Tek MDO3000 or Agilent DSOX3000

[Sal Ammoniac]

The current special offers on the Tek MDO3000 and Agilent DSOX3000 series are coming to the end very soon and I'm still sitting on the fence trying to decide which way to go. 

I already have a spectrum analyzer, so the SA in the MDO3000 is not a deciding point for me. I'm an embedded developer and spend most of my time writing and debugging firmware for various MCUs and mostly use a scope to find noise on ADC inputs, slow rise time on I2C lines, etc. I have dedicated logic analyzers too, so the mixed signal and protocol decoding capabilities of the Tek and Agilent scopes really aren't big selling points for me either.

The two differences that seem important to me boil down to the waveform update rate and the memory depth. The Agilent does 1,000,000 updates a second, while the Tek does 280,000 -- how much of a difference does that make in everyday use?

The Tek seems to have a big advantage in memory depth with 10 Mpts versus only 2 Mpts for the Agilent. Again, how important is this difference in everyday use?

I've read numerous forum and blog posts saying the Tek scope has a sluggish UI that often locks the user out while it's processing a capture -- is this still an issue, or has it been fixed in a new firmware release?

The current special offers pertain to features I really wouldn't use much, but having them anyway would probably add to the resale value of the scope if I decide to sell it and upgrade in the future. If I'm going to get one of these, I want to take advantage of these offers while they last, unless someone has some inside info on what the next special offer may be... Last year Agilent had an offer where the next higher bandwidth scope was priced the same as the lower bandwidth model (e.g. the 350 MHz model for the same price as the 200 MHz model). That kind of special offer would be more important to me than more bandwidth on the built-in SA or a free protocol analyzer.

[eurofox]

I'm a very big fan of my MSOX3001, the serial decoding work like a charm 

Do not compare the 2M segmented memory from the Agilent with the memory of the Tek

I got a Tektronix scope before, the major difference for a user is that the Agilent menus are very intuitive and that is not always the case with Tektronix.

In Europe the Agilent support is simply excellent.

[David Hess]

...

The two differences that seem important to me boil down to the waveform update rate and the memory depth. The Agilent does 1,000,000 updates a second, while the Tek does 280,000 -- how much of a difference does that make in everyday use?

I would say none unless you are glitch hunting.  The update rate per second may severely depend on horizontal time/div setting though so that should be checked if possible to see how consistent the advertised high update rate really is.

Segmented acquisition memory may be a substitute for high waveform acquisition rates.

The Tek seems to have a big advantage in memory depth with 10 Mpts versus only 2 Mpts for the Agilent. Again, how important is this difference in everyday use?

Again I would say none unless you have a specific need for high single shot sample rates at slow horizontal time/div settings which the shorter but still long record length will not satisfy.

[don]

I'm in the same boat as you after trying rigol ds4000 and siglent sds2000.  I now have an mdo3000 on order but looked at dsox 3000 ( and am intrigued by lecroy  wavesurfer 3000).  I use the mdo4000 which is very similar to mdo3000.   I guess its what you are used to and need, but I'll highlight pros of tek that are very basic but  impact my work flow:

-two very nice large cursor knobs. Very smooth, and can work in all modes (seems obvious but you'd be surprised). No button pushing to switch and has a fine / course button. Just clear menus and you have two cursor knobs
-great auto switching roll mode (down to 40ms) with zoom.  Some scopes disable zoom in roll mode. I use roll mode extensively and want it to behave like regular single aquisiton when it comes to analysis
-measurements can be gated by screen, entire waveform or cursors. There are no changes as you zoom like agilent
-pan/zoom is nice feature

I don't find the UI  sluggish with my usage profile. If you are doing something extreme perhaps (someone mentioned 50s to decode 8500 I2C transactions) but just a 10M capture with a couple measurements will not be an issue with zooming / panning and navigation. Dave's latest video on polar heart rate monitor has the mdo3000, might get a read from that.

