New Agilent scopes

[EEVblog]

I was dubious about Agilent's claim that they were "game changing" scopes. After thinking about it though, I think they are right, but possibly not for the reasons they thought. Perhaps it is an effect of the weak Pound Sterling versus the Dollar / Yen, or perhaps I have just had my nose in the 2nd hand market for too long, but for what it does, it does seem still pretty expensive. Sure, there are some stunning features. Features that people doing specific jobs may use once a year, but when you look at the core 100MHz dual trace scope, and look at what you actually use a scope for, then I can't see people queued up outside their local representative's door.

I was also initially disappointed by the prices, as I had built up visions of them being Rigol type prices.
But then I thought about and realised the prices are actually quite good (actually very impressive) when you compare with their main competitors, and even Rigol if you start talking the high end units.
Agilent haven't made scopes cheaper, they have simply added more capability for the existing price point.
And by being more open about the "upgradeable" (i.e. crippling) of the scopes then that has further helped reinforced that concept in the market that even the smaller players like Rigol will follow.

But I said it was a game changer? Absolutely. Here is a short story: In the book "The Zen of Graphics", Mike Abrash tells of how he worked with a guy who developed PC video card hardware. This was still comparatively early days, with no 3D and only a small amount of hardware acceleration. This guy had gone as far as he thought he could with the transistor count available to him on the custom silicon, but one day he heard that their competitor had added a command buffer. Now, our hero thought about what this meant, and he knew it had the possibility to free the CPU up from a wait loop, and had the potential to be a product killer, making his product look slow and outdated even before it had shipped. He knew he couldn't compete with that, he had only a couple of dozen gates spare at most, but he put in a very crude double register, which would help, but he was afraid that his competitor's hardware was going to absolutely murder them. As it turned out, the competitor's product was nothing of the sort, and our hero's product was vastly superior. All because he believed that someone else had done a better job with the same resources that he had.

I think people like Rigol are going to look at this scope and think "crikey, this isn't that far in front of where we actually are now". Big LCDs are cheap, as are FPGAs, and fast and wide ADCs are getting cheaper by the day. Add some free Linux, and they could have 90% of the scope for 25% of the price. If I were an engineer in one of these companies now, I'd be rubbing my hands at the way the market had just moved.

Yes, I agree it is pushing the market in a great direction, but I don't agree that the likes of Rigol will be able to match it and wipe the floor with a 90% solution. Unless you count that waveform update rate as being only 10% of the scope, because that's the killer feature that differentiates this scope. Of course if someone doesn't think they need that speed then they save a few hundred bucks and go buy the Rigol.
You can't just whack in an FPGA and get this kind of update rate.

The likes of Rigol are also caught in the same model/profit trap as the big players. They can't just suddenly bring out a 200MHz 4ch 2M scope for $500 and blitz the competition, as that would destroy their own optioned up market in the process. That's why none of them will make that move, as it would be suicide for the niche scope industry, and all the players know it.

My guess is the two tiered market will not change any. You'll still have the big players (Agilent, Tek, Lecroy) at one price point, and the Rigol's et.al at the lower price point with correspondingly poorer performance.

Dave.

[the_raptor]

Come on Dave where is the movie of the magic smoke? :p

[mikeselectricstuff]

You can't just whack in an FPGA and get this kind of update rate.

You pretty much can, especially at the 2000 series performance level, the problem is the FPGA is probably going to be too expensive to make it viable at the moment, but that will change in time.
The logic needed to do this sort of functionality won't have changed in a couple of years' time, but the price per gate of FPGAs will. A major bottleneck is the  bandwidth  to  aquisition memory, which HP have solved with on-chip memory, but fast memory is getting cheaper, and each new FPGA series  comes with  the interfaces (DDR2,DDR3 etc.) to talk to it.  
Probably the biggest obstacle to the higher end is the acquisition front-end, but again that is also something that is continuing to develop.
Given the basic scope architecture, it costs very little more in production cost to add MSO and a big heap of protocol decode, trigger and measurement functions, so this is an area where a competitor with  the design and software skills, or even a small group of dedicated Open Source developers, could make some good progress.
Of course the risk is Agilent could instantly increase their value for money by throwing in the premium features at the base price.

