

(because I have no need to)

You don't need to.

Connecting four 1Hz signals with transition times <2ns is sufficient.

OK, discussion has been changed from sine waves and Nyquist limits to pulses just for me... got it.

No, the discussion hasn't been changed.

Or perhaps you think the Nyquist limit only applies to sine waves? Hint: it doesn't.

[tggzzz]

It's about observing short pulses, or intermittent glitches, or debugging signal integrity issues and visualizing transitions in digital circuits in general, where having to correlate one signal (e.g. clock) to another one or several is a legit and common use case: it requires both looking at transitions and having multiple inputs enabled at the same time.

Exactly.

It repeatedly surprises me how difficult it is to get people to recognise that.

Is the Rigol's inability to properly display such waveforms fatal? No, by no means it's not: after all, it doesn't fail to display them completely, so you will see it if there is something. Will it prevent the job from being done? No it will not, at least as long as the user is aware of these edge case (pun intended!) shortcomings.

Yup. Any scope is better than none, and you always have to be aware of the limitations of your tools.

A key use case w.r.t. signal integrity is verifying that a received clock signal is monotonic and without overshoot etc. That requires bandwidth plus good probing technique, and nothing else.

For data signals... Well some people claim they aren't repetitive, but clearly such people don't understand eye diagrams. They are much better at detecting signal integrity issues than "repetitive" signals.

And once the signal integrity has been verified, it is often better to flip to the digital domain and use a logic or protocol analyser.

[awakephd]

Which square wave do you prefer? Figure7, or figure8?

<ignorance alert>Help me to understand what I am seeing in these two diagrams? I don't know enough to predict which is from which scope, or why one is better than the other.</ignorance alert>

[awakephd]

With respect to some of this recent discussion ... would it be correct to say that a year ago, when it was a choice between 8-bit, 1Gs scopes (i.e., the SDS1104X-E or the DS1054Z), neither would be able to capture / render the finer details of a rapidly rising pulse? Or is there more to this than just the sampling rate?

[nctnico]

With respect to some of this recent discussion ... would it be correct to say that a year ago, when it was a choice between 8-bit, 1Gs scopes (i.e., the SDS1104X-E or the DS1054Z), neither would be able to capture / render the finer details of a rapidly rising pulse? Or is there more to this than just the sampling rate?

Don't worry. In reality a 100MHz scope is perfectly fine for debugging a microcontroller driving some I2C and / or SPI peripherals (up to 20MHz or so). With SPI you will want to control the drive strength (add series resistors) to slow the edges down and everything will be fine. A 100MHz bandwidth is also perfectly fine for checking SPI setup / hold timing (up to 20MHz) as the interfaces are designed to have plenty of margin. Using discrete logic with flip-flops / counters is extremely rare nowadays. But even if you want to use these in a design, there are free simulators available (like Microcap) where you can check the timing behaviour of your digital design.

[KungFuJosh]

I keep wanting to like the Siglent; I really do. It seems so obviously to be the sensible option. But honestly, every time you all post screen shots of the Siglent next to screen shots of the Rigol, I feel like the Siglent is harder to read, as though I will constantly have to be peering closely to distinguish between one measurement and the next. Meanwhile, the Rigol screen seems vibrant, much easier to pick out salient numbers. Is that just my inexperience showing?

Yes. Look at shapirus' opinion on the Rigol (he owns one). He doesn't like the UI, and has pointed out some of the issues you will have to deal with using such a buggy scope.

My first scope was a Rigol, which is the only reason I hate the company. It was obnoxiously bad. My next scope was the SDS1104X-E, and my experience was AMAZING by comparison.

Since eyesight is a thing you mentioned, I will say the ONLY benefit to you for the Rigol is the HDMI out. If you connect an external monitor, it will be easier to see. However, you will need to wait a year or two for them to hopefully work out the bugs.

On the Siglent, you can connect via a web browser, and that will allow you to make it larger, however, the web UI on that model is not as good as higher models.

If you can save up a little longer, I highly recommend getting something with a larger screen. The SDS1000XHD would be a good place to look.

[shapirus]

Using discrete logic with flip-flops / counters is extremely rare nowadays. But even if you want to use these in a design, there are free simulators available (like Microcap) where you can check the timing behaviour of your digital design.