There are many other criteria, but for me I capture and analyze. I want all features to be there when analyzing. I'll take up front processing if this enables more analysis capability or consistency of operation.  I want to use 10M rolle mode with peak detect, add measurements, zoom, cursors, screen capture w/o issue.   I really just want the basics to be there with no surprises.  I only use segmented memory for decoding else I use persistence (can't do on MDO but can on agilent)  For decoding large amounts I would use a different tool, so instead prefer tek 10M to agilent  2M .  Really comes down to what is important to you.

One quick comment on SA - it has  3GHz capture bw. So in one capture you get full 3GHz spectrum.  Great for debugging noise sources compared to a traditional SA with only 10-20MHz capture bw that must be swept.

[EEVblog]

The two differences that seem important to me boil down to the waveform update rate and the memory depth. The Agilent does 1,000,000 updates a second, while the Tek does 280,000 -- how much of a difference does that make in everyday use?

You can't compare the two here really. The Agilent is superior not only on the numbers, but because it's always on and working at the fastest rate.
The Tek only gets it's claimed rate in FastAcq mode which you have to select, and you then only get a small memory size of a few K.
Indeed, the Agilent has no memory depth or fast mode settings, it "just works".
So for everyday use looking at signals without fuss, I think the Agilent is the clear winner.

But the Tek does better analysis on the full memory depth (one of the reason why it can be slow)

[HighVoltage]

After considering also the Tektronix scope, I went with the Agilent MSO 3000 X Series and I love it. I got the official option pack and one feature that I did not even expect to be useful is "Segmented Memory" and now I use it all the time. I am very happy with the Agilent scope and can only recommend it.

[don]

Regarding memory depth, the tek is up to 10M on all channels in all modes and user selectable. So you can do 4 analog and 16 digital all at 10M.

Agilent is 2M for one or two channels, 1M for three or four or channels in single acquisition.  Memory is not selectable. If in run/stop mode vs single,  memory is further  halved to 1M for two channels and 500K for four channels. If digital is added i believe it is halved yet again.  so you can be down to 250k - 500k vs 10M on tek.   You can get 4M  option to double these .

So you need to see what's important - update rate or memory depth. Also consider what you need update rate for.   Update rate is about 3x on agilent.  For my use I let a scope run and walk away.  I rarely use it and most of the time my trigger is much slower than the update rate of scope. 

[MarkL]

Agilent is 2M for one or two channels, 1M for three or four or channels in single acquisition.  Memory is not selectable. If in run/stop mode vs single,  memory is further  halved to 1M for two channels and 500K for four channels. If digital is added i believe it is halved yet again.  so you can be down to 250k - 500k vs 10M on tek.   You can get 4M  option to double these . 

The Agilent/Keysight scope under consideration does not reduce memory beyond 2x for analog channels.  One acquisition engine is shared between channels 1&2, and another between 3&4.  So, it's either 1Mpts or 2Mpts per analog channel depending which channels you have on.  Each digital channel is always 1Mpts.

All those numbers are doubled with the DSOX3MEMUP option, as you mention.

The DSOX3MEMUP and every other decode and upgrade key is being offered now as a bundle for the cost of a single option.  But that offer is only running for a few more days until Sept 30.

[MarkL]

Agilent is 2M for one or two channels, 1M for three or four or channels in single acquisition.  Memory is not selectable. If in run/stop mode vs single,  memory is further  halved to 1M for two channels and 500K for four channels. If digital is added i believe it is halved yet again.  so you can be down to 250k - 500k vs 10M on tek.   You can get 4M  option to double these . 

The Agilent/Keysight scope under consideration does not reduce memory beyond 2x for analog channels.  One acquisition engine is shared between channels 1&2, and another between 3&4.  So, it's either 1Mpts or 2Mpts per analog channel depending which channels you have on.  Each digital channel is always 1Mpts.

