NEW Keysight HD3

[EEVblog]

Even single shot capturing the noise at 1M point, it simply will not save any more binary data than 125KB

[maxwell3e10]

Let's steer the discussion from subtle design flaws and unfinished software to finding what's so revolutionary about this device, or are the key capabilities not yet unveiled?

14bit ADC, half the noise, and way faster updating. I just tested it in comparison to the MXO doing noise measurments (yes, it less than half the noise) and the measurement update was over 7 times faster on the HD3.

So about that 14 bit ADC... I can easily see the quantization steps in the raw data plotted by egonotto. For 500uV/div I estimate the step size to be about 7 uV and 4 mV full range gives about 600 steps or about 10 bits. For 5 mV/div scale the step size is about 20 uV and 40 mV full range gives 2000 steps or 11 bits. So it seems both of these ranges use digital scaling. Perhaps on larger range it uses full 14 bits, we need to see it!

[SWB]

Even single shot capturing the noise at 1M point, it simply will not save any more binary data than 125KB

Does this help? (from the manual, starting on page 264):

• Max Length — Specifies that the maximum number of waveform data points be saved. See "Length Control" on page 265.
• Length — Sets the number of data points that will be output to the file. For more information, see "Length Control" on page 265.

[...]

Length Control

The Length control is available when saving data to CSV, ASCII XY, or BIN format files. It limits the maximum number of data points that can be saved.

When Max Length is enabled, the maximum number of waveform data points is saved.

The actual number of data points saved depends on displayed data and on these factors:

• Whether acquisitions are running. When stopped, data comes from the raw acquisition record. When running, data comes from a smaller running acquisition record.
• Whether the oscilloscope was stopped using [Stop] or [Single]. Running acquisitions use less memory to provide fast waveform update rates. Single acquisitions use full memory.
• The horizontal time/div (sweep speed) setting when stopped using [Stop]. At faster settings, fewer data points appear on the display.
• Whether segmented memory is on. Acquisition memory is divided by the number of segments.
• When saving to a CSV format file, the maximum number of data points is 1 Mpts.

When necessary, the Length control performs a "1 of n" decimation of the data. For example: if the Length is set to 1000, and you are displaying a record that is 5000 data points in length, four of each five data points will be decimated, creating an output file 1000 data points in length.

[EEVblog]

Even single shot capturing the noise at 1M point, it simply will not save any more binary data than 125KB

Does this help? (from the manual, starting on page 264):

• Max Length — Specifies that the maximum number of waveform data points be saved. See "Length Control" on page 265.
• Length — Sets the number of data points that will be output to the file. For more information, see "Length Control" on page 265.

[...]

Length Control

The Length control is available when saving data to CSV, ASCII XY, or BIN format files. It limits the maximum number of data points that can be saved.

When Max Length is enabled, the maximum number of waveform data points is saved.

The actual number of data points saved depends on displayed data and on these factors:

• Whether acquisitions are running. When stopped, data comes from the raw acquisition record. When running, data comes from a smaller running acquisition record.
• Whether the oscilloscope was stopped using [Stop] or [Single]. Running acquisitions use less memory to provide fast waveform update rates. Single acquisitions use full memory.
• The horizontal time/div (sweep speed) setting when stopped using [Stop]. At faster settings, fewer data points appear on the display.
• Whether segmented memory is on. Acquisition memory is divided by the number of segments.
• When saving to a CSV format file, the maximum number of data points is 1 Mpts.

When necessary, the Length control performs a "1 of n" decimation of the data. For example: if the Length is set to 1000, and you are displaying a record that is 5000 data points in length, four of each five data points will be decimated, creating an output file 1000 data points in length.

Doesn't work even with Max Length enabled.

[EvgenyG]

Can someone confirm that 1M inputs are 135Vrms max and not Cat rated?

[Someone]

So about that 14 bit ADC... I can easily see the quantization steps in the raw data plotted by egonotto. For 500uV/div I estimate the step size to be about 7 uV and 4 mV full range gives about 600 steps or about 10 bits. For 5 mV/div scale the step size is about 20 uV and 40 mV full range gives 2000 steps or 11 bits. So it seems both of these ranges use digital scaling. Perhaps on larger range it uses full 14 bits, we need to see it!

