Thoughts on MSOS104A vs HDO6104A-MS vs MSO56?

[babryce]

I have to pick a new scope at work, I am currently considering:
MSOS104A
HDO6104A-MS
MSO56 (1 GHz version)

I don't think I will have a chance to use them all before I have to pick. I don't want to cause a flame war or bias anyone with my own research too much.

I'd say analog performance is the most important. I don't expect to need more than 16 digital channels but of course if I did the Tek would win out.

If anyone has used more multiple of these and has some insights it would be welcome. Particularly on the responsiveness with complex measurements on (FFTs, protocols, etc).

Finally I'll dig more but if anyone know if changing trace colors is possible on the scopes, that is also helpful, because Keysight in particular chooses colors for Ch1/2 that I cannot distinguish as someone who is red-green colorblind (at least on the 2000/3000/4000 scopes).

Edit: I perhaps should also consider an RTO2014 with the mixed signal option. These all cost around 25,000 USD.

[Rich@RohdeScopesUSA]

I know you didn’t mention R&S, but our RTO2000 falls in to this category and looks like it might fit your needs perfectly.  Our FFT is super fast (fastest of all the scopes you mentioned), our front-end noise performance is second to none, you can easily change the identification colors of all channels (it even updates the front panel LEDs), we have the fastest update rate for finding infrequent events (1M wfms/sec), the deepest available memory (up to 2Gpts) and the only scope of the three you mentioned that has full bandwidth at all settings (1mV/div and up).

There are lots of people on the forum that have RTOs and can vouch for their responsiveness, performance and low noise, but also don’t hesitate to let me know if you have any questions (or PM me where you’re at and we can get you a demo).

-Rich

[babryce]

I actually just edited it...

[babryce]

How different are the spectrum parts for the RTE vs the RTO. I have an old SMIQ R&S source at home. There is no R&S equipment in my workplace so it might be a harder ask but I certainly am open minded.

[Someone]

You've lumped together a group of similarly priced but radically different scopes, without knowing the specific applications, what other scopes are already in the facility, and why you were considering these its hard to give any advice. The R&S and Tek models you point to are general purpose scopes that are good at many tasks, while the Keysight and Lecroy are focused on analysis of captured data.

The responsivity, knowledge, and assistance provided by the local support for the brands should be a big differentiator as it varies wildly.

[abraxa]

I have to pick a new scope at work [...] I don't think I will have a chance to use them all before I have to pick. [...] These all cost around 25,000 USD.

So the budget is 25k and your deadline is... when? I'm just surprised your company would press you to make such a (potentially costly) decision on the whim, not even giving you time to get a demo or two.

Maybe you can tell us what kind of analysis/fault finding the scope will be used for - some examples would help us giving you more useful feedback. Emphasizing analog performance can mean any use case really, from analog switching analysis over signal integrity over RF to spectrum analysis.

[nctnico]

Lecroy doesn't have peak detect so I wouldn't pick that one. Peak detect is a really useful feature to look at a signal (for example one with bursts) at large time intervals at low samplerates. Without peak detect you'll get a bunch of nonsense on your screen.

[babryce]

You've lumped together a group of similarly priced but radically different scopes, without knowing the specific applications, what other scopes are already in the facility, and why you were considering these its hard to give any advice. The R&S and Tek models you point to are general purpose scopes that are good at many tasks, while the Keysight and Lecroy are focused on analysis of captured data.

I agree they are different. I was actually asked to pick before my first day, before I was even given a first project or had seen what was in the labs. I declined to do that so that I could figure out what piles of active probes and things were around. If anyone wants to compare to another scope that they think is more a direct competitor feel free.

I would like to answer the other questions as well, but I don't think I can with certainty because much of what I will have to remains unclear at this time. I'm certain I will have to track down a lot of EMI issues and have to do precise pulse measurement at some point. Digital stuff I have constrained to < 400 Mbit on differential buses for the foreseeable future.

Feel free to just talk about strengths and weaknesses from your experience with them even if not comparative. Its all helpful. I watched the signal path episodes for ones that existed which of course gives more of a feel than just reading the datasheet/manuals/marketing materials.

[Rich@RohdeScopesUSA]

How different are the spectrum parts for the RTE vs the RTO. I have an old SMIQ R&S source at home. There is no R&S equipment in my workplace so it might be a harder ask but I certainly am open minded.

They are basically identical.  The main differences are in the scopes themselves (like bandwidth, memory depth, etc). 

You mentioned tracking down EMI issues - here are a few resources we have on how to analyze EMI issues that may be helpful:

https://www.rohde-schwarz.com/us/applications/analyze-emi-problems-with-the-r-s-rto-r-s-rte-application-card_56279-46791.html



-Rich

[Keysight DanielBogdanoff]

The S-Series has really good ENOB and analog performance.

The GUI is also pretty intuitive, this might help give you an idea of usability, etc. :



Also, I just played around with one and you can change the channel colors to be whatever RGB value you want (see attached image).

We've also given the FFT some serious work in the last couple years, but I don't have it well documented right now. Best I can do is a grainy pic from the data sheet. But, the peak labels and cursor labels + a bunch of FFT specific measurements make it quite easy to use.
https://literature.cdn.keysight.com/litweb/pdf/5991-3904EN.pdf?id=2447379

I have no idea what this is, but you can see the FFT running:

[GlowingGhoul]

The free market playing out before my very eyes! 

[babryce]

Indeed.

People seem to break the scope market into two very clean groups: general and analysis scopes. Dave said he would not want to use an Infinium S daily/as general use in somewhere in the Tek 5 series thread that I read (where he noted the R&S RTO).

Now I have used things that were clearly high end analysis scopes. I remember using a 40 GSPS single shot Lecroy serial data analyzer in 2007 (?). It certainly wasn't going to be used for everything. It had no 1 Meg impedance settings for instance. However the S-Series does have 1 Meg settings. Is daily usability mostly the update rate for folks? Or how complex it is to setup all the details of a high end scope.

For instance Dave did not like having to pick the record length in his MDO4000 (or maybe it was 3000 video) and liked that Keysight sets it automatically. Having good presets/auto is nice but for almost all of my work I need to know the every setting to know what the trace means.

If anyone wants to articulate why some scopes are not suitable for daily use when on paper they can do everything, I'd be interested to hear the perspective. I can't tell if it is just complexity of the setup or slow update rates or other slowness, etc.

I am getting close to crossing off the Tek offering for lack of maturity. I also cannot find any decode options for certain things I might want in the new few years available. Its good that Keysight lets you change the channel colors on the Infiniium, I hope they will let you do that on the Infinivisions sometime soon (if it was there I never found it).

I don't like LeCroy's 10 GSPS claim on the HDOs, its 2.5 GSPS, interpolation is not a sample. LeCroy's interface which Tek's is starting to look like it pretty mature though, although it does sometimes use up too much screen real estate.

On the LeCroys I am also researching the Waverunner 8000 series (we already have one of these around at least, along with a HDO8000).

[KE5FX]

If anyone wants to articulate why some scopes are not suitable for daily use when on paper they can do everything, I'd be interested to hear the perspective. I can't tell if it is just complexity of the setup or slow update rates or other slowness, etc.

I won't be able to find it now, but in a prior "Which scope should I buy?" thread, there was some discussion of waveform update rate.  I mentioned the ~100K waveforms/sec rate from the MegaZoom ASIC in my MSO6054A as being a reason to consider an Agilent/Keysight model.  Someone then pointed out that the higher-end scopes in the Infinivision Infiniium S lineup -- DSOX4xxx series, maybe? -- could only capture a few hundred waveforms/sec. 

I looked at the data sheet but was unable to find any indication that they were mistaken.  That's a huge deal.  Waveform update rates in the 100K+/sec range are a non-negotiable condition that has to be satisfied before a DSO can claim to replace a lowly analog oscilloscope.

So that would be one case where you need to look very carefully at the specs, and ask pointed questions of the representatives.   A DSO with a lot of dead time between captures may be useful in specific applications, but it will miss important troubleshooting cues if you try to use it as a general-purpose bench scope.  Spending more money on higher-end scope models may not necessarily get you a better scope for your needs.

[jeremy]

I’ve used a MSOS as a daily scope (can’t remember the model, but it was the 4GHz one). There are a ton of features in the software, and it’s an extremely nice unit, but you’d want to do a heck of a lot of analysis to justify the cost of it. I wish I could, I miss that scope You would need to use the touchscreen a lot though (or a mouse), so if you’re a knob fiddler that might be an issue.

Also iirc you can pair two channels for a true differential measurement which was pretty neat.

[Neganur]

[...] Someone then pointed out that the higher-end scopes in the Infinivision lineup -- DSOX4xxx series, maybe? -- could only capture a few hundred waveforms/sec.[...]

the 4k X-series definitely has 1M wfms/s, it's a banner spec right on the first product page.

And if they were referring to segmented memory capture ("[...]segmented memory successfully capturing 1,000 events in 3.27274 seconds[...]" then I think they are nuts because analogue scopes can't even do that kind of acquisition.

[Someone]

If anyone wants to articulate why some scopes are not suitable for daily use when on paper they can do everything, I'd be interested to hear the perspective. I can't tell if it is just complexity of the setup or slow update rates or other slowness, etc.