After looking again at the Agilent datasheet, there was some squirrelly language surrounding the digital channel numbers.  I did a quick test to find out exactly what it's doing.

This is with SINGLE capture, the MEMUP option, and examining the number of points recorded for each channel via the SCPI commands.  You are correct that RUN mode vs. SINGLE will halve these numbers again, but that's already a well known behavior.

Chan1 on, chan2 off, digital off: 4Mpts on chan1.
Chan1 on, chan2 on, digital off: 2Mpts each on chan1 and chan2.

Chan1 off, chan2 off, digital on: 2Mpts on Bus1 (as the digital channels are called on SCPI).
Chan1 on, chan2 off, digital on: 2Mpts on chan1, 250k on Bus1.
Chan1 on, chan2 on, digital on: 1Mpts each on chan1 and chan2, 250k on Bus1.

What surprised me was the drop to 250k digital points if *any* analog channels channels are on.  And it doesn't seem to matter which or how many digital channels (D0 through D15) are on.

My apologies for the previous mis-statement, and thank you Agilent Marketing for the crisp specifications.

[don]

Its worth noting that while Tek architecture is often said to be old, there were clear design tradeoffs agilent and tek made. Mdo3000 has 20-40x the memory of agilent (depending on 4M option or standard)  as you add channels , or change acquisition.  That's very significant, and may be why tek is sticking with their architecture.  Hard to justify agilent 3x improvement in update rate at the expense of memory -- at least for how I use scopes

[andrewwong2000]

Before I even bought an MSOX for the speed and convenience, I'd already planned for a logic analyser (Saleae Logic Pro) to allow for embedded micro deep serial decodes.

As usual, you need more toys er tools and one size doesn't fit all MDO, MSOX or whatever


Sent from my iPhone using Tapatalk

[MarkL]

I ended up with the same combination for the same reasons: MSOX3104 + Saleae Logic 8.  They complement eachother's capabilities very well.

[trevwhite]

I have a MSO-X 3024A and I just love the speed of it. The speed at which it decodes everything and its responsive interface is great. Can always do with more memory but I agree that the PC analysers can be used for looking at a lot more information and they have a better interface as well with mouse and the screen to work through the data.

I am not sure I like the scope probes that came with the MSO-X though. The hooks do not seem to stay on that tight.

[EEVblog]

Its worth noting that while Tek architecture is often said to be old, there were clear design tradeoffs agilent and tek made. Mdo3000 has 20-40x the memory of agilent (depending on 4M option or standard)  as you add channels , or change acquisition.  That's very significant, and may be why tek is sticking with their architecture.  Hard to justify agilent 3x improvement in update rate at the expense of memory -- at least for how I use scopes

Agilent have had and still have the cheap fast real time update market to themselves since the released the X series. But the Achilles heel is that 4M on-chip ASIC memory. It was plenty 4+ years ago when released, but now it's looking long in the tooth, and they know it.

[MarkL]

I am not sure I like the scope probes that came with the MSO-X though. The hooks do not seem to stay on that tight.

I definitely don't.  They're too big and bulky for fine work.  I replaced all four with N2894A 2.5mm tip probes.  (That's a 4000X probe though, N2873A 2.5mm is the recommended one for 3000X.)

One thing I'll grant Tek is that they make much better probes.

[David Hess]

I am not sure I like the scope probes that came with the MSO-X though. The hooks do not seem to stay on that tight.

I definitely don't.  They're too big and bulky for fine work.  I replaced all four with N2894A 2.5mm tip probes.  (That's a 4000X probe though, N2873A 2.5mm is the recommended one for 3000X.)

One thing I'll grant Tek is that they make much better probes.

Agilent's 5mm probes appear to be the same ones that Texas makes or sells and Agilent's 2.5mm probes look like the 2.5mm probes that PMK makes.

[MarkL]

Agilent have had and still have the cheap fast real time update market to themselves since the released the X series. But the Achilles heel is that 4M on-chip ASIC memory. It was plenty 4+ years ago when released, but now it's looking long in the tooth, and they know it.