The hint is in the data sheet when the noise performance is identical from 2mV/div below, so just like the 3000X, MXR, and S series from HPAK:
https://www.eevblog.com/forum/testgear/new-keysight-infiniivision-hd3-oscilloscope/msg5609003/#msg5609003
On the upside the data outside the screen is usually retained!

[Anthocyanina]

and did you see this bit yet at the trigger controls? the spacing between slope and its light indicator, and mode and its light indicator isn't aligned at all with the buttons above. and the force trigger button also not centered with the knob, and not centered with the knob's text below. it looks unfortunately sloppy

THey didn't bother me. Obviously they had to move the button to fit the indicators next to those buttons.

It's not horrible, but it could be better!

a 2 minute attempt in paint(pretend they are aligned!). it may be a bit more crammed, but the distance between the slope and mode buttons and their light up indicators could be reduced, and the auto and default could be moved more toward the edge. force could be placed there, or centered near where it was originally, and a 50% round button placed between default and mode, or where i placed it. i don't know, my placement might be even worse, but that's 2 minutes in paint trying to at least get aligned buttons for free for something that will never happen. paid designers at keysight could have figured out something better for sure 

[2N3055]

The subtle awkwardness in the industrial design is striking when compared with the rest of the family.

Clearly something's missing that was either originally intended or still yet to come.

Also, even if they need(ed) to keep that space, they could have shifted the entire row to the left a bit to center it under the screen. That would have looked better. (However, they'd either have to move the power button somewhere else or increase the blank space to the left of the screen to make room for it, either of which would be unfortunate.)

For design cues,take a look at Infiniium EXR-Series and then you will see familiarity.
Infiniivision design was made many years ago, probably somebody else...

This scope is pretty much a small Infiniium EXR-Series by design.

[HighVoltage]

I wonder why the shape of the probe receptor is a square / rectangle and not the typical rounded shape that we know from the previous InfiniiVision scopes. Did Keysight release a new type of probes for this scope to fit this shape? It definitely looks weird to me.

[EEVblog]

Video shooting done, and I have 50min of edited footage!
I tried out most stuff, so there is lot in there hence the length.
Need to fiddle with the a bit more but should be able to render and release tonight.

[EEVblog]

So about that 14 bit ADC... I can easily see the quantization steps in the raw data plotted by egonotto. For 500uV/div I estimate the step size to be about 7 uV and 4 mV full range gives about 600 steps or about 10 bits. For 5 mV/div scale the step size is about 20 uV and 40 mV full range gives 2000 steps or 11 bits. So it seems both of these ranges use digital scaling. Perhaps on larger range it uses full 14 bits, we need to see it!

The hint is in the data sheet when the noise performance is identical from 2mV/div below

Correct, I demo this in my video, sofwtare scaling from 2mV. 100uV scale is coming in a firmware update.

[Neganur]

Another feature that's also quite unsettling is the AWG capability at 8k points. This is far less than my decade-old cheap Rigol AWG (which I sold recently, making it great value for money overall). On the MXO 4, it's 40M points. I needed at least a few hundred k points when I wanted to simulate a LoRa signal (at an intermediate frequency), and that was easily achieved on the MXO 4.

The 2-ch arb on the 4000 X-series is also just 8kpts, 100 MSa/s. Probably reused the design. Well, but nothing changed - who would buy the standalone waveform generators if the scope does it better.

EDIT: the 4000X Arb is 10bit, the HD3 has 14-bit 400MSa/s. I think that's alright.

[tszaboo]

Prices are on Mouser now:
https://eu.mouser.com/c/?marcom=164498230

By the looks of it you buy the 200MHz model with MSO as standard, and all bandwidth updates are options.

[nctnico]

and did you see this bit yet at the trigger controls? the spacing between slope and its light indicator, and mode and its light indicator isn't aligned at all with the buttons above. and the force trigger button also not centered with the knob, and not centered with the knob's text below. it looks unfortunately sloppy

THey didn't bother me. Obviously they had to move the button to fit the indicators next to those buttons.

It's not horrible, but it could be better!

a 2 minute attempt in paint(pretend they are aligned!). it may be a bit more crammed, but the distance between the slope and mode buttons and their light up indicators could be reduced, and the auto and default could be moved more toward the edge. force could be placed there, or centered near where it was originally, and a 50% round button placed between default and mode, or where i placed it. i don't know, my placement might be even worse, but that's 2 minutes in paint trying to at least get aligned buttons for free for something that will never happen. paid designers at keysight could have figured out something better for sure 

I disagree. Having a bit of room around the 'Force' and 'Mode'  buttons is a good thing to have as these buttons tend to get used often. And certainly you want to keep clear from the 'default' button as pushing this button can screw you up big time.

