What is the "best" Oscilliscope that I can get for $30,000?

[KE5FX]

(Shrug) When I need to capture data around here, it's measured in gigabytes at a minimum.  A scope is the wrong tool for the job.

Your mileage may vary, of course. 

[tautech]

(Shrug) When I need to capture data around here, it's measured in gigabytes at a minimum.  A scope is the wrong tool for the job.

Your mileage may vary, of course.

Ya think.
If you could save directly from your DSO to any folder on your network would that change your POV ?

[kcbrown]

The #1 recommended DSO in this thread uses 2 ADC's each with 200 Mpts of memory support !

And it might well be even if it had 2Mpts.  It certainly would be if it had 20Mpts.

What makes it the #1 recommended DSO in this thread is the combination of capabilities it possesses for the price.

However if the thought of such memory depth scares you, you can wind it back to factory default 20 Mpts or lower if that would make you more comfortable. 

This is actually something that people might do more often than you think, and not because of discomfort ( ), but rather because of the segment implementation.  Using a smaller capture depth frees up the remainder for segment use.

[Per Hansson]

I can't believe everybody here is spending all their time trying to talk him out of it instead of discussing $30k oscilloscopes.

Best comment in the thread.   LOL!!!   
So with that in mind, please discuss everything that's wrong with https://saving.em.keysight.com/en/used/oscilloscopes/msos604a-e185152669471
 

How are you gonna measure a 15kV pole transformer with that POS?
It doesn't even support transients on the inputs so you can't even shove it in the mains outlet, let alone a pole transformer like our OP intends to, quotes below.

I am also looking to monitor the sine wave from my 110V outlets

I'm considering messing with 15kV pole transformers too.  I know that it is important to wear thick rubber when messing with them.

A lot of things go bang when you come too close to a 15kV rail. There are special probes to measure these kinds of voltages, which are usually 1:1000 or higher, e.g. from Fluke. They are physically long, so you don't have to come too close to the source. All probes have maximum voltage ratings which must not be exceeded and which get MUCH lower as frequency goes up! You can easily damage your equipment or kill yourself if you apply excessive voltage to an unsuitable probe, even if it says 100:1.

As I was apparently not allowed to have fun in this thread and since nobody else mentioned it from my skimming of the thread I just want to say that I don't think Fluke has any high voltage probes designed for CAT-III or CAT-IV use that a pole transformer surely qualifies for, plus how are you even gonna carry that heavy scope up the pole?
All their HV probes that I have seen are intended only to service CRT TV's in a CAT-I environment.

I had hoped I would not need to write out that I was kidding  just as the original poster of this thread did not point out that he is trolling 

Can we stop making fun of people that got overly excited while window shopping but genuinely put an effort to learn? Also, deliberately giving bad advice is not funny. Shame on you.

When they genuinely put an effort to learn?
-Absolutely, but that sure ain't happening here!

[KE5FX]

If you could save directly from your DSO to any folder on your network would that change your POV ?

I can (see dso6000.cpp in the GPIB Toolkit.)  But I've only used that app a few times since I wrote it several years back.   I'm more likely to need to record a continuous stream, which isn't supported by any of the DSOs around here. 

Stream acquisition would be a nice feature to have, since it would let the DSO serve as a general-purpose digitizer.  But (once again) it would require only a certain amount of acquisition storage on the DSO itself, with no benefit to paying for more.

[Fungus]

The #1 recommended DSO in this thread uses 2 ADC's each with 200 Mpts of memory support !

It's just a pity it can't zoom out and see it all:

[rf-loop]

The #1 recommended DSO in this thread uses 2 ADC's each with 200 Mpts of memory support !

It's just a pity it can't zoom out and see it all:

Sometimes should perhaps somehow be distinguished from the needs of fun play, the needs of media business populism, and the real work needs of professionals.
Of course, it can require the right knowledge, experience, and skill, which often shines in their absence from these films and opinions because the need for populism passes for media business. Just like yellow media click headlines and sensational news, same approach, same earnings model.

[nctnico]

The #1 recommended DSO in this thread uses 2 ADC's each with 200 Mpts of memory support !

