First oscilloscope: DS2072A or DS1047Z?

[edavid]

Most applications can get by with just two channels if a separate trigger channel is available at the cost of some time so I would get the less expensive oscilloscope unless you have a specific need for 4 channels.

In this case the less expensive oscilloscope (DS1074Z) is the one with 4 channels 

[David Hess]

If you are relying on 300 MHz performance and still using a ground lead then the extra bandwidth you paid for is just serving to mislead.  (Haha, mislead . . .  never mind.)

I'm not an expert but I don't think you have to reach 300MHz, I already have "stuff" in my 50MHz scope when using the ground lead .

If your ground lead is long enough or there is enough flux leakage around the probe, lower frequencies can be a problem.  When I probe off-line power supplies, I get the best results using short lengths of thin coaxial cable soldered to the test points with coaxial connections to the x10 passive probes.  If I get really desperate, then I do the same thing differentially.

As a practical matter, my 300 MHz DSO is useless at high sensitivity settings with a fast passive probe and any ground lead because of local EMI.

[sgofferj]

I have been thinking quite a bit about this and I got some valuable input from this thread. As I'm mostly digital and from what I plan to do, there will likely be quite a bit of protocol analyzing involved, reliable analyzing will be important for me. On the other hand, I'm probably not going to need 4 analog channels.

Bottom line, I'm currently strongly leaning towards the MSO2072A, maybe MSO2072A-S.

[tautech]

I have been thinking quite a bit about this and I got some valuable input from this thread. As I'm mostly digital and from what I plan to do, there will likely be quite a bit of protocol analyzing involved, reliable analyzing will be important for me. On the other hand, I'm probably not going to need 4 analog channels.

Bottom line, I'm currently strongly leaning towards the MSO2072A, maybe MSO2072A-S.

https://www.eevblog.com/forum/testgear/siglent's-new-product-msosds2000-series/msg466708/#msg466708

Some comments  that may be pertinent.

[marmad]

7) DS1000Z: up to 30000 wfs/s (claimed), up to 63000 wfs/s (!!! ) (measured); DS2000A: up to 50000 wfs/s (claimed), up to 53000 wfs/s (measured).

Fagear - I watched your video, but I couldn't understand how you thought you were measuring waveform update rates. Granted, I don't speak Russian, so I could be misunderstanding something, but I didn't see any valid method of measuring. For example, at 8:53 you have the DS2000 set to 10us/div with the frequency counter showing ~54kHz while running on AUTO trigger. Did you believe that was showing that the DSO is capturing ~54000 waveforms a second? Well, it's not - as you can check with simple math:

At 10us/div (and memdepth <= screen size), it takes 140us to capture the full waveform for display (10us * 14 divs). That means the absolute maximum waveforms that ANY DSO could theoretically capture in that time would be 7142 (1/140us) - and that would be with ZERO blind time (in other words, impossible).

The DS2000 actually captures ~1000 wfrm/s @ 10us/div. See my waveform update rate chart down towards the bottom of this message. The maximum rate I've measured on the DS2000 is 52770 wfrm/s using DOTS mode @ 20ns/div - and I don't think the DS1000Z gets close to this.

Valid waveform update rates can not be measured while using the AUTO trigger - and if you're going to use the frequency counter on the DSO, you have to loop the Trigger Out back into the DSO on another channel (or External Trigger In).

12) DS1000Z: 90,5 MB internal storage; DS2000A: 64 kB internal storage.

Where did you get these numbers from? Are they published somewhere?

[Fagear]

Fagear - I watched your video, but I couldn't understand how you thought you were measuring waveform update rates. Granted, I don't speak Russian, so I could be misunderstanding something, but I didn't see any valid method of measuring.

I've answered on Youtube, but I'll do it here as well.
On 8:52 my DS2072A is measuring trigger output of the DS1104Z, not itself. So 54 kHz on hardware counter - it is waveform update rate of the DS1104Z. And if you can see later, on-screen measurements show 63 kHz between pulses. Hardware counter shows less because DS1000Z has pause in processing (10:22). "Sweet spot" settings for DS1000Z I have found: 50 ns/div, dots, auto memory.
On 12:02 the DS1104Z shows trigger output signal of my DS2072A. And "sweet spot" is 20 ns/div, dots, auto memory. So the result is ~53 kHz as you have measured.