The larger scopes with a windows backend running all the samples through the PC infrastructure (Lecroy and MSOS) are inherently limited in their sample throughput, this can be a problem when your tasks require low dead time, serial triggering (which is not in a hardware path for these), mask testing/triggering, or averaging repetitive signals. For many applications there are ways this can be worked around with complex triggers, or capturing deep memory and processing after the fact, which is where those scopes are designed for such as accumulating measurements/trends/histograms or eye diagrams from long captures, and then going back through and doing another analysis on that same data (you can buy the Keysight software as a standalone to do the analysis on a computer for instance). You do things differently than you would on a analog or realtime scope which can be limiting for some applications, and opening new possibilities for others.

The traditional reasons to avoid large format scopes of signal noise from unnecessary bandwidth and clunky UIs are dropping away, but the lack of buttons and knobs can still get in the way of some people feeling at home on a scope which is where the R&S has really stood out from this group.

[KE5FX]

[...] Someone then pointed out that the higher-end scopes in the Infinivision lineup -- DSOX4xxx series, maybe? -- could only capture a few hundred waveforms/sec.[...]

the 4k X-series definitely has 1M wfms/s, it's a banner spec right on the first product page.

And if they were referring to segmented memory capture ("[...]segmented memory successfully capturing 1,000 events in 3.27274 seconds[...]" then I think they are nuts because analogue scopes can't even do that kind of acquisition.

Correct, it wasn't the X-series.  The DSO/MSO-S-series Infiniium platform is what the earlier thread was referring to, based on Tek's 5-series comparison sheet.

These are $20K - $100K scopes, according to Shahriar's review.  I would be irritated if I sprung for one of these under the assumption that it was an overall upgrade to my earlier 6000-series Infinivision model.  But I'm sure they'd say that the Infiniiums are simply optimized for specific purposes.

[Neganur]

But I'm sure they'd say that the Infiniiums are simply optimized for specific purposes.

yeah not sure what to say about that. Different models from the same manufacturer are simply...built to serve different purposes, not all Tek are equal, nor Lecroy 
Perhaps the new x-6000 is a better replacement for the old DSO6000 if you want wfms/s, rather than the Infiniium.

[KE5FX]

Perhaps the new x-6000 is a better replacement for the old DSO6000 if you want wfms/s, rather than the Infiniium.

Sure, but the point is, they don't tell you this.  You have to somehow magically know to ask.

[babryce]

I'm still pondering if I need high update rates critically or not. I might not. Given the manufactures only really advertise the max rates, do folks know the rates as a function of what you are measuring well.

For instance; keysight makes a claim about the RTO2000's rates here: https://community.keysight.com/community/keysight-blogs/oscilloscopes/blog/2016/12/07/keysight-s-series-oscilloscopes-vs-the-rto2000-part-2

I don't have the time to borrow them all and test them even if I wanted to.

The RTO/RTE decode bundle is good value, and it has better decode abilities than x-series stuff which is the only ones with similar update rates. Outside of update rate I haven't heard much on what might make one of the let's call them OS driven scopes less usable for daily work.

[KE5FX]

I'm still pondering if I need high update rates critically or not. I might not. Given the manufactures only really advertise the max rates, do folks know the rates as a function of what you are measuring well.

For instance; keysight makes a claim about the RTO2000's rates here: https://community.keysight.com/community/keysight-blogs/oscilloscopes/blog/2016/12/07/keysight-s-series-oscilloscopes-vs-the-rto2000-part-2

I don't have the time to borrow them all and test them even if I wanted to.

The RTO/RTE decode bundle is good value, and it has better decode abilities than x-series stuff which is the only ones with similar update rates. Outside of update rate I haven't heard much on what might make one of the let's call them OS driven scopes less usable for daily work.

They all have high-level OSes.  That's a red herring.  The DSO-X models run Windows CE, one of the less well-regarded solutions when (pseudo-)realtime performance is being discussed.  But their peak waveform update rate is around 10x faster than the previous 6000-series generation.  The latter scopes run VxWorks, which was the contemporary gold standard in the RTOS business.

I'm projecting my own preferences here, which always seems to happen in these threads.   In reality I'm sure the S-series will serve well in any task you'll ever throw at it.  I just find it frustrating that unnecessary compromises like this are still being made.  It's 2018.  Everybody knows how to build fast scopes.  Pro tip to the tier-1 vendors out there: the Chinese know, too.

[babryce]

Of course they have OSes. I simply meant OS exposed to the user sort of scopes vs just an embedded "closed" as Tek put it on the 5 series OS. Sorry I'm trying to give some short hand.

You are correct though there is no real reason that one box could not do all. If I wanted to rant about scopes I'd complain about the cost of decode options. But people like my employer are willing to pay for the cost because it speed up my work.

I personally only buy used test gear for home because its 1/3 of the price of new and rarely has anything wrong with it.

[Someone]

I'm still pondering if I need high update rates critically or not. I might not. Given the manufactures only really advertise the max rates, do folks know the rates as a function of what you are measuring well.

For instance; keysight makes a claim about the RTO2000's rates here: https://community.keysight.com/community/keysight-blogs/oscilloscopes/blog/2016/12/07/keysight-s-series-oscilloscopes-vs-the-rto2000-part-2

Thats a fairly narrow example, showing that the mask testing isn't hardware driven in the RTO, similar drops in acquisition rates would be seen using serial triggers on scopes which do that in software too. There is a reason we always recommend you evaluate a scope in person for the specific applications you have, they're very diverse in their capabilities and no one scope offers it all on one device.

Of course they have OSes. I simply meant OS exposed to the user sort of scopes vs just an embedded "closed" as Tek put it on the 5 series OS. Sorry I'm trying to give some short hand.

You chose the wrong way to split them, most scopes are now processing the data in an FPGA or ASIC, except for the high end of the market which is still sticking to dumping it all to (wide and deep) memory and then processing on a commodity CPU.

[babryce]

To me it seems like the S-Series does a lot of the processing on the ASIC. It only has a single lane PCI-E I think to the CPU. At least the commentary on the signal path thought this was the case. I do recall the older LeCroy's I used doing all the calculations on the regular CPU. On paper at least the S-Series still has the best ASIC in terms of sample rates/resolution and it has an good time base in it by the looks of it as standard.

[Keysight DanielBogdanoff]

The InfiniiVision 6000 X-Series vs. the Infiniium S-Series is essentially where our two main scope lines (and R&D teams) converge. The S-Series is much more of an analysis machine, with all of the Infiniium apps, deep memory, advanced FFT GUI, etc.

The 6000 X-Series is basically for people that want an InfiniiVision oscilloscope that goes up to higher bandwidth and don't want to deal with Windows, specialty analysis applications, etc. It's great scope for more general purpose debugging and has some cool stuff (voice control, multi-touch display, fast update rate, RTE, jitter). It's also a darn cheap and very small 6 GHz scope compared to what else is available. The joke around here is that, since it's 6-inches deep, you get a GHz per inch. 

Overall, I'd recommend the Infiniium, but the 6000X is a great scope for the price - especially if you aren't looking for advanced, windows-based analysis.

I don't think anyone would argue that a Windows based scope (like the S) is quicker to use, because the baked-in UI scopes like the 6000X are designed specifically to be easy to use. Lot's of front panel buttons, etc. But, once you're used to the Infiniium UI it's still pretty quick.

[Someone]

To me it seems like the S-Series does a lot of the processing on the ASIC.

Not sure where you get that idea from? Its an advertised feature on the V series, and the infiniivision scopes (they all have hardware serial triggers too) but never mentioned for the 9000/S-series.

[nctnico]

I'm still pondering if I need high update rates critically or not. I might not. Given the manufactures only really advertise the max rates, do folks know the rates as a function of what you are measuring well.

Update rates are way overhyped. In normal use case scenarios (with deep memory) the update rate is way lower than it says in the datasheet. It is a bit like a non-turbocharged car engine: the maximum power is available in a very narrow RPM range you rarely use in normal circumstances.

[Someone]

I'm still pondering if I need high update rates critically or not. I might not. Given the manufactures only really advertise the max rates, do folks know the rates as a function of what you are measuring well.

Update rates are way overhyped. In normal use case scenarios (with deep memory) the update rate is way lower than it says in the datasheet. It is a bit like a non-turbocharged car engine: the maximum power is available in a very narrow RPM range you rarely use in normal circumstances.

The peak rate is a good single figure which keeps marketing happy and says something about the capabilities of the scope, I've not found one yet that dropped away and increased blind time at longer record lengths so you can use it to estimate the performance across the range. R&S give peak rates at each memory depth to reassure the doubters but it may be in dot rendering mode as Keysight point out in their material to push the banner spec a little higher. Even the classic Tek DPO4000 kept close to the theoretical maximum rates across most of its range:

(dotted lines are aliased or record extending beyond the visible window)

[babryce]

I remember Shahriar Shahramian making the statement in his review of it, but is it true? (90+% sure if you watch the video he will say something like that around the time that he mentions the only connection is a PCI-E slot to the computer.

Is it true or speculation? I don't know. It is one lane of PCI-E but it could be any version of PCI-E, so the bandwidth is not known.

Also if anyone knows the ambient noise level of any of these scopes that would be nice to know also.

[Someone]

I remember Shahriar Shahramian making the statement in his review of it, but is it true? (90+% sure if you watch the video he will say something like that around the time that he mentions the only connection is a PCI-E slot to the computer.

Is it true or speculation? I don't know. It is one lane of PCI-E but it could be any version of PCI-E, so the bandwidth is not known.

You're throwing out a lot of "unknowns" here that can be checked and referenced...

