Keysight 3000T and Lecroy WS3000 side by side

[JPortici]

As i wrote in another topic, lecroy is having some sort of summer sales on many products, including the WS3000..
Here in the lab we only have an old tek TPS and a picoscope 2000, untile a couple years ago there weren't high requirements and the tek was okay to use in veichles, the pico for basic protocol and waveform analysis (protocol logging is then offloaded to specific analyzers)
Since we can use a MODERN scope, i called both Keysight and Lecroy for a visit and a loaner... and here we have them, side by side.

I'm trying to do my best to evaluate them, there are lots of differences..
Hardware (analog side)
Scope (hardware interface and screen... note that the photo reflects what i am seeing)
Touch screen
Touch interface
Measurements
Responsiveness

I think i have already made the decision but i'm going to use them a couple of more days to see if i change my mind, but i don't think i will. I'll write on my thoughts and judgments but remember, they will be based on how i actually use the scope here in the job lab, may not be the same at home... or for you.

Any questions i can answer... feel free to ask

[lem_ix]

How responsive is the Lecroy against the probably benchmark Keysight? Also how does the built quality feel? Not that fond of Siglent.

[HighVoltage]

What is the main objective at your workplace? What kind of projects do you do? It looks like an automotive device on your bench?

[2N3055]

Hello,
In my work I have a need to look at decode from a large number of short bursts. For instance, every 2 seconds, a few packet burst on an SPI bus.. And for 2 hours like that..

For me it would be interesting how they handle decode from segmented memory. For example: can you see all decodes from separate segments in a common decode table (with timing marks that clearly show segments). Also what is comparison between in tools and procedure to search over segments based on decoded data, packet errors and such...

Regards,

Sinisa

[JPortici]

Three answers before i leave for the mountains :

@HighVoltage, you are corect, Aftermarket Automotive. The new scope will be used to reverse engineer parts and sensors and of course for new hardware development. to probe inside a car we will probably still use the Tek TPS a lot (battery operated, more forgiving inputs, believe it or not the best we have to look at slow signals, i LOVE scan mode. Not so much a fan of Roll)

The boards we make are not particulary high speed, due to low power contstraints, the only "high speed" part is SPI communication intra-board.. so high bandwidth scopes and probably most of lecroy's wave analysis tools would be kind of wasted with us.
First considerations: i wish, i really really wish that they both made lower bandwidth versions of these scopes: I'd be fine with 50/70 MHz, if i really need to look at higher speed signals integrity i can bring a boat anchor from home for a few days.

Too bad scope manufacturers still thinks like this: you want unusual protocols? you want histograms, waveform analysis? you must do high speed.

Unusual protocol we need: SENT.

The standard suggests 5-6 MICROSECONDS rise and fall times to contain EMI (but i've seen real world implementations with straight edges, eh.)
it is fairly slow.
Lecroy offers SENT trigger and decoding starting from the (discontinued) WS Mx44-S, 400 MHz, or the (in production) HDO4000, 200 MHz, which starts at about 10k list price in 4 channels. No need for the bandwidth and a price class that cannot be justified by us. 200 MHz to see a protocol that takes milliseconds to transmit a frame.. (end of rant)
i know, unusual means that if you need, you need to pay.. but Keysight offer this in option starting from the 3000T series and that has been a blessing.
Also, picoscopes have it available from the crappiest 2000 series, and we used that before.. but it has its drawbacks, such as inability to trigger from errors.

I had to trace an error in a device, the car wasn't happy with it. turns out, there was a sporadic error in the slow channel (hundreds of milliseconds to transmit). With the picoscope i would have never been sure to catch it because samplerate would drop to a point that the decode function wasn't guaranteed, with the keysight a quick set up and problem was found.
Could also build a quick and dirty hardware protocol decoder.. but i wasn't sure i could trust it.

Does the decoder always run at full samplerate in the keysight? it would appear so.. with SPI,I2C,SENT and CAN it would either DECODE correctly or not.
By the way, the same for LeCroy. Fantastic decoder, a bit difficult to set up at first, though.

@2N3055: I shall make some screenshots first!

@lem_ix: please, keep in mind that the main processor in the keysight is almost solely for the user interface, the interface in the keysight is always snammy because of it (and i love it). In the lecroy, considering also that it is an older machine (2014 vs 2015 i believe) the controls are not as fluid or snappy, at slower timebases (AND WITH DEEP MEMORY) it will be slower to use: rotate the encoder, the trace starts moving about 300-500 ms later. I believe this scope is to be set up and not touched, is not made for moving things around a lot

[Wuerstchenhund]

First considerations: i wish, i really really wish that they both made lower bandwidth versions of these scopes: I'd be fine with 50/70 MHz, if i really need to look at higher speed signals integrity i can bring a boat anchor from home for a few days.