Maybe they know it, but they continue releasing new scopes using the same basic design, albeit faster.  Even the brand new 6000X series has 4Mpt memory.  The target market segment maybe doesn't care so much about the next doubling or quadrupling since it's diminishing returns and then you step up to the next class instrument.  I think the target market is more sensitive to fast response and usability, and apparently BW since that's where the new products hit.  More memory might be good for churning up replacement sales, but I don't think Keysight feels pressured to do it.

Look at Tek.  They're still selling the 10kpt TDS3000 series with the same acquisition engine after 15 years.  Now *that's* long in the tooth.

[David Hess]

Agilent have had and still have the cheap fast real time update market to themselves since the released the X series. But the Achilles heel is that 4M on-chip ASIC memory. It was plenty 4+ years ago when released, but now it's looking long in the tooth, and they know it.

Maybe they know it, but they continue releasing new scopes using the same basic design, albeit faster.  Even the brand new 6000X series has 4Mpt memory.  The target market segment maybe doesn't care so much about the next doubling or quadrupling since it's diminishing returns and then you step up to the next class instrument.  I think the target market is more sensitive to fast response and usability, and apparently BW since that's where the new products hit.  More memory might be good for churning up replacement sales, but I don't think Keysight feels pressured to do it.

This might be a processing limitation and not a memory limitation.  Retaining the current fast response and usability with more memory may require faster logic that is not available on an economical ASIC.

[don]

Agilent have had and still have the cheap fast real time update market to themselves since the released the X series. But the Achilles heel is that 4M on-chip ASIC memory. It was plenty 4+ years ago when released, but now it's looking long in the tooth, and they know it.

Maybe they know it, but they continue releasing new scopes using the same basic design, albeit faster.  Even the brand new 6000X series has 4Mpt memory.  The target market segment maybe doesn't care so much about the next doubling or quadrupling since it's diminishing returns and then you step up to the next class instrument.  I think the target market is more sensitive to fast response and usability, and apparently BW since that's where the new products hit.  More memory might be good for churning up replacement sales, but I don't think Keysight feels pressured to do it.

Look at Tek.  They're still selling the 10kpt TDS3000 series with the same acquisition engine after 15 years.  Now *that's* long in the tooth.

I believe the 1M waveform update is a bit of a marketing gimmick, and most people do not realize how much memory they are leaving on the table as a result.  The sample rate is only 2GS/s.  You need a triggerable event at least ever 1us.  That means the memory depth is only 2k at best (people say tek's 10k is small).    There is a pretty limited use case in my opinion where you need or can even use the 1M waveform update.  It's not like you can update 1M waveforms per second at 1ms per division.   

Note that none of agilents high end scopes have 1M update rate.  Instead, they all have more memory and 100k or 400k update rates (and only in segmented memory...).  In other words, agilent does not market their high end scopes for waveform update rate.  Why?   There is a point of diminishing returns on update rate, and I'd guess the majority of users are much better served with a longer record length.    I guess it depends on what you do, but I know where I work record length is much more important than update rate and used daily by most engineers that are on the bench.  On the rare occasion we need a faster update rate we let it sit longer and go do something else.  But you can't do anything to create more record length.  Tek had 400k update rate almost 20  years ago and basically stopped there.  Because they couldn't do it? Doubtful.  Probably because they feel they reached the practical limit of usefulness.   So now their economical scopes have a good update rate (280k) and decent record lenght (10M on all channels, all the time).  I think that beats the agilent 3000x hands down -- unless you always work where you can trigger at 1M waveforms per second.

[MarkL]

I wasn't thinking of update rate, but I agree with you that's another area of diminishing return.  Update rate is a marketing term, and it takes very specific conditions to generate that number.  What really matters is the dead time while the scope is processing between captures.  That's the number I'd like to hear, but I guess it's not as exciting sounding because it doesn't end with a lot of zeros.

