Is Bandwidth/memory depth a waste of money in oscilloscopes?

[rstofer]

I asked, specifically, whether bandwidth would be of any use to me. Having very specifically described my application, use, and period of use.

I also asked how much memory depth would affect that situation and why.

Those have pretty much been answered.  For audio work, you don't need bandwidth so just about anything will work.  100 kHz is more than enough.  I don't know what the slowest current DSO is but probably a lot faster than you need.  It may be on the order of 50 MHz for all I know.  I can't imagine anyone designing a new DSO with a  2 MHz bandwidth like my old Dumont.

Memory depth is a little trickier if you want a large number of samples per second and a very slow sweep rate.  A couple of megabytes are probably enough.  Usually sample depth is tied to sample rate and you sure won't be sampling a low bandwidth scope at  some huge number.

Remember 2.5 - that's the magic number for number of channels * maximum bandwidth per channel * 2.5 for Shannon.  A 4 channel 100 MHz scope * 2.5 will have a sample rate of 1 GS/s (minimum).

A 2 channel 50 MHz scope * 2.5 will have a sample rate of 250 MHz (minimum).  Such a rate would fill up a 10 MB memory in 40 ms so you want to have a sweep rate of 40 ms / 10 div or 4 ms/div (slowest).  I really hope I did the math right.  The scope will simply back off on the sample rate if you run slower sweep speeds.  It doesn't matter because there are only so many horizontal pixels anyway.

Here's an article on the subject:

https://www.testandmeasurementtips.com/memory-depth-and-sampling-rate-in-oscilloscopes/

You don't need a high sample rate for audio - 44.1 kHz is adequate to represent 22 kHz audio.  Here is an article on the subject:

https://www.masteringthemix.com/blogs/learn/113159685-sample-rates-and-bit-depth-in-a-nutshell

It wasn't all that long ago that sample memory was measured in kB.  I don't think I would spend much time on this concept.  Anything will be good enough for audio.

It might be nice if the scope had a decent FFT (and the Rigol doesn't but the Siglent seems to but I don't have one so I really don't know much about it) but if I wanted precision in an FFT, I would be looking at the Analog Discovery 2 with its 14 bit ADC.  I also tend to like 27" screens.

Don't get concerned about comments re: 8 bit ADC.  Even some very high dollar scopes are no better.  How many pixels in the vertical direction?  The DS1054Z is 800x480.  It might be nice to have 1 more bit but memory tends to come in multiples of 8 bits of width.  They do some tricks with AD to effectively increase the bit width.  In any event, DSOs don't have 1024x768 screens (usually).

Limited to audio, there is no point in buying much of anything.  A low $ 50 MHz scope will be more than enough.

[vk6zgo]

 Early DSOs & very cheap modern ones have very little memory, even if their fastest sample rate
is quite reasonable.

Back in the day, the Tek & HP reps would turn up at my old work, touting the latest offering.
After looking at some other waveforms, we would ask to look at a PAL video signal at field rate.

OK, that's 50Hz, so set the thing at about  2ms/div----- that should give a nice display of 10 divisions.

What's this?---- something which looks nothing like a video signal, more like my overgrown back yard.

DSOs with poor memory reduce their sample rate at long time/div settings.
That's no problem with sine waves, but complex waves like video signals have frequency components well above the sample rate----hence "alias city".

The guys from Tek/HP couldn't get their heads around the problem, & suggested all sorts of beaut workarounds, none of which were " worth a pinch of cockatoo's poo ", & went away, convinced we were "dinosaurs" because we didn't welcome the new technology with open arms.

Modern reasonably priced DSOs like the Rigol & its direct competitors do not have this problem, as they have sufficient memory.

Some "El Cheapos" are even worse than those old  mainstream manufacturers & show obvious aliasing
on the colour bursts when displaying a line rate PAL signal.

One such device was given an enthusiastic write up in "Silicon Chip " some years back, despite their screenshots showing this effect on a video signal.

Some will say "analog video is obsolete, so who cares?", but the same problem will raise its head whenever we are looking for Mains hum perturbations of a fast pulse train.(switchmode?)

[forrestc]

Most of my work would be in the low frequency area, audio, power supplies etc. I'd consider it was relatively easy work by oscilloscope standards.

My question is; given the work I expect to do, is the significant increase in cost between the manufacturers' offerings likely to provide me any real benefits, or should I just spend the difference on cheap wine and hookers?

I'm going to try to answer your question this way:

On my bench I have a Tek MDO3024, Mixed Domain oscilloscope.   200Mhz, 2.5Gs/second, 10Mpoints of memory.    Right now it's actually in spectrum analyzer mode since I was doing some EMI compliance work.   It replaced a tek TDS2014, and that replaced a TDS1002B before it.     The TDS1002B in particular is a 60Mhz scope, with 1GS/s and 2.5K points of memory.

