Rigol vs Agilent DSO Opinions?

[amd]

Hi,
I want to buy a new 4-channel oscilloscope and I'm having a hard time deciding which one. After watching Daves review of the Agilent 3000 X-series I'm not sure anymore if a Rigol is the best choice. Here is a comparison I made (sorry, prices are in Euros):

	Rigol DS4054	Rigol DS6064	Agilent DSOX3034A
Bandwidth	500MHz	600MHz	350MHz
Max. Sample Rate	4Gsa/s	5Gsa/s	4Gsa/s
Max. Memory Depth	140Mpts	140Mpts	2Mpts
Waveform Capture Rate	110.000 wfms/s	180.000 wfms/s	1.000.000 wfms/s
Display	9“ 800x480px	10.1“ 800x480px	8.5“ 800x480px
Price	€4.795,00	€6.696,00	€ 6.015,00

One thing where the Agilent really shines is the waveform capture rate. And, well, the name. But what puzzles me is the memory depth. Do I compare apples and pears here? Is the memory of the Rigol slower? Why is it so much bigger than the one of the Agilent?

The bandwith of all three is sufficient for me, it's hard to tell if the waveform capture rate makes any difference. Ok, finding glitches, but 110.000 is not so bad either I think. Another thing is the sample rate which I think is shared with all the Rigols and the 4-channel Agilent has two 4Gsa/s channels, one for probe 1 and 2 and another for probe 3 and 4 if I'm not mistaken.
In all other aspects the Rigol seems superior to the Agilent in terms of value for the money.
I probably won't go for the signal analyzer since they are quite expensive options and not very flexible (no JTAG eg). The Agilent probably is better in this aspect since it should decode in hardware. I'd probably take a USBee RX (anyone got experience with this one?) since I should be able to implement custom protocols here (need that for SWD).

I'd be interested in your opinions.

Daniel

[EEVblog]

Remember also that the Rigol 6000 does not have an AWG (can probably get for free now?) or mixed signal LA option should you need those in the future.
The Agilent has limited memory because they decided to integrate the memory onto the main ASIC chip to get the speed that scope is capable of.
The Rigol uses standard SDRAM chips, so they can get huge memory at lost cost, but at the expense of system speed.
Yes, the Rigol is a pretty quick scope too, but still not in the same league as the Agilent which does almost everything (serial decode, masking etc) in hardware.
Yes, the Agilent has a separate ASIC (and hence memory) for each two channel pair.
Also remember that the update rate is a max value that occurs at a sweet spot on the timebase. Use that huge sample memory on the Rigol and the update rate drops to bugger-all. See this post as an example, I expect the 6000 to scale similar:
https://www.eevblog.com/forum/product-reviews-photos-and-discussion/first-impressions-and-review-of-the-rigol-ds2072-ds2000-series-dso/msg160064/#msg160064
The Agilent wins on support and firmware updates etc.
I've found the demo signals in the Agilent to be real handy is setting up complex triggers etc.
And the Agilent likely has better future options as it has a more flexible platform to implement new stuff.
Yes, the Rigol is better bang-per-buck, and kicks butt on the memory depth.
Don't know, tough call. If you value memory depth, go Rigol, otherwise I'd say the Agilent is nicer.

Wait until next week, there may possibly be a new option for you, but I'm not sure of the price range...

Dave.

[mikeselectricstuff]

Huge memry isn't a big deal for most applications - once you get over a few M, it's diminishing returns, and only really needed for things like digging though long protocol decode dumps.

[amd]

Wow, thanks Dave for this detailed answer! I will wait for the new option, whatever that is.
I think the AWG option is too expensive for what it is, I wanted to go for a DG4162 instead.
I also think I will try the USBee DX (would be a nice device for a review btw) for serial decode. I need to decode JTAG and SWD and the USBee DX allows you to program custom protocol decoders. I don't see a way of doing that neither with the Rigol nor with the Agilent.

[marmad]

Use that huge sample memory on the Rigol and the update rate drops to bugger-all. See this post as an example, I expect the 6000 to scale similar:
https://www.eevblog.com/forum/product-reviews-photos-and-discussion/first-impressions-and-review-of-the-rigol-ds2072-ds2000-series-dso/msg160064/#msg160064

When the DS2072 is capturing 14kPts @ 50kHz (20ns tb), it's converting ~700MB of analog data per second, with a blind-time of 98.6%. When it's capturing 56MPts @ 35Hz, it's converting ~1.9GB of data per second, with a blind-time of 99.9%. It's a trade-off of blind-time vs. horizontal resolution. And for all DSOs with higher update rates (>5k wfrm/s), the difference comes down to a few timebase settings (the Rigol scopes mentioned above are likely as fast as the Agilent at all but perhaps 8 or 9). I would say if you do high-end design and debugging (the 1M wfrm/s can't be beat) or write high-level instrument control software (Agilent's command set and support are superb), the 3000-X is the answer - but in most other circumstances, the Rigol will be better value for money.

once you get over a few M, it's diminishing returns

Spoken like an Agilent owner   Seriously, though, how can it possibly be diminishing returns? The more memory you have, the more horizontal resolution you can get OR the more thousands of time-slices you can capture for post-processing later. I've never failed to eventually bump up against the memory limitations of pretty much every digital device I've ever owned.