Good point btw. I, for example, tend not to use microcontrollers in my hobby projects when at all possible and prefer to have fun with discrete logic (and mixing digital and analog stuff) instead, and, frankly, I could easily get away with even a very basic handheld scope, which I actually did before I finally couldn't resist buying a decent 4-channel 12-bit scope hackable to 250 MHz (DHO804) for $342 delivered.
One could get away even without any scope at all: simulation software is mostly so good (...when it works, but that's a different topic) that for me, if my circuit works in simulation, then I can be 95% sure that it will work just the same in hardware. Of course it's always good to understand what is not being simulated, but that's second order things.

But getting better gear, as long as we're speaking of hobby applications, is not just a matter of its suitability for the immediate requirements for one's current areas of interest, but also of having capabilities sufficiently beyond that to allow exploring new areas and learning new stuff, which can be quite enjoyable on its own.

The latter, though, can be achieved with both the scopes we're discussing here. Both are good value, even without adding "for the money". Which of them is better overall? I wouldn't venture to say, even after all the comparisons of the interpolation details and all. Would I want to replace my DHO800 with SDS800X HD? No! I would now only agree to change it to something that has at least 4 Gsps and 500 MHz or better 1 GHz BW to make it a worthwhile replacement :).

But which of the two would I buy now, after my experience with the DHO800, if I had to choose one and both would have the same price? I am not sure. I would be struggling to make a choice. But I have to say that I have no problem agreeing to hack a device to change things I don't like. That's an important point when you consider the DHO800.

[KungFuJosh]

Would I want to replace my DHO800 with SDS800X HD?

Hypothetical:

Your scope is stolen, and your insurance gives you $500 to cover the loss. Do you still buy the same scope?

[shapirus]

Hypothetical:

Your scope is stolen, and your insurance gives you $500 to cover the loss. Do you still buy the same scope?

I believe I would add another something and buy a more advanced one. But if I had only those same two options, then I'm still not sure. If it were the only option, I'd buy the DHO800 again any time. But if we add the Siglent to the list... I would have to give it a test drive to decide, I think.

[tggzzz]

With respect to some of this recent discussion ... would it be correct to say that a year ago, when it was a choice between 8-bit, 1Gs scopes (i.e., the SDS1104X-E or the DS1054Z), neither would be able to capture / render the finer details of a rapidly rising pulse? Or is there more to this than just the sampling rate?

The only important parameter is the analogue bandwidth. Advertising a samples/s figure much higher than 2.5*BW figure is marketing wank; it merely eases some implementation details inside the scope.

The other issue is the number of bits. 30 years ago I used top-end HP DSOs to measure sub-nanosecond risetimes; they were 6bits at best.

The number of bits is more important if you want to look at the frequency domain version of a signal, since they translate into the noise floor and spurious interpolation products. Here again there can be a lot of marketing want (that's why my .sig is there ). The key figure is the ENOB, the effective number of bits. Here 12 is usefully more than 8.

More than 8 bits can be useful in other circumstances, but not really for signal integrity. A beginner is unlikely to require 12 bits for time domain measurements.

[awakephd]

The latter, though, can be achieved with both the scopes we're discussing here. Both are good value, even without adding "for the money". Which of them is better overall? I wouldn't venture to say, even after all the comparisons of the interpolation details and all. Would I want to replace my DHO800 with SDS800X HD? No! I would now only agree to change it to something that has at least 4 Gsps and 500 MHz or better 1 GHz BW to make it a worthwhile replacement .

But which of the two would I buy now, after my experience with the DHO800, if I had to choose one and both would have the same price? I am not sure. I would be struggling to make a choice. But I have to say that I have no problem agreeing to hack a device to change things I don't like. That's an important point when you consider the DHO800.

Those are very intriguing statements, especially in light of Josh's reminder about your dislike of the UI. In one sense, of course, this doesn't help at all! But in another sense ... I do feel a tiny bit better to hear someone with more experience say that you too would struggle with the choice.

[shapirus]

BTW @awakephd, Dave Jones has received a SDS800X HD (along with one of the 1000x HD series) for review. I would suggest to hold your purchase for a little bit more until he gets better and films that review. I belive it may help in making a decision. He also reviewed the Rigol some time last year, make sure you watch that one too.