Both DSOs are fed with 20 MHz sine wave.

So... yeah. It's very strange but DS1000Z outperforms DS2000A!?

Where did you get these numbers from? Are they published somewhere?

Just "Storage -> Disk.Manage".

[marmad]

I've answered on Youtube, but I'll do it here as well.
On 8:52 my DS2072A is measuring trigger output of the DS1104Z, not itself. So 54 kHz on hardware counter - it is waveform update rate of the DS1104Z. And if you can see later, on-screen measurements show 63 kHz between pulses. Hardware counter show less because DS1000Z has pause in processing (10:22). "Sweep spot" settings for DS1000Z I have found: 50 ns/div, dots, auto memory.
On 12:02 the DS1104Z shows trigger output signal of my DS2072A. And "sweep spot" is 20 ns/div, dots, auto memory. So the result is ~53 kHz as you have measured.

Both DSOs are fed with 20 MHz sine wave.

So... yeah. It's very strange but DS1000Z outperforms DS2000A!?

Thanks. I also answered you at YouTube, but I'll re-post here for others:

Ahh... that explains it! Sorry for my misunderstanding - one of the problems with not knowing the language I thought I had read somewhere else that a DS1000Z owner had measured the wfrm/s and never found a higher rate than ~30k, but perhaps he didn't try Dots mode. But it makes me wonder why Rigol doesn't advertise it as ~50k wfrm/s DSO?

Just "Storage -> Disk.Manage".

Ok, thanks!


EDIT: It would be handy if a DS1000Z owner would create a chart with wfrm/s speeds for all time bases, with Dots / Vectors / Sin(x).

[marmad]

So... yeah. It's very strange but DS1000Z outperforms DS2000A!?

The wfrm/s rates also seem to be fairly changeable from version to version of the firmware, at least for the DS2000. I haven't re-measured the DS2000 since FW v.3 came out, but you can see by the attached chart, some of the time base wfrm/s rates changed quite a bit from FW v.1 to v.2.

[Fagear]

a DS1000Z owner had measured the wfrm/s and never found a higher rate than ~30k, but perhaps he didn't try Dots mode

In my review of the DS1104Z-S I've showed some other measurements:
for auto, vectors:
100 ns/div = 13 kHz
50 ns/div = 22 kHz
20 ns/div = 22 kHz
10 ns/div = 20 kHz
5 ns/div = 15 kHz

dots:
50 ns/div = 53 kHz
20 ns/div = 25 kHz
10 ns/div = 25 kHz
5 ns/div = 20 kHz

Actually, I've reflashed DS1104Z to the latest FW (00.04.00.00.00, 2014-03-18) before tests.

EDIT: It would be handy if a DS1000Z owner would create a chart with wfrm/s speeds for all time bases, with Dots / Vectors / Sin(x).

Maybe I can do it. Just have to find some time for it. But how I should perform these measurements? Includind 7 ms pause in capturing (while there no impulses at all and top frequency equals 54 kHz) or without that pause, simply time between impulses (63 kHz)?

[marmad]

dots:
50 ns/div = 53 kHz
20 ns/div = 25 kHz
10 ns/div = 25 kHz
5 ns/div = 20 kHz

Actually, I've reflashed DS1104Z to the latest FW (00.04.00.00.00, 2014-03-18) before tests.

I think I understand now why Rigol marketed it as ~30k wfrm/s DSO: because in original FW, that was the highest speed they could achieve at the time it was introduced. But they've optimized the code since and, just like the DS2000, they've been able to get higher speeds at the 50ns/div (and perhaps other) time base(s).

Maybe I can do it. Just have to find some time for it. But how I should perform these measurements? Includind 7 ms pause in capturing (while there no impulses at all and top frequency equals 54 kHz) or without that pause, simply time between impulses (63 kHz)?

I think you have to include the pause - since the DSO is not capturing waveforms then - so that is the real 'effective' rate.

[rf-loop]

dots:
50 ns/div = 53 kHz
20 ns/div = 25 kHz
10 ns/div = 25 kHz
5 ns/div = 20 kHz

Actually, I've reflashed DS1104Z to the latest FW (00.04.00.00.00, 2014-03-18) before tests.