[lewis]

Try and get them on demo at the same time, and make sure you ask either the Agilent or Rigol distributor for an evaluation board, and try them. I had the choice between the Tek MSO4104, Agilent MSOX3054 and Agilent MSO7104B. I was going to choose the Tek by default because I've always used Tek (my reasoning went thus: "what's all this Agilent Packard muck anyhow? They make printers don't they? Can't be a proper oscilloscope, it hasn't got Tektronix written on it"). I was even going to just buy one without trying it first, but it was Dave's 3000X review that made me want to try the other side.

In the end, both Agilents impressed me so much I bought them both. The 3054 and the 7000 managed to capture more glitches than the Tek on Tektronix's own demo board at all equivalent timebase settings, which I found hilarious. They are fantastic, a joy to use, and I would now hesitate to use anything else. I've never used a Rigol, my feeling is that I wouldn't trust it, but that's just prejudice because the brand does not carry the same weight for me.

Anyway, my advice is get one of the manufacturers to lend you a demo board (it has a load of test signals on it specifically for showing off the scope's capabilities and is good for like-for-like comparisons), get hold of both scopes on demo, try them, test them and decide for yourself.

[aparlett]

I would Have to say Agilent, If your a student let them know you get a discount.
I got a free function Gen and training kit with my scope as well .
I have been using mainly Tek scopes at uni and I find Agilent alot more easier to uses love the interface. Also in terms of update rate and sampling rate you can tell a big difference between what you see on the screen. 
The only thing I don't like about Agilent is the cost could be a little cheaper but you can pay now for what you can afford then upgraded later.

[EEVblog]

Huge memry isn't a big deal for most applications - once you get over a few M, it's diminishing returns, and only really needed for things like digging though long protocol decode dumps.

I tend to agree.
The difference between say 1M and 4M is not much at all.
But say 1M to 40M can make a big difference in some applications, but for normal use you are going to be running a faser update smaller memory footprint anyway.
Better trigger can help overcome some issues with lack of memory if required.
But extra memory can be very useful for segmented operation and waveform replay.

Dave.

[EEVblog]

I think the AWG option is too expensive for what it is, I wanted to go for a DG4162 instead.

Agilent are throwing in the AWG for free these days I think?
Or if they aren't you can likely ask for it to be thrown in by the dealer for a $5K-$6K scope.

Dave.

[aparlett]

I think the AWG option is too expensive for what it is, I wanted to go for a DG4162 instead.

Agilent are throwing in the AWG for free these days I think?
Or if they aren't you can likely ask for it to be thrown in by the dealer for a $5K-$6K scope.

Dave.

yeah they are ATM

"For a limited time, any customer who purchases one or more new InfiniiVision 2000 or 3000 X-Series oscilloscopes is eligible to receive a free WaveGen built-in 20MHz function/arbitrary generator AND a free integrated 3-digit voltmeter (DVM) per oscilloscope."

http://www.home.agilent.com/agilent/editorial.jspx?cc=AU&lc=eng&ckey=2137356&nid=-33573.970736.00&id=2137356

But just ask the promotion I asked for wasn't for Australia and expired months ago but they still gave it to me

[ToBeFrank]

But extra memory can be very useful for segmented operation

I thought segmented memory was to give you "more memory" in the cases where it's useful. I know when I use it on my Agilent 3000 for bus decoding I have way more than enough memory.

[marmad]

I really wish I had a $1 for every time I've heard someone say over the years that XX amount of memory on XX device was more than enough - anything more was unnecessary, wasteful, irrelevant, etc. 

Granted, the majority of DSO use requires only a small amount of memory, but I believe, as with all devices, that as the capabilities and specifications of modern DSOs increase, people will find more ingenious and enhanced ways to use them. For example, I can currently do 1.3 seconds of 20us real-time data logging with the 56MPts in my DSO; if I had 16GB of memory in the scope, I could do 6.2 minutes - a much more useful period of time for some testing purposes.

