Choosing between entry-level 12-bit DSOs

[shapirus]

For an apples-to-apples comparison, could you check a 562 MHz sine at 1.25 GSa/s, or 281 MHz at 625 MSa/s, or 140 MHz @ 312.5 MSa/s with the DHO800? That should give us a good indication whether the reconstruction (interpolation) algorithms do a comparable job in both scopes.

Good point. Besides, I had not noticed that the Siglent screenshots were taken at 1 GSa/s rather than 2.

Here you go. I guess this shows enough.

[nctnico]

Perhaps a better question would be whether the Rigol's interpolation implementation is flawed and starts to suffer from aliasing too early. Well, maybe. Is the Siglent's better? Maybe, but I'd say we still don't have enough solid data. There were some decent indications that the Siglent is better, but to properly compare apples to apples we would need to capture signals that have frequencies which are equal not absolutely, but as fractions of the sampling rates of both, and that, ideally, stay within the spec'd bandwidth, for which we would need to reduce the sampling rate artificially.

Staying within the specified bandwidth does not seem critical, as long as we are not interested in measuring absolute amplitude and are not too far out (so we still get s strong-enough signal).

Performa01 has shown a nice and stable reconstruction of a 450 MHz signal with 1 GSa/s sampling rate in the Siglent, i.e. sampled at 2.22 times the signal frequency. For an apples-to-apples comparison, could you check a 562 MHz sine at 1.25 GSa/s, or 281 MHz at 625 MSa/s, or 140 MHz @ 312.5 MSa/s with the DHO800? That should give us a good indication whether the reconstruction (interpolation) algorithms do a comparable job in both scopes.

If you look carefully at the 450MHz screenshot Performa01 posted, you'll see that the displayed signal can't be right. Just look how the histogram for the frequency measurement shows a wide spread instead of a single frequency! And also look at the amplitude compared to the 408MHz signal. IOW, the screenshot doesn't show what is actually happening. Just stick to testing fs / 2.5 + 2% or so. DSOs can show fantasy signals in such circumstances!

[tggzzz]

Part of the fun of being a hobbyist (for me, at least) is finding ways of doing things that don't require a big budget

Precisely See my .sig.

Doubly so, since I'm retired and don't have access to modern professional equipment.

Besides, an engineer is someone who can do for £1 what any tool can do for £10.

[Fungus]

the fact is, if you don't need HDMI out, there's a better scope available for the same price. 🤷

What if bench space is at a premium? The Siglent is much bigger.

What it you need to use it on the go? The Rigol can run from an ordinary powerbank.

[Fungus]

No, it does not have to be periodic. "Perfect reconstruction" only requires that the original signal sampled by the ADC sampled was bandwidth-limited.

Nope.

The signal in this image is bandwidth limited (5Hz signal, 11Hz sample rate) but sin(x)/x won't reconstruct it unless the filter is infinitely wide and the signal is periodic (which can't happen in practice).

That signal has components above 5.5 Hz.

How can a 5Hz sine wave have components about 5.5Hz?

[nctnico]

Part of the fun of being a hobbyist (for me, at least) is finding ways of doing things that don't require a big budget

Precisely See my .sig.

Doubly so, since I'm retired and don't have access to modern professional equipment.

Besides, an engineer is someone who can do for £1 what any tool can do for £10.

That is true for as long as somebody is dumb enough to pay the engineer £1000 in wages where using the £10 tool costs £100 in wages. Penny wise, pound foolish.

[gf]

For an apples-to-apples comparison, could you check a 562 MHz sine at 1.25 GSa/s, or 281 MHz at 625 MSa/s, or 140 MHz @ 312.5 MSa/s with the DHO800? That should give us a good indication whether the reconstruction (interpolation) algorithms do a comparable job in both scopes.

Good point. Besides, I had not noticed that the Siglent screenshots were taken at 1 GSa/s rather than 2.
Here you go. I guess this shows enough.