It's just a pity it can't zoom out and see it all:

Sometimes should perhaps somehow be distinguished from the needs of fun play, the needs of media business populism, and the real work needs of professionals.
Of course, it can require the right knowledge, experience, and skill, which often shines in their absence from these films and opinions because the need for populism passes for media business. Just like yellow media click headlines and sensational news, same approach, same earnings model.

Well, ofcourse you want to defend your commercial interests/stay loyal to your employer but the reality is that Siglent simply has made a bade choice.  As I've written before: I won't touch a scope which can't zoom out even with a 10 foot pole. Hard pass. For some of the measurement jobs I do it would be very tedious to operate. Like using a non-auto ranging DMM to measure a wide variety of voltages. Ofcourse it is possible to get the job done but needing to change the range for every measurement is super tedious, wastes precious time and gets annoying quickly.

Now there will be a whole flurry of people chiming in who work on different kinds of projects than I am and yet are trying to force their way of working on me. Meh. I suggest they carefully read what I wrote about the GW Instek versus the Keysight scope earlier on in this thread. Sometimes seemingly little differences make for a more efficient workflow for some tasks. There is no need to rehash an old thread.

[djsb]

If money was no object, I'd buy a Tektronix MDO4104 mixed domain Oscilloscope. They are around £30k I think.

[EEVblog]

As someone with a lab full of fairly expensive scopes, it's the small simple Keysight 1000X that I grab when I just want to measure something.

Yet you'd be doing many a great disservice today recommending a DSO with just 1 Mpts mem depth.

Indeed. The OP stated protocol decoding and for that you really want deep memory.

The reality is KS 1kX is a 2 GSa/s DSO and in providing 2 GSa/s the capture length is just half that of a 1 GSa/s DSO with the same mem depth. Activate another channel and that's halved again !
The power of a DSO is severely hobbled with such limiting mem depth for capture length.

That's all a software limitation BTW, the same ASIC (with inbuilt memory) is the same from the 1000 to 4000 models.
They have increased it before with an update, and it wouldn't surprise me if they increase it again to stay relevant until the Megazoom V comes out.
But the fact that they haven't indicates that it's still selling so well that they don't have to.
Name another scope that's still selling well 12 years after release.

[EEVblog]

1054z costs the same, despite of inflation. Also, there are a few other tools that are approaching 1054's features with smaller price.
Otoh, how much was 500/350M 4ch scope 7 years ago? If tools like 1054 have established it's long-term price point, more expensive tools (like 30k scope) will probably cost less.

Actually, entry level scopes have remained fairly stable in price (actually cheaper with inflation) for at least 40 years, at least here in Aussie dollars.
When I was a kid, $600-$800 bought you a decent 20MHz dual CH analog scope. Then the Rigol DS1052E came out at $800 and you got all the DSO bells and whistles for the same price. Then the DS1054Z gave you 4 channels for a similar price.

[Sighound36]

Had both the 5 and 6 series on long term test, humm a way of the current mark of altenatives shame really so much wanted to like the 6 series but sadly not up to mark or the Keysight and Lecroy offerings and £25K for a True view probe I think not!

[EEVblog]

I've come to the conclusion that I should probably get a fully decked out SDS5104X (1 GHz, 4 CH).  Totally decked out I saw the cost is $11,308.00.  If there are options I should unselect that I could "hack" I'd be interested in hearing about it.

Why do you need 1GHz? 
You are vastly better off getting a 350MHz or even 500MHz model and a whole bunch of useful probes like HV diff probes, current probes and active probes.
Anything above about 500MHz is pretty useless with your normal passive probes anyway.
If you really need GHz range then get an older used scope in the several GHz range just for that task.

[rf-loop]

I've come to the conclusion that I should probably get a fully decked out SDS5104X (1 GHz, 4 CH).  Totally decked out I saw the cost is $11,308.00.  If there are options I should unselect that I could "hack" I'd be interested in hearing about it.

Why do you need 1GHz? 
You are vastly better off getting a 350MHz or even 500MHz model and a whole bunch of useful probes like HV diff probes, current probes and active probes.
Anything above about 500MHz is pretty useless with your normal passive probes anyway.
If you really need GHz range then get an older used scope in the several GHz range just for that task.

Do you believe that example Tektronix TPP1000 passive 1GHz probes are just useless and made for joking. Perhaps you think Tek do not know what they design and do.  No they are serious with also this product. But naturally also it need some real knowledge how to use and for what. As is with most serious T&M instruments and tools mostly, even nowdays.