Well, they could, but that doesn't mean the scope would necessarily be any cheaper than the currently lowest BW variant, it just means having to manufacture and support another SKU which is unlikely to catch many buyers.

Too bad scope manufacturers still thinks like this: you want unusual protocols? you want histograms, waveform analysis? you must do high speed.

Unfortunately in my experience they are usually right with this.

Unusual protocol we need: SENT.

The standard suggests 5-6 MICROSECONDS rise and fall times to contain EMI (but i've seen real world implementations with straight edges, eh.)
it is fairly slow.
Lecroy offers SENT trigger and decoding starting from the (discontinued) WS Mx44-S, 400 MHz, or the (in production) HDO4000, 200 MHz, which starts at about 10k list price in 4 channels.

SENT is also available in the WaveSurfer 10, although this too is a scope with larger BW (1GHz).

No need for the bandwidth and a price class that cannot be justified by us. 200 MHz to see a protocol that takes milliseconds to transmit a frame.. (end of rant)

I can only guess that there must have been little demand for SENT as otherwise I'm sure LeCroy would have offered it for the WS3000. I know that there are plans to extend its serial decode capabilities, however this is dependent on how much a feature has been asked for.

i know, unusual means that if you need, you need to pay.. but Keysight offer this in option starting from the 3000T series and that has been a blessing.

Indeed. I also often wished that the WS3k would offer more in terms of decoding (in my case it's just MIL-1553 not SENT), but there seems to be little overall demand.

@lem_ix: please, keep in mind that the main processor in the keysight is almost solely for the user interface, the interface in the keysight is always snammy because of it (and i love it). In the lecroy, considering also that it is an older machine (2014 vs 2015 i believe) the controls are not as fluid or snappy, at slower timebases (AND WITH DEEP MEMORY) it will be slower to use: rotate the encoder, the trace starts moving about 300-500 ms later. I believe this scope is to be set up and not touched, is not made for moving things around a lot

The Keysight is more fluid because it  is pretty much "cheating" in "long" memory (not sure that's a valid term for 4M in 2017  ) because in RUN mode it always uses short memory and only if you press STOP then it does the last acquisition in "long" memory. That's part of why InfiniVision scopes using the MegaZoom ASIC achieve their high waveform update rate.

The LeCroy will always use the amount of memory that you set it to use, and if you set it to use long memory (10Mpts) then it will take a certain amount of time to fill that memory at the rate the scope is sampling. This cannot be avoided. However, on the other side you have always full control over what the scope does (the DSOX Series is automatized in many areas, i.e. you have no control about memory use).

[electrolust]

I also often wished that the WS3k would offer more in terms of decoding

Too bad nobody offers a decoding API.

[Wuerstchenhund]

I also often wished that the WS3k would offer more in terms of decoding

Too bad nobody offers a decoding API.

Well, both Keysight and LeCroy offer a development option for their full-size Windows scopes (Keysight Infiniium; LeCroy WaveSurfer 10 and up) which allows you to intercept the raw sample data and do your own processing, which can be used to for decoding.

However, both scopes this thread is about are pretty much embedded platforms with very limited resources (the DSOX3kT runs on WinCE, the WS3k on Windows7 Embedded) so it's unlikely we'll ever see something similar on these scopes.

[JPortici]

about this, we'll see. daniel keeps saying he's pushing management to let them publish APIs for their megazoom scopes (see the 1000x video threads)
About memory i want to say it again os it is clear, the two other scopes we are using in the lab are a 2.5 kpts tektronix and a 48 kpts picoscope (the picoscope has at least a builtin "history" mode acquiring many hundreds acquisitions between each stop/run. 1137 in my case, that depends on the platform you are on (available ram basically, if i used the SDK instead of the main application it would be much higher.. with the SDK you like to live dangerously)

so you see, 1 MPts per channel (or even 500k) is plenty.

I also want to note that on the lecroy the setting is "max memory", at fast timebases it was using hundreds if not tens of points. Just as siglent scopes, it only acquires a window, nothing outside of it (please correct me if i'm wrong. I don't want to spread inaccuracies). Long memory is there for segments/history

[ralphrmartin]

Picoscope offer an API. I don't know whether it would do what you want, though...

[Wuerstchenhund]

about this, we'll see. daniel keeps saying he's pushing management to let them publish APIs for their megazoom scopes (see the 1000x video threads)

I know but I wouldn't hold my breath. The DSOX platform isn't really designed to be user expandable (at least if not hacked), so I doubt we'll see some internal APIs for InfiniVision scopes in the near future.

Actually, the only brand name embedded scope I am aware of that allows for user applications is the GW Instek GDS-2000E.