My guess is the two tiered market will not change any. You'll still have the big players (Agilent, Tek, Lecroy) at one price point, and the Rigol's et.al at the lower price point with correspondingly poorer performance.

A big established brand name will always attract a premium price, regardless of functionality and performance. Even within the higher end this is an issue - Agilent have been making better scopes than Tek for years, but they still suffer to some extent from the "Tek are the scope guys" attitude, partly because early Hp scopes sucked. Hopefully the current range will improve tis, unless Tek have something super-awesome up their sleeve real soon.

[Eliminateur]

"News: OH NO!!! The new Agilent 3000 series scope I have for review has FAILED! And I have it on camera, who wants to see the footage? I think I smelled some $12000 smoke..." --> HELL YEAH i want to see that!

[saturation]

me too absolutely, its failure is part of the review!

Now, it didn't fail due to faulty reassembly?   

"News: OH NO!!! The new Agilent 3000 series scope I have for review has FAILED! And I have it on camera, who wants to see the footage? I think I smelled some $12000 smoke..." --> HELL YEAH i want to see that!

[tinhead]

You can't just whack in an FPGA and get this kind of update rate.

the biggest gap in all china DSO design is the splitted functionality - on part managed by FPGA, one by µC.
To get more wfrm/s they have to move as much as possible from µC to FPGA.
Currently Rigol E/D series/UNI-T/Atten are using DSP, which is resulting in max 800wfrm/s where Hantek/Tekway is using Samsung SoC
giving them 2.5k wfrm/s. No idea what Rigol CA series is using, but seems to be still a bit better DSP giving them 2k wfrm/s.
You can't just update the µC to higehr model, sure an ARM11 with 1GHz will maybe give you 5k wfrm/s, but this is still factor 10
slower than an "proper design" - far away from what an ASIC or better FPGA can do.

Rigol seems to have enough knownledge to do it better (DS6000) but this is definitely not an entry model class of DSOs anylonger.
The other (chinese) competitiors have to learn first how to move as much as possible of the core functionality into FPGA.
The Cyclone III is definitely to small, they will have to switch over to much bigger Cyclone IV
or Stratix III - this costs time, and time is money.

For LeCroy and Tektronix this might bea bit easier, depends of course on what they already have in their lab.

On the other side, even if it will be still far away from what Agilents ASIC can do for same money
and Agilent can still update 2000x series with the crippled 3000x ASIC, to get let say 200k wfrm/s and 1Mpts.

The "biggest gap" of Agilents 2000x is the 100kpts memory, but if the ASIC is the same in both models a memory update could
be still possible - if not 2000XA will be released.

[firewalker]

[quote ]
"News: OH NO!!! The new Agilent 3000 series scope I have for review has FAILED! And I have it on camera, who wants to see the footage? I think I smelled some $12000 smoke..."[/quote]

Good News?

 

Is there an estimated time for the video?

[tinhead]

OH NO!!! The new Agilent 3000 series scope I have for review has FAILED! And I have it on camera, who wants to see the footage? I think I smelled some $12000 smoke...

outch ... hope will not Agilent kill you, but yeah as saturation said, this is part of the review so let's show us.

[firewalker]

outch ... hope will not Agilent kill you, but yeah as saturation said, this is part of the review so let's show us.

Looks like Dave to me...

[mitpatterson]

outch ... hope will not Agilent kill you, but yeah as saturation said, this is part of the review so let's show us.