But for peoples who do not have enough knowledge and exdperience about RF proping, yes they are useless... as everything is useless, even hammer is useless if do not know   how to use and for what.

If one really need GHz range then get an modern scope. Forget old boat anchors with many kind of severe limits if need seriously do something - if want playing fun or nostalgic reasons like these then situation is of course different. Exept if have special and functinally limited need and some old meet perfectly just for this "single" need.

[rf-loop]

I'm not trolling.  I'm using this thread as research as to what to buy.  I might not buy right away, but I have some good ideas as to what to start with.

Of course you are, but perhaps it is friendly trolling?  The volume and apparent sincerity of the responses is hilarious.  The answer to your question is that you should get the Siglent SDS2104X promo combo and duct tape a half kilo gold bar to the back of it.  Maybe you can squeeze a few nice probes and a matching AWG into the budget.

I've come to the conclusion that I should probably get a fully decked out SDS5104X (1 GHz, 4 CH).  Totally decked out I saw the cost is $11,308.00.  If there are options I should unselect that I could "hack" I'd be interested in hearing about it.

Why not a little higher-end (rel SDS5000X) more advanced SDS6104A. In the EU, its base price is 7,880 euros.
And if need, with SAP2500D 2.5GHz differential probe etc.

[kcbrown]

The #1 recommended DSO in this thread uses 2 ADC's each with 200 Mpts of memory support !

It's just a pity it can't zoom out and see it all:

Tell me why in principle I can't zoom out with a 10M point capture buffer when my 1 GS/s GDS-1054B scope is set to 1ms/div.  Yes, I know the technical reasons for it, but that I can zoom out under some circumstances and not others is, from a user's perspective, arbitrary.  So give me a first principles based reason that I can't zoom out under those conditions, if being able to zoom out after a capture is performed is so incredibly important.

Whether you can zoom out or not on the scopes that behave as you prefer is effectively arbitrary.  It depends on your capture length, your timebase, your base sample rate, the number of channels you have active and the distribution of those channels over the available ADCs, and possibly other things I haven't considered.


In light of that, what matters is how you control what gets captured and how much gets captured.  What is certain is that the Siglent doesn't waste any memory even when capturing less than the specified buffer size: it instead uses the additional memory for additional captures and allows you to view those captures via the history mechanism.  You can always ensure that the entire buffer is filled simply by setting your timebase appropriately.  Just like, for the scope behavior you prefer, you can always ensure that you can zoom out after a capture by setting your timebase appropriately.

It's just a difference in philosophy.  The Siglent's main advantage in this is that everything is obvious up-front.  What you see is what you're going to get, no more and no less.  You have complete and direct control of the capture length and duration.  With the mechanism you prefer, you have to do the math, or pay attention to the little indicator at the top of the display, and your control over the capture time is much more coarse, limited to those selections that the scope gives you for the capture buffer length, and using something less than the total wastes the rest unless you have the segments feature and fool around with it (segments are not a first class citizen on the Instek).

Neither approach is particularly hard to deal with as a user.  The approach you prefer is perhaps better suited to scopes with a small amount of display real estate.  The Siglent approach is, I'd argue, better when you have a large display because you can use zoom mode on such scopes without incurring much of a real estate penalty, thus getting you the control advantages the Siglent's approach brings while at the same time being able to "zoom out" at will.


If you were really interested in always being able to zoom out, you'd be arguing for something that no scope manufacturer that I'm aware of implements: a fixed ratio of displayed waveform versus captured waveform at capture time.

[nctnico]

You are going at it from the wrong perspective and make things way too complicated in your mind. More memory means that the situation in which the display shows all (and the sample rate gets reduced) happens at higher time/div settings. And it is easy to see when there is not enough memory to show a full display at the full samplerate; at that point the samplerate gets reduced. Having some signal left & right just makes life easier for some use cases; no need to mess with settings or adjusting the trigger point before taking a measurement. If it turns out the trigger point isn't optimal you just scroll left/right. Stick the probe in your circuit and get a trace without needing to think upfront or re-measure when it turns out your assumption on the signal is wrong. Very handy especially when you are hunting rare events which may take an hour to re-occur. It allows for a way more relaxed approach to using an oscilloscope. And if there isn't enough memory then so be it.