Hardware-based algorithms for accelerated drawing to display (pixel placement) enable fast pan and zoom even with deep memory

The single acquisition board is a PCIe device attached to the motherboard with a PCIe extension cable. The x4 PCIe link includes nine differential pairs: four transmit pairs, four receive pairs, and one reference clock pair

But you're falling for the emotive marketing fluff which was repeated on The Signal Path rather than making measurable performance comparisons, the speed of display may be improved compared to other similar scopes but its far behind scopes which do the majority of the processing in an FPGA or ASIC.

Waveform update rate, Up to 2,000 wfms/s

Which doesn't specify the memory depth or sample rate at the time, for all their promoting of the "fast" display Keysight are very coy when it comes to providing details about the display rate of the S-seires which should be a big red flag when they so actively promote it on many of their other products. The limited speed is clear in The Signal Path video, and people on this forum have achieved similar (and even better) realtime performance from earlier acquisition boards rendering in software.

[babryce]

I continue to do research on this spec wise as my time allows.

I already saw the Tek/S comp with 2000 wfm/s, but thanks for putting it in the thread for others to look at.

The marketing materials of all vendors need healthy doses of salt given all the "*" I have found in the research I have been able to do.

[Keysight DanielBogdanoff]

I continue to do research on this spec wise as my time allows.

I'd highly encourage you to call our technical call line if you have questions. They are all engineers manning the phones and have a very deep technical knowledge. Phone number can be found here: http://www.keysight.com/find/contactus

I *think* you might want to start with the number in the "Support Specialists" tab. It's quick and free, and I even use it from time to time when I have a question.

The marketing materials of all vendors need healthy doses of salt given all the "*" I have found in the research I have been able to do.

To be perfectly honest, the "competitive comparison" type documents are not really made for engineers. They are made to help engineers convince managers and corporate purchasers. That being said, the lawyers go through that stuff with a fine tooth comb to make sure it's all technically accurate. It may be silly, but it's accurate 

[nctnico]

I wouldn't call it accurate. It is worded so that they can get away with it legally.

[Keysight DanielBogdanoff]

I wouldn't call it accurate. It is worded so that they can get away with it legally.

I guess what I'm trying to say is that you should take those side-by-side comparisons with a grain of salt.

[srce]

Timely thread, as I'm just looking at these too. Here's a table of the main spec points I was interested in:

Vendor	Keysight	R&S	Lecroy	Tek	Lecroy
Model	MSOS104A	RTO2014	HDO6104A-MS	MSO56	WavePro 254HS-MS
List price (GBP)	£21,000.00	£19,000.00	£22,500.00	£21,618.00
CPU	i5-3550S Quad  3GHz	i5	i7-2710QE Quad 2.1/3GHz	i5-4400E dual 2.7Ghz	i5-6500 Quad 3.2GHz
RAM (GB)	8	16	16	Not specified

16
 OSWindows 10 EnterpriseWindows 7 Embedded StandardWindows 7 Pro 64-bit EmbeddedClosed Linux / Windows 10 cost optionWindows 10
 HD240GB SSD200GB SSD80GB /460GB SSD
 Screen15"12.1"12.1"15.6"15.6"
 Res1024x7681280x8001280x8001920x10801920x1080
 External monitorYesYesYesYes4096x2304
 InterfacesUSB 3, 1G EthernetUSB 3, 1G EthernetUSB 2, 1G EthernetUSB 3, 1G EthernetUSB 3, 1G Ethernet
 Dimensions330H 430W 230D249H 427W 204D291H 400W 131D309H 454W 204D345H 445W 196D
 Analog Channels4440-64
 Digital channels1616160-2416
 BW (MHz) 50OHm at 1mV/div10001000100010002500
 BW (MHz) 1MOhm5005005001000
 Digital BW (MHz)400400250500
 Max BW Upgrade (MHz)80006000100020008000
 Noise @ 1mV/div 1GHZ B/W into 50Ohm (uVrms)90100145255155
 Noise @ 100mV/div 1GHZ B/W into 50Ohm (uVrms)96016505001610889
 Noise @ 1mV/div 500MHz 1MOhm (uVrms)~268130Not specified189
 Noise @ 100mV/div 500MHz 1MOhm (uVrms)~6502300Not specified1160
 Noise @ 1mV/div 20MHz 1MOhm (uVrms)~120~29Not specified64.8
 Vertical divs81010?
 GSa/s20/10102.56.2520
 Standard memory (MPts) for 4 channels active100505062.5100
 Max memory (MPts)40010002501251000
 Waveforms per second20001000000500000
 Max res  (at what b/w?)1216151615
 ADC Res108121212
 ENOB (what BW? What mV/div?)7.878.47.67.8
 SFDR 0dBm at 1GHz (dB)7268
 Phase noise at 10kHz (dBc/Hz)-121
 Min horizontal timebase (ps/div)5252020020
 Input Coupling1MOhm, AC, DC - 50Ohm DC1MOhm, AC, DC - 50Ohm DC1MOhm, AC, DC - 50Ohm DC1MOhm, AC, DC - 50Ohm DC1MOhm, AC, DC - 50Ohm DC
 BW Limiters20MHz, 200MHz, User-defined DSP - 500MHz20MHz, 200MHz, User defined20MHz, 200MHz20MHz, 250MHz20MHz, 200MHz, 500MHz, 1GHz
 FFTStandardStandardStandardStandard
 Intrinsic Jitter(fs rms)10028045060
 Jitter analysisTIE requires EZJIT option - £1600TIE requires RTO-k12 optionTIE StandardTIE standardWPHD-JITKIT option
 Matlab / User filtersOption - £1900DFP2 option
 Compliance testsBroadR-Reach, DDR, eMMC, Ethernet, HDMI 1.4, MHL 2.0, MIPI, MOST, PCI Express Gen4, UHS, USB 2.0, HSIC, XAUIUSB, Ethernet, PCIeAutomotive Ethernet
 Decode8B/10B, ARINC, DVI, HDMI, I²C/SPI, JTAG, MIPI, SPMI, PCI Express, RS-232/UART, SATA/SAS, SVID, Ethernet, USB 2.0 3.1 , USB-PD , UFS, eSPI, CAN, LIN, FlexRay, I2S, I3C - All bundled for free for nowLIN, CAN, FlexRay, I2S, MIL-STD-1553, ARINC 429, CAN FD, SENT, MIPI RFFE, MIPI D-PHY, MIPI-M-PHY, Manchester, NRZ, 8b10b, MDIO, USB, Spacewire, PCIe, CXPII2C, SPI (SPI, SSPI, SIOP), UART-RS232, CAN1.1, CAN2.0, CAN FD, LIN, FlexRay, MIL-STD-1553, AudioBus (I2S, LJ, RJ, TDM), USB1.x/2.0I2C, SPI, RS-232, CAN, LIN, FlexRay, USB 2.0, Ethernet, I2S, LJ, RJ, TDM, MIL-STD-1553, ARINC 429
 Passive Probes1 per channel - 500MHz - 9.5pF1 per channel - 500MHz - 9.5pF1 per channel - 500MHz - 10pF1 per channel - 1GHz. 10:1 and 2:1 - 4pF
 Diff probeN2750A - 1.5GHz - 2:1 - 0.7pF - 15V- 200kOhm - £2000RT-ZD10 - 1GHz - 10:1 - 0.6pF - 8V - 1MOhmZD1000 - 1GHz - 1pFTDP1500 1.5GHz - 1:1 10:1 - 0.85V  8V - 1pF - 200kOhm - £3700
 Power probeN7020A - 2GHz - 24V - 1:1 - 50KOhm - £2000RT-ZPR20 - 2GHz - 1:1 - 60V - 50kOhmRP4030 - 4GHz - 30V -  1.2:1 -0.1uFTPP0502 - 2:1 - 12pF - 500MHz
 ARBNoOption RTO-B6 - 100MHzNoOption - 50MHz- £1000
 NotesH/W buttons for all channelsDigital connector at the back No peak detect?BW limit at <4mV/div 50Ohm


Notes: List price for MSO56 includes digital probes to make it equivalent. 10GSa/s rate for HD6104A is interpolated, actual rate appears to be 2.5. Other options are available, I've just listed the ones I'm interested in. Keysight cliam measured ENOB for LeCroy is < 8. E&OE. Corrections welcome 

Probably between the Tek and Agilent for me at the moment. Tek looks pretty good apart from the noise figures and the eye watering cost of some of the extras. The passive probes sound good. The Agilent seems to have better performance and more extensive s/w options. I'd be interested to hear more from people who say it isn't necessarily the best as a general purpose scope and why?

If anyone has an MSOS104A, I'd like to know what the noise floor is when the BW is limited to 20MHz at 1mV/div.

[nctnico]

I find the waveforms/s number on the Keysight pretty low. This probably gets worse with more stuff like math enabled. The 6 channels on the Tektronix look very appealing to me.

[Marchello]

Hi!

I have on my work HDO6104 (not A) and sometime i use it with external monitor with rezolution 2560x1080 without any problems.

p.s. error in your table - HDO6104A have a 10GSa/s (not 2.5GSa/s as HDO6104)

BR
Mark

[srce]

p.s. error in your table - HDO6104A have a 10GSa/s (not 2.5GSa/s as HDO6104)

I don't think so. 10GSa/s is 4x interpolated 2.5GSa/s. From the manual:

"On "A" models, if you choose a rate greater than 2.5 GS/s, the system will automatically enter Enhanced Sample Rate, applying Sinx/x Interpolation to prevent aliasing at the higher sample rate. An upsample factor of 2 pts. is used for 5 GS/s timebases, or 4 pts. for 10 GS/s and higher timebases"

[Eric_S]

Yap, the LeCroy is 2.5GS/s, as is the Tek if you want the ADC to be 12bits, at 6.25GS/s it's 8 irrc.

