DSO bench scopes with BIG buffer memories??

[Chris Wilson]

Some of the specialized automotive USB scopes, like the current 4000 series Pico automotive scopes, (see http://www.picoauto.com/automotive-oscilloscope.html ) boast 32 megs of buffer memory, do any DSO bench scopes have buffers approaching or exceeding this figure? Just curious as most seem to be 1 meg tops. Thanks.

[EEVblog]

Sure, but what's your budget?
All the major scope maker have one with that much memory, but you pay huge $$$$$
That's only an 80MS/s scope, that's why 32Megs is cheap and easy to add.

These go up to 1Gpoint:
http://www.home.agilent.com/agilent/product.jspx?nid=-34115.910969.00&cc=US&lc=eng

Agilent do 32Mpoint cheap in a USB scope too, but at 1GS/s:
http://www.home.agilent.com/agilent/product.jspx?nid=-34492.774931.00&cc=US&lc=eng

Dave.

[Chris Wilson]

Thanks Dave, I had failed to find bench scopes with that sort of memory depth. I am using a UK sourced Datamane 150 megs USB scope right now, but 3 rhings against it make me want to try something else.

1: Only 15k memory depth

2: Not fully isolated and floating inputs, I work on some historic race cars with early ECU's that are poorly protected, often having the original software for mapping lost or incapable of running on none bespoke hardware, and therefore I am in fear of damaging one.

3: USB scopes can be a PITA as I worry about all the cabling, my laptop getting bust, and whether the battery is going flat, or the laptop is plain somewhere else.

I thought I would try a bench scope, as 90% of the time these cars are in the workshop and portability of equipment is not really an issue. I would like to be able to connect a bench scope to a BIG monitor though, as it makes working on a loom or engine much safer and easier if you aren't having to peer at a small screen as well as fiddle with moving or hot parts. I believe quite a few have a VGA output port these days? My budget certainly won't run to the big Agilent in your link, realistically I am probably looking at used, up to a max of £1200 UK pounds, about $1800 Aus.

Finding a bench scope that meets the requirements is proving hard as I have very limited knowledge of scopes. I am unclear as to how much memory depth I need, versus bandwidth of the scope.

I am also unclear how my current, and other,  scope's "Roll Mode" works. I asked this question on the makers forum, but haven't had a reply yet:

"In automotive diagnostics we are often measuring what I call very slow waveforms, often in the sub 800 Hz region, sometimes down to 8Hz. Viewing such slow waveforms and seeing any glitches is proving tricky, so a while ago I bought the Roll Mode upgrade from you. I have been paying with it using an old signal generator I have, giving a square wave signal at 2 to 600 Hz, and capturing various sets of data at different speeds via the roll mode. By means of roll mode it is much easier to see all of the waveform over extended periods. But what effect does running out of internal memory buffer at these low frequencies have on the stored signal? When the red line has crossed the screen I believe this shows the buffer is full? After that, what is occurring? is the buffer plain full, and stays full, and the rest of the data is caught in real time on the PC? Or is the buffer emptied very fast and it refills? I assume the former as no glitch is evident in the traces. What I am getting at is when the buffer is filled, at these low frequencies will roll mode continue to store accurate traces and be reliable for seeing glitches from missed or corrupted injection or spark events, for example? To what sort of frequencies will roll mode store useful data once the buffer is filled? Apologies if the questions are naive, but the more I learn unfortunately the more questions I seem to have. Thanks."

Does Roll Mode in any way take the place of deep memory for automotive type measurements where frequencies are relatively low?

Great forum, thanks for your videos Dave, I look forward to them, and good luck with the new lab, at least you aren't paying rent on it, good move!

[wkb]

2: Not fully isolated and floating inputs, I work on some historic race cars with early ECU's that are poorly protected, often having the original software for mapping lost or incapable of running on none bespoke hardware, and therefore I am in fear of damaging one.

Historic race cars, now that is interesting!     Would you care to write a bit on that subject?  A different forum thread would be best  I think.

[alm]

I don't see the relation between long memory and roll mode. Long memory allows you to capture a bunch of data in one shot and then zoom in on a small part of this data. Roll mode only shows you a screen full of data, which is maybe a few hundred pixels wide.

Note that fully floating inputs are quite rare on bench scopes, the majority have ground referenced inputs. This means that the ground lead is attached to the earth pin in the mains plug. The only exception I can think of is the Tek TPS2000 series, but it sucks in most other regards. It tends to be more common in handheld scopes, though many of the cheap ones have a common ground between all channels.