See my autoranging DMM example. Autoranging works until you hit the limits and -for example- need to switch to the mV range manually.

And it is interesting that you bring up the history mode as being useful. In the end this also provides a varying time span of data. But not in the form of a continuous capture but a variable amount of past traces. In that perspective there is absolutely no difference between having an unknown amount of time left & right of the screen versus an unknown amount of past acquisitions. The only realisation you need to make is that both can be useful which is why nearly all of the oscilloscope manufacturers offer both and give maximum flexibility to the user.

[bdunham7]

Having some signal left & right just makes life easier for some use cases; no need to mess with settings or adjusting the trigger point before taking a measurement. If it turns out the trigger point isn't optimal you just scroll left/right. Stick the probe in your circuit and get a trace without needing to think upfront or re-measure when it turns out your assumption on the signal is wrong. Very handy especially when you are hunting rare events which may take an hour to re-occur. It allows for a way more relaxed approach to using an oscilloscope. And if there isn't enough memory then so be it.

My ancient Tek hybrid DSO capture 4 screens worth and you can scroll around in that record.  It is actually handy once in a while.  I understand not wanting to fill the full buffer every capture, but there's no reason that the capture should always be exactly one screen width.  But I doubt anything is going to change.

[kcbrown]

You are going at it from the wrong perspective and make things way too complicated in your mind. More memory means that the situation in which the display shows all (and the sample rate gets reduced) happens at higher time/div settings. And it is easy to see when there is not enough memory to show a full display at the full samplerate; at that point the samplerate gets reduced. Having some signal left & right just makes life easier for some use cases; no need to mess with settings or adjusting the trigger point before taking a measurement. If it turns out the trigger point isn't optimal you just scroll left/right. Stick the probe in your circuit and get a trace without needing to think upfront or re-measure when it turns out your assumption on the signal is wrong.

Unless the amount of time you needed to capture happens to exceed the amount of time the buffer represents at full sample rate.  Then you need to think about it up front.

Very handy especially when you are hunting rare events which may take an hour to re-occur. It allows for a way more relaxed approach to using an oscilloscope. And if there isn't enough memory then so be it.

One approach has you thinking about capture time and setting that up directly in advance, while the other has you guessing unless you do the math.

All you're really saying is that you won't have to guess most of the time.  And that's a fair point.

But you always have to set your timebase to something.  If you don't already know the characteristics of what you're capturing, then what exactly is determining what you set your timebase to in the first place?   On the Siglent, this is obvious: you set it to the timebase that gets you full buffer use in a single capture, and you're done.

See my autoranging DMM example. Autoranging works until you hit the limits and -for example- need to switch to the mV range manually.

Right.  But it's interesting to note that in both cases, if the amount of time you need to capture is in excess of the buffer's time width at full sample rate, then you have to go out of your way to set the timebase appropriately either way.

And that leads back to my question at the end: why aren't you arguing for something like a fixed (or, perhaps, minimum) ratio between displayed time and captured time, so as to ensure that you can always zoom out to see more no matter what your timebase setting is?  If the ability to zoom out and scroll around is so incredibly important as you claim, why is it that this ability isn't preserved under all conditions?

[nctnico]

But you always have to set your timebase to something.  If you don't already know the characteristics of what you're capturing, then what exactly is determining what you set your timebase to in the first place?   On the Siglent, this is obvious: you set it to the timebase that gets you full buffer use in a single capture, and you're done.

But that will lead to a situation where you can't see the details and thus needing more knob twisting to get to the interesting part of the signal.

And that leads back to my question at the end: why aren't you arguing for something like a fixed (or, perhaps, minimum) ratio between displayed time and captured time, so as to ensure that you can always zoom out to see more no matter what your timebase setting is?

That feature is already there: you can use zoom mode for that at the expense of screen real-estate but in such cases roll mode with a lot of memory + peak detect does the job just fine to catch enough details.

But all of this has been discussed at great length already so I'm going to put this subject to bed.

[kcbrown]

But you always have to set your timebase to something.  If you don't already know the characteristics of what you're capturing, then what exactly is determining what you set your timebase to in the first place?   On the Siglent, this is obvious: you set it to the timebase that gets you full buffer use in a single capture, and you're done.

But that will lead to a situation where you can't see the details and thus needing more knob twisting to get to the interesting part of the signal.