[shapirus]

A beginner is unlikely to require 12 bits for time domain measurements.

The key word here is "require". Yes, true, 12 bits are hardly required, in the strict sense.

However, having 12 bits of vertical resolution (assuming other parameters equal) simply makes the experience of using the scope more enjoyable: less pixelation and all, you know what I mean. That's important. Tools that are more pleasant to use encourage the user to spend more time with them and thus learn more and become more proficient.

[KungFuJosh]

A beginner is unlikely to require 12 bits for time domain measurements.

The key word here is "require". Yes, true, 12 bits are hardly required, in the strict sense.

However, having 12 bits of vertical resolution (assuming other parameters equal) simply makes the experience of using the scope more enjoyable: less pixelation and all, you know what I mean. That's important. Tools that are more pleasant to use encourage the user to spend more time with them and thus learn more and become more proficient.

Not only that, the scopes are so cheap there's no point in getting an 8bit scope unless there's some advanced features or specs needed. Even still, the 12 bit stuff available now is really competitive.

[2N3055]

Hypothetical:

Your scope is stolen, and your insurance gives you $500 to cover the loss. Do you still buy the same scope?

I believe I would add another something and buy a more advanced one. But if I had only those same two options, then I'm still not sure. If it were the only option, I'd buy the DHO800 again any time. But if we add the Siglent to the list... I would have to give it a test drive to decide, I think.

That is a good, smart answer. 

[tggzzz]

A beginner is unlikely to require 12 bits for time domain measurements.

The key word here is "require". Yes, true, 12 bits are hardly required, in the strict sense.

However, having 12 bits of vertical resolution (assuming other parameters equal) simply makes the experience of using the scope more enjoyable: less pixelation and all, you know what I mean. That's important. Tools that are more pleasant to use encourage the user to spend more time with them and thus learn more and become more proficient.

Pixelation is inevitable with digitizing scopes and digital displays

Someone interested in scopes will use any scope regardless of pleasantness.
Someone not interested in scopes will use it for the minimum possible time.
Only a few people are TE addicts

A beginner can be overwhelmed by complexity; without a deep interest in the subject, the overwhelmed tend to quickly move elsewhere. Simplicity and a gentle learning curve can pay dividends.

[tggzzz]

A beginner is unlikely to require 12 bits for time domain measurements.

The key word here is "require". Yes, true, 12 bits are hardly required, in the strict sense.

However, having 12 bits of vertical resolution (assuming other parameters equal) simply makes the experience of using the scope more enjoyable: less pixelation and all, you know what I mean. That's important. Tools that are more pleasant to use encourage the user to spend more time with them and thus learn more and become more proficient.

Not only that, the scopes are so cheap there's no point in getting an 8bit scope unless there's some advanced features or specs needed. Even still, the 12 bit stuff available now is really competitive.

If cost is a dominant concern, then there are much cheaper alternatives than 8bit scopes.

[KungFuJosh]

If cost is a dominant concern, then there are much cheaper alternatives than 8bit scopes.

You can get a 2 channel 12 bit scope for $390, or 4 channel for $440. True, for some that price is too high, and they'll stick with less expensive 8 bit scopes. But that's a pretty small percentage on the grand scheme of things.

Only a few people are TE addicts

A few what? Million? 🤣

I don't know how many people are on this site, but I know a lot of them are either addicts, or they're addicts in denial. 😉

[tggzzz]

If cost is a dominant concern, then there are much cheaper alternatives than 8bit scopes.

You can get a 2 channel 12 bit scope for $390, or 4 channel for $440. True, for some that price is too high, and they'll stick with less expensive 8 bit scopes. But that's a pretty small percentage on the grand scheme of things.

I was thinking of more extreme constraints and radically different solutions.

[Harrow]

Which square wave do you prefer? Figure7, or figure8?