If that matters to you is another story, again.


Also, if you've narrowed it down to Tek or Keysight. Perhaps you could invite them over for a demo? There aren't all that many scopes sold in the price bracket that you're shopping in, so I'd guess that they're going to be very accommodating towards you.

It's probably best to base that choice on the usage patterns that you have, rather than money that'll fade in the big picture of things anyway. So again, invite some sales people over, ask to loan it for a while and try to solve real world problems with them.

[Keysight DanielBogdanoff]

If anyone has an MSOS104A, I'd like to know what the noise floor is when the BW is limited to 20MHz at 1mV/div.

I'll grab one and check it. If you're going down to 1mV/div, I'd strongly encourage you to consider the 50ohm path instead of the 1Mohm. The 50ohm path is generally the lowest noise.

If you'd like a demo, it's often possible to work one out. You can PM me or call your local Keysight call center to get it rolling.

Also, it the ADC bits are a big thing for you make sure to ask about ENOB.

[srce]

If anyone has an MSOS104A, I'd like to know what the noise floor is when the BW is limited to 20MHz at 1mV/div.

I'll grab one and check it.

Thanks. 20mV/div would be interesting too.

If you're going down to 1mV/div, I'd strongly encourage you to consider the 50ohm path instead of the 1Mohm. The 50ohm path is generally the lowest noise.

Yes, that's true, but on the Tek spec sheet, they have it as lower for 1MOhm (64.8uV) input than 50Ohm (70uV). I guess probe attenuation comes in to it too, but it seems the Tek passive probes also support 2:1, which would be the same as some of the Agilent active probes.

Also, it the ADC bits are a big thing for you make sure to ask about ENOB.

Yep. That can be calculated from the noise figure though right?

Tek give their ENOB at 20MHz as 8.9 in to 50Ohm at 90% full screen, but don't state the Vertical resolution used. They do give RMS noise figures though at different vertical resolutions, so I think we can guess what it is. For 20MHz in to 50Ohm at 1mV/div, it's 70uVrms. At 20mV/div it's 102uVrms.

ENOB=log2(FS/(RMS*sqrt(12))

So:

log2((0.2*.9)/(0.000102*sqrt(12)))=8.9 (which suggests their vertical res was 20mV/div)
log2((0.01*.9)/(0.000070*sqrt(12)))=5.2

Keysight only specify ENOB=7.8 in their datasheet, but this is at 1GHz (and is slightly better than Tek, which is 7.6 at that b/w). At lower b/ws, the best I can see is on this Keysight blog, there is this graph:



Which suggests in "13-bit mode" it would be log2(0.1*0.9/(0.000076*sqrt(12)))~=8.4

It's not clear what b/w that is though. Possibly 300MHz from the associated text, so could be better at 20MHz. Also, I though high-res mode was 12-bits?

Although if your ENOB is <9, the extra bits perhaps don't matter, which is what I guess you are getting at 

[Keysight DanielBogdanoff]

Also, I though high-res mode was 12-bits?

I'm pretty sure that's at 12, not 13, and is just a weird/poor Excel axis label.

[Keysight DanielBogdanoff]

Melissa (you may have seen her on our YouTube channel) was able to take some measurements today on the S-Series and had a couple comments to add:

"
These are obviously not warranty specs or something we’d guarantee, but they are real measurements that you can try on your own.  Attached are screenshots, and below is my setup and results.   

My setup:
Nothing connected to channel one. 
BW limited channel one to 20MHz.
Added two Vrms measurements for AC and DC plus a Vpp.
(all of this is displayed in the screenshots as well)

Results:
20MHz BW, 1mV/div, 1MOhm input
Mean baseline DCVrms noise = 156 uV

20MHz BW, 1mV/div, 50Ohm input
Mean baseline Vrms noise = 180 uV

20MHz BW, 20mV/div, 1MOhm input
Mean baseline Vrms noise = 366 uV

20MHz BW, 20mV/div, 50Ohm input
Mean baseline Vrms noise = 288 uV


This is a quick measurement that anyone can verify on the spot to check noise, which they might enjoy.  But I would also strongly direct people more towards system ENOB to determine the signal integrity of their oscilloscope.  One thing to pay attention to is that we specify system ENOB for the S-Series in the datasheet. Some others only include ADC ENOB, which does not come close to telling the whole story because that only specs one single component within the system and disregards the effects of the front end, clocking, and the rest of the oscilloscope design.

For some more resources, here's an App note on SI:
http://literature.cdn.keysight.com/litweb/pdf/5991-4088EN.pdf

And a blog on SI:
https://community.keysight.com/community/keysight-blogs/oscilloscopes/blog/2016/09/01/how-does-oscilloscope-signal-integrity-impact-your-measurements-and-analysis
"

I hope this helps!

[srce]

Also, I though high-res mode was 12-bits?

I'm pretty sure that's at 12, not 13, and is just a weird/poor Excel axis label.

It's 13 elsewhere in the blog too. E.g. in this table:

Bandwidth	RTO2000 High Def Mode	S-Series High Res Mode
10 kHz – 50 MHz	16 bit	13 bit
100 MHz	14 bit	13 bit
200 MHz	13 bit	13 bit
300 MHz	12 bit	13 bit
500 MHz	12 bit	12 bit
1 GHz	10 bit	11 bit

Unfortunately in the demo of the Infiniium s/w, you can't open the Acquisition dialog to see what choices there are and it doesn't say in the help. But, I guess it doesn't actually matter, given you're not getting close to that number of bits anyway.

[srce]

Results:
20MHz BW, 1mV/div, 1MOhm input
Mean baseline DCVrms noise = 156 uV

20MHz BW, 1mV/div, 50Ohm input
Mean baseline Vrms noise = 180 uV

20MHz BW, 20mV/div, 1MOhm input
Mean baseline Vrms noise = 366 uV

20MHz BW, 20mV/div, 50Ohm input
Mean baseline Vrms noise = 288 uV

Thanks to you both. I think it's the AC RMS values from the screenshots that you want to promote though

Scope	BW	Vertical setting	Input	Noise AC RMS	ENOB limit based on noise
MSO56	20MHz	1mV/div	50 Ohm	70uVrms	5.2
MSO56	20MHz	1mV/div	1 MOhm	64.8uVrms	5.3
MSO56	20MHz	20mV/div	50 Ohm	102uVrms	9
MSO56	20MHz	20mV/div	1 MOhm	104uVrms	8.9
MSOS804A	20MHz	1mV/div	50 Ohm	48uVrms	5.4
MSOS804A	20MHz	1mV/div	1 MOhm	115uVrms	4.1
MSOS804A	20MHz	20mV/div	50 Ohm	260uVrms*	7.3
MSOS804A	20MHz	20mV/div	1 MOhm	139uVrms	8.2
RTO2014	20MHz	1mV/div	50 Ohm	33uVrms	6.3
RTO2014	20MHz	1mV/div	1 MOhm	29uVrms	6.4

* This looks a bit fishy to me, as it's far higher than the speced noise for the MSOS104A at 1GHz. (Which is 163uVrms). It does appear to be the MSOS804A the measurement was made on, from the 8.40 GHz at the top of the screen. But if the b/w is set to 20MHz, that shouldn't matter should it? Does one of the other b/w limiters need to be on too? For the 1MOhm screenshot, the 8.4GHz drops to 500MHz.

[Neganur]

There’s global bandwidth limit and per channel setting.
I can do those for you on a 1GHz model (does the time/div matter?)

[srce]

There’s global bandwidth limit and per channel setting.
I can do those for you on a 1GHz model (does the time/div matter?)

Thanks  - It could do, if you had it really as a really tiny interval. 1us as in the above pics seems OK.

I did notice you had the DSOS for sale. How did you find it overall?

[Neganur]

Noise AC RMS V/div 50 1M ------- | -------- | --------- 1mV | 42.7 uV | 135 uV 20mV | 119 uV | 151 uV Out of curiosity, since noise seems to change with sampling rate and memory depth, I fixed the amount of memory and swept the sampling rates. The data is summarised here: 50Ohm, 1 mV/div, 20 MHz BW limit | 100ns/div | 50us/div GSa/s | 20 kpts 1 Mpts | 10 Mpts ------- --------- --------- --------- 20 | 42.6 uV 42.7 uV | 44.1 uV 10 | 43.4 uV 43.8 uV | 44.9 uV 5 | 52.6 uV 52.7 uV | 55.9 uV 2.5 | 51.7 uV 51.7 uV | 54.9 uV 2 | 53.8 uV 54.0 uV | 57.3 uV 1.25 | 51.9 uV 51.9 uV | 55.3 uV 1 | 53.7 uV 53.8 uV | 57.2 uV 0.5 | 50.6 uV 50.5 uV | 53.8 uV 50Ohm, 20 mV/div, 20 MHz BW limit | 100ns/div | 50us/div GSa/s | 20 kpts 1 Mpts | 10 Mpts ------- --------- --------- --------- 20 | 135 uV 135 uV | 132 uV 10 | 151 uV 150 uV | 151 uV 5 | 331 uV 330 uV | 360 uV 2.5 | 337 uV 339 uV | 370 uV 2 | 301 uV 300 uV | 328 uV 1.25 | 364 uV 363 uV | 395 uV 1 | 310 uV 311 uV | 340 uV 0.5 | 319 uV 318 uV | 350 uV (Those odd jumps at 1.25 GSa/s are not measurement errors. I repeated those and got the same every time)