The same applies to 490 MHz. It was meant to be 49% of the sample rate, so it would correspond to 612.50 MHz @1.25GSa/s.
If you want to retry it, use 10ns/div do see at least one period of the envelope, and better capture a single shot since I don't expect a stable trigger.

[tggzzz]

Part of the fun of being a hobbyist (for me, at least) is finding ways of doing things that don't require a big budget

Precisely See my .sig.

Doubly so, since I'm retired and don't have access to modern professional equipment.

Besides, an engineer is someone who can do for £1 what any tool can do for £10.

That is true for as long as somebody is dumb enough to pay the engineer £1000 in wages where using the £10 tool costs £100 in wages. Penny wise, pound foolish.

I think that has been written too quickly and tersely to make a point effectively.

The cost of a tool includes purchase, learning/training, familiarisation, as well as use.

For beginners, which is the immediate context and also the OP's context, the learning/training and familiarisation may well dominate.

[shapirus]

The same applies to 490 MHz. It was meant to be 49% of the sample rate, so it would correspond to 612.50 MHz @1.25GSa/s.
If you want to retry it, use 10ns/div do see at least one period of the envelope, and better capture a single shot since I don't expect a stable trigger.

600+ MHz isn't really an option, because the amplitude gets too low at this point. We can try 612.5 * 0.49 = 300.1 MHz at 612.5 MSa/s, though. Here you go. And a bonus: 312.5 MHz at 100 us/div.

edit: mistake! that was supposed to be 625, not 612.5. I'll keep that screenshot and attach a new one: 625*0.49 = 306.25 MHz at 625 MSa/s. That's the same thing in principle, though.

[wasedadoc]

No, it does not have to be periodic. "Perfect reconstruction" only requires that the original signal sampled by the ADC sampled was bandwidth-limited.

Nope.

The signal in this image is bandwidth limited (5Hz signal, 11Hz sample rate) but sin(x)/x won't reconstruct it unless the filter is infinitely wide and the signal is periodic (which can't happen in practice).

That signal has components above 5.5 Hz.

How can a 5Hz sine wave have components about 5.5Hz?

I wrote "above", not "about".  The wave as shown is not a continuous sine wave.  It is a continuous 5Hz sine wave gated on by a pulse 1 second long.  That is Amplitude Modulation which produces additional frequencies above and below 5Hz.

[gf]

No, it does not have to be periodic. "Perfect reconstruction" only requires that the original signal sampled by the ADC sampled was bandwidth-limited.

Nope.

The signal in this image is bandwidth limited (5Hz signal, 11Hz sample rate) but sin(x)/x won't reconstruct it unless the filter is infinitely wide and the signal is periodic (which can't happen in practice).

That signal has components above 5.5 Hz.

How can a 5Hz sine wave have components about 5.5Hz?

I wrote "above", not "about".  The wave as shown is not a continuous sine wave.  It is a continuous 5Hz sine wave gated on by a pulse 1 second long.  That is Amplitude Modulation which produces additional frequencies above and below 5Hz.

If it is meant to be a one-time burst of 5 cycles (and zero outside the burst), then it is not properly bandwidth-limited and the spectrum has even an infinite extent. The attached plot shows the spectrum up to 100 Hz. Since it violates the sampling theorem it is not reconstructibe exactly from the samples.

[gf]

If you look carefully at the 450MHz screenshot Performa01 posted, you'll see that the displayed signal can't be right. Just look how the histogram for the frequency measurement shows a wide spread instead of a single frequency! And also look at the amplitude compared to the 408MHz signal. IOW, the screenshot doesn't show what is actually happening. Just stick to testing fs / 2.5 + 2% or so. DSOs can show fantasy signals in such circumstances!

I also have the feeling that the interpolation already suffers a bit in the 450 MHz screenshot.