Looks like Dave to me...

nice, i can't tell, did you shop a person in there(or someone else shop it in)

and in regards to this, i'm supprised that none of the other companies he has either: 1. bad mouthed 2. given bad reviews

Havent gotten to him

[snes]

Should I get the 2000 100MHz at once? Or upgrade the 70MHz later? I don't really know if I need it.

[EEVblog]

Should I get the 2000 100MHz at once? Or upgrade the 70MHz later? I don't really know if I need it.

There isn't much difference between 70MHz and 100MHz, is it even worth the money now or later?
200MHz would be nice step up of course.

Dave.

[mikeselectricstuff]

..of course it may depend on what's actually available - demand is sure to be high so you may find the higher spec models are available earlier.

[EEVblog]

..of course it may depend on what's actually available - demand is sure to be high so you may find the higher spec models are available earlier.

I read somewhere that sales have been double what was expected.

Dave.

[mikeselectricstuff]

Has anyone enquired about availability ?
Looks like they are planning on selling via distributors, but no confirmed signs of stock anywhere yet.
My guess is most/all the first production will go to loan & demo units. IME Agilent are pretty good at providing eval kit - when I was looking at the 6000 series I got one on loan for a couple of weeks - when I ordered, they let me keep the loan unit for a couple more weeks until mine arrived!

[Zad]

Does all Agilent test gear have its BNC sockets bayonet pins orientated with their pins vertical now? Its a stupidly small thing, but it is bugging me that they aren't horizontal.

[tesla500]

Does all Agilent test gear have its BNC sockets bayonet pins orientated with their pins vertical now? Its a stupidly small thing, but it is bugging me that they aren't horizontal.

The powered probe interface was designed with the BNC pins vertical, so they must stay that way to maintain compatibility.

Is it just an aesthetic thing, or do some of your probes end up sideways because of it?

David

[djsb]

Has anyone enquired about availability ?
Looks like they are planning on selling via distributors, but no confirmed signs of stock anywhere yet.
My guess is most/all the first production will go to loan & demo units. IME Agilent are pretty good at providing eval kit - when I was looking at the 6000 series I got one on loan for a couple of weeks - when I ordered, they let me keep the loan unit for a couple more weeks until mine arrived!

Here

http://www.aspen-electronics.com/page.asp?page=3000X-Oscilloscope&gclid=CNPw5cHJlKcCFUgTfAodsVsfdQ

and (from Germany)

http://www.datatec.de/lshop,showdetail,2004g,d,,,msox3024a,,,.htm

I know what I'll be asking Santa for this Christmas (the MSOX3024A)

David

[snes]

Has anyone enquired about availability ?

Got an estimate of 9 weeks for the 2002.
It's still better then the year I waited for my TI ez430 :p

[saturation]

Finally had time to watch the whole show, and it was a show.  I hope Agilent gives you credit & recognition for the fine work, your review was so good, I felt like buying one right away. I was sorry too see it end, but luckily there was a tear down as Part II.

 I also like that is was natural, and your disassembly real, in many reviews the cases magically come apart quickly, but realistically, opening a new box we all need to subtly look at how cases, board and parts fits to know which screws need come off, any sequence if needed, or which clips need to be freed.  I do those screwdriver lift techniques all the time.

I'd love to see more contrasts of its performance against any scopes you have around, and small comparisons with the Rigol was perfect; you do get more for your money with the Agilent.

However, your diplomatic way of saying its faults tempered my enthusiasm.  It does have more virtues than faults.  If I were a pro EE, I could amortize the cost of equipment, but for home labs, this is sadly, overkill.  The Rigol or Tekway is bang for buck for home labs.

Function generator: based on the spec and capabilities, it isn't good value for $400+ they want for it.
LAN card: useful, and its desktop software fantastic, but very pricey.
Digital Probes: pricey, but if you need them, you need them.
Unused buttons on your model: only activated with the updated models? you now have 'useless' buttons on a face.  Its probably better to provide covers for them.

..of course it may depend on what's actually available - demand is sure to be high so you may find the higher spec models are available earlier.