[Rich@RohdeScopesUSA]

Hi srce - a couple tweaks for the RTO2000 portion of your table (looks like you might have had an old datasheet - here is the latest version:

https://cdn.rohde-schwarz.com/pws/dl_downloads/dl_common_library/dl_brochures_and_datasheets/pdf_1/RTO2000_dat-sw_en_3607-2684-22_v1601.pdf

1.  We run a Core i5 with up to 16GB of RAM.  The majority of our analysis work is done in a custom ASIC so we don't rely on the CPU as much.
2.  We have USB 3.0 ports
3.  In addition to the standard analog filters (20MHz and 200MHz), we also offer digital filters that are handled in our custom ASIC with a wide number of user-definable choices.
4.  In addition to USB 2.0, PCIe 1.1/2.0 and Ethernet (10G, 100Base-T1, 2.5/5G, EEE, 1000Base-T1) compliance, we also offer MIPI D-PHY and eMMC.
5.  We offer a very broad amount of trigger/decodes (not just decodes like some other scopes):
I2C/SPI Serial Decoding
UART/RS-232/RS-422/RS-485 Serial Decoding
CAN/LIN Serial Triggering and Decoding
FlexRay™ Serial Triggering and Decoding
I²S Serial Triggering and Decoding
MIL-STD-1553 Serial Triggering and Decoding
ARINC 429 Serial Triggering and Decoding
Ethernet Serial Decoding
CAN-FD Serial Triggering and Decoding
SENT Serial Triggering and Decoding
MIPI RFFE Serial Triggering and Decoding
MIPI D-PHY Serial Triggering and Decoding
MIPI M-PHY Serial Triggering and Decoding
Manchester and NRZ Serial Triggering and Decoding
8b10b Serial Decoding
MDIO Serial Triggering and Decoding
IEEE 100BASE-T1 Serial Triggering and Decoding
USB 1.0/1.1/2.0/HSIC Serial Triggering and Decoding
USB 3.1 Gen 1 Serial Triggering and Decoding
USB Power Delivery Serial Triggering and Decoding
USB 3.1 SSIC Serial Triggering and Decoding
SpaceWire Serial Triggering and Decoding
PCI Express 1.1/2.0 Serial Triggering and Decoding
CXPI Serial Triggering and Decoding
6.  For the diff probe, you'll want our RT-ZD-series, not the RT-ZS-series.

Sounds like you are leaning towards Tek or Keysight already (both good scopes), but if you'd like to try out a RTO2000 our team in the UK can get you one easily - just PM me.

-Rich

[Rich@RohdeScopesUSA]

(Those odd jumps at 1.25 GSa/s are not measurement errors. I repeated those and got the same every time)

In general, there are two components that dominate noise - the front end and the ADC.  The front end dominates at smaller volt/div settings and the ADC dominates at larger volt/div settings. 

I'm guessing the reason you are seeing a large jump in noise when you go from 10GS/s (and above) to 5GS/s (and below) is I believe the S-Series is only 10-bits at 10GS/s and above.  Below that it runs at 8-bits.  This would explain why you see the biggest jump in your second table (the larger volt/div setting).

Perhaps Daniel can clarify or explain if I have misunderstood how the S-Series operates.

-Rich

[Neganur]

I did notice you had the DSOS for sale. How did you find it overall?

Very happy with it, it is very nice to use and compared to the 9000-series the USB3 and fast processor is most welcome.
I mainly capture waveforms and process them in matlab so most of the decodes etc are not so super critical.

However, zone triggers (InfiniiScan option N5415B ), advanced math (user defined function option N5430B) are not in the scope by default. Of course thw most basic triggers (I2C, RS232, SPI) are missing too...can understand that the decode is out but the triggers...doh!

I can totally understand that the advanced options are costly and indeed optional but I would have assumed serial bus trigger, zone trigger and something where I can multiply with a Log(x) and Cosine using a free-form equation entry standard in such a scope.

Probe selection available is good too. I would have loved if some of the probes that worked with the 9000-series scope worked with this one too (namely 1153A, 1152A and 1155A.

It's fairly heavy though and tends to blow a lot of hot air.

One really cool feature is the offline Infiniium tool that basically runs the scope software on your computer for post processing (love it for reporting!!) but unfortunately it is a paid option (because it can run some of the paid options that the scope also uses). It would be so incredibly nice if I could just send a data capture to a customer and have them use the offline tool to see what I am talking about. Dear Keysight, release a stripped down version that can do only the basic scope functions (fft is a must)?

[Neganur]

[...] Also, I though high-res mode was 12-bits?[...]

High Resolution does 13 bits at 1.25GSa/s and below.

I'm guessing the reason you are seeing a large jump in noise when you go from 10GS/s (and above) to 5GS/s (and below) is I believe the S-Series is only 10-bits at 10GS/s and above.  Below that it runs at 8-bits.  This would explain why you see the biggest jump in your second table (the larger volt/div setting).

You're right, it switches to 8-bit below 10 GSa/s

[srce]

V/div		50		1M
-------	|	--------	|	---------
1mV	|	42.7 uV	|	135 uV
20mV	|	119 uV	|	151 uV

Thanks. That seems more like what I'd expect.

Out of curiosity, since noise seems to change with sampling rate and memory depth, I fixed the amount of memory and swept the sampling rates.

Interesting. Looks like there's a bit of a step <10GSa/s. I think I read on another thread that the resolution at lower sampling rates is 8-bit rather than 10-bit, so perhaps related to that. (Although that seems the wrong way round to me!)

edit: Ah, too slow! Already covered.

[srce]

Hi srce - a couple tweaks for the RTO2000 portion of your table (looks like you might have had an old datasheet - here is the latest version:

https://cdn.rohde-schwarz.com/pws/dl_downloads/dl_common_library/dl_brochures_and_datasheets/pdf_1/RTO2000_dat-sw_en_3607-2684-22_v1601.pdf

Thanks Rich, I've updated it.

[Rich@RohdeScopesUSA]

Hi srce - a couple tweaks for the RTO2000 portion of your table (looks like you might have had an old datasheet - here is the latest version:

https://cdn.rohde-schwarz.com/pws/dl_downloads/dl_common_library/dl_brochures_and_datasheets/pdf_1/RTO2000_dat-sw_en_3607-2684-22_v1601.pdf

Thanks Rich, I've updated it.

And I missed a couple other items:

The RTO2000 runs W7 Embedded Standard.  It comes with at least a 200GB SSD. 

For the 1mV/div setting with the 20MHz BW limit on, here is what we get (keep in mind these are also at 10div - so they will appear about 20% worse than scopes with only 8div):
20MHz BW, 1mV/div, 1MOhm input:  29uV AC-RMS
20MHz BW, 1mV/div, 50Ohm:  33uV AC-RMS

-Rich

[Neganur]

How is there less noise with 1Mohm than 50Ohm?

[Rich@RohdeScopesUSA]

How is there less noise with 1Mohm than 50Ohm?

When we were designing the front end for the RTO (and RTE/RTA/RTM for that matter), the team focused on low noise for both paths as they are general purpose scopes.  Traditionally, scopes have only focused on improving (and talking about) noise performance on the 50Ohm path so I think people are used to seeing 1MOhm have significantly more noise. 

-Rich

[Someone]

How is there less noise with 1Mohm than 50Ohm?

Rather than the worthless marketing speak above, I'd guess that the two paths have different v/div gain structure/grouping/settings. If you could get noise data plotted from multiple v/div settings you'd likely see some step changes in the channels at different points...

Actually checking the RTO2000 data sheet they have listed the 50 ohm input impedance noise for the different bandwidth models separately, but the noise for the 1M input is just a single table. There are no step changes in the data they show there. Perhaps there is a low pass filter on the 1M input path that restricts its bandwidth to less than on the 50 ohm path? But that shouldn't be making much effect at the 20MHz bandwidth limited measurement Rich presented.

[srce]

Actually checking the RTO2000 data sheet they have listed the 50 ohm input impedance noise for the different bandwidth models separately, but the noise for the 1M input is just a single table. There are no step changes in the data they show there. Perhaps there is a low pass filter on the 1M input path that restricts its bandwidth to less than on the 50 ohm path? But that shouldn't be making much effect at the 20MHz bandwidth limited measurement Rich presented.

The 1MOhm input b/w is 500MHz, whereas 50Ohm is up to 6GHz. Presumably they can use some lower noise components / different design to optimise noise given the lower b/w requirement on that path.

[Keysight DanielBogdanoff]

Given where this thread has gone, this talk from one of our designers might be interesting for you all. You may or may not want to skip the first video, it's pretty basic.
https://www.youtube.com/playlist?list=PLzHyxysSubUmxGOMVpiKLxouweh2AAlG1

[babryce]

I actually evaluated the S-Series and the RTO on site and am thankful to R&S and Keysight for providing me loaner units. I will write up a burb in a few weeks after the acquisition is over, which I think will occur just after April 1st.

[srce]

I've gone for an MSOS204A in the end. Nearly went for the Tek, as there are some quite nice things about it, but was swayed in the end by some of the better specs for the Keysight scope as well as its maturity (there were a few issues when I saw a demo of the 5 series). Looking forward to having a play 

[babryce]

I will now attempt to write my blurb as I made my recommendation for a scope.

This post will be the background and selection choice.