[Fungus]

edit: mistake! that was supposed to be 625, not 612.5. I'll keep that screenshot and attach a new one: 625*0.49 = 306.25 MHz at 625 MSa/s. That's the same thing in principle, though.

You get the same "AM" effect we've been discussing all along, no surprises there. The trick is not to pass 625MHz/2.5

[Fungus]

If it is meant to be a one-time burst of 5 cycles (and zero outside the burst)

I don't think it is. It's just illustrating where sample points lie on a sine wave as you approach Nyquist.

[gf]

You get the same "AM" effect we've been discussing all along, no surprises there.

My primary aim why I asked Performa01 for 490 MHz @1GSa/s was to provoke this effect on the Siglent, too.
It is expected to be most pronounced for frequencies slightly below sample_rate/2.

[gf]

If it is meant to be a one-time burst of 5 cycles (and zero outside the burst)

I don't think it is. It's just illustrating where sample points lie on a sine wave as you approach Nyquist.

You don't think? Isn't it your diagram?

[awakephd]

machine would be overkill for the projects he or she typically pursues. A "mini-lathe" or "mini-mill" are rightly regarded by the professional as toys compared to the 250K 8,000 lb. behemoth - but those "toys" may be the far better choice for the hobby user, and the hobby user might actually be producing superior work - in terms of precision, or in terms of beauty, or so on - because that is what the hobby user is focused on, without being constrained by the demands of the clock.

But also keep in mind that good tools are more enjoyable to work with. So even for hobby use, the professional tools are worth considering. Basically it all boils down to budget. If you have little to spend then going for cheap options can be a way to get some form of tooling but don't expect great results. But if you have the budget to aim higher, then do so. Cry once and smile forever. Every now and then I do woodworking; the tools from Festool and Makita put a smile on my face every time I put them to use. You can't really put a price on that. Especially since I've tried to use cheap tools for so long and know how mediocre the end result is despite best efforts.

Thus the reason I am aiming for one of these two scopes rather than for the Hantek or so on. Let me extend the analogy a bit - over 35 years of woodworking, I have found that there is no substitute for good tools ... but also that there are good tools at different price points. For the past 25 years or so, I've used a Craftsman "contractor style" saw. In its original form, it was a vast improvement on the little direct-drive saw that I had ... but after I made a simple dust-collection chute and a t-square style fence for it, it became an exceptional tool. In the past year, I traded it out for a Powermatic 66 - a cabinet saw that was long considered the best-of-breed. It has way more power than the Craftsman ... it's a really nice saw ... but honestly, I haven't found it to be any sort of revolutionary step up. In fact, it is far less efficient at dust collection, despite the enclosed cabinet.

Here's the key: after working with wood for 35 years, I know what makes a saw really, really good ... and what makes it unbearable. With that experience, I can make informed choices about tools that will serve me really, really well, regardless of the brand or price. Sadly, I don't yet have that experience with respect to DSOs.

A "mini-lathe" or "mini-mill" are rightly regarded by the professional as toys compared to the 250K 8,000 lb. behemoth - but those "toys" may be the far better choice for the hobby user, and the hobby user might actually be producing superior work - in terms of precision, or in terms of beauty, or so on - because that is what the hobby user is focused on, without being constrained by the demands of the clock.

I disagree with this analogy. Professionals look for quality, size is not relevant in that way. The goal of the professional is to have the higher quality tool for the job, regardless of size. If a "mini table saw" is the appropriate tool, then that's the tool they will use. I have a full size table saw, and a tiny mini table saw that can sit on top of it, without even getting in the way for medium size projects. Both of them are valuable tools with their places. If either one of them was limited in their expected range of usefulness because of quality issues, then that would be a toy the professional wouldn't accept.