So, lets start with Frequency response:  If you're only doing audio, and maybe some digital, you probably don't need more than a few tens of Mhz.  I don't think you really probably need more than 50Mhz or so.   I'm assuming you didn't bump into the 60Mhz response of your old Tek so getting a similar Mhz DSO should be fine.

As far as memory depth - If you're using the scope exactly as a replacement for your analog scope, you don't need more memory than you can see on the screen.   That TDS1002B would be a good example - I used it for years to measure intermittent one-time events.   So any scope would have enough memory depth for that.

Where the  extended memory depth helps is where you want to capture a period of time and then zoom in and around the data.   For instance, I can do a single shot trigger on my MDO3024 and then zoom in and out and look at different portions of the waveform, even those off the screen.   Someone else mentioned measuring time between events - for example, if you had two events say a second apart, you could use the deep memory to capture data for that period and then use cursors to measure the time difference.   In my MDO3024 I have the protocol decoders enabled, and occasionally use them.  For this it's very helpful to have deep memory since you need to capture the entire waveform of the protocol you're decoding, which can be over an extended period if you're talking long messages.   

However, none of the above applications seems applicable to "low frequency, audio, power supplies, etc."

The one area I can think of which a long sample  depth might be useful for you is if you are wanting to run a FFT (audo spectrum analysis) or other analysis on a signal.   But even then an extreme sample depth might not be all that helpful.  Note that even the cheapest DSO's today see to all have at least a few million points, so that really what you're needing to decide between is 'reasonable' and 'extreme' as opposed to 'too small' and 'reasonable'.

Maybe someone else can suggest some other areas where you might want a long sample depth - or longer than the cheapest of the scopes provide.

I'll add one more note:   I recently was in the market for a scope for the automated test system in the manufacturing plant.  Just needed something cheap to do some basic signal measurements.  Needed a mask test, some way to control the device via LAN or USB, etc.    After a side-trip with a siglent that didn't work for the application, I ended up buying a Rigol DS1054Z.    I'm impressed enough with the scope that I'm pretty certain that if I was wanting one for hobby work, this is the one I'd get.     I'm assuming that many of the other scopes out there will work just as well.  But for the type of work you're looking at I'd be more interested in what features you might want and whether there's any gotchas you're likely to encounter with what you will be doing.

One more thing I just thought of:  If you're interested in FFT, you may want to look at eevblog #845, being mindful that some of this might have changed since it was filmed, and may not include the scopes you're talking about.

[nctnico]

Input filtering is also a useful feature to get rid of unwanted noise.

[nfmax]

For audio work, you need a frequency response that goes up above the highest likely frequency of oscillation of the power output transistors in a class B amplifier, or enough to let you look at the switching times in a class D amplifier. Say 100 to 200 MHz. A wider bandwidth scope generally means a faster ADC sample rate, which for audio work means that in 'hi-res' mode (where successive ADC samples are averaged/combined together to give the effect of more bits & less noise at a lower sample rate) you get a wider bandwidth. You should do fine with a 1 GS/s 8-bit ADC for two channels (or two of them for four channels)

Memory depth is not particularly important for audio (except if it's so short the FFT resolution is too low to be useful). For audio, 50,000 to 100,000 samples per channel is fine.

What does matter, especially if you are used to an analogue scope, is waveform update rate. If you are playing audio (not sinewaves)  into a digital scope, too low a waveform update rate produces a 'freeze-frame' effect where what you see on the screen is a series of delayed, momentary snapshots of what you are hearing. Most disconcerting. A waveform update rate of 10,000 updates/second at the timebase rates you will be using is about the minimum I'd accept.

Of course, more is always better but as you are concerned about wasting money you may not want to go there...

[Fungus]

You should do fine with a 1 GS/s 8-bit ADC for two channels (or two of them for four channels)

That's a point worth mentioning: To look at amplifier outputs, etc. you need to connect a pair of probes and look at the difference between them. For stereo that means a four-channel oscilloscope.

[David Hess]

Since it has delayed timebase capability and peak detection, it does not need a large record length as much as a modern DSO which lacks these features.

All modern DSOs have delayed time bases! That has been pointed out before. You can set the trigger point outside the screen and there you have it.

It has nothing to do with the screen or display record.  None of the single timebase DSOs I have evaluated allowed the trigger to be moved outside of the waveform record.  But if the waveform record is long enough for the application, then this is not a problem except perhaps for performance when processing a long record length.