I think I understand now why Rigol marketed it as ~30k wfrm/s DSO: because in original FW, that was the highest speed they could achieve at the time it was introduced. But they've optimized the code since and, just like the DS2000, they've been able to get higher speeds at the 50ns/div (and perhaps other) time base(s).

Maybe I can do it. Just have to find some time for it. But how I should perform these measurements? Includind 7 ms pause in capturing (while there no impulses at all and top frequency equals 54 kHz) or without that pause, simply time between impulses (63 kHz)?

I think you have to include the pause - since the DSO is not capturing waveforms then - so that is the real 'effective' rate.

I recommend that use example 1 second average (1 second gate time if use  conventional pulse counting principle working frequency counter. ) for fast time/div settings. If want measure slow timebases then need example 10s gate time. 

If we measure very short time peak speeds it may lead to situation where some scope with some settings may give unexpected and not useful fast peak rates.

[Fagear]

I haven't re-measured the DS2000 since FW v.3 came out

I've done all of the measurements and did not find any differences in wfs/s performance between FWs v.2 and v.3.
Here is the comparison:


More details about performance of FW 00.03.00.01.03:


I've also have similar data for the DS1000Z series:



And comparison DS1000Z vs DS2000A up to the limits:


Found some interesting details: DS2000A has enough power to process all data:


But DS1000Z feels to be limited, it goes to full memory only after some steps of sampling speed:


Overlay comparison:

[marmad]

I've done all of the measurements and did not find any differences in wfs/s performance between FWs v.2 and v.3.

Nice! Very thorough - thanks for doing these!

[tom66]

Your comparison of memory length vs sampling rate is inaccurate on the DS1000Z. The auto mode does have those selections, but in all memory lengths you can get 1GSa/s, regardless of timebase. It's only when the distance between the points is less than about 1ns (1GSa/s) does the memory depth drop.

[marmad]

Your comparison of memory length vs sampling rate is inaccurate on the DS1000Z. The auto mode does have those selections, but in all memory lengths you can get 1GSa/s, regardless of timebase. It's only when the distance between the points is less than about 1ns (1GSa/s) does the memory depth drop.

Tom - can you be more specific about which chart you mean? As far as I can tell, all of his charts are focused exclusively on measuring the AUTO mode (so memory length would be set automatically), except for #3 - and that one isn't dealing with the sampling rate. So I'm confused.

[tom66]

Your comparison of memory length vs sampling rate is inaccurate on the DS1000Z. The auto mode does have those selections, but in all memory lengths you can get 1GSa/s, regardless of timebase. It's only when the distance between the points is less than about 1ns (1GSa/s) does the memory depth drop.

Tom - can you be more specific about which chart you mean? As far as I can tell, all of his charts are focused exclusively on measuring the AUTO mode (so memory length would be set automatically), except for #3 - and that one isn't dealing with the sampling rate. So I'm confused.

He is using Auto mode but the chart is making it seem like those are the limits when they aren't. You can get 12Mpt or 24Mpt (licensed or hacked) at all timebases except where there isn't enough data to acquire (below about 20ns/div.) The auto mode just selects (presumably) the best memory depth and sample rate for high wfms/s or better intensity grading but you're free to get maximum memory in virtually all timebases. I've noticed a significant drop in wfm/s from 12->24Mpts, so I only use 24Mpts if absolutely required.

Not sure why it jumps to 7.5Mpts for only 5ms/div - it seems a little odd.

[marmad]

He is using Auto mode but the chart is making it seem like those are the limits when they aren't. You can get 12Mpt or 24Mpt (licensed or hacked) at all timebases except where there isn't enough data to acquire (below about 20ns/div.) The auto mode just selects (presumably) the best memory depth and sample rate for high wfms/s or better intensity grading but you're free to get maximum memory in virtually all timebases. I've noticed a significant drop in wfm/s from 12->24Mpts, so I only use 24Mpts if absolutely required.

I'm not sure the charts are making it seem that way. Chart #3 specifically shows other memory selections at all timebases - and chart #4 focuses on the AUTO setting - which, if it functions as it does on the DS2000, optimizes memory length to get, in order of priority:
1) the fastest sampling rate possible
2) the highest wfrm/s

You will always get the fastest sample rate possible at any timebase setting when using AUTO - but it doesn't mean, of course, that other memory depth settings aren't possible to achieve the same sample rate.