Well ... I'm going to have to respectfully disagree as well. When it comes to metalworking, size really does matter - or more precisely, rigidity matters, and that is tied to mass, which is tied to size. Lack of size/mass/rigidity not only limits the size of work that can be done, but also the speed and precision of the work. Of course, not every lathe or mill has to be 8,000 lbs; there are some exquisitely made (and exquisitely priced) machine tools, that would be seen as professional grade, that are, relatively speaking, small - but small in this context means "only" 1,000 lbs or so. And even so, no matter how exquisitely made, the smaller size is going to limit the depth of cut; there simply is no way around mass and rigidity.

Of course, there are much lighter mills and lathes available ... they are called "hobby grade." That doesn't necessarily mean poorly made, nor does it necessarily mean inexpensive; it means smaller and lighter and inescapably less capable ... and also typically easier for the hobby user to start with, far easier and less expensive to run, far easier to purchase and install, and on the list goes. No professional machinist would ever describe a Sherline as a professional tool, but it might be the best tool for some hobby users.

I grant that the term "professional" can get fuzzy - someone might make a profession of building very small models on a very small metal lathe & mill ... or for that matter, small wooden projects on a mini table saw. But in general, those who machine for a living are working with larger projects where speed of results matter. Even when it comes to woodworking, I would think professionals would generally not describe a mini-table-saw as a professional tool - though again it might be perfect for the needs of a hobby user. Note that I am thinking of a mini-table-saw as something using a 4" blade as an apt analogy for comparing the mini-lathe to the smallest of professional lathes.

Of course, both metal- and wood-working include a range of smaller tools that would certainly be considered professional. A 7" "skill saw" and a 4.5" angle grinder would be examples from each field. There, I agree that size is not the issue; there are professional-grade, exquisitely made 6" saws that would beat out a multitude of crappy 7" saws, and likewise some 4" grinders that are superior to a multitude of crappy 4.5" grinders. But these are not the tools to use when you are needing to rip a 10' 6/4 oak board perfectly straight and to size, or when you need to shape a piece of metal to within .0005". Yes, a skilled craftsman could do it, with enough time and patience, but for these tasks the larger stationary tools are the best option, and for these size matters.

Sorry ... I think I'm drifting way off topic at this point!

[KungFuJosh]

Sorry ... I think I'm drifting way off topic at this point!

That didn't disagree with my statement, you simply preferred working on larger stuff. And you know the difference in quality in either case. 😉

I don't think you would use an 8000lb lathe to make a fancy toothbrush. Well, not for a human anyway.

[Harrow]

I WILL make a choice, and soon ... but first I'm still wading through some of the lengthier threads to try to get a better feel for how each machine would be to use.

"In ancient Rome, there was a poem
About a dog who found two bones
He picked at one, he licked the other
He went in circles, he dropped dead."
- Freedom of choice, DEVO, 1980.

Don't take too long, LOL. 

[Mechatrommer]

A "mini-lathe" or "mini-mill" are rightly regarded by the professional as toys compared to the 250K 8,000 lb. behemoth - but those "toys" may be the far better choice for the hobby user, and the hobby user might actually be producing superior work - in terms of precision, or in terms of beauty, or so on - because that is what the hobby user is focused on, without being constrained by the demands of the clock.

I disagree with this analogy. Professionals look for quality, size is not relevant in that way.

from what i observed, not everytime. pro usually look for time saving, quality maybe "okayish" or "acceptable", but anyone who look for quality will need more time. this what imho make apart name like Porche or Lamborghini with like Ford or Tesla.

I am genuinely struggling between the options. If I had more experience, I would have a better idea of which things I might regret the most if I went with this one vs.

to make it simple, from demonstration made here, if you want a better scope (only looking at, analyzing and debugging signal traces) sds800x will be better. but if you think you want to expand your scope into basic MSO mixed signal and AFG with bode plot (and LA probe ready) at cheap cost (people are working on it. not so easy with siglent), and can accept the fact that you need to probe frequency at lesser or equal to Sr / 2.5 at any configuration, then its dho804. at full channel on and MSO, 156.25MSps meaning to probe 62.5MHz signal. without LA turned on, 312.5MSps ie 125MHz and so on... with sds800x, all 4 channels on, 500MSps ie 200MHz, but if you want to add LA, prepare to pay another half or more the scope's price... want to add bode plot? another half or more the scope's price. we are looking at 2-3X the scope's price if we need full MSO+AFG capability...