I also want to note that on the lecroy the setting is "max memory", at fast timebases it was using hundreds if not tens of points. Just as siglent scopes, it only acquires a window, nothing outside of it (please correct me if i'm wrong. I don't want to spread inaccuracies). Long memory is there for segments/history

The WS3000 will acquire what you set it up to acquire. By default it's just enough to fill the memory to keep the waveform update rates high, but if you set the memory size to something else it will acquire exactly that - and for every single acquisition. Which is useful for a lot more than just segmented memory, i.e. for glitch hunting via WaveScan. it can also help to reduce the blind time.

It's important to remember that both scopes you have there represent completely different design goals:

- Keysight's DSOX3kT is a very compact scope which has been optimized for very high update rates and which uses (not user interceptible) automatization for things like memory management, interpolation and so on to reduce the number of things a user has to worry about.

- The LeCroy WaveSurfer 3000 is, like all LeCroy X-Stream scopes, an instrument that has been designed for deeper analysis. You get better math/FFT and analysis tools and access to things like memory management, plus a larger screen. However, this also means that its up to the user to make the right settings for the specific task.

Which means the deciding factor here should be how *you* use the scope. If you spend most of the time quickly probing some signals and use the scope mostly in a visual way (i.e. a bit like an analog scope) then the Keysight is clearly a better option. If however you often find yourself having to drill further down into a signal then the toolset of the LeCroy would be better suited.

BTW: this is one reason why at work we just don't buy scopes from one vendor but from 2 1/2 (mostly Keysight and LeCroy, rarely R&S) 

[nctnico]

about this, we'll see. daniel keeps saying he's pushing management to let them publish APIs for their megazoom scopes (see the 1000x video threads)
About memory i want to say it again os it is clear, the two other scopes we are using in the lab are a 2.5 kpts tektronix and a 48 kpts picoscope (the picoscope has at least a builtin "history" mode acquiring many hundreds acquisitions between each stop/run. 1137 in my case, that depends on the platform you are on (available ram basically, if i used the SDK instead of the main application it would be much higher.. with the SDK you like to live dangerously)

so you see, 1 MPts per channel (or even 500k) is plenty.

I also want to note that on the lecroy the setting is "max memory", at fast timebases it was using hundreds if not tens of points. Just as siglent scopes, it only acquires a window, nothing outside of it (please correct me if i'm wrong. I don't want to spread inaccuracies). Long memory is there for segments/history

IMHO there has to be a setting which tells the scope to use all the memory for a normal acquisition mode. Also Keysight's 1Mpts (250kpts in real scenarios) is way too small. If you want to look at long traces (using segmented recording) then you will struggle for sure because the decoding needs a significant amount of oversampling to work. I used to own an DSO7000A series (8 Mpts) and it was barely doing the job when I needed deep memory. It wouldn't surprise me if it turns out the WS3000 has twice the memory per channel in segmented mode but check the datasheet and do the math. Many scopes use a double buffering system and depending on the marketing department they include that memory in the memory length on the badge (Keysight) or not.

[JPortici]

The WS3000 will acquire what you set it up to acquire. By default it's just enough to fill the memory to keep the waveform update rates high, but if you set the memory size to something else it will acquire exactly that - and for every single acquisition.

so what do i have to do? or does it acquire at the desired memory depth only in single acquisition or with wavescan activated? because the only option i could find for the memory is "max memory depth"

It's important to remember that both scopes you have there represent completely different design goals:

[...]

Which means the deciding factor here should be how *you* use the scope. If you spend most of the time quickly probing some signals and use the scope mostly in a visual way (i.e. a bit like an analog scope) then the Keysight is clearly a better option. If however you often find yourself having to drill further down into a signal then the toolset of the LeCroy would be better suited.

i think i said exactly that in the beginning, comparison is useless without the bigger picture, otherwise it's just marketing wank trying to make the other look bad.. this is NOT what i want to do.

[Wuerstchenhund]

The WS3000 will acquire what you set it up to acquire. By default it's just enough to fill the memory to keep the waveform update rates high, but if you set the memory size to something else it will acquire exactly that - and for every single acquisition.

so what do i have to do? or does it acquire at the desired memory depth only in single acquisition or with wavescan activated? because the only option i could find for the memory is "max memory depth"

That's it, Max Memory. This setting allows you to limit the used amount of sample memory (from the full 10Mpts), and it affects every acquisition, no matter what mode. So if you set Max Memory to 1Mpts then it will use 1Mpts on every acquisition, if you set it to 2M then it uses 2M and so on.

At least this is what I remember (I haven't spend much time with the WS3000 aside from having one for a day back then when it came out in 2014, also some things on these X-Stream Lite scopes work slightly different to normal X-Stream scopes I work with), however the User Manual (p.33) seems to agree.

That however means you have to think about what you want to do and setup the memory accordingly, which you don't have to (because you can't) on the Keysight as it will always give you the minimum required sample size during repetitive acquisitions and will only give you "full" memory (whatever this is, as the scope won't tell you and it's heavily dependent on what you do so you might end up with less than 1Mpts) on the last acquisition.