<ignorance alert>Help me to understand what I am seeing in these two diagrams? I don't know enough to predict which is from which scope, or why one is better than the other.</ignorance alert>

I'll take a guess that the two diagrams represent the displays of two oscilloscopes with slightly different bandwidths...Figure 8 being the higher. The point being that since the Fourier series of a square wave is 1 x fundamental + 1/3 x 3rd harmonic + 1/5 x 5th harmonic + 1/7 x 7th harmonic...etc., then you need a very high bandwidth to accurately display a square wave (or step function) of only modest frequency. When you don't pick up those higher frequencies, that shows in the corners of the trace and the difference in what is displayed by an oscilloscope that samples 60% faster than another is not worth worrying about. If you want an accurate display of a square wave or pulse I'd say you want to talk about how many MULTIPLES faster your sampling is....2X, 5X, 10X, etc, as 60% extra just isn't going to make that much of a difference.

[tautech]

A beginner is unlikely to require 12 bits for time domain measurements.

The key word here is "require". Yes, true, 12 bits are hardly required, in the strict sense.

However, having 12 bits of vertical resolution (assuming other parameters equal) simply makes the experience of using the scope more enjoyable: less pixelation and all, you know what I mean. That's important. Tools that are more pleasant to use encourage the user to spend more time with them and thus learn more and become more proficient.

Pixelation is inevitable with digitizing scopes and digital displays

Someone interested in scopes will use any scope regardless of pleasantness.
Someone not interested in scopes will use it for the minimum possible time.
Only a few people are TE addicts

A beginner can be overwhelmed by complexity; without a deep interest in the subject, the overwhelmed tend to quickly move elsewhere. Simplicity and a gentle learning curve can pay dividends.

As with any new technology to them, analog or digital.

A scope is a scope is a scope and a DSO is no different to drive yet offers a massive amount of additional features which is entirely up to the user whether they use them or not.

To not have a decent feature set is a significant handicap, if not immediately certainly in the future when skill and knowledge develops.

[Harrow]

Or perhaps you think the Nyquist limit only applies to sine waves? Hint: it doesn't.

I think a better way to describe it is that Nyquist is about sinewaves and that all other waveforms can be constructed by a Fourier series of sinewaves. Then you think about what those other waveforms look like when the higher frequency components of the Fourier series are attenuated or removed. Having someone understand this goes a long way to them realising why their traces look like they do.

[Mechatrommer]

I agree that the machine tool analogy proved to be less than helpful.
...
Maybe I should give up on trying to generate analogies ... I think I'm going from bad to worse!

That frequently happens with analogies Analogies are often dangerous

analogies is good to broaden our perspective. imho awakephd's analogy is valid from his perpective. try to learn to see the world from other people's eye.

The other issue is ensuring t_hold requirements are met. Those are fixed for a given logic family, are <<t_clock, and reducing the clock frequency doesn't help (unlike t_setup).

whats your point? those are specifications available in datasheet. they can add variables into the equation. better set the scew setting in scope right first. with BW limited scope, risetime will be "smoothen" anyway.

You don't need to.
Connecting four 1Hz signals with transition times <2ns is sufficient.

OK, discussion has been changed from sine waves and Nyquist limits to pulses just for me... got it.

No, the discussion hasn't been changed.
Or perhaps you think the Nyquist limit only applies to sine waves? Hint: it doesn't.

practically for square wave, Sinc interpolation failure is to plot only on transition edge. square wave misplot (Gibb's artefacts) is a lot less confusing than sin/analog misplot. if you know what i mean.

It's about observing short pulses, or intermittent glitches, or debugging signal integrity issues and visualizing transitions in digital circuits in general, where having to correlate one signal (e.g. clock) to another one or several is a legit and common use case: it requires both looking at transitions and having multiple inputs enabled at the same time.

Exactly.
It repeatedly surprises me how difficult it is to get people to recognise that.

Is the Rigol's inability to properly display such waveforms fatal? No, by no means it's not: after all, it doesn't fail to display them completely, so you will see it if there is something. Will it prevent the job from being done? No it will not, at least as long as the user is aware of these edge case (pun intended!) shortcomings.

Yup. Any scope is better than none, and you always have to be aware of the limitations of your tools.

A key use case w.r.t. signal integrity is verifying that a received clock signal is monotonic and without overshoot etc. That requires bandwidth plus good probing technique, and nothing else.

For data signals... Well some people claim they aren't repetitive, but clearly such people don't understand eye diagrams. They are much better at detecting signal integrity issues than "repetitive" signals.