I just tried with my GW Instek GDS2204E and it does exactly what you describe is not possible. I could try some of my other scopes as well but I'm pretty sure I get the same result. I get the feeling you have tried a few very limited DSOs and now map that information on all other DSOs. Doing a delayed trigger is pretty easy to implement in a DSO so there is no reason not to have it.

The GW Instek GDS2204E fits in the exception I deliberately mentioned in my post which you apparently did not bother to read or did not understand:

The only way I have found to reliably know what a DSO can do is look for the B trigger capability; if it is lacking, then the DSO is limited to pan and zoom within the waveform record.

Guess what, the GW Instek GDS2204E has B trigger capability although they do not call it that on page 145 of the manual.  So good for GW Instek and thanks for supporting my point.

Are there any less expensive DSOs which have this capability?


As far as record length and sample rate, I have been disappointed with the display fidelity of every DSO compared to an analog oscilloscope.  (1) I would have expected large record lengths to improve the situation but if there has been improvement, then it has been small for general purpose applications.  I do not consider a large record length to be advantageous for display fidelity especially if it slows down the display processing speed.

My compromise has been to rely on peak detection or DPO (digital phosphor oscilloscope) operation and just live with it.  My 2230/2232 has a terrible display in this respect since it lacks index grading but most DSOs which support index grading handle it poorly and the 2230/2232's very high display resolution compared to modern DSOs somewhat makes up for it.

Rohde & Schwarz make some DSOs which apparently address this issue but I have not had a chance to evaluate them.  They wrote a great application note on the subject which described what they did not long ago but I read it once and have not been able to find it since.

Basically, I would ignore this issue because there is little to be done about it except insofar as the instrument you pick supports peak detection which avoids the worse problems.  Then learn to use the DSO to best effect within its display limitations.  Ignore marketing's record length wars.

For audio work, an old oscilloscope I might consider is the Tektronix 2225 because of its 5MHz bandwidth limit and 500uV/div sensitivity.  The Tektronix 2210/2211 is the same oscilloscope but with digital storage capacity.  But I do not recommend these because they are old and the DSOs lack peak detection.

If there was a modern low cost DSO which supported network analysis, I might recommend it.  Keysight has something but it is neither inexpensive nor does it have acceptable performance; it is a toy.  The closest I am aware of are from Cleverscope and Syscomp Electronic Design but they are USB based instruments:

https://cleverscope.com/products/
https://www.syscompdesign.com/products/

I would not consider 8-bit DSO FFT capability as useful for audio; at best it is only sufficient for low fidelity.

(1) Try doing tangential measurement on a DSO; the ones which support index grading are actually *worse*.  This just demonstrates that they are not making a serious effort to duplicate an analog display; they only make a serious effort to say they are.

I do not even think DSOs *should* duplicate an analog display as there are better ways but as it is, they are not doing either.

[nctnico]

Since it has delayed timebase capability and peak detection, it does not need a large record length as much as a modern DSO which lacks these features.

All modern DSOs have delayed time bases! That has been pointed out before. You can set the trigger point outside the screen and there you have it.

It has nothing to do with the screen or display record.  None of the single timebase DSOs I have evaluated allowed the trigger to be moved outside of the waveform record.  But if the waveform record is long enough for the application, then this is not a problem except perhaps for performance when processing a long record length.

I just tried with my GW Instek GDS2204E and it does exactly what you describe is not possible. I could try some of my other scopes as well but I'm pretty sure I get the same result. I get the feeling you have tried a few very limited DSOs and now map that information on all other DSOs. Doing a delayed trigger is pretty easy to implement in a DSO so there is no reason not to have it.

The GW Instek GDS2204E fits in the exception I deliberately mentioned in my post which you apparently did not bother to read or did not understand:

The only way I have found to reliably know what a DSO can do is look for the B trigger capability; if it is lacking, then the DSO is limited to pan and zoom within the waveform record.

Guess what, the GW Instek GDS2204E has B trigger capability although they do not call it that on page 145 of the manual.  So good for GW Instek and thanks for supporting my point.

Are there any less expensive DSOs which have this capability?

Even the Rigol DS1054Z has this feature according to the manual. Sorry but you are really  here. I can't imagine why a DSO wouldn't have delayed triggering.

Edit: Delayed trigger also works on my MicSig TO1104

[David Hess]

The GW Instek GDS2204E fits in the exception I deliberately mentioned in my post which you apparently did not bother to read or did not understand:

The only way I have found to reliably know what a DSO can do is look for the B trigger capability; if it is lacking, then the DSO is limited to pan and zoom within the waveform record.

Guess what, the GW Instek GDS2204E has B trigger capability although they do not call it that on page 145 of the manual.  So good for GW Instek and thanks for supporting my point.

Are there any less expensive DSOs which have this capability?