I read somewhere that sales have been double what was expected.

Dave.

[Eliminateur]

i thought about the pricing and neutered-stuff and one way to look at it is that they charge what they charge because they need to cover the extensive R&D costs AND the ASIC fab itself(like Dave said, that can't be cheaper).
So looking at it from that way, it makes sense they don't provide a hardware-neutered entry-level as it would actually cost the same or more at first.

maybe after some time, or after a new revision(who knows, smaller lithography, cheaper fpgas, etc etc) they could provide a 2ch HW (dispensing the 2 non-used channels would lower some cost) without the unneesary buttons as well(all the 3xxx functionality, lopgic buttons, etc, that all adds in the end) that would be only BW upgradeable at a very competitive cost.

But then again, it's Agilent, it will never be "Rigol" cost-effective, like Fluke will never be cost-effective compared to a similar specced Extech DMM....

I think that for the price, it's prohibitive for garage/home-labs and only relegated to companies buying them for their labs/production. I don't know about the rest of the world, but 1.2K is luxury-expensive over here(and with customs and shipping, that thing will cost upwards of 2.5K here, for the base model...)

For the price, you can get the usual rigol plus a decent CRO and you get "infinite" waveform update plus DSO functionality , albeit not together but oh well...., beggars can't be chosers

[jahonen]

For the price, you can get the usual rigol plus a decent CRO and you get "infinite" waveform update plus DSO functionality , albeit not together but oh well...., beggars can't be chosers

Actually, analog scope does not solve the problem as infrequent single glitches are too dim to be seen (or they fade away too quickly away for one to see), even if they are drawn more often. Besides, even analog scope needs some time to do the retrace & trigger rearm (X-channel amplifier has limited slew-rate), thus waveform "capture" rate is very well finite.

Regards,
Janne

[alm]

Actually, analog scope does not solve the problem as infrequent single glitches are too dim to be seen (or they fade away too quickly away for one to see), even if they are drawn more often.

This is why the microchannel plate CRT was invented.

Besides, even analog scope needs some time to do the retrace & trigger rearm (X-channel amplifier has limited slew-rate), thus waveform "capture" rate is very well finite.

But it is (mostly) proportional to the time base, so it can be very high at fast time bases. It is obviously finite, though.

[saturation]

I've not had too much experience with glitches, but what I've used on low end scopes is the roll mode.  It gives me a rough idea that a glitch is actually happening, and roughly the frequency; when I get some idea of its periodicity, I can estimate the timing and trigger characteristics I need try to 'zero' in on so I can better see the type of waveform it is with as a faster timebase.

[mikeselectricstuff]

A few comments on the 3000 teardown

The ASICs nearest the BNCs are clearly the ADCs - could be that they use the same ADC on the 2000, multiplexed 4 ways, and two ways on the 3000 for the full bandwidth.

Re. the master/slave thing - the slave MegaZoom ASIC is clearly handing the additional acquisition width for the extra channels, with the master still handling triggering etc., hence the diff traces on the underside going to all 4 channels. My guess is the two display buffers are overlaid  to combine the fast-update data - can't think how else this could be done.
Like the 2000, there seem to be two sets of diff traces from each channel on the PCB underside - my guess  is there are fast comparators in the cans, with 2 seperate programmable thresholds, one for trigger & one for time measurements?

My guess is the 2-channel version will only have the master channel populated.

An interesting difference on the 3000 is the extra DRAM chip near each ASIC. According to Micron's Part number decoder, D9JLR is a MT47H32M16HR-3:F 8Mx16 DDR2 SDRAM, which is where the larger memory depth is clearly occurring.

Re. the function gen and lack of ARB, I wonder if 'mask' on that ASIC drawing could be maskrom for wave data rather than mask test..? Mask ROM would certainly use less silicon than dedicated ARB RAM.

Any chance of some high-res pics soon so we can do some further digging...?