I really appreciated SRCE and other's comments on this thread including the company reps.

All of the scope discussed here are good. Of the biggest vendors of test gear I think we have only really left out Yokagowa.

As I mentioned elsewhere I evaluated the RTO2000 and S-Series on site. I also had access to a LeCroy WaveRunner 8000 series. I own an MDO4000 at home (that I got surplus with the SA-3 option).

So the scopes considered were basically: LeCroy WR 8000, HDO6000; Tek MSO56, MDO4000; Keysight X-Series, S-Series; R&S RTO2000, RTE1000.

As was noted at the beginning of this thread these scopes (the big 3 in the title that I was at first focusing on) are not the same. Of course they have all sorts of design choices in them.

I'm very happy that today we have a lot of choice in scopes. This seems a healthy and competitive market. I hope it is not so competitive that too much consolidation occurs as we have seen in the IC market as of late. New companies growing from the bottom as well have a very competitive T&M space.

First what do I need? Some people asked at the start of this thread. I gave the answers I could then, but I had just joined the organization. While I have 15+ years electronics experience, I did not know details of what the organization was like. Thankfully it being large, it took a bit of time for the money to become available and I had time to get some on site to look at. After a few weeks I realized that I would probably be using this scope for everything and that I would probably have it for 5-10 years. I also realized that I would probably always mostly be asked to do analog focused design with some mixed signal stuff thrown in. Most of it not to high a frequency work.

My focus being analog came up with the initial list as these all have higher resolution ADCs in them. My group only has two vendors of scopes at present: Tek and LeCroy (technically we have one Yokagowa) Neither is surprising. LeCroy has long had good relationship with particle physics folks.  LeCroy himself having a degree himself in that area. The folks that use the LeCroys are the analog designers mostly and they lean toward particle type work. The digital focused people all seem to have Tek scopes. If you need a ton of digital channels the new 5-Series might be a good fit.

Given my analog leanings my focus would be on signal integrity. Some of my choices would be arbitrary but with finite time you have to sometimes cut on something. For instance SpaceWire decode only exists on LeCroy, R&S and Keysight. Triggering only exists on R&S and Keysight.

How could I eliminate some of the scopes without a full multi-day test of each? Well given I would be using this thing effectively forever (in EE terms), I would want something that could cover any foreseeable task. That meant digital and analog work.

Although the Tek 5-Series looks impressive in some ways I think it has the following major downsides right now:

* Being very new it is not mature
* Protocol decodes are lacking
* Discounts (bang for buck) would almost certainly be lower given it is the newest thing
* Sample rate at full resolution is only just over 3 GSPS; 6 GSPS is only at 8-bits

On the positive side the MSO56 would have given me 2 extra analog channels most of the time given I will mostly do analog work even if the MSO56 is only the same mix as a normal 4 + digital if you choose to use it that way, the 2 extra analog channels look like a very good thing. If I had at more time I would have evaluated it. My biggest two reasons for not evaluating the 5-Series were the: lack of decodes, and the simple fact that with so much Tek equipment around and with quite a few new hires picking equipment someone would certainly bring this capability to the group.

I played with the Waverunner 8000 we already have to get a sense of the most up to date LeCroy software and looked at the pricing structure and got quotes for the HDO. I must say that of all the marketing I felt the 10 GSPS advertised rate being an interpolation on 2.5 GSPS was the line I was the most annoyed with. Interpolation is not sampling. For a while I was focused heavily on LeCroys because of the analog designers support of them in the group. This could have meant shared active probes and other items. But even more than with Tek, there already are new LeCroy's in the group so it did not provide any new capabilities I could not borrow on the rare occasion I might need a unique feature.

This leaves us with Yokagowa, Keysight and R&S offerings. I must admit I didn't investigate Yokagowa carefully I probably should have but I did not. On the Keysight front I've use lower end X-Series a lot. Very fine for quick bench work but not really for signal integrity and analog work. One of my biggest annoyances with them is that for me being colorblind channel 1 and 2 are the same color, and you cannot change it. One would think colorblind safe color schemes that map makers have known about forever would have reached UI designers but time and again this is not so.

So in the end the two scopes I tired out on site were the Keysight S-Series and the R&S RTO 2000.

In the end my recommendation was to select the RTO2000. Details on the next post which some might find the more interesting without the background bit in this one hence the separate post. 

[babryce]

Having given some fairly stream of consciousness background on my scope down-select I will now go into the final two scopes: The MSOS104A and the RTO2014. I will try to be terse.

Signal Integrity
As analog signal performance was core to this exercise I really focused on this when I evaluated the scopes.

Keysight has a very fancy ASIC at the core of their S-Series, that while older is still a very excellent performer. The S-Series comes fully calibrated up to the full 8-GHz (though you only can do this on 2 channels). It runs at 20 GSPS on 2 channels and 10 GSPS if you are using 4 channels. The RTO just runs at 10 GSPS for any number of channels. It is calibrated to the bandwidth you order though it is still just a software limit I believe.

I measured things like the SFDR with the S-Series and the RTO at low signal levels. I focused on repetitive signals that could be averaged as well to find the best DANL etc. The S-Series basically did better in these tests. The RTO showed spurs above the noise floor in many of these tests while the S-Series did not. Note these were not at high levels like the animation Keysight showed. The difference in the triggering systems was very noticeable. Trigger jitter while undesirable would have effectively spread the power of any component out. I did not really get to look too deeply at this in relation to phase noise measurements.

As shown in the Signal Path video, the S-Series is good at demodulating tasks. These are optional extras and not something I would be doing but a lot of care looks like it went into the S-Series's ASICs and analog design. This focus I felt was born out in the SFDR performance.

The RTO of course is well designed also but I did not feel the signal chain was as good as the S-Series. Of course it is an 8-bit converter which can process a lot more waveforms/s. They are different.

R&S make lot of their HD option which increases the ENOB of the system via digital filtering. I would call this emblematic of their approach to design. While I felt the heart of the Keysight was a very excellent set of analog and mixed signal ASICs, I felt the heart of the R&S was software/digital. In some ways this is more like LeCroy.

The HD option increases resolution for signals under 1 GHz, by applying a digital low pass filter after sampling before feeding the samples to the rest of the chain (see: High-Resolution Measurements with R&S Oscilloscopes; which compares it to a simple moving average filter).

As I am focused on < 1 GHz here this is a big deal. With HD mode on while SFDR and other tests were slightly worse than the S-Series it wasn't very dramatic. Everything on both scopes was of course inline with their datasheets.

If you had to go faster than 1 GHz and needed the best signal integrity, I would say: stop looking; go buy an S-Series. The new 5-Series doesn't have the bandwidth or sample rate. The LeCroys don't in their regular HDO series (though plenty of super fast fancy LeCroys out there that I have used in the past).

For a time I was very focused on the fact that the S-Series had better signal integrity by my testing at least.

Looking at the core specs, the upgrade path is unmatched for bandwidth at this price and the time base was excellent. I feel many people don't think about time bases enough. An error in "x" goes into "y" though the differential at the point. A good time base is super important.

User Interface

The Keysight user interface is fine. You can find most things in the menus without too much trouble and once you know where it is it is not hard to use. It has dedicated knobs for each of the 4 main channels and the digital signal bundle is on the front (the R&S MSO option is on the back if you get it).

I lot of people want dedicated knobs for each channel, but I think that might be a bad use of space if there are more important functions. I found the lack of other knobs and buttons compared to the front panel of the RTO made the S-Series harder to use.

The software for the Keysight is very mature but with that maturity comes a lot baggage. In general things are configured though dialog boxes. While you can use the touch screen to use these dialogs, they are text driven and on the smaller side. I think the UI works for the most part better with a mouse. I would say this probably is because of the legacy of the software and where it came from with touch being only added recently. Pinch and cursors work well though.

R&S have less time in the scope market (though I don't know how much Hameg is in the RTO line). The software they have created while not beautiful (looks like the 5-Series will be easier on the eyes), is clearly designed with using the touch screen most places. I appreciated the combination of most features into visual depictions of signal chains and acquisition settings together.

Fragmentation of averaging, filtering, etc is annoying. LeCroy has long put this into one big channel menu which is good. R&S is almost as good at putting it all together, perhaps better if you like graphical representation of the signal chain. While we all know what the options are for, no one gives all text slides for a presentation, more people are visual. Keysight and others should take note.

As I noted in the previous post I am colorblind (red-green), this affects perhaps 2-8% of the user market for Oscilloscopes.

R&S uses shared knobs for channels. While I doubt they had the colorblind in mind when they made the design choice, this means they backlit everything with RGB leds to match the selected channel color. This means for perhaps the first time ever, all of the colors (not just the ones I can change on the screen via menus which often is a feature on analysis scopes) actually match and I can tell them apart (the backlight matches screen color). Those with full color vision take color coding as a standard way to make things easy. Often the over reliance on it makes my life miserable. Here at least I have the ability to make it work well for me: a first. Others please do this if you cannot make things colorblind safe (which you could if you tried at all).

This is not a deciding feature, because I always have worked around this by moving the vertical offset a little to figure out which pattern is which, but it still matters.

Overall the UI for the RTO is better than the S-Series. The visual driven system with fewer text menus and drop boxes is just nicer to use.

There are some oddities in the RTO UI like channel math functions that use only one channel showing two channel inputs. Some colors in dialogs did not update with the channel color. This later issue is just a bug which I am sure they will eventually fix (I don't remember where it was).