Even the Rigol DS1054Z has this feature according to the manual. Sorry but you are really  here. I can't imagine why a DSO wouldn't have delayed triggering.

The Rigol DS1000Z series is what brought this issue to my attention.  It is my prime example of a modern DSO which does not.  Check the link I gave earlier which describes the sample rate falling as the time/div decreases.

[nctnico]

I guess the DS1000Z is an exception then but I find it hard to understand why it can't do it because it is so simple to implement.

[rf-loop]

Since it has delayed timebase capability and peak detection, it does not need a large record length as much as a modern DSO which lacks these features.

All modern DSOs have delayed time bases! That has been pointed out before. You can set the trigger point outside the screen and there you have it.

It has nothing to do with the screen or display record.  None of the single timebase DSOs I have evaluated allowed the trigger to be moved outside of the waveform record.  But if the waveform record is long enough for the application, then this is not a problem except perhaps for performance when processing a long record length.

Where this becomes a problem, and this was discussed on these forums more than a year ago, is things like measuring jitter in a 1 pulse per second GPS signal.  20M points over 1 second is only 20MS/s yielding 50 nanosecond resolution which is worse than the jitter of many GPS 1 PPS signals.

Here is example with cheap  Siglent SDS1104X-E

It allow trigger to be placed outside of the waveform record as can see in attached images.
What is difficult here. Just very normal perhaps with most modern oscilloscopes.

1pulse/s, pulse width 100ns and around 15ns rise and fall times and around 3V amplitude.
Trigger normal and rising edge. Acquisition mode normal. Sinc iterpolation and display mode vectors.
Trigger position far out from display and also far out from acquisition lenght.
Acquisition length here is 14ms (14M memory) and sample speed 1GSa/s
Whole 14M is between screen leaft and right border, no overlapp.
image a
Then just scope running with this 1s delay
selected zoom and zoomed in so that pulse is well visible. It can zoom also more but it do not give any advantage for example pulse to pulse jitter measurements.

I can tell that this scope is not suitable for characterize example GPS 1pps jitter. Scope own reference  have too much jitter. (just for random readers thinking: 1us in 1s period is 1ppm what is far away what need if want really characterize good GPS 1pps jitter. We need far better. Here in zoomed window one time division is 50ppb from delay and scoppe own short time jitter is roughly around this class...not enough for this purpose. And scope can not use external reference clock)

Also even my old ancient Hewlett-Packard boat anchor 545xx series DSO's can do it (even better - least amount of max delay is even more and due to better reference clock also jitter is less ). 

Note, both images are snap shots from live running scope "on the fly"

[RoGeorge]

The GW Instek GDS2204E fits in the exception I deliberately mentioned in my post which you apparently did not bother to read or did not understand:

The only way I have found to reliably know what a DSO can do is look for the B trigger capability; if it is lacking, then the DSO is limited to pan and zoom within the waveform record.

Guess what, the GW Instek GDS2204E has B trigger capability although they do not call it that on page 145 of the manual.  So good for GW Instek and thanks for supporting my point.

Are there any less expensive DSOs which have this capability?

Even the Rigol DS1054Z has this feature according to the manual. Sorry but you are really  here. I can't imagine why a DSO wouldn't have delayed triggering.

The Rigol DS1000Z series is what brought this issue to my attention.  It is my prime example of a modern DSO which does not.  Check the link I gave earlier which describes the sample rate falling as the time/div decreases.

Not sure if I properly understood the complaint, but Rigol DS1054Z can do something very similar with delay triggering, and still keep the high sampling rate.

Here is a video for a signal made out of 10ns pulses spaced half a second apart, and watched at 1 GSa/s at 1 second or 0.5 seconds after triggering. The video was made to show how phase noise can lead to jitter that accumulates with the distance in time between the triggering point and the observed point. As it can be seen, the sampling rate is kept at 1GSa/s all the time, and the delayed is varied to 0s, 1s and 0.5s (the number in the upper right corner of the display indicates the delay after trigger).

[nctnico]

So it seems the DS1054Z can do delayed triggering.

[emece67]

.

[Fungus]

So it seems the DS1054Z can do delayed triggering.

Yep. It's not a special thing you need to enable (which is maybe why it's not mentioned in the manual), you simply move the trigger point to the left and keep on going.

There's an on-screen box dedicated to it (top-right, labelled 'D' for 'delay'). The farthest I could get it to go was 500s.



(Update: It works, too. I left it there and stuff started appearing on screen after 8 minutes or so)

The farthest I could get it to go at 1GS/s and full memory was 1 second. That's way outside the 24MSample memory but maybe not a massive delay in real terms. What delays can other 'scopes manage? 

[Fungus]

Then just scope running with this 1s delay

Is one second the maximum delay on the Siglent, too?