[jusaca]

Does serial decoding only work when enabled before the capture? Or is the new ASIC now able to decode an already captured signal?
Espescially interesting when serial settings are not known in advance and you want to play with the settings until the decoding makes sense.

[simone.pignatti]

Does serial decoding only work when enabled before the capture? Or is the new ASIC now able to decode an already captured signal?
Espescially interesting when serial settings are not known in advance and you want to play with the settings until the decoding makes sense.

Just made a try, it works only with DECODE ON.
When pressing DECODE after single shot it won't work.

[EEVblog]

Video shooting done, and I have 50min of edited footage!
I tried out most stuff, so there is lot in there hence the length.
Need to fiddle with the a bit more but should be able to render and release tonight.

Video complete, rendering now. 52min total, but I have all the time stamps for those only interested in certain things.
Best I could do 24 hours!
Can always do any 2nd channel video on specific things people want to see.
Sorry, didn't have time to systematically measure the update rate on every time base, nor the noise on every setting. I did shoot a brutal update rate comparison with the R&S MXO4 though.

[simone.pignatti]

Video shooting done, and I have 50min of edited footage!
I tried out most stuff, so there is lot in there hence the length.
Need to fiddle with the a bit more but should be able to render and release tonight.

Video complete, rendering now. 52min total, but I have all the time stamps for those only interested in certain things.
Best I could do 24 hours!
Can always do any 2nd channel video on specific things people want to see.
Sorry, didn't have time to systematically measure the update rate on every time base, nor the noise on every setting. I did shoot a brutal update rate comparison with the R&S MXO4 though.

Thanks!

[EEVblog]

00:00 - NEW Keysight MD304MSO 14bit ADC Oscilloscope with Megazoom V ASIC!
08:19 - Boot up
10:24 - New User Interface
14:12 - A few bugs
15:40 - 100MHz Signal Gen
16:28 - USB & Quick Action
18:13 - Intensity Graded Display DOWNGRADE?
19:30 - Ethernet remote and VNC
21:01 - Screen Scaling?
22:59 - Measurements
23:44 - Coming soon... X-Y & Roll
24:52 - 100uV/DIV ?
25:38 - Measurement on full data or screen?
27:34 - Trigger level range limitation, another coming soom feature
28:02 - Comparison to R&S MXO4, Half the noise, and 14bit vs 12bit
29:57 - Update rate comparison
30:44 - Noise measurement at 1V/DIV
31:17 - Custom ADC speculation
31:55 - BRUTAL Update rate comparison, Megazoom Magic
35:13 - Lack of bandwidth filters, maybe another coming soon option...
36:53 - 16 HD mode vs 18bit HD
37:10 - 14bit ADC TESTED
38:14 - Window docking & Tiling & Full Screen
39:08 - Zone triggering
40:31 - Serial decode
41:43 - NEW FEATURE: Fault Hunter
45:50 - 1GHz Passive Probe!
46:43 - Conclusion

[tv84]

00:00 - NEW Keysight MD304MSO 14bit ADC Oscilloscope with Megazoom V ASIC!

Did you reviewed a KS or a Tek?

[Someone]

45:50 - 1GHz Passive Probe

That doesn't come with the scope.

[2N3055]

00:00 - NEW Keysight MD304MSO 14bit ADC Oscilloscope with Megazoom V ASIC!

Did you reviewed a KS or a Tek?

HD, MD whatever... :-)

[ArdWar]

I actually like the UX more than I thought.

That 3 touchscreen click to access "Quick action" is hilarious tho...

[tggzzz]

Ah yes, the "high impedance" *10 1GHz passive probes.

Their impedance at 1GHz is actually j40 // 10M . Note the irrelevance of the 10M and the lack of zeros in 40 , due to the 4pF tip capacitance

For comparison my 1.5GHz *10 HP10020A passive probes (introduced in 1978!) are >j230 // 500 Note the significantly reduced loading due to the <0.7pF tip capacitance

(And yes, with content more than simple unboxing plus a ToC, I will skim yootoob videos )

[EEVblog]

That 3 touchscreen click to access "Quick action" is hilarious tho...

They need to fix that.