oscilloscope sampling rate vs record length??

[SnakeBite]

Hello,

i want to buy an oscilloscope soon and i saw some DSO that has 500MS/S sampling and HUGE 1,000,000 pt/sec max record length.
i wanted to buy a min of 1GS/S DSO but i really can't decide between sampling rate and record length - whats more important??

i know that basicly sampling rate is the DSO "dots" that it sample and later connects them and create the wave you see on the screen , which means the more you have the more accurate waveform you get, and the record length is the max "dots" that the DSO can save in a sec  - which means the more you have the more accurate and fast waveform you can save.

i starting to think that having 500MS/S sampling and 1,000,000 pt/sec record length is kind of a bottle neck but i'm not sure.
can you help me?

thanks
Ido

[jahonen]

It depends what you are measuring. If you want to just sample a long segment and then afterwards look at the data at maximum precision, there is no substitute for large record length. I have done a good amount of measurements in the past years at work with a HP 54645D scope, which has only 200 MS/s and 100 MHz bandwidth (but good sin x/x reconstruction) without having a feeling I'm missing anything important. In practice, 54645 has much greater sample rate for lower timebase settings compared to Tek 3000-series, which has only 10k of acquisition memory (54645 has 1 Msample), but 2.5 GS/s maximum sample rate. So for example at 10 ms/div setting, Tek 3000 has only 100 kS/s sampling due to memory limitation, despite the much higher sampling rate. From that, 500 MS/s is not necessarily a great bottleneck (especially in hobbyist use) if acquisition is otherwise good.

So basically, the memory amount dictates the sample rate, i.e. maximum sample rate = record length/(time/div*10) with saturation to maximum banner sample rate. Maximum sample rate matters only when you are measuring high-bandwidth single events in short timebase settings. So in the end, it is your call

Regards,
Janne

[alm]

It depends on the bandwidth you require. Sample rate determines the maximum single shot bandwidth, and memory depth determines at which timebase settings you can use the extra sampling rate. I think the Rigol DS1052E is limited to 500MS/s with long memory in single channel mode, and 250MS/s in dual channel mode. I believe the Instek GDS-1062A can do 1GS/s with long memory enabled.

So basically, the memory amount dictates the sample rate, i.e. maximum sample rate = record length/(time/div*10) with saturation to maximum banner sample rate. Maximum sample rate matters only when you are measuring high-bandwidth single events in short timebase settings.

Actually it's also the other way around: long memory is useless on a fast timebase without a high sample rate to fill it all. You need both for single shot acquisitions of signals with high-frequency components. You can work around short memory length on repetitive signals by triggering on the right part of the signal, and work around sample rate by equivalent time sampling (if it's well implemented).

I agree that it depends on your application. I recently was observing a ~1ns rising edge in a circuit: don't really need more than a screen full of data, but do need fast sampling rate to sample the fast edge. I've used a TDS-220 for a number of years, and I rarely had problems with the 2.5kpoints/channel memory. An analog scope doesn't have more than one screen of data either . But for some applications, especially if it's hard to trigger on the phenomena (or if the scope has crappy triggering), long record length is great.

[jahonen]

The long memory is used entirely in short timebase settings so that more than one screenful is sampled, at least on the 54645 and my MSO6034. In fact on the MSO6034 you can zoom out quite a bit and still it looks like it has been just captured. That actually means it does not matter what the actual timebase setting is, as long the sample rate has reached its maximum.

I guess it is what kind of working habit one has, but I like to zoom the waveform around back and forth after the singleshot trigger and I almost always curse if record length is very short (or nothing is stored beyond visible screen). Sometimes having several screens is nice if you notice that the interesting event was just after/before the trigger screenful.

Regards,
Janne

[alm]

I guess it is what kind of working habit one has, but I like to zoom the waveform around back and forth after the singleshot trigger and I almost always curse if record length is very short (or nothing is stored beyond visible screen). Sometimes having several screens is nice if you notice that the interesting event was just after/before the trigger screenful.

If it is a periodic signal, adjusting trigger delay is just as easy. For single shot, you're correct, it can be a pain.

I don't disagree with you that long record length is a desirable feature, but so is a fast sample rate, both can be critical for some applications. I don't think one is clearly more important. A 2.5kpoints scope would be a pain for analyzing slow signal with fast edges (the edge will be undersampled), and a 200MS/s scope would be a bad match for a 3.5ns single shot edge (undersampled, again).

[SnakeBite]

thanks for all the answers guys.

my question refers to a specific DSO model -"Lecroy lt342"

http://www.google.co.il/url?sa=t&source=web&cd=3&ved=0CCcQFjAC&url=http%3A%2F%2Fcobweb.ecn.purdue.edu%2F~aae520%2Fwavepro-learnerinfo-sps.pdf&ei=t1hgTYyLCMWp8APA64Ra&usg=AFQjCNHZ9Z3606AsWp45QsgMMDW_Q7ShRA&sig2=KygOOjyKCdFh61qEDvkR3w



can you refer to this model?

[mikeselectricstuff]

A HUGE  advantage of deep memory is it can save a lot of time setting up triggers - in many cases you can do without triggering altogether  -just grab a big bunch of data and zoom in on what you're interested in.

Another thing you need to look at carefully when comparing specs is how the memory is actually used.
For example, the MSO6034 originally shipped with 1MPts which sounds a lot, but...
use more than 3 channels (or 1+2 instead of 1+3)  and that's 500K
Use peak acquisition mode to make sure you don't miss glitches, that's 125K (This is not mentioned in the manual)
Use Continuous capture instead of single, that's 62.5K *
Turn on digital channels, and that's 31K
So the 8M upgrade suddenly doesn't seem so extravagant..!
 
* in continuous mode, memory is double-buffered to maximise display update speed, single-shot mode is the only way to get the full memory. A 'repeated single-shot' mode would be a useful addiiton.