NEW Keysight HD3

[ArdWar]

It's a matter of preference I guess.

I like new(er) R&S the most, close by but still distinct with no random crap between/around them.
The one that put all V+H in one line is fast, but may need a bit of getting used to not accidentally grab the wrong control.
Tek MDO3 is bad because the control is too cluttered in general.
HPAK DSOX is alright I guess. The placement and distinction between clusters are fine. Horz at the top, Vert at the bottom, random crap in between. This HDO3 however went a bit too far.

All things considered, it probably will be just fine once you get used to. Most people will only use one scope at a time.

[egonotto]

Hello,

422 uVpp corresponds with 1 mV/div and 8 div and 14 bit ADC about 844 pp. With 12 bit you would have 211 pp. Does 14 bit ADC really help much?

Best regards
egonotto

[electr_peter]

http://www.jays.co.kr/12_Files/Debug%20Automotive%20Serial%20Buses%20Quickly.pdf
http://jays.co.kr/12_Files/How%20to%20Make%20Ripple%20and%20Noise%20Measurements.pdf

This is about the only interesting stuff from the appnotes :

See What You’ve Been Missing
4x the resolution and up to 10x less noise
The HD3 Series brings Keysight’s industry-leading capabilities from high-performance scopes to the high-volume level, making precision portable from 200 MHz to 1 GHz. Leveraging custom hardware technology from the UXR Series, the HD3 boasts the most impressive resolution on the market with 4x the vertical accuracy and up to 10x less noise than the competition. Paired with our fast, uncompromised waveform update rate and 25x more memory, the HD3 Series is truly set apart from other oscilloscopes in this class.

The HD3 Series oscilloscopes can update over 1.3 million waveforms per second with Keysight’s exclusive custom ASIC technology. HD3 oscilloscopes remain responsive even when capturing long waveforms while using automatic deep acquisition memory (100 Mpts per channel), which automotive serial bus applications often require.

Descriptions in app note imply that HD3 design philosophy is very similar to that of DSOX3/4/6k series - emphasis on waveform update speed, hardwave based measurements & decoding. IMO HD3 has a long waited "MegaZoom V" tech.

[nctnico]

Descriptions in app note imply that HD3 design philosophy is very similar to that of DSOX3/4/6k series - emphasis on waveform update speed, hardwave based measurements & decoding. IMO HD3 has a long waited "MegaZoom V" tech.

But this also means that you really need to set the scope to single shots to use full memory. Like on the old Megazoom scopes. With the deep memory it would be so much nicer if Keysight (finally) put the memory usage under user control instead of guessing what the user wants and being wrong. For a lot of measurements I do, I don't care about waveform update speed at all. As I typically have other ways to trigger the DUT to output a waveform, I want my scope to be on automatic trigger AND use full memory. It is just inconvenient having to setup both scope AND DUT to make a measurement.

[electr_peter]

Yes, it seems very like MegaZoom workflow - sample rate and memory set automatically to screen width while STOP/SINGLE gives full memory length. I am fine with that. Zoom mode allows forcing memory depth if you need more memory in RUN mode.

DSOX3000T/G has digitiser mode allowing manual sample rate and memory setting, but that disables most of the functions (it is intended for automated setups, not hand-ons work). I do not expect for HD3 to be much different in this aspect.
DSOX6000 can allocate more memory for measurement data buffer (but at slower update rate).

[EEVblog]

Yes, it seems very like MegaZoom workflow - sample rate and memory set automatically to screen width while STOP/SINGLE gives full memory length. I am fine with that. Zoom mode allows forcing memory depth if you need more memory in RUN mode.

Always been fine for me, you get the best of both worlds IMO.

[mikeselectricstuff]

Yes, it seems very like MegaZoom workflow - sample rate and memory set automatically to screen width while STOP/SINGLE gives full memory length. I am fine with that. Zoom mode allows forcing memory depth if you need more memory in RUN mode.

Always been fine for me, you get the best of both worlds IMO.

It would  sometimes be nice to have a repeated-single mode though - for infrequent events, the normal mode effectively wastes half the memory with unnecessary double-buffering.

[Martin72]

OK, the thing probably has 14 native bit resolution, with the corresponding principle “advantages”, how much of that remains to be seen.
What else does it have going for it, apart from its somewhat unfortunate appearance?

[nctnico]

Another interesting thing to find out is whether the HD3 uses actual data or decimated data for measurements.

[shabaz]

That photo of it is very unfortunate, as if the photographer was focussing on its feet, with the instrument towering above.

The style has an almost retro feel. Just needs the polished wood effect.

[EEVblog]

What else does it have going for it, apart from its somewhat unfortunate appearance?

It has new feature I'm interested in trying out.

[2N3055]

Another interesting thing to find out is whether the HD3 uses actual data or decimated data for measurements.

Probably decimated. Again, it has focus on speed.
And not only that, Megazoom IV  would take only single measurement of each type per trigger.
For instance, If you measure Rise Time, it would find first edge from the left and measure that one and that is it. So if you needed to make a statistics on that measurement, you would stretch signal with timebase so you only have one edge on the screen and get statistics by letting it run (trigger) repetitively for some time. That would also get best accuracy (because of decimation).
On my other scopes, if you have 10000 edges in a capture,  you get 10000 edge measurements from single capture and stats too.
Both ways work except in some edge cases where one or the other proves to be better.

We'll see how it is implemented on this new one..

[electr_peter]

Another interesting thing to find out is whether the HD3 uses actual data or decimated data for measurements.