Boot time
The RTO booted in 70-80 seconds
The S-Series booted in 220-240 seconds

The RTO and S-Series reached windows at about the same time, but the software for the Keysight is just bigger I guess. It loaded things like a Matlab kernel. I'm not sure if that occurs if you do not pay for the Matlab option or not but for these fully loaded demos this was what I saw. Boot times aren't ultra critical but that sort of time difference is obvious and probably could get annoying with time. Instrument boot time to ready varies a lot depending on what calibrations it must do etc. Worth noting though for a workhorse.

Noise
Neither scope was loud in testing under the loads I could give them, but the fans on the RTO were louder than the S-Series on my units. My old LeCroy LT344 was a lot louder than either and I did at times find it tiresome to listen to. I doubt most would be bothered by either the RTO or S-Series noise levels.

Math
As I mentioned above the S-Series has a Matlab option. That's cool but it costs extra. Normal channel math like +-*/, etc are there. Not as fancy as a LeCroy. The Matlab option looks from the datasheet nicely integrated into the traces though.

I personally started as an Engineer with Matlab, but I have mostly stopped using it in favor of Python. Python works for more things than Matlab and unless I needed a very special toolbox I would not choose to use it. While I can still write code in Matlab I try not too these days. Julia looks interesting and is like Matlab style wise maybe its fast execution would be helpful for scope like applications. Anyway... Matlab for advanced math and pay more for the S-Series.

The RTO's math can do more functions built-in and do compound expressions without chaining channels. This is good. Though I would take my data off the scope and use Python to look at it in detail for full analysis, for quick functions in real time the better math of the RTO is very valuable. This feature looks like it is built-in for the base model. It looks like the 5-Series might also have this though try as hard as I could I could not find any screen shot of it...

It is a little unclear how deep the external hooks are for the RTO. I found a Python tool, and saw the .NET/C# API along with the normal VISA like stuff. I'd love to see a direct realtime API for custom decoding and math in Python or Julia or something open and free. It might crimp the decode software market though...

Probes
Keysight has more probes than R&S but it looks like R&S has been expanding. They have the most common probes I might want, so its a draw really. Costs are almost the same.

Documentation
R&S has a proper manual for the RTO which is a few thousand pages in length. Keysight has the help file you would see in software. Well written documentation is undervalued and is frankly rare. I like single monolithic documents rather than fragmented documentation.

Option, options, options
To make the RTO2014 closer to the MSOS104A requires options. I would say the biggest one is the ovened reference (RTO-B4). I would not bother to get a scope costing this much without an excellent time base. With the time base option they are similarly good. You also need the HD mode (RTO-K17) and the Mixed signal module (RTO-B1).

Final bit
If you are going to stay below 1 GHz with these options the RTO has some advantages: faster waveform update rates, faster boot times, friendlier UI.

Above 1 GHz if signal integrity is your prime motive I would probably just select the S-Series.

For me although I was very focused on signal integrity, I felt like the easy built in math (without writing an m file), and UI would make my daily tasks faster with the RTO. Signal integrity was still good, and for the most difficult things I will still have to use RF techniques regardless.

I got some other options on my RTO quote that matched the current offers from Keysight. Some of these options on the R&S list for nearly 2x what they do on the S-Series. So think through your future need carefully! I'm not sure you have as much leverage for getting an option at a discount later.

Just ask sales folks for what you need at the price you need it, they will probably help you if they can. The Keysight and R&S sales staff were both helpful.

[Rich@RohdeScopesUSA]

Thanks babryce for your analysis and posting up your results.  There are lots of good scopes on the market these days and I've always been a fan of trying them out to see which one best fits your exact needs. 

-Rich

[exmadscientist]

Thanks for posting your thoughts on these models. It's always useful to see a head-to-head comparison from someone independent.

Although the Tek 5-Series looks impressive in some ways I think it has the following major downsides right now:

* Being very new it is not mature
* Protocol decodes are lacking
* Discounts (bang for buck) would almost certainly be lower given it is the newest thing
* Sample rate at full resolution is only just over 3 GSPS; 6 GSPS is only at 8-bits

On the positive side the MSO56 would have given me 2 extra analog channels most of the time given I will mostly do analog work even if the MSO56 is only the same mix as a normal 4 + digital if you choose to use it that way, the 2 extra analog channels look like a very good thing. If I had at more time I would have evaluated it. My biggest two reasons for not evaluating the 5-Series were the: lack of decodes, and the simple fact that with so much Tek equipment around and with quite a few new hires picking equipment someone would certainly bring this capability to the group.

You weren't missing much by skipping the 5 Series. We recently had one in for a test and it was unimpressive. "Not mature" is really probably the best way to summarize it. A kind-of-ranty summary of the 5 Series:

The good:

• The medium-resolution screen is really nice. Not quite bright enough, and not quite sharp enough (seriously, my several-years-old phone has more pixels), but it's so far ahead of everything else in this space that it's kind of sad.
• Eight channels! This is really the only thing that's truly special about this scope. Everything else is secondary.
• Good, fairly snappy performance at default settings (1Mpts/channel record length)
• Good zoom features (not, you know, any better than other recent Tek scopes, but at least they didn't regress here.)
• Dedicated memory for each channel (so enabling/disabling channels is almost entirely a display thing). Nice, but hardly essential.
• Eight! Channels!

The bad:

• Trivial to crash. It crashed a lot during testing, and I wasn't trying that hard. It only cleanly failed once (with a "please reboot me" message); the rest of the time, it just shut itself down a few seconds after hanging. If this is how bad it is now, with the latest firmware, I can only shudder to think how awful it must have been at launch.
• Terrible UI. Whoever thought up "single-tap to access useless controls you already have on the front panel, double-tap to get to the real menu" should be flogged. Just bring up the double-tap menu already, I'm always going to need it.... I could go on about the UI, but it's obvious after a few minutes' use that it's crap and they really should just sit down and watch someone try to use the damn thing.
• Insane horizontal (record length) controls. The UI ties together sample rate, horizontal scale, and record length, so that adjusting one changes the other two. It turns out to be extremely difficult and annoying to increase the record length above the default ~1Mpts/channel. It's an absolutely bizarre system, but I can see why they had to do it this way, because...
• ...the scope's deep memory performance is pathetically awful. Yes, I know, cranking the capture length up to the full 125Mpts is not something you're supposed to do often. But I had a problem where the device under test was going weird sometime after startup, so I needed to grab the full startup sequence to see exactly when it went weird. So I wanted a long, high-resolution capture. That's what the memory is for, right? But it should never take fifteen seconds to capture one single-shot waveform! (And before anyone asks, the horizontal scale was somewhere around 4ms/div, so 40ms per capture. Maybe 10ms/div, I don't remember for sure. But short enough that the capture period itself was very quick.) And the thing doesn't have the decency to give you partial updates or any kind of feedback that it's processing... it just sits there, its little green "Triggered" light on solidly. The first time I did this I thought I'd crashed it again. I'm grateful I only needed two or three of these painful captures to figure out where in the startup sequence I needed to be looking and could set up a better trigger.

The ugly:

• Working with 8 traces is pretty annoying. There's a "stacked" mode and an "overlay" mode, but that doesn't help much: if they're stacked, each trace is too small. If they're overlaid, something's always blocking something else. (Why can't we have transparent traces? It's 2018!) You end up playing the same old games as usual with offset and scale knobs.
• Crazy long boot time. But that might have been affected by crashing so often; maybe a reboot-after-crash takes longer?
• The combination of default sinc interpolation and low sample rates is not a good one. Sinc interpolation goes nuts when you have frequency content above Nyquist... and with the rates this thing likes to sample at, you're going to have a lot of aliasing if you like to use the zoom button. I'm really glad I noticed this setting before trying to figure out why the signal I was looking at seemed to be bouncing below ground! Interpolation settings should not lie to you nearly as badly as this scope's do.
• The ludicrous price tag. Enough said!

So,yeah: it's immature and not worth the current asking price. At half the price I'd be interested. The above comes across more negative than I intended, but it's really just not a revolutionary product in any respect. Certainly it's not worthy of Tek's marketing hype, but I guess Danaher is just really proud that they finally spent money on something? I think it'll actually be more interesting to see how the other manufacturers respond in their next refreshes.

Does anyone know how other scopes in this class (S-Series, RTO, etc) stack up in waveform rate when you need a deep memory capture? I'm guessing they do better than 0.07 waveforms per second, but I haven't actually tried them myself and I couldn't come up with anything when searching around. The S-Series is probably next on our tryout list, but that requires dealing with the local salesmen and that's never pleasant, not when there's real work to be done.

[Eric_S]

Working with 8 traces is pretty annoying. There's a "stacked" mode and an "overlay" mode, but that doesn't help much: if they're stacked, each trace is too small. If they're overlaid, something's always blocking something else. (Why can't we have transparent traces? It's 2018!) You end up playing the same old games as usual with offset and scale knobs.

Too bad you can't combine the stacked and overlayed modes. So that you'd have (say) four windows with two traces in each.

[nctnico]

Working with 8 traces is pretty annoying. There's a "stacked" mode and an "overlay" mode, but that doesn't help much: if they're stacked, each trace is too small. If they're overlaid, something's always blocking something else. (Why can't we have transparent traces? It's 2018!) You end up playing the same old games as usual with offset and scale knobs.

Too bad you can't combine the stacked and overlayed modes. So that you'd have (say) four windows with two traces in each.

IMHO that should be possible. Dunno for Tektronix specifically but on other scopes (Yokogawa) you can usually combine more traces in one window.