And once the signal integrity has been verified, it is often better to flip to the digital domain and use a logic or protocol analyser.

what you type is true but inapplicable if you want to use it on BW crippled scope like both dho800 and sds800x. dho800 maybe less crippled, but can show incorrect signal. with sds800x, you wont see any glitches or overshoots on transition edge due to "proper" BW limiting. anyway, trying to do signal integrity at sub ns risetime, you need more expensive tool. signal integrity is about risetime specification is met, and also no excessive overshoot or reflection, and also prog. delay and jitter etc... imho you need at least 10 points for each ns, and thats 10-20GSps, none dho800 nor sds800x can provide. they only suitable for checking 10-20ns risetime logic, i guestimate 5-10Mbps, "signal integrity" wise. dont think nor expect too much.

fyi, recently i managed to figure out (rev.eng) which one is which on 14 bit DAC lane from FPGA clocked at 150-170MHz with just 4CH scope Rigol DS1054Z. with all the theories, opinions and philosophicals here, i shouldnt be able to. ymmv.

The only important parameter is the analogue bandwidth. Advertising a samples/s figure much higher than 2.5*BW figure is marketing wank; it merely eases some implementation details inside the scope.

are you talking about sds800x? you opinion is partly true but saying its marketing wank is too much imho. trying to ease implementation also can means trying to represent true signal on screen and trying to be honest.

The other issue is the number of bits. 30 years ago I used top-end HP DSOs to measure sub-nanosecond risetimes; they were 6bits at best.

The number of bits is more important if you want to look at the frequency domain version of a signal, since they translate into the noise floor and spurious interpolation products. Here again there can be a lot of marketing want (that's why my .sig is there ). The key figure is the ENOB, the effective number of bits. Here 12 is usefully more than 8.

More than 8 bits can be useful in other circumstances, but not really for signal integrity. A beginner is unlikely to require 12 bits for time domain measurements.

i cannot brain this... afaik signal integrity, you need more bits and more sample rate because we try to look in analog domain, not digital domain. and super many overlapping traces. if you want to look at digital domain alone, 1 bit scope / LA is enough. and 12 bits is not just for your logic family's application, other people use it for other things like audio? and FFT, so dont assume we are the only batch of breed on earth. and err... whats the cost of your top-end HP DSOs back then? and its equivalent price today?

If cost is a dominant concern, then there are much cheaper alternatives than 8bit scopes.

You can get a 2 channel 12 bit scope for $390, or 4 channel for $440. True, for some that price is too high, and they'll stick with less expensive 8 bit scopes. But that's a pretty small percentage on the grand scheme of things.

I was thinking of more extreme constraints and radically different solutions.

is this what's in your mind? £6 24MHz 8 Channel USB Logic Analyzer 8 CH Logic Analyzer for Arduino MCU this imho is good enough for logic families (but not for signal integrity mind you). please try to be specific instead of open pop quiz statement. this is electronics forum, not philosophical forum.

here you demo that at some setup Sr / 2.22, sds800x is unable to reconstruct signal properly? from the look of it, yes it is.

Not correct. What I've shown is the behavior at SR / 2.04, not 2.22. The input signal was a 490 MHz sine.

but you have the capture title saying 450MHz (now i see you edited it), anyway in my earlier post above you, i said i stand corrected...cheers.

A beginner is unlikely to require 12 bits for time domain measurements.

The key word here is "require". Yes, true, 12 bits are hardly required, in the strict sense.

However, having 12 bits of vertical resolution (assuming other parameters equal) simply makes the experience of using the scope more enjoyable: less pixelation and all, you know what I mean. That's important. Tools that are more pleasant to use encourage the user to spend more time with them and thus learn more and become more proficient.

Not only that, the scopes are so cheap there's no point in getting an 8bit scope unless there's some advanced features or specs needed. Even still, the 12 bit stuff available now is really competitive.

not to mention he keep confusing beginners that will read this thread with his 70's era's opinion. no insult nor ill-mean intended to him, i only reply to what are ambiguous/confusing, not just from him. cheers.

[DimitriP]

To not have a decent feature set is a significant handicap, if not immediately certainly in the future when skill and knowledge develops.

"Not owning a car is a significant handicap, if not immediatelly certainly in the future when you are old enough to reach the pedals and get a license to drive"

Got it!