Differential probes are used to convert grounded inputs to floating, but they tend to be somewhat pricey.

[Chris Wilson]

I don't really understand the relationship between roll mode and memory depth, to be honest, I kind of thought roll mode was slow acquisition, whilst buffer memory was a short but intense saving of fast data.  I don't believe I really have my head around it yet

[IanB]

I don't really understand the relationship between roll mode an memory depth, to be honest, I kind of thought roll mode was slow acquisition, whilst buffer memory was a short but intense saving of fast data.  I don't believe I really have my head around it yet

I'm only just getting a grasp of this myself. I'm probably going to do an experiment with my new Rigol scope to explore its capabilities and try to put what I've read and learned about sample depth into practice (and to see how well the Rigol works compared to demonstrations I've seen on more expensive scopes). I'll post the results here when I have them.

[Chris Wilson]

IanB: Yes, please let us know how you find it. I am about to purchase a little gizmo that imitates an engine's sensor outputs, including crank and cam position sensors, and the various temperature and pressure sensors. Firstly to allow me to bench test new ECU and loom installations, and secondly so I can feed semi realistic data to my `scope so I can get a handle on its controls with the sort of data going to it I expect to be working with. I am not sure how one would introduce a transient glitch though, if someone came up with a way of electronically saving real world data from cars one could fiddle with your own test equipment to see how it handled the data in real time, it would be a useful training aid. Running engines for hours on end and trying to mess with the `scope is not very environmentally friendly, and I would rather be sat at my desk looking at healthy data is not as exciting as looking for "issues" either.

[IanB]

OK, here are some tests I did with the long sample memory on the Rigol DS1102.

I am looking at a pulsed signal with a 2 second period:



Superimposed on this signal is a higher frequency signal looking here like a dense block of noise. If I use the normal 16k sample memory, then when I zoom in to see the detail, it just looks like random noise:



However, if I capture the same trace using the large 1M sample memory, then when I zoom in I can see an actual sawtooth waveform:



I can measure this waveform to get its frequency (presumably from a DC-DC converter):



Another way to see more detail is to use the delayed capture, or "picture-in-picture" feature. This allows you to select a small sample window within the main scan to be captured in more detail:



In the image above I captured a window near the trigger point, but I didn't have to. I could have moved the position and size of the detail window to anywhere within the main sample interval.

Overall it seems the Rigol might not be as sophisticated as more expensive scopes, but the basic capability is there and functional.

[0xdeadbeef]

The better (and more expensive) LeCroy scopes had a very large buffer memory for years. An older (5years or so) WaveRunner scope I often use at work has 12MPts (no options installed, so it was probably not the maximum).
The current WaveSurfer line can have up to 32Mpts (default: 16Mpts) and the WaveRunner line even 64Mpts (default: 32Mpts).

BTW: as nobody mentioned it yet: even the very affordable Owon SDS scopes have 10MPts (full speed up to 100MHz). Of course you can't compare them to even an older LeCroy in any other aspect, but the sample memory is really great.

[chscholz]

The current WaveRunner 6Zi scopes (released about a year ago) go up to 128 Mpts.
You can actually USE all the memory that you buy, LeCroy does not distinguish between "acquisition" and "processing" memory like other vendors do.

[0xdeadbeef]

The current WaveRunner 6Zi scopes (released about a year ago) go up to 128 Mpts.

Sure about that? The LeCroy page says 64 Mpts/Ch ...
Anyway, our 6Zi at work has "only" 32MPts and I never came into a situation where I needed more. Even the 12MPts of its predecessor where more than sufficient.

[saturation]

Nice demo IanB of the value of 'long memory' and how to use the Rigol's features.  The more a user is familiar with all its features, you can squeeze performance out of it.   I use the same techniques you showed quite often to zoom into regions of interest and tease out the details.

OK, here are some tests I did with the long sample memory on the Rigol DS1102.
... seems the Rigol might not be as sophisticated as more expensive scopes, but the basic capability is there and functional.

[chscholz]

Yes, option M64 gives you 64 Mpoints on 4 channels (or 128 Mpoints interleaved on 2 channels, sorry I was not very clear previously).
BTW, the WaveRunner HRO (12 bit scope) has an option XL-256 that gives you 265 Mpoints/Channel.

If, for example you are looking at low frequency jitter or need to decode long strings of digital data, the long memory is really helpful but of course YMMV.

Chris

Sure about that? The LeCroy page says 64 Mpts/Ch ...
Anyway, our 6Zi at work has "only" 32MPts and I never came into a situation where I needed more. Even the 12MPts of its predecessor where more than sufficient.