[M. András]

Use that huge sample memory on the Rigol and the update rate drops to bugger-all. See this post as an example, I expect the 6000 to scale similar:
https://www.eevblog.com/forum/product-reviews-photos-and-discussion/first-impressions-and-review-of-the-rigol-ds2072-ds2000-series-dso/msg160064/#msg160064

When the DS2072 is capturing 14kPts @ 50kHz (20ns tb), it's converting ~700MB of analog data per second, with a blind-time of 98.6%. When it's capturing 56MPts @ 35Hz, it's converting ~1.9GB of data per second, with a blind-time of 99.9%. It's a trade-off of blind-time vs. horizontal resolution. And for all DSOs with higher update rates (>5k wfrm/s), the difference comes down to a few timebase settings (the Rigol scopes mentioned above are likely as fast as the Agilent at all but perhaps 8 or 9). I would say if you do high-end design and debugging (the 1M wfrm/s can't be beat) or write high-level instrument control software (Agilent's command set and support are superb), the 3000-X is the answer - but in most other circumstances, the Rigol will be better value for money.

once you get over a few M, it's diminishing returns

Spoken like an Agilent owner   Seriously, though, how can it possibly be diminishing returns? The more memory you have, the more horizontal resolution you can get OR the more thousands of time-slices you can capture for post-processing later. I've never failed to eventually bump up against the memory limitations of pretty much every digital device I've ever owned.

what are these blind time numbers for the 3000x series?

[marmad]

what are these blind time numbers for the 3000x series?

The blind-time numbers are for the Rigol DS2072 - the DSO associated with the table of update rates which Dave linked to. Given the Agilent 3000X series reported wfrm/s rate of 960k @ 20ns/div, it would have a blind-time of 80.8% with a 50k sample depth.

[EEVblog]

I thought segmented memory was to give you "more memory" in the cases where it's useful. I know when I use it on my Agilent 3000 for bus decoding I have way more than enough memory.

Kinda, more like "less wasted memory".
Say you are capturing a dozen 1ms packets of data that are widely spaced with a 1 second period between packets. If you have segmented memory then the memory isn't wasted on that 99ms of useless time that has no real info you want to capture. So in this case you'd effectively get 99 times more memory. That lets you sample at a higher clock rate and zoom in your packet and capture more packets that you couldn't do otherwise.

Dave.

[M. András]

that seems poor in my terms it still misses the 80% of the events. do the old analog scopes have blind time too?

[marmad]

that seems poor in my terms it still misses the 80% of the events. do the old analog scopes have blind time too?

Only the retrace time of the beam. That's why triggers are even more important on digital scopes.

But yes, blind-time is a difficult technological problem to crack because it takes much more time to process and display data than just to capture it.

[marmad]

that seems poor in my terms it still misses the 80% of the events. do the old analog scopes have blind time too?

One other point: blind-times drop significantly at slower timebase settings, as you would imagine - because the ratio of total time needed for a sweep versus the time it takes to process/display becomes much greater. So at the slower settings, all DSOs eventually reach parity.

[EEVblog]

that seems poor in my terms it still misses the 80% of the events.

But that figure can be deceptive because it does not take into account the actual probability math.
http://cp.literature.agilent.com/litweb/pdf/5989-7885EN.pdf
Agilent's example shows a 90% dead time 3000X (1,000,000wfs) has a probability of 92% of capturing a signal within a set time period.
The same probability of a 99.97% dead time scope (2600wfs) of capturing the same signal is only 0.65%

Dave.

[vvanders]

Something else to keep in mind, not sure what the markup is like on Rigol/Tek but I've been getting quotes on a DSOX 3000 series and I've seen ranges as much as 20% off of list price.

[paudav]

Or if they aren't you can likely ask for it to be thrown in by the dealer for a $5K-$6K scope.

Gawd.  And to think I've spent the last two weeks hemming and hawing between a $900 scope and a $400 scope...

It's time for the ultimate face-off:  Rigol 1052 vs. <The most expensive scope Agilent will loan you>.   

[EEVblog]

Gawd.  And to think I've spent the last two weeks hemming and hawing between a $900 scope and a $400 scope...

There is no contest, if you can in any way afford the $900 scope, get it. The difference in capability between the DS2000 and DS1052E is many many times more than the price difference.

Dave.

[paudav]

  I went with the 1102E.

It's my first scope, and I just couldn't find my way around to spending the $900.  I was thinking that if I need something I'm not getting from the 1102E, I'll sell it local for about what I paid, and reinvest in the DS2000 series.

Wrong choice?

[EEVblog]

Wrong choice?

Not at all if that's all you can afford now, it will serve you well.
The DS2000 would be a better long term investment. But in a year or two's time it'll likely be the price of the 1052E.
I paid >$700 for my DS1052E!

Dave.

[amd]

I really love this forum already, super-interesting discussions!
I got some specs about the new Agilent, definately will wait for that and have a look at it next week at the Electronica in Munich. This one was built for me, thanks for the tip Dave!