[nctnico]

I think I start to understand where the confusion is coming from. David Hess is looking for (delayed) B-trigger. This effectively means having 2 different time bases (time/div). On a DSO you can mimic that behaviour by using the zoom mode which will show the same signal using two differen time bases but the amount of time you can move left/right is indeed limited by the record length and sample rate. Another way around it would be using a reference trace at a long time/div setting and then trigger on a specific part of a signal using the delay and a much shorter time/div setting. On my GDS2204E I can use this trick to have traces with different time/div settings on screen at the same time. It allows to do cursor measurements on the reference trace taken at 500ms/s and do the same for the acquired trace at much shorter time/div settings. This however depends on how well the reference traces are implemented. I think this may not work on Keysight oscilloscopes because as far as my experience goes the references traces are just pixels on the screen.

[Pete F]

Without wanting to OT the OT, thanks for the contributions on the questions I had.

The bandwidth I feel I understand pretty well (better than some here it seems; oscilloscopes don't suddenly stop seeing inputs just because they've reached their specced bandwidth limit!), but didn't know if maybe the DSO situation is different to an analogue 'scope. It seems it's just the same so happy days there.

It was the memory depth that was very confusing for somebody like myself trying to get my head around the modern DSO and my own particular use. This is an area I've never been exposed to, and while I understand the whole concept of the sampling etc, it's transferring the theory to my practical applications where without practical experience I wasn't sure of the effects of different memory sizes.

The FFT could definitely be something that I would find useful if it's actually a serious tool and not just a bit of a gimmick, I suspect the latter. I used to have access to a spectrum analyser, but that was more for RF work anyway, but yes maybe it would be good. From what I have read it's not really suitable to serious audio work where THD is being chased etc, but hey if it's a function that's there, why not!

"Where this becomes a problem, and this was discussed on these forums more than a year ago, is things like measuring jitter in a 1 pulse per second GPS signal.  20M points over 1 second is only 20MS/s yielding 50 nanosecond resolution which is worse than the jitter of many GPS 1 PPS signals."

Ok I do have to ask, out of curiosity, why would you be measuring the jitter of a GPS signal? Many years ago my apprentice built a master clock oscillator for our workshop and we used the Omega signal as the phase master to drive the local clock. However I'm interested in why you'd be measuring that GPS jitter outside of navigation needs.

I have no idea why somebody would think working on a stereo audio system requires 4 inputs, that makes no sense. In fact most of my work was done by probing with a single channel, and that put bread on the table for many years without the world spinning off its axis. I've never really had much (or any I can think of) need for more than 2 channels.

While "more is better" in terms of bandwidth, I'm sometimes concerned that at the bottom end of the market (ie where I'm looking) models may be pushed beyond their real practical limits in order to provide better specs. In other words, a series of oscilloscopes, say 50, 100, 200 MHz. While the 200 MHz 'scope may indeed meet the -3db spec for its front end, the rest of the machine may not really be a 200 MHz oscilloscope and is just pushed up there because it looks good on paper. I don't know if that's a fact, but is definitely a concern. I typically find the middle of a manufacturer's product lineup tends to provide the best value, but that's hardly a hard and fast rule and simply a broad generalisation.

It may well be that the final arbiter on the decision will simply be the local availability of a brand/machine. While most of the mentioned brands are available, I'm not interested in paying (literally) double the price for a local bricks 'n' mortar company to courier out a purchase, when I can get the same thing through a purely online retailer. That may however limit my options. At the end of the day it's not like this is a $100,000 decision. The difference one way or the other may be a few hundred dollars from one extreme to the other, and from the advice here it seems for my application it won't make all that much difference in any case. I'm very busy with work, so I'll probably just go with something I think looks like reasonable value and move on. I'm now very keen to have the opportunity to use a DRO, as has been mentioned by others they really are a powerful piece of kit.

[tautech]

It was the memory depth that was very confusing for somebody like myself trying to get my head around the modern DSO and my own particular use. This is an area I've never been exposed to, and while I understand the whole concept of the sampling etc, it's transferring the theory to my practical applications where without practical experience I wasn't sure of the effects of different memory sizes.

It's all about the ability to convert data points into an accurate representation of the signal, the more there are the less chance of interpolation inaccuracies. DSO's with good amounts of memory depth don't even need a vector display and can show a good replication from just the data points displayed.

"Where this becomes a problem, and this was discussed on these forums more than a year ago, is things like measuring jitter in a 1 pulse per second GPS signal.  20M points over 1 second is only 20MS/s yielding 50 nanosecond resolution which is worse than the jitter of many GPS 1 PPS signals."