I'm also not saying that long capture memories aren't useful either.  There's really no substitute in some applications.  It's obviously important in your work.

My point was that Agilent seems to have found a sweet spot with this shared 4Mpts ASIC design, and I think they're going to milk it for as much as they can by expanding in areas other than just making the record length longer.  There's a heck-of-a-lot of measurement and compliance capability you can bring to the lower cost market that will fit in 4Mpts that was once the exclusive domain of much more expensive scopes.  And the fast response of the UI inherent in the design makes it that much more attractive.

[don]

All true.  And the x series is  over 3 years old.  Will be interesting to see if they  merge the design with their high end scopes as tek has done or keep their current approach for this segment. Thinking about it,  it's funny because  the scope is very responsive. And user sees 1M waveforms per second spec.   One almost reinforces the  other to show how good/fast it is, and slower scopes processing 20-40x the data are classified as old and out of date in comparison even though apple to oranges. Just different design targets and customers.   Hopefully customer understands needs and if they don't know might as well take the faster scope for general useability.

[Wuerstchenhund]

Note that none of agilents high end scopes have 1M update rate.  Instead, they all have more memory and 100k or 400k update rates (and only in segmented memory...).  In other words, agilent does not market their high end scopes for waveform update rate.  Why?   There is a point of diminishing returns on update rate, and I'd guess the majority of users are much better served with a longer record length.

Indeed. The waveform rate is an important factor when the aim is to "see" things like rare glitches on the screen (i.e. using a DSO in a similar way as an old CRO). But no-one buys a high end scope for merely *looking* at a signal, and these scopes have other (more advanced) ways to capture glitches and such than relying on persistence mode and waiting for the user to see it.

[don]

Note that none of agilents high end scopes have 1M update rate.  Instead, they all have more memory and 100k or 400k update rates (and only in segmented memory...).  In other words, agilent does not market their high end scopes for waveform update rate.  Why?   There is a point of diminishing returns on update rate, and I'd guess the majority of users are much better served with a longer record length.

Indeed. The waveform rate is an important factor when the aim is to "see" things like rare glitches on the screen (i.e. using a DSO in a similar way as an old CRO). But no-one buys a high end scope for merely *looking* at a signal, and these scopes have other (more advanced) ways to capture glitches and such than relying on persistence mode and waiting for the user to see it.

Yes, and a  simple deep memory approach to glitch hunting is to trigger on a high level system failure (reset or such) and monitor all activity on suspected signals leading up to.  This gives you system insight of all activity vs looking at a 100ns slice of time.   

Does anyone have usage patterns where they glitch hunt (and successfully find glitches ) more than you would use deep memory?   Trying to think of real world uses cases beyond demo board.  I do look for glitches and such, but they are usually on signals synchronous to much slower triggerable events and cannot make use of 1M waveform update rate.

[nctnico]

I don't see the point of having an ultra high waveform capture rate to look for glitches. If I assume there is a narrow pulse I set the trigger condition for a narrow pulse. Wide pulses vice versa. If I want to see if a signal stays within boundaries I use the infinite persistance mode and let is run for a while. No need to stare at a screen without blinking my eyes. IMHO the whole waveform capture update rate has become a magic marketing number.

[David Hess]

I don't see the point of having an ultra high waveform capture rate to look for glitches. If I assume there is a narrow pulse I set the trigger condition for a narrow pulse. Wide pulses vice versa. If I want to see if a signal stays within boundaries I use the infinite persistance mode and let is run for a while. No need to stare at a screen without blinking my eyes. IMHO the whole waveform capture update rate has become a magic marketing number.

It is more useful when you suspect a problem but do not know what to specifically trigger on.  A specialized trigger is always going to win this race.

I agree it has become a marketing number but there is a big difference between the 45 to 90 waveforms per second which used to be common in DSOs and the fast rates available now even at the low end.