[fonograph]

One important thing that was not mentioned in this thread is DC accuracy,I believe its limited by ADC non linearity.I think you can oversample 8 bit ADC so it have 16 bit ADC like noise levels,but you cant improve linearity/DC accuracy by oversampling.The LeCroy 12bit ADC is relatively slow at 2.5 Gs/s,but its true 12 bit,without oversampling.

Even better is 14 bit Native Instruments PXIe scope,its only 1 Gs/s,but its DC accuracy is best and they have by far most detailed datasheet about signal integrity.

I am too lazy to read it again but I believe most 8 bit scopes,including RTO2000 and MSO 56 are 2% DC accuracy,Keysight S is 1%,HDO is 0.5% and  NI 14 bit scope is 0.25%

[babryce]

Interesting. I had not even considered crashes. 0 crashes of either S-Series or RTO in my testing FWIW.

If the S had crashed that boot time would have been very painful.

I probably should have looked harder at the HDO9000 not just the HDO6000 and WR 8k. We already have a HDO9k though.

Maybe an HDO8000A would work for you given you like the channel count high? LeCroy's UI is very mature at this point I feel like its not radically different since at least 2008 time frame. Sample rate is not that different from 5-Series.

The S-Series is a beast though.

[exmadscientist]

Working with 8 traces is pretty annoying. There's a "stacked" mode and an "overlay" mode, but that doesn't help much: if they're stacked, each trace is too small. If they're overlaid, something's always blocking something else. (Why can't we have transparent traces? It's 2018!) You end up playing the same old games as usual with offset and scale knobs.

Too bad you can't combine the stacked and overlayed modes. So that you'd have (say) four windows with two traces in each.

IMHO that should be possible. Dunno for Tektronix specifically but on other scopes (Yokogawa) you can usually combine more traces in one window.

I'd have thought that too, but we couldn't figure out how to do it during our demo period.

Still, we didn't explore every corner of the crap UI and so it might be buried somewhere. The Tek salesdroid's pitch had a lot of fluff about how stacked mode "maximizes ADC usage" for each channel, but the DUT on our bench right now doesn't call for much precision, so we didn't waste much time trying to get that to work out. Traditional trace stacking was just fine (if tedious).

[srce]

Interesting. I had not even considered crashes. 0 crashes of either S-Series or RTO in my testing FWIW.

In the ~2-3 hour demo I had with Tek, it crashed twice     Obviously going through all the different features of the scope in that period though.

[exmadscientist]

I probably should have looked harder at the HDO9000 not just the HDO6000 and WR 8k. We already have a HDO9k though.

Maybe an HDO8000A would work for you given you like the channel count high? LeCroy's UI is very mature at this point I feel like its not radically different since at least 2008 time frame. Sample rate is not that different from 5-Series.

The S-Series is a beast though.

We do a lot of different things. On the bench right now is a board with an interesting power stage, so voltage and current probe pairs everywhere. The high channel count scopes really shine for that kind of work. (In fact isn't that how LeCroy markets many of them?) Next month it'll be something different, so who knows?

It's also worth noting that I wasn't the one who called in the 5 Series, another engineer here was. I just got to use it for a while. I think the Tek sales reps were fairly aggressive in trying to demo it; perhaps it's not selling well? (Again, no surprise at their asking prices.) It's also the showiest scope around, and we were hoping that would help with upper management. Alas, they seem to continue keeping their heads in the usual places and being stingy. Guess we'll just have to try again in a few months!

But a "beast" scope sounds really good... right now our "beast scope" for the most demanding tasks is a mid-2000s MSO4000 (yep, the original), which is... not quite what I'd like us to have. But that's my problem, not the scope's!

[babryce]

You can improve anything that is a random error by averaging or low-pass filters etc. You cannot remove any systematic error from the signal path that way. That could be INL on a DC level or other things. The linearity of the converter is captured in the SFDR or THD numbers certainly. R&S talks about this in their white paper about their HD mode technique. You cannot exceed the number of actual bits of the converter for SFDR of course but none of the converters reach this limit. It is perfectly possible to have a noisy but totally linear 8-bit converter have better linearity than a 12-bit, though in general that wouldn't be the case.

SFDR were < 1-bit lower on the RTO in my limited testing than the S-Series. Both the RTO and S-Series are good. The HDO9k and WR8k are both nice too. The MDO4000 with the spectrum is a useful thing too. The 5-series strength on paper would either be the 8 channels or just having gobs of digital channels. There are probably situations where you want to look at more than 16-lines. They will eventually get the software sorted though if there are bad design choices like how to click that I wouldn't expect to see them change for a long time; just the crashing and adding features they didn't have time to have at the start, probably mostly as pay options.

For myself I never by new equipment, better value to buy last generation on the surplus market.

[srce]

Results:
20MHz BW, 1mV/div, 1MOhm input
Mean baseline DCVrms noise = 156 uV

20MHz BW, 1mV/div, 50Ohm input
Mean baseline Vrms noise = 180 uV

20MHz BW, 20mV/div, 1MOhm input
Mean baseline Vrms noise = 366 uV

20MHz BW, 20mV/div, 50Ohm input
Mean baseline Vrms noise = 288 uV

Thanks to you both. I think it's the AC RMS values from the screenshots that you want to promote though

Scope	BW	Vertical setting	Input	Noise AC RMS	ENOB limit based on noise
MSO56	20MHz	1mV/div	50 Ohm	70uVrms	5.2
MSO56	20MHz	1mV/div	1 MOhm	64.8uVrms	5.3
MSO56	20MHz	20mV/div	50 Ohm	102uVrms	9
MSO56	20MHz	20mV/div	1 MOhm	104uVrms	8.9
MSOS804A	20MHz	1mV/div	50 Ohm	48uVrms	5.4
MSOS804A	20MHz	1mV/div	1 MOhm	115uVrms	4.1
MSOS804A	20MHz	20mV/div	50 Ohm	260uVrms*	7.3
MSOS804A	20MHz	20mV/div	1 MOhm	139uVrms	8.2
RTO2014	20MHz	1mV/div	50 Ohm	33uVrms	6.3
RTO2014	20MHz	1mV/div	1 MOhm	29uVrms	6.4

* This looks a bit fishy to me, as it's far higher than the speced noise for the MSOS104A at 1GHz. (Which is 163uVrms). It does appear to be the MSOS804A the measurement was made on, from the 8.40 GHz at the top of the screen. But if the b/w is set to 20MHz, that shouldn't matter should it? Does one of the other b/w limiters need to be on too? For the 1MOhm screenshot, the 8.4GHz drops to 500MHz.

My MSOS204A arrived It's noisy and hot! Surprisingly it comes with Windows 10, rather than the advertised 7.

There are two b/w limiters. A global one, that be manually set in steps of 500MHz, and a per channel limiter, that can be set to 20MHz, 200MHz or a custom frequency. It seems to get the lowest noise on the 50Ohm path, you need to set both. (I'm guessing the global limit is pre-ADC and the per channel filter is just done digitally). It seems the global limit can't be adjusted manually in high res mode.

Scope	Sample rate	Resolution	BW Limit	Vertical setting	Input	Noise AC RMS	ENOB limit based on noise
MSOS204A	20GSa/s	10-bit	2GHz/20MHz	1mV/div	50 Ohm	41uVrms	5.8
MSOS204A	20GSa/s	10-bit	2GHz/20MHz	1mV/div	1 MOhm	119uVrms	4.2
MSOS204A	20GSa/s	10-bit	500MHz/20MHz	1mV/div	50 Ohm	40uVrms	5.8
MSOS204A	20GSa/s	10-bit	500MHz/20MHz	1mV/div	1 MOhm	118uVrms	4.3
MSOS204A	20GSa/s	10-bit	2GHz/20MHz	20mV/div	50 Ohm	148uVrms	8.3
MSOS204A	20GSa/s	10-bit	2GHz/20MHz	20mV/div	1 MOhm	145uVrms	8.3
MSOS204A	20GSa/s	10-bit	500MHz/20MHz	20mV/div	50 Ohm	87uVrms	9
MSOS204A	20GSa/s	10-bit	500MHz/20MHz	20mV/div	1 MOhm	147uVrms	8.3
MSOS204A	5GSa/s	11-bit	1.14GHz/20MHz	20mV/div	50 Ohm	194uVrms	7.9
MSOS204A	5GSa/s	11-bit	1.14GHz/20MHz	20mV/div	1 MOhm	203uVrms	7.8
MSOS204A	1.25GSa/s	13-bit	277MHz/20MHz	20mV/div	50 Ohm	104uVrms	8.8
MSOS204A	1.25GSa/s	13-bit	277MHz/20MHz	20mV/div	1 MOhm	153uVrms	8.2
MSOS204A	100MSa/s	13-bit	22.1MHz/20MHz	20mV/div	50 Ohm	62uVrms	9.5
MSOS204A	100MSa/s	13-bit	22.1MHz/20MHz	20mV/div	1 MOhm	101uVrms	8.8
MSOS204A	100MSa/s	13-bit	22.1MHz/20MHz	500mV/div	50 Ohm	2mVrms	9.2
MSOS204A	100MSa/s	13-bit	22.1MHz/20MHz	500mV/div	1 MOhm	2.5mVrms	8.8
MSOS204A	10MSa/s	13-bit	2.1MHz/20MHz	500mV/div	50 Ohm	985uVrms	10.2
MSOS204A	10MSa/s	13-bit	2.1MHz/20MHz	500mV/div	1 MOhm	1.14mVrms	10