Probably decimated. Again, it has focus on speed.

Agree on decimation. But if measurement memory buffer is increased from MegaZoom IV levels, that would be improvement making this decimation step almost the same as full memory in many cases.

And not only that, Megazoom IV  would take only single measurement of each type per trigger.
For instance, If you measure Rise Time, it would find first edge from the left and measure that one and that is it. So if you needed to make a statistics on that measurement, you would stretch signal with timebase so you only have one edge on the screen and get statistics by letting it run (trigger) repetitively for some time. That would also get best accuracy (because of decimation).
On my other scopes, if you have 10000 edges in a capture,  you get 10000 edge measurements from single capture and stats too.
Both ways work except in some edge cases where one or the other proves to be better.

Interesting point on one measurement per trigger. Sure, both methods work to get ~same results in the end.

To test these aspects on HD3, very detailed study of user manual is required and/or careful test setup is needed. As these are subtle points, they are not likely to be discussed on YouTube review.

[2N3055]

It all depends on decimation ratio, and how it is done. If you decimate down to 64k points (MZ IV used 32kpts) from 1 MPts buffer, or to same 64kpts from 20MPts buffer we could argue you lost even more data.

For the two methods results are not exactly the same because I can use method of "enlarging the feature to measure" as I have to do on MSOX3000T also on on Siglent SDS3000xHD and get even better results, but I cannot do the other way around.
By making individual measurements less dependent to specialized individual setup, you can measure more of them simultaneously and get good results for all of them.

With both ways you can characterize some signal and get good data. But ironically, with very fast (in user experience and WFMs/s) you are slower because you do measurements one by one and get stats by spending user time.
Other way of doing things gets you more good data in parallel. That is why that type of scope is called "analytic scope". As opposed to what I would call "interactive scope"...
First type excels when doing R&D and R.E. to get deep insight into signals. Second type is "service" and "tweak" type, when you are looking specific things, and is closer to workflow of CRT analog scopes..

Tool for the job.

As the time goes by the two types are starting to merge in the middle a bit.

[DaneLaw]

I'm not revealing much now by saying it's 10.1"

Looks like a 16:10 screen aspect ratio..
whats the resolution?   [1280×800, 1440×900, 1680×1050, 1920×1200, or 2560×1600] 

[electr_peter]

I'm not revealing much now by saying it's 10.1"

Looks like a 16:10 screen aspect ratio..
whats the resolution?   [1280×800, 1440×900, 1680×1050, 1920×1200, or 2560×1600] 

1280×800 with sharp waveforms on screen. Text and images are anti-aliased. Display can be split into multiple sections.

[Martin72]

It has new feature I'm interested in trying out.

And that would be?
Let me guess, you're not allowed to say...

[adam4521]

I wonder if it is the decoding features? The application note mentioned two channel time correlated decoding, in hardware.

That feels a bit niche, although will be important for someone. I wonder if there is something else? Being able to trigger on something bigger than one byte of message (cooler still, a filter or regexp expression) would be useful, if it were possible. Can triggers stay fast and have that level of expressive complexity?

[nctnico]

I wonder if it is the decoding features? The application note mentioned two channel time correlated decoding, in hardware.

That feels a bit niche, although will be important for someone. I wonder if there is something else? Being able to trigger on something bigger than one byte of message (cooler still, a filter or regexp expression) would be useful, if it were possible. Can triggers stay fast and have that level of expressive complexity?

This feature set is pretty much standard on every serious DSO. 2 decoding channels is a bit meager though. Typically you'd get at least 4 decoding channels nowadays.

[BillyO]

I'm sure it's a great scope, but..

It's got a face only it's mother could love..

And..

Even if I could justify the cost, being an amatuer Keysight would gladly remove any warranty for me.

So, albeit a little interesting, not really anything I'd ever get too excited about. 

[CRTbrain]

14-bits is a market ploy that many engineers accept.  THe noise levels aren't anywhere near the 14-bit level...so the scope is digitizing several bits of noise.  the claim still sells as 99% of customers never do the simple math.

[Martin72]

At some point in the future, it will be possible to advertise more obviously with ENOBs; at the moment, most manufacturers are still reluctant to do so.

[pdenisowski]

At some point in the future, it will be possible to advertise more obviously with ENOBs; at the moment, most manufacturers are still reluctant to do so.

We did listen and published an educational note on ENOB a couple of months ago, including some ENOB values for MXO4 and MXO5

https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_application/application_notes/1td12/1TD12_0e_Selecting_Your_Next_Scope_Understanding_Oscilloscope_ENOB_Values.pdf

The paper also addresses the question as to why manufacturers often don't publish ENOB values / the challenges in comparing reported ENOB values

[EEVblog]

At some point in the future, it will be possible to advertise more obviously with ENOBs; at the moment, most manufacturers are still reluctant to do so.

Keysight have published the ENOB figures for the HD3 for both 50 ohm and 1Mohm for 20MHz through to 1GHz (500MHz for 1M) using: "Normal Sample Mode 100mV/div, on a 10MHz 90% Full-screen Sine Wave"
As opposed to R&S figures using: "50 Ω path, 50 mV/division, 80% full screen, filters."

Obviously I can't tell you what they are right now, sorry.

[EEVblog]

14-bits is a market ploy that many engineers accept.  THe noise levels aren't anywhere near the 14-bit level...so the scope is digitizing several bits of noise.  the claim still sells as 99% of customers never do the simple math.

I have not tested it yet, but from screen shots I've seen, it can pull out signals it claims other scopes can't.