during ds1054z age, the choice is easy, even though not easily hackable to full MSO+AFG, but siglent's cheapest price is still 2-3X the rigol's price, so its quite no brainer, but now, its quite challenging not just for you. a piece of advice, please stick to your topic, dont think about getting used ds1054z because its not 12bit, so we have to be clear about it. and you'll risk non 100% functional used scope or short lived one only for saving like $100 or so? you can end up spending more money in the end... cheers.

[KungFuJosh]

A "mini-lathe" or "mini-mill" are rightly regarded by the professional as toys compared to the 250K 8,000 lb. behemoth - but those "toys" may be the far better choice for the hobby user, and the hobby user might actually be producing superior work - in terms of precision, or in terms of beauty, or so on - because that is what the hobby user is focused on, without being constrained by the demands of the clock.

I disagree with this analogy. Professionals look for quality, size is not relevant in that way.

from what i observed, not everytime. pro usually look for time saving, quality maybe "okayish" or "acceptable", but anyone who look for quality will need more time. this what imho make apart name like Porche or Lamborghini with like Ford or Tesla.

You're right, of course, but we're talking about different types of professionals now. Contractors or laborers vs. specialty stuff like furniture makers, or artiiiists. Some furniture people buy chisel sets that cost more than a full DeWalt set of contractor power tools.

These analogies are getting out of hand lol.

[nctnico]

If you look carefully at the 450MHz screenshot Performa01 posted, you'll see that the displayed signal can't be right. Just look how the histogram for the frequency measurement shows a wide spread instead of a single frequency! And also look at the amplitude compared to the 408MHz signal. IOW, the screenshot doesn't show what is actually happening. Just stick to testing fs / 2.5 + 2% or so. DSOs can show fantasy signals in such circumstances!

I also have the feeling that the interpolation already suffers a bit in the 450 MHz screenshot.

The 450MHz signal in the 450MHz screenshot definitely is a fantasy signal. The amplitude suddenly gets higher (compared to the 408MHz screenshot) while you'd expect it to be lower. I don't know what the signal amplitude is at the input but likely it is quite high to at least see some signal. When increasing the sensitivity, there will be more gain so likely the front-end is now distorting due to clipping / overdrive recovery / etc. When low-pass filtering the distorted signal from the front-end you still get a sine wave but it has nothing to do with whatever sine wave signal is at the input. And there may be other elements in the signal processing chain as well. I've seen similar behaviour when testing the R&S RTM3004 under similar conditions.

When doing these tests, you need to start with setting the oscilloscope to a v/div setting which does not distort / clip the input signal in the analog front-end and use that v/div setting as the lower limit you can set the sensitivity to at the input signal level. That way you ensure you are always looking at a real signal (if you can see any).

Edit: on closer inspection it is clear Performa01 did take care of not overdriving the front-end. Having different sample rates, v/dviv and t/div settings didn't make the posting very clear though.

[Fungus]

If it is meant to be a one-time burst of 5 cycles (and zero outside the burst)

I don't think it is. It's just illustrating where sample points lie on a sine wave as you approach Nyquist.

You don't think? Isn't it your diagram?

No, it's from here:
https://dsp.stackexchange.com/questions/60859/conclusions-of-sampling-around-nyquist-rate

[nctnico]

The answers over there are quite wrong.

[Fungus]

The answers over there are quite wrong.

Am I going to be blamed for that, too?

FWIW I never even read the page, I just found the image on google image search.

It's a 5Hz sine wave sampled at 11Hz intended to show where the "AM" effect comes from.

That's it... why would anybody even try to complicate things beyond that?