Unless you're using zoom mode.  Then you get the advantages of both.

Like I said, I think the standard approach (really, the "minimum ratio" approach) is better for scopes with small screens, and the Siglent approach is better for scopes with larger screens.

And that leads back to my question at the end: why aren't you arguing for something like a fixed (or, perhaps, minimum) ratio between displayed time and captured time, so as to ensure that you can always zoom out to see more no matter what your timebase setting is?

That feature is already there: you can use zoom mode for that at the expense of screen real-estate but in such cases roll mode with a lot of memory does the job just fine.

But then you have to zoom in (via zoom mode or after the fact) to see the details, which is precisely why you argue against the Siglent approach in the first place!

So this is no answer to my question.

You can't have it both ways.  Either you prefer being able to zoom out under all conditions irrespective of the timebase, in which case the standard approach only gets you an approximation of that, or you use zoom mode to get you a detail view, in which case your objection to the Siglent approach falls apart.

The Siglent wins for consistency, at least.  You do the same thing regardless of the time characteristics of the signal: set the timebase so that the screen shows the entire capture you're interested in.

[Anthocyanina]

I have an 1104z, and I have not felt i need something more advanced, more responsive, definitely, but, were i to have an unlimited budget for an oscilloscope, i would go with the keysight 4000x  series mostly because of its many protocol decoders. It seems to be the cheapest to decode USB 2.0 at 8.5k usd for the 4 channel version, and being able to easily debug USB 2.0 would be really nice, should the need ever arise! Have you considered something like that? which protocols would you like to have decoded for you by the oscilloscope? of course that's not the most important thing about the oscilloscope, just seems to be something that hasn't been talked about much in this thread, where, with 30k$, QoL things like that could probably make the difference all else being equal

[Someone]

[epic trolling]

You can't have it both ways.

Dont feed the troll! This has been going on for years? now.

There is no scope that has full control of acquisition memory vs screen display when not in zoomed mode. None. Either the user makes use of zoom to get full control, or they put up with whatever approximation/estimation/setting the manufacturer uses in normal/default mode.

I think the approach used by all keysight/lecroy/siglent makes most sense, less time spent adjusting memory depth to keep the scope in its highest throughput mode.

I understand not wanting to fill the full buffer every capture, but there's no reason that the capture should always be exactly one screen width.

There is a good reason that the capture should not overflow the sides of the display, it slows down the acquisition rate and increases blind time. Ideally it should be user controlled, which the zoom mode of just about every single scope does.

I have an 1104z, and I have not felt i need something more advanced.... which protocols would you like to have decoded for you by the oscilloscope? of course that's not the most important thing about the oscilloscope, just seems to be something that hasn't been talked about much in this thread, where, with 30k$, QoL things like that could probably make the difference all else being equal

Not much in the way of actual requirement/use cases from the OP, so hard to make any specific suggestions! Love how this thread has been the classic eevblog fanboys + trolling while the OP pays little attention and just does their own thing.

[nctnico]

I have an 1104z, and I have not felt i need something more advanced, more responsive, definitely, but, were i to have an unlimited budget for an oscilloscope, i would go with the keysight 4000x  series mostly because of its many protocol decoders. It seems to be the cheapest to decode USB 2.0 at 8.5k usd for the 4 channel version, and being able to easily debug USB 2.0 would be really nice, should the need ever arise! Have you considered something like that? which protocols would you like to have decoded for you by the oscilloscope? of course that's not the most important thing about the oscilloscope, just seems to be something that hasn't been talked about much in this thread, where, with 30k$, QoL things like that could probably make the difference all else being equal

IMHO decoding high speed busses with complicated protocols on top makes not much sense unless you are rolling your own hardware implementation. Interfaces like USB, MIPI, PCIe, etc are usually implemented using proven hard-IP blocks in the chip and have specific (matched impedance) routing requirements. If you have a problem you can use a HF (several GHz) oscilloscope or use board layout simulation to see whether you have adhered to the layout rules. Beyond that it makes more sense to monitor the protocol at the software side (for example Wireshark).

[KaneTW]

Sometimes you have an obscure signaling issue on a highspeed bus that's otherwise using proven building blocks. Arguably it makes more sense to loan a high-bw scope for a few days than to buy one in those cases unless you have a continuous need.