Ok I do have to ask, out of curiosity, why would you be measuring the jitter of a GPS signal? Many years ago my apprentice built a master clock oscillator for our workshop and we used the Omega signal as the phase master to drive the local clock. However I'm interested in why you'd be measuring that GPS jitter outside of navigation needs.

This could be a concern if you ever go down this rabbit hole:
https://www.eevblog.com/forum/metrology/an-advanced-question-sampling-an-oscillators-signal-for-analysis/

I have no idea why somebody would think working on a stereo audio system requires 4 inputs, that makes no sense. In fact most of my work was done by probing with a single channel, and that put bread on the table for many years without the world spinning off its axis. I've never really had much (or any I can think of) need for more than 2 channels.

Yeah I get that but there's at least a couple of examples why
Differential measurements
Data protocol decoding.

While "more is better" in terms of bandwidth, I'm sometimes concerned that at the bottom end of the market (ie where I'm looking) models may be pushed beyond their real practical limits in order to provide better specs. In other words, a series of oscilloscopes, say 50, 100, 200 MHz. While the 200 MHz 'scope may indeed meet the -3db spec for its front end, the rest of the machine may not really be a 200 MHz oscilloscope and is just pushed up there because it looks good on paper. I don't know if that's a fact, but is definitely a concern. I typically find the middle of a manufacturer's product lineup tends to provide the best value, but that's hardly a hard and fast rule and simply a broad generalisation.

Sorry to be unkind: 
The modern instrument that is part of a series is HW and/or SW BW limited, the many brands that are now BW cracked are proof of this.

It may well be that the final arbiter on the decision will simply be the local availability of a brand/machine. While most of the mentioned brands are available, I'm not interested in paying (literally) double the price for a local bricks 'n' mortar company to courier out a purchase, when I can get the same thing through a purely online retailer. That may however limit my options. At the end of the day it's not like this is a $100,000 decision. The difference one way or the other may be a few hundred dollars from one extreme to the other, and from the advice here it seems for my application it won't make all that much difference in any case. I'm very busy with work, so I'll probably just go with something I think looks like reasonable value and move on. I'm now very keen to have the opportunity to use a DRO, as has been mentioned by others they really are a powerful piece of kit.

Only you can make the decision to buy abroad and then risk possible substandard warranty support.
Your money, your risk.

[Pete F]

I buy a ton of stuff abroad ... well actually tonnes if a literal interpretation is made I can honestly say I've had more hassles with local retailers/distributers than the foreign distributers. In my experience the Australian warranty is generally a joke, where the distributer will use all manner of BS technicalities to either claim it was my fault (hence no warranty) or little more than semantics to weasel out of the claim. Obviously that's not always the case, and I've had good experiences too, very good in some cases, but my experience is that's the exception rather than the rule. I wish that wasn't the case, but that's just been my life's experience. In both cases I consider the warranty zero on a product like this, hence that is how much the premium is I'd be prepared to pay. It's a disposable consumer product and nothing more.

I have no idea why somebody would think working on a stereo audio system requires 4 inputs, that makes no sense. In fact most of my work was done by probing with a single channel, and that put bread on the table for many years without the world spinning off its axis. I've never really had much (or any I can think of) need for more than 2 channels.

Yeah I get that but there's at least a couple of examples why
Differential measurements
Data protocol decoding.

While "more is better" in terms of bandwidth, I'm sometimes concerned that at the bottom end of the market (ie where I'm looking) models may be pushed beyond their real practical limits in order to provide better specs. In other words, a series of oscilloscopes, say 50, 100, 200 MHz. While the 200 MHz 'scope may indeed meet the -3db spec for its front end, the rest of the machine may not really be a 200 MHz oscilloscope and is just pushed up there because it looks good on paper. I don't know if that's a fact, but is definitely a concern. I typically find the middle of a manufacturer's product lineup tends to provide the best value, but that's hardly a hard and fast rule and simply a broad generalisation.

Sorry to be unkind: 
The modern instrument that is part of a series is HW and/or SW BW limited, the many brands that are now BW cracked are proof of this.

Yes I absolutely agree, there would be many occasions when somebody working in specific areas would want 4 (or many more channels), I wasn't one of them, but that was just me. But the person was specifically referring to the requirement for a 4 channel oscilloscope in order to work on stereo audio, and I'm afraid that's just complete crap.

Just because the front end of an oscilloscope is capable of xxx MHz according to the definition of that, doesn't mean it is truly an xxx MHz in a practical sense. Whether they've got to that figure by hacking or by original spec. I have used a number of oscilloscopes where up around their specified bandwidth they were all but useless from a practical perspective. So no, it wasn't BS, and if you want to believe otherwise knock yourself out. This is one area where I've found there is a big difference between a high quality manufacturer and the bottom end of the market that I'm currently looking at. The quality manufacturers sell their equipment based on a fitness for purpose, and if it's sold as being capable of xyz, chances are that's what it will do. The manufacturers of cheap products know many people buy products on paper spec and will often grossly exaggerate their claims or use nonsense "specs" to try to win a supposed advantage over competitors. I have no idea if that's the case with these DSOs, but it certainly wouldn't surprise me if it was the case. Indeed I'd be more surprised if it wasn't.

[tautech]

I buy a ton of stuff abroad ... well actually tonnes if a literal interpretation is made I can honestly say I've had more hassles with local retailers/distributers than the foreign distributers. In my experience the Australian warranty is generally a joke, where the distributer will use all manner of BS technicalities to either claim it was my fault (hence no warranty) or little more than semantics to weasel out of the claim. Obviously that's not always the case, and I've had good experiences too, very good in some cases, but my experience is that's the exception rather than the rule. I wish that wasn't the case, but that's just been my life's experience. In both cases I consider the warranty zero on a product like this, hence that is how much the premium is I'd be prepared to pay. It's a disposable consumer product and nothing more.

I find that truly sad that you have experienced such bad support. In some way this forum is a good place to address such poor experiences, by naming and shaming !

Just because the front end of an oscilloscope is capable of xxx MHz according to the definition of that, doesn't mean it is truly an xxx MHz in a practical sense. Whether they've got to that figure by hacking or by original spec. I have used a number of oscilloscopes where up around their specified bandwidth they were all but useless from a practical perspective. So no, it wasn't BS, and if you want to believe otherwise knock yourself out.

With a complex waveform at near rated BW there's no doubt some scopes will be better than others but rated BW is only using a pure sine wave to where no more than 3 dB attenuation occurs. That's universal to all brands.

This is one area where I've found there is a big difference between a high quality manufacturer and the bottom end of the market that I'm currently looking at. The quality manufacturers sell their equipment based on a fitness for purpose, and if it's sold as being capable of xyz, chances are that's what it will do. The manufacturers of cheap products know many people buy products on paper spec and will often grossly exaggerate their claims or use nonsense "specs" to try to win a supposed advantage over competitors. I have no idea if that's the case with these DSOs, but it certainly wouldn't surprise me if it was the case. Indeed I'd be more surprised if it wasn't.

There's a poultice of info on all manner of models and brands right here in this forum that demonstrates what some DSO's are capable of and others aren't. You of course don't have any idea if the replies you get are from members who don't have a clue or really know there stuff. 
I've had a good # of CRO's in days gone by and when I got my first decent DSO (TDS2012B (was then)) and fixed up a couple of other Teks, TDS210 and TSD1002 I was then sold on DSO's and just knew they'd take over the hobbyist market before much longer.
Shortly after I looked at Hantek and many others and kept looking for something better until a few years back I was offered Siglent.
You surprise me when you state you consider a cheap DSO a throw away consumable and in this day and age I get that, however not in my book and marketplace experience and product reliability has shown me this is indeed not the case.
For many of my customers a DSO is a big expense and the models I choose to market (not bottom of the barrel) are all dearer than the prices you've indicated previously and yet with all the scope experience you have I wonder why you've set your sights at real entry level equipment. 
I would urge you to set your sights higher.

[Specmaster]

Sitting reading through all of these postings you have been given loads of good advice but no one in my opinion seems have got it right. Having myself only been used to analogue scopes for many years , I have just acquired a really good combiscope that now gives me the best of both worlds and is perhaps the best way into the world of digital scopes as you chose which ever mode seems to be most effective for what you want to do at the time without a major investment.

Personally if I was you I'd spend a little time on checking out what has gone wrong with your old scope and it is more than likely that simply checking the caps and replacing the bad ones would bring it back into life again and also spending a little more replacing all the caps would more than likely produce a scope that would once again be a reliable and useful tool on your bench and thus leaving the way forward to perhaps looking at and buying the best DSO that you can afford and then you would again have as mentioned above, the best of both worlds.

[nctnico]

Ok I do have to ask, out of curiosity, why would you be measuring the jitter of a GPS signal? Many years ago my apprentice built a master clock oscillator for our workshop and we used the Omega signal as the phase master to drive the local clock. However I'm interested in why you'd be measuring that GPS jitter outside of navigation needs.

Very precise time distribution is one of the applications to measure jitter on 1PPS outputs. There are lots of applications for this. For example having several A/D converters 10km apart and be able to timestamp their samples with less than a nanosecond uncertainty.

I have no idea why somebody would think working on a stereo audio system requires 4 inputs, that makes no sense. In fact most of my work was done by probing with a single channel, and that put bread on the table for many years without the world spinning off its axis. I've never really had much (or any I can think of) need for more than 2 channels.

You don't have to use all channels on a 4 channel scope. Once you have the 2 extra channels you will start using them and it will be more convenient.

[Fungus]

I have no idea why somebody would think working on a stereo audio system requires 4 inputs, that makes no sense.

To measure many components of an audio system you need to look at the difference between input and output voltage, not the difference between input voltage and ground (for example).

eg.  Loudspeakers, the outputs of an amplifier.

That means a single measurement needs two channels.

In a stereo system you might want to look at both channels simultaneously - to make sure they're equal and electrically isolated. That's four channels.

Or you might want to look at the input and output of a crossover circuit simultaneously on screen. That's four channels.

In fact most of my work was done by probing with a single channel, and that put bread on the table for many years without the world spinning off its axis. I've never really had much (or any I can think of) need for more than 2 channels.

Sure, but that doesn't mean you can't do it much better.

[Pete F]

Yes it is very sad that's my experience in Australia, and why I have no sympathy for the Australian retail chain when it comes to online shopping. Don't get me wrong, I'm not one of those people who picks the brains of a retailer and then goes off and buys it cheaper online. I don't have anything to do with them and do my own research, part of which is why I'm here, and am effectively invisible to the local retailers. As far as "naming and shaming", that's not appropriate as firstly my experiences are from many different sectors and secondly I have witnessed, first hand, the enthusiasm with which some companies will commence legal action for defamation to bully critical comments to silence, despite the fact they have no hope of winning the case. Others should seek their own legal advice if presented with this situation, however I believe one of the tenets of a defamation action is that as the defendant you are forced to defend the case. If you do not you will, by default, lose the case. In Australia we have no SLAPP legislation or equivalent. All very much OT and not something I want to get in to, but trust me when I say, again in my opinion, it's generally not worth going there. Take it up with the party, and if they continue to behave in a way you feel is inappropriate then vote with your feet. ... as I have done. That's my opinion and other's opinion may well be different.

I can have a functional DSO landed on my desk for a little over A$300. Many people would spend that on a couple of nights out. Less than 3 tanks of fuel. Half of a weekend away ... shall I go on? One of what I consider an outright scam I have witnessed locally is that it is my responsibility to pay for a courier for the faulty product to and from the distributer. At the very least to them. So they've sold me something that has failed within "warranty" and I'm expected to pay to courier the product to them? On a $3-400 purchase that is a significant percentage of the price. Now whether some here think a $350 DSO is good or bad is not the point, that is their opinion and thousands of people have coped just fine with their buying decision. Yes if I was you I'd "set my sights higher" and buy a more expensive oscilloscope. However I'm not you, I'm me So I can be equally aghast as to why somebody would want to buy something that is far more expensive than another one that will be perfectly adequate for the job. I've said many times that this is no longer my occupation and quite possibly won't get much use. It's also not especially important to me. I do want to ensure the purchase will indeed do the job, but I think pretty much anything will.

While I'm not a wealthy person, nor am I destitute! If this was an important thing for me I would be quite happy to spend a significant amount of money. But it's not, so I won't!  I'm surprised that some people have difficulty in empathising with other people's perspective and appreciating that what may be important to them may not be important to others.

As far as jitter, yes I appreciate what may be a theoretical use, I was involved with this area in regard to jitter in clocks, hence why I asked. But I wonder how many people ACTUALLY have a need to date stamp in this case an A/D converter to "nanosecond accuracy" off a GPS signal.

If of interest to where I'm at with all this, I'm currently considering a Siglent SDS1202X-E. The irony of that is despite all I've said above, it's waaay over specced for basic audio/PS work I mentioned. It's also from an Australian retailer, one in fact I just bought another piece of test equipment from. Whether that eventuates we'll have to see, but I'd like to get something here before much longer. Despite being seen as an odd candidate considering my comments, I always like to keep an open mind and try not to be biased one way or the other. One of the significant considerations was the fact that many of the models popular with members here are actually quite old models. They may well have been updated, but the "bones" still seem quite ancient to me. The Siglent is a relatively new design, offers serial decodng, and I was doing some I2C work a while back where that may have been handy (probably not, but hey it's available), it has relatively generous memory for an oscilloscope in that price range, etc etc etc We'll see. The bottom line is the most basic model will cost me around A$320. The 1202 will be around twice the price. However in real terms the difference is as above; 3 tanks of fuel etc. I take the train and am a general tight-wad with money, so don't begrudge spending money if I can see the value in doing so.

As I mentioned I have been around the block a few times here and had relatively good equipment available to me, that included 4 channel oscilloscopes. I didn't use 4 channels then so see no reason why I would suddenly see a need to have one now.